2024 A Practical Guide To MR Linac

A Practical
Guide
to MR-Linac
Technical Innovation and Clinical
Implication
Indra J. Das
Filippo Alongi
Poonam Yadav
Bharat B. Mittal
Editors
123
A Practical Guide to MR-Linac
Indra J. Das • Filippo Alongi
Poonam Yadav • Bharat B. Mittal
Editors
A Practical Guide
to MR-Linac
Technical Innovation and Clinical
Implication
Editors
Indra J. Das Filippo Alongi
Department of Radiation Oncology Department of Advanced Radiation
Northwestern Memorial Hospital, Oncology
Northwestern University Feinberg Ospedale Sacro Cuore Don Calabria
School of Medicine Negrar di Valpolicella, Verona, Italy
Chicago, IL, USA
Bharat B. Mittal
Poonam Yadav Department of Radiation Oncology
Department of Radiation Oncology Northwestern Memorial Hospital,
Northwestern Memorial Hospital, Northwestern University Feinberg
Northwestern University Feinberg School of Medicine
School of Medicine Chicago, IL, USA
Chicago, IL, USA
ISBN 978-3-031-48164-2 ISBN 978-3-031-48165-9 (eBook)

https://doi.org/10.1007/978-3-031-48165-9
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Switzerland AG 2024
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher,
whether the whole or part of the material is concerned, specifically the rights of translation,
reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any
other physical way, and transmission or information storage and retrieval, electronic adaptation,
computer software, or by similar or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are
exempt from the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in
this book are believed to be true and accurate at the date of publication. Neither the publisher nor
the authors or the editors give a warranty, expressed or implied, with respect to the material
contained herein or for any errors or omissions that may have been made. The publisher remains
neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Paper in this product is recyclable

My personal dedication goes to my late grandfather Jagdeo
Das who provided me dream of aspiration through education.
Additionally, I extend my loving gratitude to my family
(Sununta-my wife, Avanindra-son, Anita-daughter, and twin
grandchildren Avan and Aska who were born during inception
of this book). Their unwavering support and patience have
been invaluable, allowing me the space and time on this
book.—Indra J Das
Special thanks to all my patients. All cases, their feelings, the
fear, the trust are in my mind. All efforts to cure cancer remain
to me as experience for improving the next.—Filippo Alongi
A great many individuals deserve credit and recognition from
whom I have the opportunity to learn and grow. Especially, the
two who enriched my career in academia are Dr. Bhudatt
R. Paliwal, whose deep knowledge for therapeutic physics
helped me find innate delight in work and modesty, and Dr.
Bharat B Mittal, whose unprecedented guidance turned work
into an adventure.—Poonam Yadav
I’m grateful to my patients for their trust in me.—Bharat B
Mittal
Preface
Over the years, significant strides have been made in the field of radiation
oncology, with cutting-edge innovations driving improved precision treat-
ment that is improving the patient outcomes. One recent revolutionary
advancement that is transforming the landscape of radiation treatment is
MR-Linac, an amalgamation of magnetic resonance imaging (MRI) and lin-
ear accelerator (Linac) technology, which has joined forces to create a para-
digm shift in cancer care. This novel approach offers an unprecedented level
of precision, enabling real-time visualization of the tumor and surrounding
healthy tissues during treatment delivery. Unlike conventional radiotherapy,
where imaging is performed separately from treatment, MR-Linac technol-
ogy offers the unprecedented ability to capture high-resolution images of the
tumor and surrounding healthy tissues immediately before and even during
the radiation treatment process. This real-time imaging (MR-guided Radiation
therapy, MRgRT) capability has the potential to reshape the way treatment
can be planned and executed.
The utilization of imaging to optimize precision and accuracy throughout
the entire process of radiation therapy delivery has been called “image-guided
radiation therapy” (IGRT). By comparing current images to those obtained
during simulation procedures, the patient’s positioning and/or the geometry
of the radiation beams may be adjusted to more precisely address the radia-
tion dose to the target. Radiation therapy has long been guided by X-ray
images; in fact, historically, the portal films or, later on, electronic megavolt-
age images (obtained with electronic portal imaging devices, EPID) repre-
sented an early type of IGRT. Portal films and EPID images are two-dimensional
evaluations that, lacking 3-D visualization of soft tissue, provide limited
information with a poor image quality. The EPID images can indicate reason-
ably well the location of the isocenter and of the field margins relatively to the
bony landmarks. However, the treatment target could be a mobile soft tissue
mass within the body, and patient repositioning based on bony landmarks
alone is subject to significant uncertainties. Due to these inaccuracies, it is
inevitable to include a larger volume of normal tissue in the planning target
volume to ensure target coverage.
Different types of IGRT modalities are now available. The target location
may be identified by several methods for localization of surrogates, including
implanted fiducial markers, external surface markers, or anatomical features,
by planar imaging, fluoroscopy, kV or MV computed tomography (CT),
vii
viii Preface
ultrasound and X-ray imaging, electromagnetic localization, or optical

surface imaging. Previously, for precise treatment, weekly portal film was
developed and changes were made. However, in the late 1990s, the concept of
adaptive therapy was developed. Based on changes seen in imaging of the
treatment area, treatment shifts were made in the next fraction. This led to
adaptive therapy using daily CT simulation, CT-on rail, and cone beam CT. In
most of these efforts, imaging had time delays from several days up to a
week. However, a real-time adaptive approach is needed where tumors can be
visualized before treatment. The recent introduction of MR-Linac systems
into clinical practice can be interpreted as the latest step in this IGRT
sequence, opening new perspectives in radiation treatment delivery. The on-
board MR imaging allows clinicians to improve the quality of image guid-
ance compared to conventional CT-based IGRT, by increasing the soft tissue
contrast visualization of the anatomy without the addition of further radiation
doses. Thus, the advent of these hybrid machines may represent a game-
changer for the radiation oncology community.
During the development of this book, we collaborated with leading
experts, researchers, and clinicians who have been at the forefront of adopt-
ing and refining MR-Linac technology, especially MRgRT. Their invaluable
insights provide a comprehensive overview of the challenges and opportuni-
ties that lie ahead in harnessing the full potential of this transformative tech-
nology and are presented in various chapters in this book. Additionally, we
examine the evolving clinical evidence surrounding MR-Linac technology,
including its applications in various cancer types and treatment scenarios.
Furthermore, this book endeavors to address potential concerns and limi-
tations of MR-Linac technology. As with any innovative medical technology,
it is vital to assess the feasibility, safety, training requirements, and cost-
effectiveness of implementation. Our exploration offers a balanced perspec-
tive on the practical challenges faced by medical institutions and the ongoing
efforts to overcome them.
In conclusion, MR-Linac technology is a testament to human ingenuity
and our unwavering commitment to advancing patient care. As this transfor-
mative technology continues to mature, we are poised on the precipice of a
new era in radiation treatment, one where personalized, precise, and real-time
therapies become the norm rather than the exception. While the promise of
MR-guided radiation therapy is undeniable, we strive to offer a well-rounded
perspective on the potential obstacles faced by healthcare institutions and the
ongoing efforts to overcome them. It is expected that this book will support
the dissemination of essential MR-Linac clinical and technical knowledge.
Chicago, IL, USA Indra J. Das

Negrar di Valpolicella, Italy Filippo Alogi
Chicago, IL, USA Poonam Yadav
Chicago, IL, USA Bharat B. Mittal
Acknowledgment
A book like this A Practical Guide to MR-Linac: Technical Innovation and

Clinical Implication requires the support, encouragement, and contributions
from institution, collaborators, and many authors. Our gratitude extends to
them for their expertise in the field of MRI who shared their knowledge and
passion that is reflected in respective chapters.
ix
Contents
1 Introduction to MR-Linac�� 1

Bharat B. Mittal, Indra J. Das, Poonam Yadav,
and Filippo Alongi
2
Role of MRI in Radiation Oncology�� 5
Indra J. Das, Poonam Yadav, Filippo Alongi,
and Bharat B. Mittal
3
Clinical Necessity and Patient Selection in MR-Linac �� 21
Poonam Yadav, Filippo Alongi, Bharat B. Mittal,
and Indra J. Das
4
Education and Training in MR-Based Radiation Treatment�� 35
Luca Nicosia, Riccardo Filippo Borgese, Kerryn Brown,
and Filippo Alongi
5 Introduction to MR Sequences Used in Oncology with
An
Application to MR-Linac�� 45
Hersh Sagreiya, Karthik Sundaram, Jooyoung (James) Sohn,
and Poonam Yadav
6 Motion Management and Tracking�� 69
Elizabeth Huynh, Eenas A. Omari, Sara N. Lim,
and Raymond H. Mak
7
sCT and Dose Calculation �� 89
Emilie Alvarez-Andres, Fernanda Villegas, Anaïs Barateau,
and Charlotte Robert
8
Treatment Planning Considerations for an MR-Linac�� 123
Justine M. Cunningham, Jennifer L. Dolan, Kate Aldridge,
and Ergys Subashi
9
The First Low-Field MRI-Guided Radiation Therapy Hybrid
Integrated System: MRIdian�� 149
Poonam Yadav, Jooyoung (James) Sohn,
and Ahtesham U. Khan
10 Elekta Unity System�� 175
Jochem W. H. Wolthaus, Eenas A. Omari, Xinfeng Chen,
and Bram van Asselen
xi
xii Contents
11 The
Alberta Rotating Biplanar Linac-MR, a.k.a.,
Aurora-RT™�� 193
B. Gino Fallone, Satyapal Rathee, Nicola de Zanche,
Eugene Yip, Keith Wachowicz, and Jihyun Yun
12 Clinical
Applications of MR-Linac in Oligometastatic
Disease�� 217
Katie N. Lee, Martijn Intven, Filippo Alongi,
and Mai Anh Huynh
13 Management
of Pancreatic Cancer with MRI-Guided
Adaptive Radiotherapy�� 243
Ritchell van Dams, Sylvia S. W. Ng, John Hayes,
and Tarita O. Thomas
14 The
Role of MR-Guided Radiation Therapy
in the Management of Liver Malignancies�� 257
Nicki Karimi-Mostowfi, Luca Boldrini, Cihan Gani,
and Ann Raldow
15 Magnetic Resonance Imaging-Guided Radiation Therapy:
Advances and Opportunities for Thoracic Malignancies�� 273
Julius Weng, Jinzhong Yang, Raymond H. Mak,
Mohamed Abazeed, and Percy Lee
16 MRI-Guided
Radiation Therapy for Prostate Cancer:
Less Is More When Technology and Evidence Intersect�� 285
Luca Valle, Rosalyne Westley, Filippo Alongi, Sean Sachdev,
and Amar Kishan
17 Breast
Radiotherapy Using MR-Linac�� 307
Chiara De-Colle, Anna M. Kirby, Desiree van den Bongard,
and Maureen Groot Koerkamp
18 Considerations
for Using MR Linac for the Treatment
of Patients with Gynecologic Cancer:
A Practical Guide and Early Clinical Experience�� 323
Lorraine Portelance, Ina Jürgenliemk-Schulz, Kyle R. Padgett,
Rosa Patricia Castillo, and Astrid van Lier
19 MR-Guided
Radiotherapy in the Pediatric
and Adolescent Patient�� 355
Margaret M. Kozak, Sean Mahase, Bryan Traughber,
Mitchell Machtay, and John Buatti
20 MR-Linac-Guided
Adaptive Radiotherapy for Brain Tumors��375
Tugce Kutuk, Jay Detsky, Arjun Sahgal, and Rupesh Kotecha
21 Treatment
of Head and Neck Cancers with MR-Linac�� 395
Julia Khriguian, Laila Gharzai, Jolien Heukelom,
Brigid McDonald, and Clifton D. Fuller
Contents xiii
22
The Australian MRI-Linac: A 10-Year Journey�� 425
Gary Liney
23 Low-Field MR Imaging �� 433
Yingli Yang, Shaolei Li, Daniel A. Low, Zhimin Li,
and Peng Hu
24
Image-Based Biomarkers in Magnetic Resonance-Guided
Radiotherapy (MRgRT)�� 459
Andrea D’Aviero, Haidy Nasief, Charlotte Robert,
and Davide Cusumano
25 Epilogue and Outlook�� 469
Filippo Alongi, Bharat B. Mittal, Indra J. Das,
and Poonam Yadav
Index�� 473
Editors and Contributors
About the Editors
Indra J. Das, PhD is Director of Medical Physics, Professor and Vice Chair of Physics at
Northwestern Memorial Hospital and Northwestern University Feinberg School of
Medicine in Chicago. He is internationally acclaimed medical physicist with expertise in
radiation dosimetry, treatment planning, small fields, proton beam, and MR-Linac. His
profile can be viewed at: https://www.feinberg.northwestern.edu/sites/rad-oncology/fac-
ulty/profile.html?xid=48210.
Filippo Alongi, MD is Head of Advanced Radiation Oncology Department at IRCCS

Sacro Cuore Don Calabria Cancer Care Center of Negrar in Verona, Italy and full Professor
at the University of Brescia, Italy. He has been involved with 1.5 T MR-Linac over 7 years
and has written extensively. He is highly visible in European societies and on editorial
board of many journals. His profile can be viewed at:
Poonam Yadav, PhD is Co-director of MR-Linac program, Chief of Clinical Physics and
Associate Professor at Northwestern Memorial Hospital and Northwestern University
Feinberg School of Medicine in Chicago. She has been active with MR-Linac over its
inception and has published extensively on this technology. She is also expert on radiation
dosimetry and treatment planning. Her profile can be viewed at: https://www.feinberg.
northwestern.edu/sites/rad-oncology/faculty/profile.html?xid=52112.
Bharat B. Mittal, MD is Willian N. Brand Professor and Chair of Radiation Oncology at

Northwestern University Feinberg School of Medicine in Chicago. He specializes primar-
ily in Head and Neck and Lymphoma. He has keen interest in MR-Linac who made it pos-
sible to acquire the technology at Northwestern Memorial Hospital. He had been the chair
of the department for nearly 30 years and have published extensively on every aspect of the
radiation oncology. His profile can be viewed at: https://www.feinberg.northwestern.edu/
sites/rad-oncology/faculty/profile.html?xid=12553.
Contributors
Mohamed Abazeed Department of Radiation Oncology, Northwestern

University Feinberg School of Medicine, Chicago, IL, USA
Kate Aldridge Department of Radiation Oncology, Henry Ford Health,
Detroit, MI, USA
xv
xvi Editors and Contributors
Filippo Alongi University of Brescia and Advanced Radiation Oncology

Department, IRCCS Sacro Cuore Don Calabria Hospital, Negrar di
Valpolicella, Italy
Università di Brescia, Direttore del Dipartimento di Radioterapia Oncologica
Avanzata, Brescia, Italy
Department of Advanced Radiation Oncology, IRCCS Sacro Cuore Don
Calabria, Cancer Care Center, Negrar di Valpolicella, Italy
University of Brescia, Brescia, Italy
Advanced Radiation Oncology Department, Cancer Care Center, IRCCS
Sacro Cuore Don Calabria Hospital, Negrar di Valpolicella, Italy
Department of Advanced Radiation Oncology, IRCCS Ospedale Sacro Cuore
Don Calabria, Negrar di Valpolicella, Italy
Advanced Radiation Oncology Department, IRCCS Sacro Cuore Don
Calabria Hospital, Negrar di Valpolicella, Italy
Emilie Alvarez-Andres Department of Radiotherapy and Radiation Oncology,
Faculty of Medicine and University Hospital Carl Gustav Carus of Technische
Universität Dresden, Dresden, Germany
OncoRay—National Center for Radiation Research in Oncology, Faculty of
Medicine and University Hospital Carl Gustav Carus, Technische Universität
Dresden, Dresden, Germany
Helmholtz-Zentrum Dresden-Rossendorf, Dresden, Germany
Bram van Asselen Department of Radiotherapy, University Medical Center
Utrecht, Utrecht, the Netherlands
Anaïs Barateau Univ Rennes, CLCC Eugène Marquis, Inserm, LTSI -
UMR 1099, F-35000 , Rennes, France
Luca Boldrini Department of Bioimaging, Radiation Oncology and
Hematology, Fondazione Policlinico Universitario “Agostino Gemelli”
IRCCS, Rome, Italy
Desiree van den Bongard Department of Radiation Oncology, Amsterdam
University Medical Centers, Cancer Center Amsterdam, Cancer Treatment
and Quality of Life, Cancer Biology and Immunology, Amsterdam, The
Netherlands
Riccardo Filippo Borgese Department of Advanced Radiation Oncology,
IRCCS Sacro Cuore Don Calabria, Cancer Care Center, Negrar di Valpolicella,
Italy
Kerryn Brown Department of Radiation Oncology, ONJ Centre, Austin
Hospital, Heidelberg, VIC, Australia
John Buatti Department of Radiation Oncology, University of Iowa
Hospitals and Clinics, Iowa City, IA, USA
Editors and Contributors xvii
Rosa Patricia Castillo Radiology Department, Miller School of Medicine,

University of Miami, Miami, FL, USA
Xinfeng Chen Department of Radiation Oncology, Medical College of
Wisconsin, Milwaukee, WI, USA
Justine M. Cunningham Department of Radiation Oncology, Henry Ford
Health, Detroit, MI, USA
Davide Cusumano Fondazione Policlinico Universitario “Agostino
Gemelli” IRCCS, Rome, Italy
UO Fisica Medica e Radioprotezione, Mater Olbia Hospital, Olbia, Italy
Andrea D’Aviero UO Radioterapia Oncologica, Mater Olbia Hospital,
Olbia, Italy
Ritchell van Dams Department of Radiation Oncology, Dana-Farber
Brigham Cancer Center, Harvard Medical School, Boston, MA, USA
Indra J. Das Department of Radiation Oncology, Northwestern Memorial
Hospital, Northwestern University Feinberg School of Medicine, Chicago,
IL, USA
Chiara De-Colle Department of Radiation Oncology, University Hospital
and Medical Faculty, Eberhard Karls University Tübingen, Tübingen,
Germany
Jay Detsky Department of Radiation Oncology, Odette Cancer Centre,
Sunnybrook Health Sciences Centre, University of Toronto, Toronto, ON,
Canada
Jennifer L. Dolan Department of Radiation Oncology, Henry Ford Health,
Detroit, MI, USA
B. Gino Fallone Department of Medical Physics, Cross Cancer Institute,
Edmonton, AB, Canada
Department of Oncology, Medical Physics Division, University of Alberta,
MagnetTx Oncology Solutions, Edmonton, AB, Canada
Clifton D. Fuller Department of Radiation Oncology, The University of
Texas MD Anderson Cancer Center, Houston, TX, USA
Cihan Gani Department of Radiation Oncology, University Hospital
Tübingen, Tübungen, Germany
German Cancer Research Center (DKFZ) Heidelberg and German Consortium
for Translational Cancer Research (DKTK), Partner Site Tübingen, Tübingen,
Germany
Laila Gharzai Department of Radiation Oncology, Northwestern Memorial
IL, USA
xviii Editors and Contributors
John Hayes Department of Radiation Oncology, Northwestern Memorial

IL, USA
Jolien Heukelom Department of Radiation Oncology (Maastro), GROW
School for Oncology and Reproduction, Maastricht University Medical
Centre, Maastricht, the Netherlands
Peng Hu School of Biomedical Engineering, ShanghaiTech University,
Shanghai, China
Elizabeth Huynh Department of Physics and Engineering, London Regional
Cancer Program, London Health Sciences Centre, London, ON, Canada
Mai Anh Huynh Harvard Radiation Oncology Program, Brigham and
Women’s Hospital/Dana Farber Cancer Center/Massachusetts General
Hospital/Boston Children’s Hospital, Boston, MA, USA
Department of Radiation Oncology, Dana-Farber/Brigham and Women’s
Cancer Center and Harvard Medical School, Boston, MA, USA
Martijn Intven Division of Imaging and Oncology, Department of
Radiotherapy, University Medical Centre Utrecht, Utrecht, the Netherlands
Ina Jürgenliemk-Schulz Department of Radiation Oncology, University
Medical Center Utrecht, University Utrecht, Utrecht, the Netherlands
Nicki Karimi-Mostowfi Lewis Katz School of Medicine at Temple
University, Philadelphia, PA, USA
Ahtesham U. Khan Department of Radiation Oncology, Northwestern
Julia Khriguian Department of Radiation Oncology, The University of
Anna M. Kirby Department of Radiotherapy, Royal Marsden NHS
Foundation Trust and Institute of Cancer Research, Sutton, United Kingdom
Amar Kishan Department of Radiation Oncology, David Geffen School of
Medicine, University of California Los Angeles, Los Angeles, CA, USA
Maureen Groot Koerkamp Department of Radiation Oncology, University
Medical Centre Utrecht, Utrecht, the Netherlands
Rupesh Kotecha Department of Radiation Oncology, Miami Cancer
Institute, Baptist Health South Florida, Miami, FL, USA
Margaret M. Kozak Department of Radiation Oncology, University of
Iowa Hospitals and Clinics, Iowa City, IA, USA
Department of Radiation Oncology, UT Southwestern Medical Center,
Dallas, TX, USA
Tugce Kutuk Department of Radiation Oncology, Miami Cancer Institute,
Baptist Health South Florida, Miami, FL, USA
Editors and Contributors xix
Katie N. Lee Harvard Radiation Oncology Program, Brigham and Women’s

Hospital/Dana Farber Cancer Center/Massachusetts General Hospital/Boston
Children’s Hospital, Boston, MA, USA
Department of Radiation Oncology, Dana-Farber/Brigham and Women’s
Cancer Center and Harvard Medical School, Boston, MA, USA
Percy Lee Department of Radiation Oncology, City of Hope National
Medical Center, Los Angeles, CA, USA
Astrid van Lier Department of Radiation Oncology, University Medical
Center Utrecht, University Utrecht, Utrecht, the Netherlands
Sara N. Lim Department of Radiation Oncology, Northwestern Memorial
IL, USA
Gary Liney Sydney, NSW, Australia
Shaolei Li Institute for Medical Imaging Technology, Ruijin Hospital,
Shanghai Jiaotong University School of Medicine, Shanghai, China
Department of Radiology, Ruijin Hospital, Shanghai Jiaotong University,
School of Medicine, Shanghai, China
Zhimin Li Department of Radiology, Northwestern University, Chicago, IL,
USA
Daniel A. Low Department of Radiation Oncology, University of California,
Los Angeles, CA, USA
Mitchell Machtay Department of Radiation Oncology, Penn State Cancer
Institute, Hershey, PA, USA
Sean Mahase Department of Radiation Oncology, Northwestern University
Feinberg School of Medicine, Hershey, PA, USA
Raymond H. Mak Department of Radiation Oncology, Brigham and
Women’s Hospital, Dana-Farber Cancer Institute, Boston, MA, USA
Brigid McDonald Department of Radiation Oncology, The University of
Bharat B. Mittal Department of Radiation Oncology, Northwestern
Memorial Hospital, Northwestern University Feinberg School of Medicine,
Chicago, IL, USA
Haidy Nasief Department of Radiation Oncology, Medical College of
Sylvia S. W. Ng Department of Radiation Oncology, Sunnybrook Odette
Cancer Centre, University of Toronto, Toronto, ON, Canada
Luca Nicosia Department of Advanced Radiation Oncology, IRCCS Sacro
Cuore Don Calabria, Cancer Care Center, Negrar di Valpolicella, Italy
Eenas A. Omari Department of Radiation Oncology, Medical College of
xx Editors and Contributors
Kyle R. Padgett Radiation Oncology Department, Sylvester Comprehensive

Cancer Center, Miller School of Medicine, University of Miami, Miami, FL,
USA
Lorraine Portelance Radiation Oncology Department, Sylvester
Comprehensive Cancer Center, Miller School of Medicine, University of
Miami, Miami, FL, USA
Ann Raldow Department of Radiation Oncology, University of California,
Los Angeles, Los Angeles, CA, USA
Satyapal Rathee Department of Medical Physics, Cross Cancer Institute,
Charlotte Robert UMR 1030 Radiothérapie Moléculaire et Innovation
Thérapeutique, ImmunoRadAI, Université Paris-Saclay, Institut Gustave
Roussy, Inserm, Villejuif, France
Department of Radiation Oncology, Gustave Roussy, Villejuif, France
Sean Sachdev Department of Radiation Oncology, Northwestern University
Feinberg School of Medicine, Chicago, IL, USA
Hersh Sagreiya Division of Abdominal Imaging, Department of Radiology,
Hospital of the University of Pennsylvania, Philadelphia, PA, USA
Arjun Sahgal Department of Radiation Oncology, Odette Cancer Centre,
Sunnybrook Health Sciences Centre, University of Toronto, Toronto, ON,
Canada
Jooyoung (James) Sohn Department of Radiation Oncology, Northwestern
Ergys Subashi Department of Medical Physics, Memorial Sloan Kettering
Cancer Center, New York, NY, USA
Karthik Sundaram Division of Abdominal Imaging, Department of
Radiology, Hospital of the University of Pennsylvania, Philadelphia, PA,
USA
Tarita O. Thomas Department of Radiation Oncology, Northwestern
Memorial Hospital, Northwestern University Feinberg School of Medicine,
Chicago, IL, USA
Bryan Traughber Department of Radiation Oncology, Penn State Cancer
Institute, Hershey, PA, USA
Luca Valle Department of Radiation Oncology, David Geffen School of
Medicine, University of California Los Angeles, Los Angeles, CA, USA
Veterans Affairs Greater Los Angeles Healthcare System, Los Angeles, CA,
USA
Editors and Contributors xxi
Fernanda Villegas Radiotherapy Physics and Engineering, Medical Radiation

Physics and Nuclear Medicine, Karolinska University Hospital, Stockholm,
Sweden
Department of Oncology-Pathology, Karolinska Institute, Stockholm,
Sweden
Keith Wachowicz Department of Medical Physics, Cross Cancer Institute,
Julius Weng Division of Radiation Oncology, The University of Texas MD
Anderson Cancer Center, Houston, Texas, USA
Rosalyne Westley Clinical Oncology Academic Urology Unit, Institute of
Cancer Research, The Royal Marsden Hospital, London, UK
Jochem W. H. Wolthaus Department of Radiotherapy, University Medical
Center Utrecht, Utrecht, the Netherlands
Poonam Yadav Department of Radiation Oncology, Northwestern Memorial
IL, USA
Jinzhong Yang Division of Radiation Oncology, The University of Texas
MD Anderson Cancer Center, Houston, TX, USA
Yingli Yang Institute for Medical Imaging Technology, Ruijin Hospital,
Shanghai Jiaotong University School of Medicine, Shanghai, China
Department of Radiology, Ruijin Hospital, Shanghai Jiaotong University,
School of Medicine, Shanghai, China
Eugene Yip Department of Medical Physics, Cross Cancer Institute,
Jihyun Yun Department of Medical Physics, Cross Cancer Institute,
Nicola de Zanche Department of Medical Physics, Cross Cancer Institute,
Introduction to MR-Linac
1
Bharat B. Mittal , Indra J. Das , Poonam Yadav ,
and Filippo Alongi
Historical Perspective Becquerel in 1896 and Marie and Pierre Currie

in 1898.
Radiation therapy is an integral part of cancer Within a year after the discovery of X-rays,
treatment where a multimodality approach has Emil Herman Grubbe, a chemist and homeopath
proven to have the superior outcome. The from Chicago, claimed to have used X-rays to
advances in radiation oncology are intertwined treat a patient with breast cancer. However, the
with innovations in engineering, radiological and first verifiable priority claim can be credited to
physical sciences, and computer technology, two general practitioners, Tage Sjögren and Tor
along with a better understanding of cellular Stenbeck, who reported two successful cancer
interactions, the effects of radiation on biological treatments at the Swedish Society of Medicine in
tissues, and the roles genetics and epigenetics 1899 [2]. Since then, advances in radiologic sci-
play in determining radiation response. The dis- ences, both diagnostic and therapeutic, combined
coveries made over the past several decades have with innovations in computer science, have
been instrumental in the development of preci- played a critical role in the development of
sion radiotherapy through advances in technol- sophisticated technology to deliver radiation for
ogy and cancer biology. cancer treatment.
The use of radiation for the treatment of X-Ray tube technology to image in two
malignant tumors has experienced a technologi- dimensions (2-D) evolved into three-dimensional
cal revolution since the discovery of X-rays by (3-D) imaging with the invention of computed
the German physicist Wilhelm Conrad Röntgen tomography (CT) scans in 1972 by the British
on November 8, 1895 [1], and the subsequent engineer Godfrey Hounsfield and physicist Alan
discovery of radioactive elements by Henri Cermak of Tufts University [3, 4]. CT scanners
became widely available for diagnostic purposes
in 1980. In the 1990s, the use of CT scans was the
B. B. Mittal (*) · I. J. Das · P. Yadav first step in 3-D image reconstruction via a CT
Department of Radiation Oncology, Northwestern
simulator [5], and this spurred advances in radia-
Memorial Hospital, Northwestern University
Feinberg School of Medicine, Chicago, IL, USA tion planning and delivery which evolved into
e-mail: [email protected]; 4-D planning and delivery, taking tumor motion
[email protected]; [email protected] into account. This was the first step towards adap-
F. Alongi tive radiotherapy.
University of Brescia and Director of Advanced In 1969 Raymond Vahan Damadian, an
Radiation Oncology Department, IRCCS Sacro
American physician, hypothesized that benign
Cuore Don Calabria Hospital, Negrar-Verona, Italy
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 1

I. J. Das et al. (eds.), A Practical Guide to MR-Linac, https://doi.org/10.1007/978-3-031-48165-9_1
2 B. B. Mittal et al.
cells had longer relaxation times than malignant biomarkers that could potentially advance preci-
cells [6]. Further work by Paul Christian sion radiotherapy and outcome assessment using
Lauterbur, an American chemist, and Sir Peter image texture (radiomics). Integrating MRI tech-
Mansfield, an English physicist, resulted in the nology with linear accelerators is a paradigm
invention of the magnetic resonance imaging shift in delivering radiation based upon the ana-
(MRI) scanner, for which they received Nobel tomic and physiologic changes in a tumor to opti-
Prize in Medicine in 2003. In 1977, MRI was first mize therapeutic ratio.
used to scan the whole body.
Superficial X-ray technology to treat cancer
gave way to the use of deeply penetrative MR-Guided Radiation Therapy
radiation from radioactive elements like cobalt-60
or high-energy X-rays produced by linear accel- The first clinically active MR-guided radio-
erators. The cobalt-60 teletherapy machine was therapy machine was installed at Washington
invented in the 1950s, and in 1951 physicist University in St. Louis, and the first patient
Harold Johns, from the University of was treated in January 2014. The machine con-
Saskatchewan, became the first researcher to sisted of three cobalt-60 sources mounted on a
treat a patient with radiation emitted by cobalt-60. ring integrated with a 0.35-Tesla magnet with
In parallel, linear accelerator technology evolved. the same isocenter [13]. It included an adaptive
In 1928, R. Wideroe built the first research accel- treatment planning system to efficiently mod-
erator that emitted positive ions at about ify the treatment plan and deliver radiation
50 keV. In 1952, Henry Kaplan and Edward based upon MRI information. The next devel-
Ginzton began building a medical linear acceler- opment was integrating MRI with linear accel-
ator. In 1953, the first patient was treated using a erators (MR-Linac). In North America, the first
medical linear accelerator in London. In 1989, MR-Linac consisting of a 0.35-T magnet with
the first hospital-based particle therapy center a 6-MV photon linear accelerator received
became operational. Advances in hardware, elec- FDA approval in 2017 and was installed at
tronics, and computer technology quickly Henri Ford Hospital in Michigan. The first
resulted in intensity-modulated radiotherapy and patient was treated in July 2017. Simultaneously,
image-guided radiotherapy [7, 8]. A detailed his- a prototype system combining high-field MRI
torical perspective has been provided by with a linear accelerator was being developed
Rammohan et al. [9]. in Europe. The first such system was installed
The next evolution in adaptive radiotherapy at UMC Utrecht with the first patient with
was the integration of online MRI with external spine metastases treated in May 2017. This
beam radiotherapy [10–12]. This became possi- MR-Linac consisted of a 1.5-T magnet inte-
ble once the technical problem of interference grated with a ring-based gantry including a
between the magnetic field (which distorted the 7-MV standing wave linear accelerator [14]. In
accelerator) and the metals in the linear accelera- the last 5 years, there has been an exponential
tor (which distorted the magnetic field) was growth in the number of patients treated using
solved by active shielding of the magnet. The MR-Linac worldwide. There are two other
advantages of MRI over CT-based images include types of MR-Linac units under development,
better soft tissue definition, the ability to monitor one in Canada and the other in Sydney,
tumor motion, and real-time treatment adaptation Australia, that have been described in the lit-
taking interfraction and intrafraction changes in erature [12]. It is expected that MR-Linac tech-
the shape and location of tumors and surrounding nology will create a paradigm shift in the
organs into account. This technology will eventu- management of cancer that will result in
ally lead to MRI-based physiologic changes and improved therapeutic ratios, tumor dose esca-
1 Introduction to MR-Linac 3
lation, and improved tumor control, represent- References

ing a step towards personalized radiotherapy.
1. Roentgen W. Ueber eine neue Art von Strahlen. In:
Sitzungsberichte der Physikalisch-Medizinischen
Gesellschaft zu Wuerzburg, vol. 9; 1895. p. 132–41.
Outline of the Book 2. Mould RF. The early years of radiotherapy with
emphasis on x-ray and radium apparatus. Br J Radiol.
This book consists of 25 chapters divided into 1995;68:567–82.
three sections. The first 11 chapters are devoted 3. Hounsfield GN. Computerized transverse axial scan-
ning (tomography): part I. Description of system. Br J
to the various aspects of MR-Linac. Chapters 1–8 Radiol. 1973;46:1016–22.
describe clinical necessities, education and train- 4. Hounsfield GN. Nobel award address. Computed
ing, imaging sequences, motions and tracking, medical imaging. Med Phys. 1980;7:283–90.
synthetic CT, and treatment planning. The next 5. Coia LR, Schultheiss TE, Hanks GE. A practical
guide to CT simulation. Madison, WI: Advanced
three chapters, 9–11, are devoted to in-depth cov- Medical Publishing; 1995.
erage of three systems which are commercially 6. Damadian R. Tumor detection by nuclear magnetic
available, i.e., the ViewRay MRIdian system, the resonance. Science. 1971;171:1151–5.
Elekta Unity system, and the Alberta Aurora 7. Keall PJ, Sawant A, Berbeco RI, Booth JT, et al.
AAPM Task Group 264: the safe clinical implemen-
Magneton system. tation of MLC tracking in radiotherapy. Med Phys.
Chapters 12–21 cover disease-specific clinical 2021;48:e44–64.
implementation of MR-Linac without bias 8. Das IJ, Sanfilippo NJ, Fogliata A, Luca CL. Intensity
towards any specific vendor. The treatments cov- modulated radiation therapy: a clinical overview.
Bristol: IOP Publishing; 2020.
ered in this book include common types and sites 9. Rammohan N, Randall JW, Yadav P. History of
of cancer, such as oligometastatic, pancreas, technological advancements towards MR-Linac: the
liver, lungs, prostate, breast and gynecological future of image-guided radiotherapy. J Clin Med.
tumors, pediatrics, brain, and head and neck. The 2022;11:4730.
10. Hall WA, Paulson E, Li XA, Erickson B, et al.
chapters on clinical sites have common outlines, Magnetic resonance linear accelerator technology and
e.g., an overview of disease, need for MRI, con- adaptive radiation therapy: an overview for clinicians.
touring, dose-volume histogram (DVH) con- CA Cancer J Clin. 2022;72:34–56.
straints, dose escalation, online adaptation, 11. Das IJ, Yadav P, Mittal BB. Emergence of MR-Linac
in radiation oncology: successes and challenges
evaluation, and outcomes. Clinical chapters proof riding on the MRgRT bandwagon. J Clin Med.
vide unbiased views of the most common tech- 2022;11:5136.
nologies (Unity and MRIdian) wherever 12. Das IJ, McGee KP, Tyagi N, Wang H. Role and
applicable. future of MRI in radiation oncology. Br J Radiol.
2019;92:20180505.
At the end, the book covers the Sydney project 13. Mutic S, Dempsey JF. The ViewRay system: magnetic
MR-Linac (Chap. 22). There is significant inter- resonance-guided and controlled radiotherapy. Semin
est in low-field imaging for various disease sites, Radiat Oncol. 2014;24:196–9.
which is covered in Chap. 23. Chapter 24 covers 14. Raaymakers BW, Jürgenliemk-Schulz IM, Bol GH,
et al. First patients treated with a 1.5 T MRI-Linac:
artificial intelligence, radiomics, texture analysis, clinical proof of concept of a high-precision, high-
and biomarkers in the context of MRI. The final field MRI guided radiotherapy treatment. Phys Med
chapter provides an epilogue on the future of MR Biol. 2017;62(23):L41–50. Published 2017 Nov 14.
technology in the management of cancer. https://doi.org/10.1088/1361-6560/aa9517.
Role of MRI in Radiation Oncology
2
Indra J. Das, Poonam Yadav, Filippo Alongi,
and Bharat B. Mittal
Introduction location of the tumor is a critical determinant in

planning the radiation dose. Optimal treatment
The evolution of radiation oncology over planning is achieved by selecting adequate imag-
125 years has been surprisingly very positive, ing modalities to discern the tumor and OARs
with the ultimate goal of achieving improved sur- and by adding appropriate margins to optimally
vival and reduced normal tissue toxicity. More irradiate tumors. In early 2000, image-guided
and more patients are living longer, as shown in radiotherapy (IGRT) led the way in imaging and
the American Cancer Society (ACS) report that treating patients in a way that avoids radiation
cancer mortality has been dropping since 1990 in exposure to OARs. Imaging became a corner-
every disease site [1]. This success story is due to stone of radiation therapy for volume delineation
the multimodality approach to cancer manage- from pre-treatment verification to post-treatment
ment, in which radiation therapy plays an impor- evaluation.
tant role. In the early days of radiation therapy, it was
Radiation does not distinguish between tumor mandatory to do weekly portal filming to verify
and normal tissues; hence, it is the user who has patient setup and significant anatomical changes
to choose the proper tools with which to apply [2]. Eventually it was realized that a week is too
radiation. Multiple tools (planning and treat- long and a better system should be adapted,
ment) are used to optimize radiation so that it is which was the beginning of the era of adaptive
focused on the tumor. The goal in radiation ther- radiation therapy (ART) [3–5]. Standard clinical
apy is to maximize the dose to the tumor while practice involved compensating for changes or
minimizing the dose to normal tissues, also anatomical shifts in the following day’s treat-
termed organs at risk (OARs). In this context, the ment. Additionally, adaptive treatment was
undertaken with the help of daily computed
tomography (CT) scans [6, 7]. This led to CT-on-
I. J. Das (*) · P. Yadav · B. B. Mittal rail [8, 9] and then cone-beam CT [10–12] for
University Feinberg School of Medicine, localization, but in most cases, there was still a
Chicago, IL, USA time lag. Therefore, a technology with real-time
e-mail: [email protected]; visualization and adaptation was needed. Leaping
[email protected]; forward to the present day, this is possible with
[email protected]
magnetic resonance-guided linear accelerator
F. Alongi (MR-Linac) radiation therapy, which provides
Università di Brescia, Direttore del Dipartimento di
Radioterapia Oncologica Avanzata, Brescia, Italy visualization of tumors, minimizes registration

6 I. J. Das et al.
errors, and adapts treatment plans in real time in visualize the position of the tumor, accounting
a method called MR-guided radiotherapy for any movement or changes in the patient’s
(MRgRT). anatomy during treatment. This precise targeting
Since the advent of the X-ray, the question of of the tumor site minimizes damage to surround-
delineation of organs and tumor has dominated, ing healthy tissues and organs, reducing
and a balanced view is needed. In radiation treatment-related side effects.
oncology, magnetic resonance imaging (MRI) is MRI is not a perfect system and has many dif-
commonly used as a secondary imaging modality ficulties. One of the main problems with MRI is
to identify and define the extent of soft tissue that the images of air and bones are alike, and
tumors. MRI is a key imaging modality utilized differentiation is often difficult, especially when
in radiation oncology to guide treatment planning there is an air/bone interface. The MR signal is
and delivery. The images produced by MRI allow also not unique, as it varies with MR sequences
radiation oncologists to accurately target the that are dependent on the magnetic field strength.
tumor site, sparing healthy tissues and organs Unlike CT data where the Hounsfield unit (mea-
from unnecessary radiation exposure. In addi- sure of tissue attenuation) is only dependent on
tion, MRI provides important information about the kilovoltage of the machine [13], the MR sig-
the response of tumors to treatment, helping cli- nal is multifactorial and requires a good under-
nicians determine whether additional treatment is standing of the MR theory. In this context,
necessary. MRI is a noninvasive imaging tech- radiation treatment planning that is dependent on
nique that uses a magnetic field and radio waves the electron density cannot be easily performed
to produce detailed images of the internal struc- with MR images. This problem is being addressed
tures of the body. This imaging modality is par- using synthetic CT (sCT), which is discussed in
ticularly important in radiation oncology because greater detail in Chap. 7.
it provides high-quality, three-dimensional The use of sCT is growing and is an active
images of soft tissue that are not available with area of research. Most of the initial approaches
other imaging modalities such as CT or X-rays. have used an atlas- or bulk density-based Dixon
One of the advantages of MRI in radiation method, but now they are based on machine
oncology is its ability to differentiate between tis- learning and artificial intelligence (AI) as
sue types based on their relaxation time in a mag- described in various references [14–23]. In the
netic field. MRI tends to provide tissue bulk model, air, water, fat, and soft tissue and
characterization for anatomical and disease type. bone densities are mapped with the MR signal
This makes it possible to differentiate between and a correlation is used to create a sCT [14, 15].
normal and abnormal tissue and help delineate In general, soft tissue CT numbers do not change
tumor boundaries. This is achieved by selecting much against water, and hence, the dose calcula-
imaging sequences (see Chap. 5) by tuning the tion using the Dixon method usually provides
magnetic field and radiofrequency waves used satisfactory results [24, 25].
for imaging to provide the optimal view of the This chapter deals with MR-only approaches
tumor and surrounding tissues. The diagnostic and provides various aspect of CT and MR-Linac
capabilities are appealing, as it does not require technology in comparison.
ionization radiation exposure. This accurate
imaging is critical for treatment planning, help-
ing clinicians to determine the optimal radiation Margin
dose and delivery method to achieve the best pos-
sible outcomes for patients. Difficulty in delivering a tumoricidal dose is
MRI also plays an important role in IGRT, a compounded by uncertainty in location and posi-
highly precise radiation delivery technique that tion, and hence, the concept of margin is intro-
uses real-time imaging to accurately target the duced. It is also due to the notion that recurrences
tumor. During IGRT, MR images are used to could be avoided if large margins are given to the
2 Role of MRI in Radiation Oncology 7
target. A certain amount of margin is required to tumor volumes that could produce recurrence.
cover a tumor for radiation treatment; however, it For small margins, the relationship for extra vol-
is at the expense of an extra dose to the normal ume irradiated is a nearly linear function
tissues. The International Commission on (Fig. 2.1), but as margin increases, the functional
Radiation Units and Measurements, in publica- form becomes steeper. For example, for a 3-cm-
tions ICRU-50 and ICRU-62 [26, 27], provided radius tumor, a margin of 5 mm versus 2 mm
international guidelines for defining the target, adds 2.8-fold volume, whereas a 10-mm margin
gross tumor volume (GTV), clinical target vol- adds a 6.4-fold volume, thus increasing the prob-
ume (CTV), planning target volume (PTV), and ability of the complication rate.
OARs. Whereas the GTV is defined as gross The question is how to reduce the margin while
tumor volume, the CTV is based on subclinical minimizing the cost, e.g., not covering the target
disease, and the PTV accounts for setup and volume adequately and thus allowing recurrence.
organ motion. The latter two volumes have mar- This is a complex question and requires under-
gins with a significantly greater amount of irradi- standing of the available imaging tools. Current
ated tissue, which in turn may lead to more radiographic pre-post imaging is not suitable for
treatment-related complications. soft tissue contrast, and visibility is relatively
The simple calculation of volume and expan- poor; thus, discerning the tumor versus OARs is
sion is shown in Fig. 2.1. This provides a measure often difficult and the margin is hard to reduce.
of the extra volume treated for a hypothetical Additionally, the most current tools are passive
spherical tumor with radiation. The extra volume and cannot be used in real time, which is needed
is computed based on the PTV margin. Margin is due to intra- and inter-fraction target motion and
a safety net to make sure that the entire target is more so with organs which are susceptible to
treated and reduces the probability of missed physiological changes such as breathing and gas-
Fig. 2.1 Extra volume irradiated for a given tumor size and irradiation margin
8 I. J. Das et al.
tric and intestinal movements. ART and IGRT tissue complications and improving overall sur-
have evolved over 25 years with the help of fidu- vival. Pancreatic cancer, which has a universally
cial markers and better imaging tools, but still poor prognosis due to critical structures in the
lack real-time implementations. Better imaging vicinity, has been shown to have promising
and gating could help reduce the margin, which is results when treated with adaptive MR-Linac that
where MRgRT can play important role [28]. provides real-time imaging with a reduced mar-
Figure 2.1 is theoretical in nature, but has been gin [35, 36].
proven in clinical practice. To manage inter-
fractional setup error, large margins are advo-
cated. For example, in prostate cancer, CTV to Multimodality Imaging
PTV expansion varies from 5 mm to 15 mm
depending upon the institution. This margin X-Rays, MRI, positron emission tomography
could be significantly reduced, as shown in the (PET), and ultrasound are the basic imaging tools
MOMENTUM trial [29] using MR-Linac. that are used in clinical practice; here we com-
Christiansen et al. [30] analyzed the data using a pare and contrast imaging via X-rays and MRI.
normal tissue complication probability (NTCP)
model for standard (IMRT) and adaptive
MR-Linac treatment with a small margin (CTV X-Ray Imaging
to PTV expansion; were 3 mm R-L, 4 mm S-I,
and 5 mm A-P with additional 2-mm uniform The most common imaging method is X-ray,
margin and 5, 7, and 12 mm in intermediate- acquired in either planar (radiograph) or 3D
grade prostate cancer. In the standard arm the (CT). Planar images have been used since the
total margins from CTV were 12 mm, 14 mm, beginning of X-ray radiography, but CT data
and 19 mm in R-L, AP, and SI directions, respec- became available with the development of the CT
tively). It was shown that the incidence of com- scanner in 1972 [37, 38]. A revolutionary change
plications, including pain, incontinence, took place with the advent of digitally recon-
hematuria, rectal bleeding, proctitis, and GI tox- structed radiographs (DRRs) [39]. This allowed
icity, was significantly lower in the MR-Linac the visualization of the PTV and OARs on a
arm. Additionally, margin reduction was so effec- reconstructed image based on axial CT data. In
tive in reducing toxicity that the MIRAGE trial 1990, the voxel processor called a CT simulator
for prostate cancer was closed much earlier than became popular. It had three components: a CT
planned, as shown by Kishan et al. [31] where scanner, virtual simulation, and laser alignment
grade 2 toxicity was significantly lower with a [40]. The CT simulator created a paradigm shift
2-mm margin with MR-Linac compared to a in the way radiation oncologists view and delin-
4 mm with CT. Alongi et al. [32] analyzed the eate tumors and normal structures. However, CT
data from 100 prostate cancer patients treated images are based on the attenuation properties of
with MR-Linac for quality of life (QOL). They the tissue and do not provide soft tissue contrast.
evaluated a 5-mm uniform margin from the CTV Other imaging modalities are thus required for
and showed a significant reduction in toxicity real-time visualization of non-bone structures.
that was favorable for QOL. Additionally, X-rays are ionized radiation that
The margin reduction related to ART also has adds an extra dose to the entire body. A few addi-
clear dosimetric advantages in other disease sites tional features of X-irradiation should be
such as the liver, lung, pancreas, and lymph nodes considered:
[33]. Michalet et al. [34] showed that for the
treatment of the adrenal gland, the use of ART • Dose escalation is needed for ablative
with a smaller margin provided superior PTV treatment.
dose coverage and nearly 50% reduction in duo- • In general, the higher the dose, the higher the
denum and stomach dose, thus reducing normal probability of complications.
• Visualization in real time is needed to provide protons of the water, additional contrast can be
confidence in beam delivery. created, which is exploited clinically. Another
• Reduce margin to avoid OAR dose and related advantage of MRI is that one can create different
toxicities. images based on various relaxation times in terms
of T1 and T2, as has been described by Das et al.
[43]. Additionally, MRI parameters and sequenc-
Superiority of MRI ing that are useful in radiation oncology have
been described by Jacob et al. [44]. Soft tissue
With the advent of MRI in 1971 [41], various characterization by MRI is so superior that it is
types of images became popular as described by considered the gold standard imaging modality
Bitar et al. [42] and in Chap. 5 of this book. MRI for brain tumors and all soft tissue such as the
is based on the nuclear magnetic resonance of pancreas, liver, prostate, etc., which are discussed
water molecules in the different tissues of the in other chapters of this book. In a recent article,
body and their relaxation time in a magnetic field. Srinivasan et al. [45] described the usefulness of
It is used for clinical-level diagnosis and for pro- MRI in visualizing brain, prostate, and gastroin-
viding anatomical details. MRI contrast comes testinal sites.
from the differing water content in different tis- A few examples of the superiority of MRI are
sues. Also, with extrinsic additives such as gado- shown in the following figures. Figure 2.2 shows
linium chelates or iron oxide particles that can CT and MRI data for a prostate cancer patient in
change the relaxation properties of the mobile the axial and sagittal plane. The prostate is diffi-
a c
b d
Fig. 2.2 Axial and sagittal images of a prostate cancer patient with CT (a, b) and MRI (c, d). Please note the superior
soft tissue contrast in MRI, where visibility of prostate is much better (arrow indicates the prostate gland)
10 I. J. Das et al.
cult to delineate in CT images, whereas it is rela- of soft tissue contrast and tissue visualization is
tively easy in MRI due to soft tissue contrast. shown for a case of liver lesion (Fig. 2.3) with CT
When the magnetic field is increased, the soft tis- and MRI. Once again, the visibility of the lesion
sue visualization is even better. Another example is better in the MR images. Figure 2.4 shows a
a b
Fig. 2.3 Imaging of a liver cancer patient by CT (a) and MRI (b). The arrows show the location of the lesion
a c
b d
Fig. 2.4 A case of pancreatic cancer; (a, b) CT with clips and (c, d) MR images. Note that the lesion is not visible in
the CT image (arrows), whereas it is clearly visible in the MR images
case of pancreatic cancer. CT and MRI images are made of high-Z metal and have very high
are shown in the axial and sagittal planes and attenuation [46–50]. This creates problems in CT
indicate a superior soft tissue contrast with MR images due to photon starvation and streaking
images that can be used for visualization and artifacts (Fig. 2.5). It is very hard to delineate any
adaptive therapy which is not possible with CT structure with these artifacts, as the CT numbers
images and will be even worse in cone-beam CT are distorted and normal anatomic boundaries are
images. lost, leading to poor dose calculation [51–55].
MR images can improve the artifacts; however,
there is a certain amount of distortion depending
High-Z Artifacts on the prosthesis that have blooming effect: MR
effect as shown in Fig. 2.6. Additionally, dose
Prosthetic devices are common in the elderly calculations from MR images are still in infancy
population in whom the probability of cancer with prosthetic devices due to the uncertainty
incidence is also higher. Most prosthetic devices around electron density.
a b
Fig. 2.5 Image artifacts due to a dental filling. (a) CT, (b) T2-weighted MRI (adapted from Das et al. [56])
a b c
Fig. 2.6 Images of hip prosthesis by (a) CT and (b) T1-weighted and (c) T2-weighted MRI
12 I. J. Das et al.
MR-Only Approach radiation therapy to supplement CT data, mainly

for improved soft tissue definition in the treat-
With the wide array of MR sequences suitable for ment of the prostate, brain, liver, and pancreas.
characterizing tissues and measuring tumor vol- However, MR images require fusion and registra-
ume and treatment response, it is generally tion, processes that create additional uncertainty
agreed that MR offers a high signal-to-noise ratio in volume delineation. As we move to reduce
and contrast, with submillimeter resolution. This margin and rely on high-contrast and
contributes to better flexibility and greater utility high-resolution MR images, the need for MR
in MR images over CT that leads to improved tar- simulation has become a priority in radiation
get delineation, which is a critical component in oncology. In this context the MR simulator was
radiation oncology [57, 58]. Tenhunen et al. [59] introduced for the treatment of prostate cancer at
showed that MR can be used by itself for treat- Fox Chase Cancer Center with a 0.23T magnet
ment planning in 92% of prostate cancer patients, [68]. Similar efforts were also carried out by
and only 8% patients needed supplementary CT Kapanen et al. [69] to have MR-only treatment
data. planning for prostate cancer.
With better soft tissue contrast and MRgRT, it There has been significant momentum towards
is obvious that an MR-only option for patient the use of MR simulators, which has been
treatment could be chosen. There are many pub- described by AAPM Task Group No. 284 [70].
lications advocating an MR-only philosophy [40, Owing to the superiority of MR images over CT,
60–62]. Due to its appeal, many institutions are many institutions have adopted diagnostic MR
looking into installing an MR simulator, similar images in radiation oncology planning, but due to
in many ways to a CT simulator [40], that can time constraints, immobilization requirements,
provide DRRs. Early attempts at MR-based and selection of sequences in radiation oncology,
DRRs [63, 64] were relatively inferior but gave the necessity for a dedicated MR simulator sys-
way to the concept which is being explored with tem was realized. Glide-Hurst et al. [71] described
modern technology in conjunction with sCT (see an open 1.0T magnet for MR simulation that has
Chap. 7 in this book). The rosy picture of the very similar features to that of a CT simulator
MR-only approach is not shared by everyone, as such as a flatbed, a laser device, and incorpora-
indicated in a point/counterpoint article by tion of immobilization devices. Moore-Palhares
Bayouth et al. [65]. However, there is adequate et al. [57] described their first 5 years of experi-
sentiment that an MR-based approach will pro- ence with an MR simulator. Many vendors are
vide superior treatment outcomes, as it can miti- now providing MR simulators. Tyagi et al. [72]
gate inter- and intra-fractional motion in real time described the Philips MR simulator in detail.
[58]. Significant difficulties still exist in terms of full
implementation of MR simulators in clinical
practice, such as contouring, body size, bloom
MR Simulation artifacts, sCT issues, the air-bone interface, and
the training of staff [72–74].
Treatment simulation is a prerequisite for the Figure 2.7 provides images of two MR simu-
radiation therapy process and has been used since lators of different design and magnetic field
the first simulator was devised in 1966 in strength. The basic components of immobiliza-
Philadelphia [66]. Currently, every radiation tion devices, laser, and flatbed are similar to that
oncology department has a CT simulator that of a CT simulator. MR simulators have evolved
consists of a CT scanner, a virtual simulator for from 0.23T to 3T, so institutions can tailor their
DRR, and a laser marking device; this setup has needs for MR simulators depending on budget
been described in detail by AAPM Radiation and functionality. One caveat is that MR simula-
Therapy Committee Task Group No. 66 [67]. tors cannot be used in lieu of CT simulators,
Now there is increasing use of MR images in since MR simulation is only capable of covering
Fig. 2.7 MR simulators. (Left) Philips open high-field et al. [71]. (Right panel) 3T Philips Ingenia scanner. Note
MR simulator with (a, b) a laser system, (c) vertical field similar components to that of a CT simulator. Adapted
magnet, (d) MR laser, (e) body coil, (f) MR-compatible from Tyagi et al. [72]
flatbed, and (g) motor drive. Adapted from Glide-Hurst
approximately 1/3 of the total patient population tumor volume variability among clinicians is sig-
with soft tissues, as shown by Moore-Palhares nificant, even for regularly shaped tumors. It was
et al. [57]. noted that on average GTV by MRI was 69.7 cm3
compared to 59.5 cm3 by CT (a 17% difference, p
value <0.05). For the pancreas and liver, the vari-
Contouring ability between CT and MRI was 5–15 cm3 and
20–60 cm3, respectively [88].
Target volume and OAR delineation is a prereq- In the era of MR-Linac and MR simulation, it
uisite for radiation treatment planning. Accurate is expected that contour delineation variability
imaging with tissue contrast is required for tissue will be minimized, as anatomical structures can
segmentation and contouring. As we move be seen with a high degree of confidence. These
towards an MR-only approach for treatment developments will lead us into a new era of auto-
planning, it is imperative that volume delineation contouring, which is discussed in the following
be accurate. However, there is a large amount of section.
literature demonstrating that variability among
planners and modalities is significant. Since MRI
provides better soft tissue contrast, it was used in Autocontouring and Segmentation
prostate cancer volume delineation as early as
1990. However, the volumes drawn from CT and Segmentation and contouring have been highly
MRI are not identical, as shown by several inves- variable in most disease sites, and this is an issue
tigators who advocate a multimodality imaging for the radiation oncology community to resolve
approach [75–77]. In general, for prostate, CT [78–84, 89–93]. Thanks to advances in technol-
provides a 27% larger volume than MRI. The tar- ogy with machine learning and AI, most OARs
get volume variability between CT and MRI is can be efficiently contoured with a high degree of
also observed in other disease sites including the accuracy, as shown by numerous publications
brain, lung, breast, and head and neck, as shown [94–99]. Virtually, all radiation planning vendors
in the literature [78–87]. This is primarily due to are attempting to enter this market and provide
the better soft tissue contrast in MRI, and thus, autosegmentation software. We expect that within
volume delineation can be done accurately. In 5 years, autocontouring will be available on most
brain tumors, Weltens et al. [84] showed that CT simulators and treatment planning systems.
14 I. J. Das et al.
Automated image segmentation is essential in other system is very limited, as this cannot be
MR-Linac treatment, especially for adaptive handled by most non-cine treatment planning
therapy, in whose rapid workflow time is of the systems. MR-Linac vendors are also not prepared
essence. MR-based segmentation, especially in to transfer these images outside their domain.
MR-Linac, has challenges due to the choice of Kim et al. [107] provided an intuitional approach
magnetic field and imaging sequence. Fast et al. in managing the cine images especially for four-
[100] compared MR-based manual and autocon- dimensional MRI use in contouring, evaluation,
touring in locally advanced lung cancer. It was and motion management.
shown that interobserver variability in target At the present time, use of cine MR images in
delineation was significantly small when evalu- MR-Linac is more for motion management,
ated in terms of the Dice similarity Index and which is discussed in Chap. 6 of this book. One
Hausdorff distance. For autocontouring, Dice example of cine MRI is also shown in Fig. 2.8,
similarity was 0.95 and Hausdorff distance where the sagittal images show the tracking
<3 mm with p < 0.05. Such encouraging findings structure in red and the corresponding expansion
will lead to early adaptation of autocontouring. volume in yellow. When these two structures are
Liang et al. [101] provided a novel algorithm that in phase, the beam is turned on automatically in a
yields high accuracy in delineating the liver, the ViewRay system and manually in an Elekta Unity
kidney, and the spinal canal on T1-weighted MRI system. Details of motion management in the
acquired on-board of radiation therapy context of adaptive therapy have been described
platforms. by Winkel et al. [108] for the Unity system.
Role of Ciné Imaging Dose Calculation
One of the advantages of the MR signal is that it Dose calculation is dependent on the electron
carries motion information that can be used for density, which is acquired from the CT number.
any dynamic imaging. This has been exploited in However, when the MRI-only approach is used
cardiac and many other sites, where motion is an and inhomogeneity correction is needed, a totally
important aspect [102–104]. To complement different approach is required. This is performed
four-dimensional CT in liver cancer, Akino et al. using sCT, which is discussed in Chap. 7 of this
[105] proposed orthogonal ciné imaging for book. For soft tissues with minimal high-density
respiratory motion. They successfully presented material like the rectum and prostate, the differ-
data for internal target volume (ITV) and motion ence in dose between the MR-only approach and
management. Cine MRI is a special MR sequence CT is minimal. For the lung and rectum, Wang
with very fast data acquisition (milliseconds) to et al. [24, 25] showed that with simplified sCT
look for motion that can be viewed in any orthog- using the Dixon approach, the dose differences
onal plane. Cine MRI has been used mainly for are <3%. However, more advanced methods are
cardiac motion for 20 years as shown by Ho et al. needed to create accurate sCT data for accurate
[106]. Multiple images that are acquired in rapid dose calculations in the MR-only approach. This
succession are strung together like in cinema to is discussed in detail in Chaps. 7 and 8 in this
give the visual perception of motion. These book.
images are called frames. Cine imaging is now
exploited in MRIdian view ray system for
MRgRT. In the MRIdian system, 4 frames/sec or MR Education and Training
8 frames/sec are available that can be selected by
the user for motion visualization. The utility of Apart from the above issues, the MR-only
cine MRI images and their transportability to any approach to radiation therapy requires specific
a b
Fig. 2.8 (a) Cine image of a pancreas under treatment on expansion structure usually kept uniformaly to 3 mm. (b)
a ViewRay MRIdian system using 4 frames/sec. The red When red and yellow structures are in phase based on cine
contour is the tracking stucture (a subset of the lesion) MR images, the beam is turned on
being monitored for treatment. The yellow contour is the
training and education of staff including physi- 61% of the radiology residents did not feel com-
cians, physicists, dosimetrists, therapists, and fortable with MRI, due to lack of resident train-
others. Medical residents in the field of radiology ing. It is clear that MR simulation and MR-Linac
have a poor perception of MRI training, as shown will require robust training to improve knowl-
by Wald et al. [109]. As radiation oncology edge and understanding in this area.
becomes its own discipline and splits from radi- The role of MRI is growing globally, and
ology, MR training in radiation oncology is rela- global training and involvement in MR is urgently
tively scant. Education and training were a sought [112]. Upon the formation of an MR con-
significant barrier to the early adoption of MR sortium (Unity and ViewRay), Hall et al. [113]
simulators, as stated by Moore-Palhares et al. provided the views of members that training in
[57]. This topic is addressed in detail in Chap. 4 MR is urgently needed. They cited lack of train-
of this book, but a few facts are worth noting ing as one of the negative points in MR-Linac
here. implementation. Cahoon et al. [114] provided
Despite widespread use and exponential views from Australia on the training of technolo-
growth of MRI in radiation oncology, didactic gists in MR, especially MR simulation for radia-
education and training for radiation oncology tion oncology. It was felt that most training is
staff is lacking. The problem is that expertise in given by the vendors, whereas training would be
MR falls largely within radiology and its transla- more effective if provided by clinicians. The lack
tion to radiation oncology is very slow. This fact of comprehensive education in basic MR phys-
was felt by the AAPM, which decided to form an ics, MR safety, image acquisition, image inter-
ad hoc committee to explore the issue and pro- pretation, and knowledge of MR anatomy could
vide guidance (see McGee et al. [110]). Singer be a long-term problem if not resolved by
et al. [111] described survey data showing that improved training mechanisms.
16 I. J. Das et al.
Summary 10. Gayou O, Miften M. Commissioning and clini-

cal implementation of a mega-voltage cone beam
CT system for treatment localization. Med Phys.
MRI is a valuable tool in radiation oncology, pro- 2007;34:3183–92.
viding critical information to guide treatment 11. Gayou O, Miften M. Comparison of mega-voltage
planning and delivery. Its high-quality, three- cone-beam computed tomography prostate local-
ization with online ultrasound and fiducial markers
dimensional images allow for precise identifica- methods. Med Phys. 2007;35:531–8.
tion and delineation of tumor boundaries, aiding 12. Nielsen M, Bertelsen A, Westberg J, Jensen HR,
in the accurate delivery of radiation therapy. MRI et al. Cone beam CT evaluation of patient set-up
has significantly advanced the field of radiation accuracy as a QA tool. Acta Oncol. 2009;48:271–6.
13. Das IJ, Cheng CW, Cao M, Johnstone PAS. CT
oncology by improving treatment outcomes and imaging parameters for inhomogeneity correc-
reducing side effects with the aid of MR simulation in radiation treatment planning. J Med Phys.
tion and MR-Linac. Integration of education and 2016;41:1–11.
interdisciplinary training should allow this tech- 14. Cusumano D, Lenkowicz J, Votta C, Boldrini L,
et al. A deep learning approach to generate synthetic
nology to reach its full potential. CT in low field MR-guided adaptive radiotherapy
for abdominal and pelvic cases. Radiother Oncol.
2020;153:205–12.
References 15. Cusumano D, Placidi L, Teodoli S, Boldrini L,
et al. On the accuracy of bulk synthetic CT for MR-
guided online adaptive radiotherapy. Radiol Med.
1. ACS. American Cancer Society. Cancer facts and
2020;125:157–64.
figures 2023. https://www.cancer.org/content/dam/
16. Farjam R, Tyagi N, Deasy JO, Hunt MA. Dosimetric
cancer-o rg/research/cancer-f acts-a nd-s tatistics/
evaluation of an atlas-based synthetic CT generation
annual-cancer-facts-and-figures/2023/2023-cancer-
approach for MR-only radiotherapy of pelvis anat-
facts-and-figures.pdf. 2023.
omy. J Appl Clin Med Phys. 2019;20:101–9.
2. Byhardt RW, Cox JD, Hornburg LG. Weekly local-
17. Han X. MR-based synthetic CT generation using
ization films and detection of field placement errors.
a deep convolutional neural network method. Med
Int J Radiat Oncol Biol Phys. 1978;4:881–7.
Phys. 2017;44:1408–19.
3. Yan D, Vicini F, Wong J, Martinez A. Adaptive radi-
18. Kim J, Garbarino K, Schultz L, Levin K, et al. Dosi-
ation therapy. Phys Med Biol. 1997;42:123–32.
metric evaluation of synthetic CT relative to bulk
4. Yan D, Lockman D, Brabbins D, Tyburski L, et al.
density assignment-based magnetic resonance-only
An off-line strategy for constructing a patient-
approaches for prostate radiotherapy. Radiat Oncol.
specific planning target volume in adaptive treat-
2015;10:239.
ment process for prostate cancer. Int J Radiat Oncol
19. Lei Y, Harms J, Wang T, Tian S, et al. MRI-based
Biol Phys. 2000;48:289–302.
synthetic CT generation using semantic random
5. Jaffray DA. Image-guided radiotherapy: from cur-
forest with iterative refinement. Phys Med Biol.
rent concept to future perspectives. Nat Rev Clin
2019;64:085001.
Oncol. 2012;9:688–99.
20. Tyagi N, Fontenla S, Zhang J, Cloutier M, et al.
6. Lattanzi J, McNeeley S, Pinover W, Horwitz E, et al.
Dosimetric and workflow evaluation of first com-
A comparison of daily CT localization to a daily
mercial synthetic CT software for clinical use in pel-
ultrasound-based system in prostate cancer. Int J
vis. Phys Med Biol. 2017;62:2961–75.
Radiat Oncol Biol Phys. 1999;43:719–25.
21. Spadea MF, Maspero M, Zaffino P, Seco J. Deep
7. Lattanzi J, McNeely S, Hanlon A, Das I, et al. Daily
learning based synthetic-CT generation in radiother-
CT localization for correcting portal errors in the
apy and PET: a review. Med Phys. 2021;48:6537–66.
treatment of prostate cancer. Int J Radiat Oncol Biol
22. Spadea MF, Pileggi G, Zaffino P, Salome P,
Phys. 1998;41:1079–86.
et al. Deep convolution neural network (DCNN)
8. Wong JR, Grimm L, Uematsu M, Oren R, et al.
multiplane approach to synthetic CT generation from
Image-guided radiotherapy for prostate cancer by
MR images-application in brain proton therapy. Int J
CT-linear accelerator combination: prostate move-
Radiat Oncol Biol Phys. 2019;105:495–503.
ments and dosimetric considerations. Int J Radiat
23. Tang B, Wu F, Fu Y, Wang X, et al. Dosimetric evalu-
Oncol Biol Phys. 2005;61:561–9.
ation of synthetic CT image generated using a neural
9. Wong JR, Gao Z, Uematsu M, Merrick S, et al. Inter-
network for MR-only brain radiotherapy. J Appl Clin
fractional prostate shifts: review of 1870 computed
Med Phys. 2021;22:55–62.
tomography (CT) scans obtained during image-
24. Wang H, Chandarana H, Block KT, Vahle T, et al.
guided radiotherapy using CT-on-rails for the treat-
Dosimetric evaluation of synthetic CT for magnetic
ment of prostate cancer. Int J Radiat Oncol Biol
resonance-only based radiotherapy planning of lung
Phys. 2008;72:1396–401.
cancer. Radiat Oncol. 2017;12:108.
25. Wang H, Du K, Qu J, Chandarana H, et al. Dosi- 39. Das IJ, McGee KP, Desobrey GE. The digitally
metric evaluation of magnetic resonance generated reconstructed radiograph. In: Coia LR, Schultheiss
synthetic CT for radiation treatment of rectal cancer. TE, Hanks GE, editors. A practical guide to CT sim-
PLoS One. 2018;13:e019088. ulation. Madison, WI: Advanced Medical Publish-
26. ICRU Report 50. Prescribing, recording, and ing; 1995. p. 39–50.
reporting photon beam therapy. Bethesda, MD: 40. Coia LR, Schultheiss TE, Hanks GE. A practical
International Commission on Radiation Units and guide to CT simulation. Madison, WI: Advanced
Measurements; 1993. Medical Publishing; 1995.
27. ICRU Report 62. Prescribing, recording, and report- 41. Damadian R. Tumor detection by nuclear magnetic
ing photon beam therapy (supplement to ICRU resonance. Science. 1971;171:1151–5.
report 50). International Commission on Radiation 42. Bitar R, Leung G, Perng R, Tadros S, et al. MR pulse
Units and Measurements: Bethesda, MD; 1999. sequences: what every radiologist wants to know but
28. ICRU Report 97. MRI-guided radiation therapy is afraid to ask. Radiographics. 2006;26:513–37.
using MRI-linear accelerators. Bethesda, MD: Inter- 43. Das IJ, Sagreiya H, Yadav P, Allen BD. Basics of
national Commission on Radiation Units and Mea- MR imaging for the radiation oncologist. In: Ozyar
surements; 2022. E, Onal C, Hackett SL, editors. MR Linac radiother-
29. de Mol van Otterloo SR, Christodouleas JP, Blezer apy, a new personalized treatment approach. Lon-
ELA, Akhiat H, et al. The MOMENTUM study: don: Academic Press; 2022. p. 5–32.
an international registry for the evidence-based 44. Jacobs MA, Ibrahim TS, Ouwerkerk R. AAPM/
introduction of MR-guided adaptive therapy. Front RSNA physics tutorials for residents: MR imaging:
Oncol. 2020;10:1328. brief overview and emerging applications. Radio-
30. Christiansen RL, Dysager L, Hansen CR, Jensen graphics. 2007;27:1213–29.
HR, et al. Online adaptive radiotherapy potentially 45. Srinivasan S, Dasgupta A, Chatterjee A, Baheti A,
reduces toxicity for high-risk prostate cancer treat- et al. The promise of magnetic resonance imaging
ment. Radiother Oncol. 2022;167:165–71. in radiation oncology practice in the management
31. Kishan AU, Ma TM, Lamb JM, Casado M, et al. of brain, prostate, and GI malignancies. JCO Glob
Magnetic resonance imaging-guided vs computed Oncol. 2022;8:e2100366.
tomography-guided stereotactic body radiotherapy 46. Baissalov R, Sandison GA, Donnelly BJ, Saliken
for prostate cancer: the MIRAGE randomized clini- JC, et al. Suppression of high-density artefacts in
cal trial. JAMA Oncol. 2023;9:365–73. x-ray CT images using temporal digital subtraction
32. Alongi F, Rigo M, Figlia V, Nicosia L, et al. 1.5T with application to cryotherapy. Phys Med Biol.
MR-guided daily-adaptive SBRT for prostate can- 2000;45:N53–9.
cer: preliminary report of toxicity and quality of life 47. Wei J, Sandison GA, Hsi WC, Ringor M, et al. Dosi-
of the first 100 patients. J Pers Med. 2022;12:1982. metric impact of a CT metal artifact suppression
33. Nierer L, Eze C, da Silva MV, Braun J, et al. Dosi- algorithm for proton, electron and photon therapies.
metric benefit of MR-guided online adaptive radio- Phys Med Biol. 2006;51:5183–97.
therapy in different tumor entities: liver, lung, 48. Lewis M, Reid K, Toms AP. Reducing the effects of
abdominal lymph nodes, pancreas and prostate. metal artefact using high keV monoenergetic recon-
Radiat Oncol. 2022;17:53. struction of dual energy CT (DECT) in hip replace-
34. Michalet M, Bettaïeb O, Khalfi S, Ghorbel A, et al. ments. Skelet Radiol. 2013;42:275–82.
Stereotactic MR-guided radiotherapy for adrenal 49. Lell MM, Meyer E, Kuefner MA, May MS, et al.
gland metastases: first clinical results. J Clin Med. Normalized metal artifact reduction in head and
2022;12:291. neck computed tomography. Investig Radiol.
35. Chuong MD, Bryant J, Mittauer KE, Hall M, 2012;47:415–21.
et al. Ablative 5-fraction stereotactic magnetic 50. Andersson KM, Nowik P, Persliden J, Thunberg
resonance- guided radiation therapy with on-table P, et al. Metal artefact reduction in CT imaging
adaptive replanning and elective nodal irradiation of hip prostheses—an evaluation of commercial
for inoperable pancreas cancer. Pract Radiat Oncol. techniques provided by four vendors. Br J Radiol.
2021;11:134–47. 2015;88:20140473.
36. Chuong MD, Herrera R, Kaiser A, Rubens M, et al. 51. Ojala J, Kapanen M, Sipila P, Hyodynmaa S, et al.
Induction chemotherapy and ablative stereotactic The accuracy of Acuros XB algorithm for radiation
magnetic resonance image-guided adaptive radiation beams traversing a metallic hip implant - compari-
therapy for inoperable pancreas cancer. Front Oncol. son with measurements and Monte Carlo calcula-
2022;12:888462. tions. J Appl Clin Med Phys. 2014;15:162–76.
37. Hounsfield GN. Computerized transverse axial scan- 52. Higashigaito K, Angst F, Runge VM, Alkadhi H,
ning (tomography): part I. Description of system. Br et al. Metal artifact reduction in pelvic computed
J Radiol. 1973;46:1016–22. tomography with hip prostheses: comparison of vir-
38. Hounsfield GN. Nobel Award address. Computed tual monoenergetic extrapolations from dual-energy
medical imaging. Med Phys. 1980;7:283–90. computed tomography and an iterative metal artifact
18 I. J. Das et al.
reduction algorithm in a phantom study. Investig 68. Mah D, Steckner M, Palacio E, Mitra R, et al. Char-
Radiol. 2015;50:828–34. acteristics and quality assurance of a dedicated open
53. Mullins JP, Grams MP, Herman MG, Brinkmann 0.23 T MRI for radiation therapy simulation. Med
DH, et al. Treatment planning for metals using an Phys. 2002;29:2541–7.
extended CT number scale. J Appl Clin Med Phys. 69. Kapanen M, Collan J, Beule A, Seppälä T, et al.
2016;17:179–88. Commissioning of MRI-only based treatment plan-
54. Carrasco P, Jornet N, Duch MA, Panettieri V, et al. ning procedure for external beam radiotherapy of
Comparison of dose calculation algorithms in slab prostate. Magn Reson Med. 2013;70:127–35.
phantoms with cortical bone equivalent heterogene- 70. Glide-Hurst CK, Paulson ES, McGee K, Tyagi N,
ities. Med Phys. 2007;34:3323–33. et al. Task group 284 report: magnetic resonance
55. Yadav P, Chang SX, Cheng CW, DesRosiers CM, imaging simulation in radiotherapy: considerations
et al. Dosimetric evaluation of high-Z inhomogene- for clinical implementation, optimization, and qual-
ity used for hip prosthesis: a multi-institutional col- ity assurance. Med Phys. 2021;48:e636–70.
laborative study. Phys Med. 2022;95:148–55. 71. Glide-Hurst CK, Wen N, Hearshen D, Kim J, et al.
56. Das IJ, McGee KP, Tyagi N, Wang H. Role and Initial clinical experience with a radiation oncology
future of MRI in radiation oncology. Br J Radiol. dedicated open 1.0T MR-simulation. J Appl Clin
2019;92:20180505. Med Phys. 2015;16:5201.
57. Moore-Palhares D, Ho L, Lu L, Chugh B, et al. Clin- 72. Tyagi N, Fontenla S, Zelefsky M, Chong-Ton M,
ical implementation of magnetic resonance imaging et al. Clinical workflow for MR-only simulation and
simulation for radiation oncology planning: 5 year planning in prostate. Radiat Oncol. 2017;12:119.
experience. Radiat Oncol. 2023;18:27. 73. Tyagi N, Zelefsky MJ, Wibmer A, Zakian K, et al.
58. Otazo R, Lambin P, Pignol JP, Ladd ME, et al. Clinical experience and workflow challenges with
MRI-guided radiation therapy: an emerging paramagnetic resonance-only radiation therapy simula-
digm in adaptive radiation oncology. Radiology. tion and planning for prostate cancer. Phys Imaging
2021;298:248–60. Radiat Oncol. 2020;16:43–9.
59. Tenhunen M, Korhonen J, Kapanen M, Seppala T, 74. Cao M, Padgett KR, Rong Y. Are in-house diagnos-
et al. MRI-only based radiation therapy of prostate tic MR physicists necessary for clinical implementa-
cancer: workflow and early clinical experience. Acta tion of MRI guided radiotherapy? J Appl Clin Med
Oncol. 2018;28:1–6. Phys. 2017;18:6–9.
60. Pollard JM, Wen Z, Sadagopan R, Wang J, et al. The 75. Kagawa K, Lee WR, Schultheiss TE, Hunt MA,
future of image-guided radiotherapy will be MR et al. Initial clinical assessment of CT-MRI image
guided. Br J Radiol. 2017;90:20160667. fusion software in localization of the prostate for 3D
61. Chandarana H, Wang H, Tijssen RHN, Das conformal radiation therapy. Int J Radiat Oncol Biol
IJ. Emerging role of MRI in radiation therapy. J Phys. 1997;38:319–25.
Magn Reson Imaging. 2018;48:1468–78. 76. Roach M 3rd, Faillace-Akazawa P, Malfatti C, Hol-
62. Nousiainen K, Santurio GV, Lundahl N, Cronholm land J, et al. Prostate volumes defined by magnetic
R, et al. Evaluation of MRI-only based online adap- resonance imaging and computerized tomographic
tive radiotherapy of abdominal region on MR-linac. scans for three-dimensional conformal radiotherapy.
J Appl Clin Med Phys. 2023;24:e13838. Int J Radiat Oncol Biol Phys. 1996;35:1011–8.
63. Ramsey CR, Oliver AL. Magnetic resonance imag- 77. Guo L, Shen S, Harris E, Wang Z, et al. A tri-modality
ing based digitally reconstructed radiographs, virtual image fusion method for target delineation of brain
simulation, and three-dimensional treatment plan- tumors in radiotherapy. PLoS One. 2014;9:e112187.
ning for brain neoplasms. Med Phys. 1998;25:1928– 78. Basson L, Jarraya H, Escande A, Cordoba A, et al.
34. Chest magnetic resonance imaging decreases inter-
64. Ramsey CR, Arwood D, Scaperoth D, Oliver observer variability of gross target volume for lung
AL. Clinical application of digitally-reconstructed tumors. Front Oncol. 2019;9:690.
radiographs generated from magnetic resonance 79. Batumalai V, Burke S, Roach D, Lim K, et al. Impact
imaging for intracranial lesions. Int J Radiat Oncol of dosimetric differences between CT and MRI
Biol Phys. 1999;45:797–802. derived target volumes for external beam cervical
65. Bayouth JE, Low DA, Zaidi H. MRI-linac systems cancer radiotherapy. Br J Radiol. 2020;93:20190564.
will replace conventional IGRT systems within 15 80. den Hartogh MD, Philippens ME, van Dam IE,
years. Med Phys. 2019;46:3753–6. Kleynen CE, et al. MRI and CT imaging for preoper-
66. Kramer S, Kusner D, Gunn WG. Clinical experience ative target volume delineation in breast-conserving
with the Jefferson Hospital radiotherapy simulator. therapy. Radiat Oncol. 2014;9:63.
Radiology. 1966;87:134–6. 81. Lee E, Park W, Ahn SH, Cho JH, et al. Interobserver
67. Mutic S, Palta JR, Butker EK, Das IJ, et al. Qual- variation in target volume for salvage radiotherapy
ity assurance for computed-tomography simulators in recurrent prostate cancer patients after radical
and the computed-tomography-simulation process: prostatectomy using CT versus combined CT and
report of the AAPM Radiation Therapy Committee MRI: a multicenter study (KROG 13-11). Radiat
Task Group No. 66. Med Phys. 2003;30:2762–92. Oncol J. 2018;36:11–6.
82. White I, Hunt A, Bird T, Settatree S, et al. Interob- inter-observer variability in radiotherapy planning.
server variability in target volume delineation for CT/ Radiother Oncol. 2020;144:152–8.
MRI simulation and MRI-guided adaptive radiother- 95. Wong J, Huang V, Wells D, Giambattista J, et al.
apy in rectal cancer. Br J Radiol. 2021;94:20210350. Implementation of deep learning-based auto-
83. Zhang H, Fu C, Fan M, Lu L, et al. Reduction of segmentation for radiotherapy planning structures: a
inter-observer variability using MRI and CT fusion workflow study at two cancer centers. Radiat Oncol.
in delineating of primary tumor for radiother- 2021;16:101.
apy in lung cancer with atelectasis. Front Oncol. 96. Gooding MJ, Smith AJ, Tariq M, Aljabar P, et al.
2022;12:841771. Comparative evaluation of autocontouring in clini-
84. Weltens C, Menten J, Feron M, Bellon E, et al. cal practice: a practical method using the Turing test.
Interobserver variations in gross tumor volume Med Phys. 2018;45:5105–15.
delineation of brain tumors on computed tomog- 97. Loap P, De Marzi L, Kirov K, Servois V, et al. Devel-
raphy and impact of magnetic resonance imaging. opment of simplified auto-segmentable functional
Radiother Oncol. 2001;60:49–59. cardiac atlas. Pract Radiat Oncol. 2022;12:533–8.
85. Aoyama H, Shirato H, Nishioka T, Hashimoto S, 98. Rhee DJ, Akinfenwa CPA, Rigaud B, Jhingran A,
et al. Magnetic resonance imaging system for three- et al. Automatic contouring QA method using a deep
dimensional conformal radiotherapy and its impact learning-based autocontouring system. J Appl Clin
on gross tumor volume delineation of central ner- Med Phys. 2022;23:e13647.
vous system tumors. Int J Radiat Oncol Biol Phys. 99. Kawula M, Hadi I, Nierer L, Vagni M, et al. Patient-
2001;50:821–7. specific transfer learning for auto-segmentation in
86. Aslian H, Sadeghi M, Mahdavi SR, Babapour adaptive 0.35 T MRgRT of prostate cancer: a bi-
Mofrad F, et al. Magnetic resonance imaging-based centric evaluation. Med Phys. 2023;50:1573–85.
target volume delineation in radiation therapy treat- 100. Fast MF, Eiben B, Menten MJ, Wetscherek A, et al.
ment planning for brain tumors using localized Tumour auto-contouring on 2d cine MRI for locally
region-based active contour. Int J Radiat Oncol Biol advanced lung cancer: a comparative study. Radio-
Phys. 2013;87:195–201. ther Oncol. 2017;125:485–91.
87. Raman S, Chin L, Erler D, Atenafu EG, et al. Impact 101. Liang F, Qian P, Su KH, Baydoun A, et al. Abdomi-
of magnetic resonance imaging on gross tumor vol- nal, multi-organ, auto-contouring method for online
ume delineation in non-spine bony metastasis treated adaptive magnetic resonance guided radiotherapy:
with stereotactic body radiation therapy. Int J Radiat an intelligent, multi-level fusion approach. Artif
Oncol Biol Phys. 2018;102:735–43. Intell Med. 2018;90:34–41.
88. Dalah E, Moraru I, Paulson E, Erickson B, et al. 102. Feng L, Grimm R, Block KT, Chandarana H, et al.
Variability of target and normal structure delineation Golden-angle radial sparse parallel MRI: combi-
using multimodality imaging for radiation therapy nation of compressed sensing, parallel imaging,
of pancreatic cancer. Int J Radiat Oncol Biol Phys. and golden-angle radial sampling for fast and flex-
2014;89:633–40. ible dynamic volumetric MRI. Magn Reson Med.
89. Vorwerk H, Beckmann G, Bremer M, Degen M, 2014;72:707–17.
et al. The delineation of target volumes for radio- 103. Nepal P, Bagga B, Feng L, Chandarana H. Respira-
therapy of lung cancer patients. Radiother Oncol. tory motion management in abdominal MRI: radiol-
2009;91:455–60. ogy in training. Radiology. 2023;306:47–53.
90. Cazzaniga L, Marinoni M, Bossi A, Bianchi E, et al. 104. Curtis AD, Cheng HM. Primer and historical review
Interphysician variability in defining the planning on rapid cardiac CINE MRI. J Magn Reson Imaging.
target volume in the irradiation of prostate and semi- 2022;55:373–88.
nal vesicles. Radiother Oncol. 1998;28:293–6. 105. Akino Y, Oh RJ, Masai N, Shiomi H, et al. Evalu-
91. Tsang Y, Hoskin P, Spezi E, Landau D, et al. Assess- ation of potential internal target volume of liver
ment of contour variability in target volumes and tumors using cine-MRI. Med Phys. 2014;41:111704.
organs at risk in lung cancer radiotherapy. Tech 106. Ho VB, Foo TK. Impact of “Cine MR imaging:
Innov Patient Support Radiat Oncol. 2019;10:8–12. potential for the evaluation of cardiovascular func-
92. Unkelbach J, Bortfeld T, Cardenas CE, Gregoire V, tion”. AJR Am J Roentgenol. 2006;187:605–8.
et al. The role of computational methods for auto- 107. Kim T, Wu Y, Ji Z, Gach HM, et al. Technical note:
mating and improving clinical target volume defini- Institutional solution of clinical cine MRI for tumor
tion. Radiother Oncol. 2020;153:15–25. motion evaluation in radiotherapy. J Appl Clin Med
93. Das IJ, Compton JJ, Bajaj A, Johnstone PA. Intra- Phys. 2022;23:e13650.
and inter-physician variability in target volume 108. Winkel D, Bol GH, Kroon PS, van Asselen B, et al.
delineation in radiation therapy. J Radiat Res. Adaptive radiotherapy: the Elekta Unity MR-linac
2021;62:1083–9. concept. Clin Transl Radiat Oncol. 2019;18:54–9.
94. Wong J, Fong A, McVicar N, Smith S, et al. Com- 109. Wald C, Luchs J, Davila J, Lozano K, et al. Resi-
paring deep learning-based auto-segmentation of dents’ preceptions of MRI training in the United
organs at risk and clinical target volumes to expert States. J Am Coll Radiol. 2004;1:331–7.
20 I. J. Das et al.
110. McGee KP, Tyagi N, Bayouth JE, Cao M, et al. Find- status and the role of medical physicists. Phys Med.
ings of the AAPM Ad Hoc committee on magnetic 2022;103:46–58.
resonance imaging in radiation therapy: Unmet 113. Hall WA, Paulson ES, van der Heide UA, Fuller CD,
needs, opportunities, and recommendations. Med et al. The transformation of radiation oncology using
Phys. 2021;48:4523–31. real-time magnetic resonance guidance: a review.
111. Singer L, Rosenberg SA. The impact of MRI on Eur J Cancer. 2019;122:42–52.
radiation oncology graduate medical education. J 114. Cahoon G, Skehan K, Elwadia D, Rai R. The current
Am Coll Radiol. 2019;16:859–63. and future role of the MRI radiographer in radiation
112. Hasford F, Mumuni AN, Trauernicht C, Ige TA, et al. oncology: a collaborative, experiential reflection on
A review of MRI studies in Africa with special focus the Australian rollout of dedicated MRI simulators. J
on quantitative MRI: historical development, current Med Radiat Sci. 2023;70(Suppl 2):107–13.
Clinical Necessity and Patient
Selection in MR-Linac
3
Poonam Yadav, Filippo Alongi, Bharat B. Mittal,
and Indra J. Das
Introduction imaging technology to enhance tumor definition,

and minimize treatment uncertainties.
The monumental technological improvements in Even with modern technology such as onboard
radiation oncology have ushered in a new para- imaging, target and normal structures cannot be
digm in patient management. In recent times, clearly visualized with cone beam computed
radiation therapy has benefitted from significant tomography (CBCT), which has become a routine
technological advancements that have trans- process and standard of care. This is due to its
formed the field. These advancements, such as lack of soft tissue contrast and real-time imaging.
intensity-modulated radiation therapy (IMRT), CBCT is used before treatment to ensure that
adaptive radiation therapy (ART), and image- treatment is delivered as planned; however, it can-
guided radiation therapy (IGRT) [1–3], have not confirm if the position is accurate during treat-
allowed clinicians to administer precise and ment. Multiple image acquisitions with X-rays
highly conformal radiation doses. To achieve the also add radiation exposure and are of concern
necessary dose conformity and steeper dose gra- especially for long-term survivors [4]. Magnetic
dients used in modern radiation therapy to spare resonance imaging in combination with a linear
normal tissues, accurate localization and beam accelerator (MR-Linac) provides an opportunity
targeting techniques are critical. However, the to visualize the target and normal structures dur-
reproducibility of target positions during radia- ing treatment, continuously and in real time, for
tion therapy can be affected by the motion of optimal coverage. Additionally, beam gating has
internal organs during radiation and the displace- been used on an ad hoc basis to keep the target in
ment of internal organs between radiation ses- the field of view, but due to lack of continuous
sions. Fortunately, IGRT leverages modern monitoring capability, it has not been widely
applied. In contrast, in MR-Linac continuous
monitoring is automated and provides seamless
visualization without radiation exposure.
P. Yadav (*) · B. B. Mittal · I. J. Das
Northwestern Memorial Hospital, Northwestern
University Feinberg School of Medicine,
Chicago, IL, USA Onboard Imaging
e-mail: [email protected];
[email protected]; [email protected]
Modern radiation therapy machines are equipped
F. Alongi with onboard imaging systems. While these sys-
Università di Brescia, Direttore del Dipartimento di
Radioterapia Oncologica Avanzata, Brescia, Italy
tems have varying designs, the majority of them

22 P. Yadav et al.
share one common element: using a kV source imaging and digitally reconstructed radiographs
and a flat-panel imager. The purpose of these in prostate imaging [8]. Other studies have also
systems is to capture volumetric images with reported the geometric reliability of
enough detail to visualize soft tissues properly. It CBCT. However, there is a risk of unwanted radi-
is truly remarkable to see how far technology has ation being delivered to the tissues adjoining the
come to improve imaging. Imaging of soft tis- tumor due to the poor image quality of CBCT
sues is crucial for planning of radiation treat- imaging. Radiation exposure may also increase
ment. CT alone or along with other imaging the risk of radiation-induced cancer, as reported
techniques is routinely used for soft tissue delin- by some studies. The CBCT imaging dose also
eation for the planning of radiation therapy. depends on several attributes such as field of
There has been continuous improvement in the view, number of projections, and anatomical con-
sensitivity and specificity of CT-based systems. siderations. Additionally, skin doses within the
New contrast agents have improved the delinea- imaging field are often 2–5 cGy [2, 4], and mul-
tion of boundaries of tumors in CT. CT simula- tiple rounds of imaging will certainly increase
tion with three-dimensional (3D) planning is the this dose.
standard of care in most centers. Dedicated CT CT imaging and linear accelerators can be
simulators have now totally replaced conven- integrated in various ways. Helical tomotherapy
tional simulators. Jaffrey et al. [5] provided a is a Linac on a CT ring gantry. Megavoltage
rationale for IMRT and IGRT and how they will (MV) Linac beams can be used for MVCT scans
affect the outcome. In a similar manner Ling with the help of MV image detectors for recon-
et al. [6] provided an editorial on these technolo- structing CT projections. MVCT prevails over
gies, but lamented that the full potential has not planar MV portal radiographs due to improved
been achieved. soft tissue contrast and 3D projections and is
The kV CBCT helps target localization by comparable to planning CT. From a performance
acquiring 3D images that provide additional perspective, flat-panel detectors are suitable, and
details such as size, shape, and organs in proxim- a handful of rotations can scan an entire region of
ity. These details are vital for ART procedures, interest. However, increased scatter at the detec-
which rely on regular imaging and treatment tor surface impairs the image contrast, further
modification based on imaging results. ART is requiring imaging dose escalation or an increase
resource-intensive due to the need for frequent in slice thickness [9].
imaging, but the benefits are undeniable. One A CT scanner can also be installed on a tradi-
study reported the feasibility of dose reduction in tional C-arm Linac gantry, usually using elec-
organs at risk (OAR) and an average dose escala- tronic portal imaging systems. In a single rotation,
tion of >10% to the prostate when the treatment several CT slices can be obtained. However, the
plan was modified based on daily fractional scans patient’s daily radiation dose is significant and
[7]. Thus, ART has excellent potential to deliver has limited use. Future advances such as employ-
precise radiation doses as was advocated initially ing multirow detector systems, increasing the
by Yan et al. [3]. number of CT rows on CT scanners, and increas-
Apart from its success, CBCT suffers from ing data processing speed may help. CBCT on
various drawbacks such as low image quality, conventional accelerators has drawbacks, such as
blurred images mainly due to organ motion, and accommodating gantry sag and slower movement
artifacts due to scattering of X-rays. Despite of the C-arm gantries [10]. Figure 3.1 shows the
these constraints, CBCT is still used routinely. inferior image quality (low contrast) where struc-
Van Herk and colleagues reported good accuracy tures are hard to visualize even with optimum
of CBCT as compared with electronic portal techniques and proper filters.
3 Clinical Necessity and Patient Selection in MR-Linac 23
a b c
Fig. 3.1 Pretreatment imaging using CBCT of a head and neck (a), prostate (b), and liver (c) showing poor resolution
and contrast of images which make target alignment difficult
Fig. 3.2 Motion artifact of a 4DCT lung patient on axial, sagittal, and coronal views. Arrows show the motion
artifact
4D Computed Tomography and the radiation dose is administered during the
period of breath hold. Examples of breath-hold
External body movement and the motion of techniques include active breathing control
organs may not precisely correlate and may cause (ABC), deep inspiration breath hold, and self
inaccuracies in imaging. There might be varia- breath hold. ABC requires shallow breathing and
tions in amplitude proportions between fiducial tries to freeze the movement of the tumor for
and tumor displacement. These differences may short intervals of time [12, 13].
follow a pattern between cycles, amplitude ratios Respiratory gating methods monitor the natu-
between fiducial and tumor displacement may ral breathing cycle of patients. The radiation
vary from cycle to cycle, and thoracic and beam can be periodically turned on in a predeter-
abdominal points may involve relative phase mined phase of the patient’s breathing cycle.
displacement. Such considerations are essential Generally, the beam is turned off at the end of
for treating tumors in the vicinity of the lungs. inhalation or exhalation. These respiratory gating
The breathing cycle can produce significant methods help offset the motion of the patient.
motion artifacts for large organs such as the liver However, they require patient cooperation. The
due to temporal averaging [11]. CT images utilized for treatment should repre-
Additionally, respiratory motion may create sent the actual situation. It has been observed that
artifacts and cause errors during radiation treat- gating centered on the end of expiration can
ment. Breath-hold techniques are mainly used to increase reproducibility, since the post-expiration
counteract these respiratory motion artifacts. period may comprise up to 30% of the breathing
Different active and passive techniques are period, giving an adequate treatment time to
employed depending upon the tumor’s location, leverage delivery in this phase [14]. Figure 3.2
24 P. Yadav et al.
shows an example of motion-related challenges Many radiotherapy procedures are nowadays

and the adaptive nature of treatment in lung ste- MRI-assisted, where CT scans and MR images of
reotactic body radiation therapy. the tumor site are used in combination for radio-
therapy treatment planning. CT scans have
remained at the heart of treatment planning, not
agnetic Resonance Imaging
M just for image guidance but also as a prerequisite
Systems for electron density calculations, while MR scans
help in superior soft tissue isolation.
The characteristics of magnetic resonance imag- Uncertainties in gross tumor volume delinea-
ing (MRI) for the superior distinction of soft tis- tion have remained a key constraint in precision
sue for the abdomen, pelvis, and central nervous treatment delivery [16]. Many factors contribute
system are not serendipitous. They are frequently to variations in tumor delineation, such as target
used for diagnostic and tumor segmentation pur- motion, setup errors, and interobserver variabil-
poses. Refining various imaging processes ity, but tumor visibility remains one of the top
makes it intuitive to explore MRI-based treat- concerns. Several studies from 2000 to 2010 have
ment planning simulations for radiotherapy pro- characterized the role of tumor delineation in
cedures. In the presence of a magnetic field, the compromising treatment delivery [8, 17–20].
bulk magnetization of the nuclei in a given voxel In contemporary high dose rate procedures,
through the radiofrequency coils is quantified dose escalation to the tumor and simultaneous
during the MRI scan. A fusion of MRI and CT sparing of normal tissues are required for better
scans is evaluated to understand the areas of locoregional control and to reduce the probability
interest. The diagnostic use and understanding of recurrence. With constrained tumor visibility,
of MRI have gone through a series of techno- adequately delineating the regions of interest
logical developments in the past four decades. becomes impossible, and the associated uncer-
Until the last decade, most of the research on tainties hinder steep dose gradients needed for
improving MRI image quality was directed optimal tumor dose and better OAR sparing.
towards increasing the strength of the magnetic With reduced tumor visibility, the delineation
field, as is recognized by the installation of 3T protocols cannot be aggressively set up, and sys-
and higher field strength scanners in clinics. tematic variations are found across treatment
Organ motion, especially breathing- induced, planning. The resulting compromise in treatment
abdominal, heartbeat-related, and temporal, has delivery has been adversely associated with
always remained a concern for clinicians, and tumor control probability [16, 21].
fast cine MRI has evolved as a prevalent solution Inter- and intraobserver variability also has a
for such situations. negative effect on treatment plan quality and is
The success of MRI with soft tissues and ner- especially noticeable in sites involving soft tissue
vous systems has also encouraged the exploration anatomy and deformable organs. If MR scans are
of special MRI procedures such as functional utilized for treatment planning, such observer-
MRI, cardiac MRI, MR venography, non-contrast related discrepancies can be significantly
and contrast MRI, MR angiography, MR spec- reduced, mainly due to enhanced soft tissue con-
troscopy, and open-field MRI. Functional MRI trast [18, 22]. Additionally, imaging is crucial in
has gained popularity by detecting small and brain tumor management. MRI can recognize
measurable signal changes in the brain and apply- and characterize edema [23] and provides accu-
ing diffusion tensor imaging techniques, which rate contouring of brain tumors compared to CT
measure diffusion in different directions with the images [24]. Functional MRI can be used to eval-
help of fractional anisotropy. A functional map of uate the radiotherapy treatment response. MR-CT
the brain captures the dynamic neural activity registration, i.e., mapping of one set of images to
and allows for improved diagnosis of brain the other, can successfully increase the dose to
tumors [15]. the target and reduce radiation-related severe tox-
icity due to increased accuracy in the localization Early MR-based IGRT procedures relied on
of gross tumor volume and neighboring struc- tumor and region of interest definition via the
tures [25–27]. For oropharyngeal cancer radio- assistance of appropriate image registration algo-
therapy, MRI scans can be acquired in the rithms porting the structures from MRI to CT
treatment setup position with the help of the same images. MR-CT co-registration can lead to geo-
immobilization device. Such setup reproducibil- metric uncertainties in the treatment planning
ity is beneficial in reducing MR-CT registration process, and studies have reported geometrical
errors and can improve target definition, dose uncertainties of approximately 2 mm for the
coverage, and dose homogeneity. brain and up to 3 mm for the prostate and gyne-
Multiparametric MRI scans have been found cological treatment sites [38, 39]. Though the
to be useful in prostate cancer. MR-generated MRI setup taken in the treatment position helps
anatomical and functional imaging has increased reduce discrepancies, there are still geometric
accuracy in segmenting subvolumes of the misalignments throughout the process of CT-MR
prostate and maintained the functionality of sen- co-registration, which raises concerns regarding
sitive tissues in the apical posterior region, an geometric miss [8].
area of relatively high incidence and hence one Though CT remains the standard for contem-
for which it is critical to delineate such volumes porary radiotherapy, parallel MRI-based co-
precisely [28–30]. Haider et al. reported that registration has gained significant attention,
functional MRI is useful in the treatment plan- especially for sites involving soft tissue anatomy.
ning of recurrent prostate cancer [31]. Quality of It has paved the way for research on incorporat-
life improvements have been reported with ing MRI simulators in radiotherapy treatment
MR-based registration, mainly due to conformal planning to address the geometrical mismatch in
coverage [32]. Pötter et al. reported that survival the co-registration of images from the two modal-
gains of up to 20% could be achieved by dose ities. Recent research on MRI-based simulations
escalation in cervical cancer cases with a simul- has shown that MR simulators can be an appeal-
taneous reduction in gastrointestinal and urinary ing option for radiotherapy treatment planning
morbidity [33]. MRI-based tumor delineation [40–42]. This topic is addressed in Chap. 2 of this
may lead to optimal dosimetry and improvement book. MRI simulators can address various aspects
in therapeutic ratio, as improved delineation can of diagnostic MRI for treatment planning pur-
support tumor dose escalation (~7 Gy) and spare poses. The simulators can resolve the mismatch
the rectal wall [34]. in bore sizes, flat tabletop instead of immobiliza-
The Radiation Therapy Oncology Group con- tion devices, in-room alignment systems, and
sensus panel on pelvis contouring identified MRI imaging algorithms, and a high-precision patient
as an appropriate modality for the adnexa, utero- model can be constructed. Therefore, MR-only
cervix, and bladder [35]. Functional MRIs are treatment planning can be accomplished where
superior for gynecological malignancies due to synthetic CT data is constructed based on MR
inherent metabolic profiling and monitoring of scans.
tissue microarchitecture and tumor characteris- So, the radiotherapy workflow can accommo-
tics and are also suggested by ESTRO guidelines date MRI in several ways, like MRI co-
for delineating gynecological tumors [33]. MRI’s registration with CT scans, MR scans in the
predictive capabilities have already been proved; treatment position, MR-only treatment planning,
treatment response can also be evaluated with and others. In most of these cases, the primary
MRI [36]. Dynamic contrast-enhanced MRI per- rationale is the superior soft tissue contrast of
fusion studies during different phases of radio- MRI. Hence, MR-assisted soft tissue delineation
therapy treatment can detect cervical tumor is preferred, while CT scans often confirm treat-
response and help adapt treatment courses in ment setup. Thus, MRI can be a crucial part of
early or mid-radiotherapy phases [37]. the pretreatment process, though systematic
26 P. Yadav et al.
errors in setup remain a concern, as highlighted importance of patient selection cannot be over-
in previous sections [43–45]. stated and should be started well in advance of
There are some barriers to MR-based plan- initiating the treatment process. Some patients
ning, and we can see some considerable limita- might require support such as relaxation tech-
tions. It cannot be overlooked that CT data has a niques. Preparing patients with knowledge of the
one-to-one relationship with tissue attenuation workflow and overcoming certain situations has
and, subsequently, dose calculations and setup been found to be very effective for successful
reproduction purposes. This makes CT often the patient compliance with the instructions of the
best choice for geometric accuracy in treatment procedure. The patient should be made aware of
setup. Any MRI scan process lasts longer and the MR screening forms, the noise level that is
raises the probability of motion and patient generated from the MR scanner, the need for
distress-related issues, while CBCT-based data wearing headphones or earplugs for hearing pro-
collection is relatively faster [8]. tection, the burning effect that maybe caused by
The clinically available MR-Linacs utilize a the loop generated by connected body parts, and
deformable image registration process. This pro- most importantly, that the magnetic field is on all
cess is integrated with the treatment planning the time.
system and facilitates the propagation of struc-
tures from the planning CT scan to the MR scan
for the treatment session. This reduces the uncer- MRI Safety
tainties, as the deformable registration process is
integral to the system. At the same time, a wide MRIs are routinely utilized for better target defi-
range of deformable image registration algo- nition and are mostly done in diagnostic imaging
rithms has resulted in disparities in results. departments. With the evolution of hybrid
MR spectroscopy is another powerful MR-Linac, careful planning by the interdisciplin-
MR-based tool for pinpointing the location of ary team, made up of radiation oncologists, med-
localized prostate cancers during brachytherapy, ical physicists, nurses, and radiation therapy
as noted by Dibase et al. [29]. This technique technologists, is important from various perspec-
allows for precise delivery of ionizing radiation tives such as developing standardized treatment
and has quickly become a valuable addition to procedures, MR safety and training, logistics,
the medical field. The kV CBCT images acquired and familiarity with treatment and control rooms.
during this process are also of exceptional qual- Despite the relatively lower field strength of
ity, offering precise delineation of soft tissue hybrid MR-Linacs compared to diagnostic scan-
structures. Numerous studies have demonstrated ners, the safety criteria are the same. Therefore,
the clinical usability of these images within a all patients undergoing treatment should be
short amount of time, with analysis times being screened. To standardize the MR safety proce-
reduced with experience. In one study, analysis dures, everyone on the care team is trained by
time was even reduced from 10 min to just 2 min. experts in coordination with the vendor. The
These advancements are truly remarkable and importance of training cannot be overstated. An
have the potential to revolutionize the way we American Association of Physicists in Medicine
treat localized prostate cancers. (AAPM) task group, TG-203 [46], has pointed
out that the radiotherapy community did not fully
utilize measures of education and training with
Patient Selection clinical teams on the recommendations of TG-34
[46, 47].
Despite all the advantages of MR as discussed in One of the initial MR safety checks involves
previous sections, not every patient can be completion of a checklist or questionnaire, not
treated with MR-Linac. This section provides only for the patients but for everyone entering the
some information on patient selection. The treatment room, such as volunteers and visitors.
Existing MR safety checklists from diagnostic evaluated before the start of the treatment, and
imaging departments can serve as a baseline for the clinical team should make due preparations
developing such a document and should include for such accommodations. A light sedation may
questions about neurostimulators, claustropho- help, especially for breath-hold compliance and
bia, pacemakers, intracranial clips, ferromagnetic other sudden movements. Other solutions involve
materials, and other electronic hardware installing a mirror in the bore and audio assur-
(Fig. 3.3). The installation of ferromagnetic ances. A well-vetted preselection screening pro-
detectors can be helpful in elevating MR safety cedure could reduce patient dropout due to
checks. Recently developed electronic devices MR-related discomfort.
such as pacemakers, prostheses, and other
implants are designated as MR safe. A thorough
check of compliance with MR safety measures Metallic Implants
should be a routine part of MR-guided radiation
therapy (MRgRT) workflows. Certain metallic objects that were previously
considered contraindicated are now observed as
sufficiently safe. Along with reduction in the use
Claustrophobia of ferromagnetic material in implants, better
understanding and identification of implants with
Proper patient selection is key to the success of clear labeling in distinguishable categories such
both diagnostic MRI scans and MRgRT treat- as MR safe, MR conditional, and MR unsafe
ment procedures. Certain existing conditions helps in the screening process (Fig. 3.3).
such as arrhythmias, claustrophobia, and prob- Figure 3.4 provides additional images of metallic
lems with following breath-hold instructions implants where in MR images it produces a
should be properly addressed. Challenges posed blooming effect with loss of anatomical
by claustrophobia and anxiety are well docu- information.
mented for MRI procedures, and numerous solu- An item with no known hazard in any MR
tions have been explored [48]. The probability of environment and basically nonconducting, non-
a patient developing claustrophobia should be metallic, and/or nonmagnetic can be considered
Fig. 3.3 Figure showing the MR signs and unsafe objects in MRI
28 P. Yadav et al.
a b c
Fig. 3.4 Sagittal images of a patient with tracheal implant fact) is observed in the MR image. Arrows indicate the
via (a) planning CT, (b) CBCT, and (c) 0.35T MR. As position of the trachea
noted, a total loss of image information (blooming arti-
as MR safe. If an item has no known hazards in a artifacts on cardiac implantable electronic

specified MRI environment with particular con- devices (CIEDs) [50]. The American Association
ditions of use like spatial magnetic gradient, of Physicists in Medicine TG-203 acknowledges
radiofrequency fields, related systemic absorp- MRI procedures do not significantly increase the
tion rates, and static magnetic field strength, it is probability of CIED-related issues, but still
known as MR conditional. An item is considered requires that well-documented MR safety proto-
as MR unsafe if it could be hazardous in MR cols should be developed involving such devices
environments. with due consideration of multidisciplinary
Most coronary stents are either MR safe or aspects of both devices and procedures [46].
MR conditional. Prosthetic valves and sternal While developing radiotherapy treatment pro-
wires are usually MR safe. Specific exceptions grams for treating cases with CIEDs, radiation
and explicit declaration should be included on the oncologists should coordinate their activities
screening forms [49]. Studies have suggested that with cardiology teams with well-defined respon-
MR procedures for pacemaker-dependent sibilities for the management of such cases. Staff
patients should not be performed unless the and departmental requirements should be out-
potential benefits surpass the risks. Nowadays, lined and categorized as low risk, medium risk,
MR conditional pacemakers are available. and high risk for the radiation dose exposure and
Likewise, MR procedures in patients with MR field strength. It is important to ascertain a
implantable cardioverter defibrillators are dis- patient’s risk category early in the process. A
couraged. For situations involving pacemakers copy of the patient’s device card should be
and other cardiac devices, if needed, screening obtained, and it is advisable to consult the manu-
should be monitored by competent imaging and facturer to understand the device’s tolerance for
electrophysiology specialists. Such devices the magnetic field and radiation dose rate [51,
might require reprogramming or some sort of 52]. Emergency response devices should be eas-
recalibration once the MR procedure is com- ily available to address the probability of medical
pleted and may require a device-related clinic interventions. Patients should be asked about
visit. CIEDs during the initial appointment, and such
The Heart Rhythm Society 2017 report pro- information should be documented in their medi-
vides comprehensive information on various cal file and verified during consultation with the
effects of magnetic fields such as unplanned radiation oncologist. Clinics should preemptively
stimulation, mechanical displacement, electrode disclose such details during weekly chart round
heating, RF injected disturbances, and imaging meetings and discuss management plans.
Devices such as implantable loop monitors, treatment fractions and avoidance of patient-
cochlear implants, epidural pumps, nerve stimu- induced early termination. Patient weight may be
lators, insulin pumps, and programmable intra- a consideration in some cases as some manufac-
thecal drug delivery devices should be examined turers have a maximum couch weight limit of
for MR safety concerns. 400 lb. (181.4 kg).
Breath-hold imaging and treatment delivery
are readily used to overcome challenges posed by
Patient Comfort respiratory motion. The MRgRT systems from
both of the current manufacturers are capable of
Traditionally, patient comfort is given high prior- monitoring two-dimensional target movement
ity in radiation therapy and is usually achieved by during treatment delivery and this real-time
providing a comfortable immobilization system. tumor imaging is utilized for modulating radia-
However, patients undergoing treatment in an tion dose during breathing. Patients should be
MR-Linac do not have comfortable immobiliza- able to coordinate with the breath-hold instruc-
tion due to eddy currents and the proximity of tions so that not only motion artifacts and plan-
coils. MR-Linac patients have to spend consider- ning margins are constrained but also treatment
ably longer treatment time in an enclosed bore, can be optimally delivered.
which is a well-known concern of patients as Physiological motion during setup and treat-
shown in Fig. 3.5. The narrower bores can further ment can adversely impact the overall procedure,
impact such unpleasant experiences. Some as the former results in unrepresentative setup
patients cannot stay in MR-Linac’s bore for an images and the latter forces the team to increase
extended period of time. Also, large body habitus planning margins. Breath-hold techniques are
can be a constraint in patient selection if coil size reported to have shorter treatment time [55].
or instrument bore size is relatively small. Patient’s visual feedback can help facilitate the
MR-Linac’s longer and narrower bore size breath-hold workflows. Also, breath-hold instruc-
and requirements for immobilization along with tions can be provided through audio systems, but
tolerance for coil placement can increase the visual coaching can be a great help. With the help
patient’s discomfort [53]. Klüter et al. have of equipment such as custom-designed prismatic
reported such anxiety and distress-related com- glasses allowing sight of a computer monitor on
plaints from patients undergoing MRgRT [54]. the wall behind the MRI bore, patients can
Adequate consideration and remedies of such actively synchronize the breath-hold and release
concerns is crucial for safe administration of all phases of the respiratory cycle.
a b
Fig. 3.5 An example of the resources to provide a sooth- animals, flowers, etc. that a patient can choose. (b) Gantry
ing environment to the patient during treatment is shown. with selected view
(a) The library of the display includes mountains, oceans,
30 P. Yadav et al.
Breathing Fatigue tions, ensuring multiple means of communicat-

ing with the patient, avoiding loops in coil wires
Modern treatment pathways consider 17 s as the and on patients, and use of medical gowns to
minimum breath-hold duration for adequate avoid any metals in clothing (even underwires in
imaging, simulation, and reduced treatment bras can be a hazard). Contrast administration
duration. This usually involves sites where inter- only in the presence of a physician is a good
organ and intra-organ mobility is a challenge patient safety measure. Patient safety also
such as thoracic and abdominal sites. Also, includes screening the patient, rechecking medi-
breath hold can facilitate the delineation of the cal records, and verification of implants for MR
target with OARs, especially in soft tissue anat- safety or compatibility considering the field
omy and sites where organ displacement is a strength of the scanner. The simple rule here is
huge concern. In many patients, such a long trust but verify the information. For situations
breath hold is not possible, and hence, it should when a patient is unable to communicate or is
be tested upfront during patient selection. An under anesthesia, protocols should entail extra
optimal breath-hold technique can address the cautionary measures.
overlap of critical OARs with the target, and Practicing emergency safety is also equally
with superior soft tissue contrast and real-time important. It covers proper signage of emergency
gating, not only is treatment duration reduced response equipment, ascertaining emergency
but dose escalation is also feasible. Coordinated response plans, and training and drilling on
breathing patterns and enhanced visualization removing patients from the couch. There should
and delineation of tumor and healthy tissues be an emergency cart with instant access to safety
facilitates better treatment adaptation, even for equipment, and monitoring systems should be
cases with complex contours with challenging functional at all times.
treatment objectives. These considerations are
key to the success of high dose rate single frac-
tion and stereotactic deliveries. So, the benefits Thermal Effects
of breath hold and motion management should
be evaluated during patient selection. This also When undergoing MR imaging, it is essential to
applies for assessing or screening the cases for keep in mind that there is a risk of localized
online adaptative MRgRT. heating if patient-created current loops are pres-
ent. To avoid this, patients should ensure that
different parts of their body do not touch during
Magnetic Susceptibility the procedure and that they are not wearing any
wet or damp clothing. It is also important to be
Magnetic fields might temporarily affect various cautious when using transdermal patches, as
materials. The magnetic susceptibility varies they may cause burns to the skin. Modern scan-
with the material composition and shape, but ners have safety measures in place to determine
more importantly with the rotating gantry as in the amount of RF energy a patient may receive
MR-Linac [56–58]. This effect is widely known in terms of specific absorption rate, and clinical
as magnetic susceptibility, and implants may be limits are set by International Electrotechnical
impaired due to this; the scan quality may also be Commission standards (IEC 60601-2-33). The
affected. Magnetic susceptibility can be reduced FDA recommends limiting core temperature
or enhanced depending on the sequence selected rise to 1 °C. Patients with medical implants
for the procedure. require special attention as they may be at a
Patient safety is of paramount importance and higher risk for local heating injury. Chances of
includes many aspects such as awareness of the device failure are higher in MR, and challenges
specific absorption rate, checking hearing protec- increase as some devices may be compatible
with one MR strength and not others. To prevent 6. Ling CC, Yorke E, Fuks Z. From IMRT to IGRT:
injury, proper patient preparation guidelines and frontierland or neverland? Radiother Oncol.
2006;78(2):119–22.
avoiding noncompliant patient monitoring 7. Ghilezan M, Yan D, Liang J, Jaffray D, Wong
leads, looping wires, and clothing is important J, Martinez A. Online image-guided intensity-
[49]. modulated radiotherapy for prostate cancer: how
much improvement can we expect? A theoretical
assessment of clinical benefits and potential dose
escalation by improving precision and accuracy
Summary of radiation delivery. Int J Radiat Oncol Biol Phys.
2004;60(5):1602–10.
MR-Linac provides a unique opportunity for 8. van Herk M. Errors and margins in radiotherapy.
Semin Radiat Oncol. 2004;14(1):52–64.
real-time tumor visualization that can be used for 9. Siewerdsen JH, Jaffray DA. Cone-beam computed
gated therapy. However, not all patients can tomography with a flat-panel imager: magnitude and
undergo such treatment. Owing to different ana- effects of x-ray scatter. Med Phys. 2001;28(2):220–31.
tomical sites with different tissue characteristics 10. Mackie TR, Kapatoes J, Ruchala K, Lu W, Wu C,
Olivera G, et al. Image guidance for precise con-
in MRI and CT imaging, a proper patient s election formal radiotherapy. Int J Radiat Oncol Biol Phys.
protocol should be adopted. Patient selection 2003;56(1):89–105.
should be based on magnetic susceptibility, ther- 11. Shirato H, Seppenwoolde Y, Kitamura K, Onimura
mal issues, MR safety, claustrophobia, high-Z R, Shimizu S. Intrafractional tumor motion: lung and
liver. Semin Radiat Oncol. 2004;14(1):10–8.
implants like pacemakers and CIEDs, and most 12. Dawson LA, Brock KK, Kazanjian S, Fitch D,
importantly, breathing irregularity and fatigue. McGinn CJ, Lawrence TS, et al. The reproducibil-
Additionally, it is a time-consuming technology ity of organ position using active breathing control
with treatment sessions lasting over an hour (ABC) during liver radiotherapy. Int J Radiat Oncol
Biol Phys. 2001;51(5):1410–21.
where patient comfort can be a critical factor. It 13. Wong JW, Sharpe MB, Jaffray DA, Kini VR, Robertson
has been estimated that only 1/3 of the patient JM, Stromberg JS, et al. The use of active breathing
population is suitable for this technology [59]. control (ABC) to reduce margin for breathing motion.
Thus, prudent and proper selection of patients is Int J Radiat Oncol Biol Phys. 1999;44(4):911–9.
14. Xing L, Thorndyke B, Schreibmann E, Yang Y, Li TF,
critical for the success of this technology to pro- Kim GY, et al. Overview of image-guided radiation
vide optimum care to the patients. therapy. Med Dosim. 2006;31(2):91–112.
15. Hamilton RJ, Sweeney PJ, Pelizzari CA, Yetkin FZ,
Holman BL, Garada B, et al. Functional imaging in
treatment planning of brain lesions. Int J Radiat Oncol
References Biol Phys. 1997;37(1):181–8.
16. Njeh CF. Tumor delineation: the weakest link in the
1. Supanich M, Siewerdsen J, Fahrig R, Farahani K, search for accuracy in radiotherapy. J Med Phys.
Gang GJ, Helm P, et al. AAPM Task Group Report 2008;33(4):136–40.
238: 3D C-arms with volumetric imaging capability. 17. Weiss E, Hess CF. The impact of gross tumor vol-
Med Phys. 2023;50(8):e904−945. ume (GTV) and clinical target volume (CTV) defini-
2. Ding GX, Alaei P, Curran B, Flynn R, Gossman tion on the total accuracy in radiotherapy theoretical
M, Mackie TR, et al. Image guidance doses deliv- aspects and practical experiences. Strahlenther Onkol.
ered during radiotherapy: quantification, manage- 2003;179(1):21–30.
ment, and reduction: report of the AAPM Therapy 18. Rasch C, Steenbakkers R, van Herk M. Target defini-
Physics Committee Task Group 180. Med Phys. tion in prostate, head, and neck. Semin Radiat Oncol.
2018;45(5):e84–99. 2005;15(3):136–45.
3. Yan D, Georg D. Adaptive radiation therapy. Z Med 19. Vorwerk H, Beckmann G, Bremer M, Degen M, Dietl
Phys. 2018;28(3):173–4. B, Fietkau R, et al. The delineation of target volumes
4. Murphy MJ, Balter J, Balter S, BenComo JA Jr, Das IJ, for radiotherapy of lung cancer patients. Radiother
Jiang SB, et al. The management of imaging dose dur- Oncol. 2009;91(3):455–60.
ing image-guided radiotherapy: report of the AAPM 20. van Mourik AM, Elkhuizen PH, Minkema D, Duppen
Task Group 75. Med Phys. 2007;34(10):4041–63. JC, Dutch Young Boost Study Group, van Vliet-
5. Jaffray DA, Das S, Jacobs PM, Jeraj R, Lambin Vroegindeweij C. Multiinstitutional study on target
P. How advances in imaging will affect precision volume delineation variation in breast radiotherapy
radiation oncology. Int J Radiat Oncol Biol Phys. in the presence of guidelines. Radiother Oncol.
2018;101(2):292–8. 2010;94(3):286–91.
32 P. Yadav et al.
21. Chen AM, Farwell DG, Luu Q, Chen LM, assisted dose volume adaptation and dose escalation
Vijayakumar S, Purdy JA. Marginal misses after in brachytherapy of locally advanced cervix cancer.
postoperative intensity-modulated radiotherapy for Radiother Oncol. 2007;83(2):148–55.
head and neck cancer. Int J Radiat Oncol Biol Phys. 34. Steenbakkers RJ, Deurloo KE, Nowak PJ, Lebesque
2011;80(5):1423–9. JV, van Herk M, Rasch CR. Reduction of dose deliv-
22. Jolicoeur M, Racine ML, Trop I, Hathout L, Nguyen ered to the rectum and bulb of the penis using MRI
D, Derashodian T, et al. Localization of the surgi- delineation for radiotherapy of the prostate. Int J
cal bed using supine magnetic resonance and com- Radiat Oncol Biol Phys. 2003;57(5):1269–79.
puted tomography scan fusion for planification of 35. Gay HA, Barthold HJ, O’Meara E, Bosch WR, El
breast interstitial brachytherapy. Radiother Oncol. Naqa I, Al-Lozi R, et al. Pelvic normal tissue con-
2011;100(3):480–4. touring guidelines for radiation therapy: a Radiation
23. Chang EL, Akyurek S, Avalos T, Rebueno N, Spicer Therapy Oncology Group consensus panel atlas. Int J
C, Garcia J, et al. Evaluation of peritumoral edema Radiat Oncol Biol Phys. 2012;83(3):e353–62.
in the delineation of radiotherapy clinical target vol- 36. McVeigh PZ, Syed AM, Milosevic M, Fyles A, Haider
umes for glioblastoma. Int J Radiat Oncol Biol Phys. MA. Diffusion-weighted MRI in cervical cancer. Eur
2007;68(1):144–50. Radiol. 2008;18(5):1058–64.
24. Just M, Rosler HP, Higer HP, Kutzner J, Thelen 37. Mayr NA, Wang JZ, Zhang D, Grecula JC, Lo SS,
M. MRI-assisted radiation therapy planning of brain Jaroura D, et al. Longitudinal changes in tumor per-
tumors—clinical experiences in 17 patients. Magn fusion pattern during the radiation therapy course
Reson Imaging. 1991;9(2):173–7. and its clinical impact in cervical cancer. Int J Radiat
25. Khoo VS, Joon DL. New developments in MRI for Oncol Biol Phys. 2010;77(2):502–8.
target volume delineation in radiotherapy. Br J Radiol. 38. Ulin K, Urie MM, Cherlow JM. Results of a multi-
2006;79 Spec No 1:S2–15. institutional benchmark test for cranial CT/MR
26. Mardor Y, Pfeffer R, Spiegelmann R, Roth Y, Maier image registration. Int J Radiat Oncol Biol Phys.
SE, Nissim O, et al. Early detection of response to 2010;77(5):1584–9.
radiation therapy in patients with brain malignan- 39. van Herk M, Kooy HM. Automatic three-dimensional
cies using conventional and high b-value diffusion- correlation of CT-CT, CT-MRI, and CT-SPECT using
weighted magnetic resonance imaging. J Clin Oncol. chamfer matching. Med Phys. 1994;21(7):1163–78.
2003;21(6):1094–100. 40. Devic S. MRI simulation for radiotherapy treatment
27. Garcia-Alvarez R, Liney GP, Beavis AW. Repeatability planning. Med Phys. 2012;39(11):6701–11.
of functional MRI for conformal avoidance 41. Kapanen M, Collan J, Beule A, Seppala T, Saarilahti
radiotherapy planning. J Magn Reson Imaging. K, Tenhunen M. Commissioning of MRI-only
2006;23(2):108–14. based treatment planning procedure for external
28. Nguyen PL, Aizer A, Assimos DG, D’Amico AV, beam radiotherapy of prostate. Magn Reson Med.
Frank SJ, Gottschalk AR, et al. ACR appropriate- 2013;70(1):127–35.
ness criteria(R) definitive external-beam irradiation in 42. Paulson ES, Erickson B, Schultz C, Allen
stage T1 and T2 prostate cancer. Am J Clin Oncol. LX. Comprehensive MRI simulation methodology
2014;37(3):278–88. using a dedicated MRI scanner in radiation oncology
29. Debois M, Oyen R, Maes F, Verswijvel G, Gatti G, for external beam radiation treatment planning. Med
Bosmans H, et al. The contribution of magnetic reso- Phys. 2015;42(1):28–39.
nance imaging to the three-dimensional treatment 43. Johnstone E, Wyatt JJ, Henry AM, Short SC, Sebag-
planning of localized prostate cancer. Int J Radiat Montefiore D, Murray L, et al. Systematic review of
Oncol Biol Phys. 1999;45(4):857–65. synthetic computed tomography generation meth-
30. Chen ME, Troncoso P, Johnston DA, Tang K, odologies for use in magnetic resonance imaging-
Babaian RJ. Optimization of prostate biopsy only radiation therapy. Int J Radiat Oncol Biol Phys.
strategy using computer based analysis. J Urol. 2018;100(1):199–217.
1997;158(6):2168–75. 44. Khoo VS, Padhani AR, Tanner SF, Finnigan DJ,
31. Haider MA, Chung P, Sweet J, Toi A, Jhaveri K, Leach MO, Dearnaley DP. Comparison of MRI
Menard C, et al. Dynamic contrast-enhanced mag- with CT for the radiotherapy planning of pros-
netic resonance imaging for localization of recurrent tate cancer: a feasibility study. Br J Radiol.
prostate cancer after external beam radiotherapy. Int J 1999;72(858):590–7.
Radiat Oncol Biol Phys. 2008;70(2):425–30. 45. Tyagi N, Fontenla S, Zelefsky M, Chong-Ton M,
32. McLaughlin PW, Narayana V, Meirovitz A, Troyer S, Ostergren K, Shah N, et al. Clinical workflow for
Roberson PL, Gonda R Jr, et al. Vessel-sparing pros- MR-only simulation and planning in prostate. Radiat
tate radiotherapy: dose limitation to critical erectile Oncol. 2017;12(1):119.
vascular structures (internal pudendal artery and cor- 46. Miften M, Mihailidis D, Kry SF, Reft C, Esquivel C,
pus cavernosum) defined by MRI. Int J Radiat Oncol Farr J, et al. Management of radiotherapy patients
Biol Phys. 2005;61(1):20–31. with implanted cardiac pacemakers and defibrillators:
33. Potter R, Dimopoulos J, Georg P, Lang S, Waldhausl a report of the AAPM TG-203(dagger). Med Phys.
C, Wachter-Gerstner N, et al. Clinical impact of MRI 2019;46(12):e757–e88.
47. Marbach JR, Sontag MR, Van Dyk J, Wolbarst 52. Harden SP. MRI conditional pacemakers: the start of a
AB. Management of radiation oncology patients with new era. Br J Radiol. 2011;84(1005):773–4.
implanted cardiac pacemakers: report of AAPM Task 53. Botman R, Tetar SU, Palacios MA, Slotman BJ,
Group No. 34. American Association of Physicists in Lagerwaard FJ, Bruynzeel AME. The clinical intro-
Medicine. Med Phys. 1994;21(1):85–90. duction of MR-guided radiation therapy from a RTT
48. Munn Z, Jordan Z. The effectiveness of interven- perspective. Clin Transl Radiat Oncol. 2019;18:140–5.
tions to reduce anxiety, claustrophobia, sedation and 54. Kluter S, Katayama S, Spindeldreier CK, Koerber
non-completion rates of patients undergoing high SA, Major G, Alber M, et al. First prospective clini-
technology medical imaging. JBI Libr Syst Rev. cal evaluation of feasibility and patient acceptance of
2012;10(19):1122–85. magnetic resonance-guided radiotherapy in Germany.
49. Levine GN, Gomes AS, Arai AE, Bluemke DA, Strahlenther Onkol. 2020;196(8):691–8.
Flamm SD, Kanal E, et al. Safety of magnetic 55. Sahin B, Zoto Mustafayev T, Gungor G, Aydin G,
resonance imaging in patients with cardiovas- Yapici B, Atalar B, et al. First 500 fractions delivered
cular devices: an American Heart Association with a magnetic resonance-guided radiotherapy sys-
scientific statement from the Committee tem: initial experience. Cureus. 2019;11(12):e6457.
on Diagnostic and Interventional Cardiac 56. Wachowicz K, Stanescu T, Thomas SD, Fallone
Catheterization, Council on Clinical Cardiology, BG. Implications of tissue magnetic susceptibility-
and the Council on Cardiovascular Radiology and related distortion on the rotating magnet in an
Intervention: endorsed by the American College MR-linac design. Med Phys. 2010;37(4):1714–21.
of Cardiology Foundation, the North American 57. Jackson S, Glitzner M, Tijssen RHN, Raaymakers
Society for Cardiac Imaging, and the Society for BW. MRI B (0) homogeneity and geometric dis-
Cardiovascular Magnetic Resonance. Circulation. tortion with continuous linac gantry rotation
2007;116(24):2878–91. on an Elekta Unity MR-linac. Phys Med Biol.
50. Kusumoto FM, Schoenfeld MH, Wilkoff BL, 2019;64(12):12NT01.
Berul CI, Birgersdotter-Green UM, Carrillo R, 58. Stanescu T, Wachowicz K, Jaffray
et al. 2017 HRS expert consensus statement on DA. Characterization of tissue magnetic
cardiovascular implantable electronic device susceptibility-induced distortions for MRIgRT. Med
lead management and extraction. Heart Rhythm. Phys. 2012;39(12):7185–93.
2017;14(12):e503–e51. 59. Moore-Palhares D, Ho L, Lu L, Chugh B, Vesprini
51. Bhandiwad AR, Cummings KW, Crowley M, D, Karam I, et al. Clinical implementation of mag-
Woodard PK. Cardiovascular magnetic resonance netic resonance imaging simulation for radiation
with an MR compatible pacemaker. J Cardiovasc oncology planning: 5 year experience. Radiat Oncol.
Magn Reson. 2013;15(1):18. 2023;18(1):27.
Education and Training
in MR-Based Radiation Treatment
4
Luca Nicosia, Riccardo Filippo Borgese,
Kerryn Brown, and Filippo Alongi
Introduction MRIdian by Viewray (Fig. 4.1) consist of a

0.35T MR scanner with a split magnet design
Magnetic resonance image (MR)-guided radio- (detail in Chap. 9 of this book) [3]. In this system
therapy (MRgRT) is one of the most innovative the force lines of the magnetic field are oriented
developments of image-guided radiotherapy along the cranio-caudal axis of the patient. The
(IGRT) [1]. The acquisition of MR images imme- ring gantry is positioned between the two mag-
diately before and after radiotherapy treatments nets, and it is equipped with a 6 MV flattening
permits daily online planning adaptation with the filter free linear accelerator (Linac). The Linac is
aim to improve treatment accuracy and tolerabil- hosted into a dedicated shielded compartment
ity. This new therapeutical approach introduces mounted on the gantry. The beam is collimated
new specific issues like preparation before device with a double-focused and stacked multileaf col-
installation, MRI safety program and specific limator. With this design, the radiation does not
training, development of online adaptive work- pass through the magnet, but it is absorbed by the
flow, and treatment delivery. Firstly, a deep 5-mm-thick connecting fiberglass placed between
knowledge of the available treatment systems the two magnets [4–6].
and their peculiar differences is necessary to The original project of the MR-Linac system
exploit the potential of MRgRT [2]. of Elekta, Unity (Fig. 4.2), was started in 1999 in
L. Nicosia (*) · R. F. Borgese

Department of Advanced Radiation Oncology,
IRCCS Sacro Cuore Don Calabria, Cancer Care
Center, Negrar di Valpolicella, Italy
[email protected]
K. Brown
Deparment of Radiation Oncology, ONJ Centre,
Austin Hospital, Heidelberg, VIC, Australia
e-mail: [email protected]
F. Alongi
Department of Advanced Radiation Oncology,
IRCCS Sacro Cuore Don Calabria, Cancer Care
Center, Negrar di Valpolicella, Italy
e-mail: [email protected] Fig. 4.1 MRIdian System by Viewray

36 L. Nicosia et al.
procedures requiring specific training consist of

the following:
• MR safety training for all personnel involved

in the MR procedures or working in the MR
environment with clear guidance, authoriza-
tion, and responsibilities (i.e., RTTs, radiation
oncologist, medical physicists, cleaning staff,
visitors).
• Vendor-specific training to learn the different
Fig. 4.2 Unity system by Elekta aspects of the online workflow (i.e., alterna-
tive workflows in case of system errors) is
UMC Utrecht [7]. The system mounts a 1.5T required to safely administer adaptive plans.
MRI system with the gantry positioned in the Special training might be offered by vendors
mid-transversal plane around the isocenter. The on dedicated consoles for planning and simu-
Linac is a 7 MV flattening filter free [8]. There is lation of adaptive treatment to speed up the
an active magnetic shield to magnetically decou- learning curve and identify potential chal-
ple the MRI and the Linac allowing them to oper- lenges and pitfalls. An initial experience with
ate independently without mutual interaction. these procedures before clinical implementa-
Moreover, the lines of magnetic field force are tion is essential.
oriented cranio-caudally to the patient [9]. Details • Specific training for each role, based on the
of the Elekta Unity system can be found in Chap. traditional role and responsibilities including
10 of this book. the following: RTTs are a component of the
Both systems have a model within the treat- adaptive team and may need dedicated train-
ment planning system that considers the impact ing to improve their skills (i.e., MRI training
of magnetic field on the secondary electrons to ensure patient safety, understand MR
which can be compensated by dedicated optimi- sequences, and troubleshoot imaging prob-
zation strategies [10, 11]. The commissioning lems). Radiation oncologists, or dedicated
and acceptance procedures are similar to those of RTTs, must be familiar with the online con-
the traditional RT systems, except for those touring tools and algorithms and understand
uniquely linked to the MRgRT aspects. Other the time for contouring edits is limited during
than the innovative equipment, one of the most online adaptive procedures, since longer treat-
peculiar aspects of this new technology is the ment sessions might increase treatment uncer-
possibility of the online adaptive workflow [12]. tainties such as patient fatigue and movement.
Guidelines for contouring should be provided
to maintain consistency between team mem-
Preparation Before Starting bers, for example, defining treatment margins
or which contours are to be amended during
The online adaptive workflow procedure differs the online adaptive procedures.
from the typical treatment workflow on conven- • Medical physicists and dosimetrists special-
tional RT machines and requires dedicated teams, ized in treatment planning are responsible for
including radiotherapists (RTTs), radiation the online treatment planning procedures.
oncologist, and medical physicists. The three Dedicated training should include the knowl-
major components of a MR-Linac workflow are edge of the different online treatment planning
(1) MRI safety and imaging acquisition, (2) alternatives, pre-planning offline actions, cre-
imaging registration and contouring or recon- ation of treatment planning templates, and
touring of the organs at risk (OAR), and (3) treat- specific plan quality assurance (QA) proce-
ment plan adaptation and verification. The dures. The training can be provided in many
4 Education and Training in MR-Based Radiation Treatment 37
ways and can consist of online modules and workflow with conventional radiotherapy, adap-
onsite training sessions. It should include all tive workflow requires daily recontouring and
the different aspects of the respective online approval of target and treatment planning, either
workflows such as different solutions for at the treatment console or remotely.
adaptive planning, possible solution in case of To reduce potential errors and mitigate risks, a
errors or system failures, software modules risk analysis for a safe and efficient workflow
including treatment calendar, MRI console may be useful [15]. For example, the dedicated
use, tools for data transfer, image registration, multidisciplinary team might implement a pro-
contouring and treatment planning, and stor- cess failure mode and effect analysis (P-FMEA)
age and actionability of data. Dedicated visita- to improve the quality and safety of the treatment
tion of experienced centers might increase the and reduce risks [16–18]. This framework com-
learning curve on a peer-to-peer basis and prises a first step in which the clinical team devel-
offer details of clinical and technical insights. ops a process map for MR-guided adaptive
• Foster close collaboration between medical radiotherapy. In a second step, the team identifies
physicists, MR imaging scientists, and physi- possible failure mode and pitfalls and risk prob-
cians to leverage complementary expertise in ability numbers are assigned based on the proba-
order to address those unmet needs related to bility of occurrence, the severity of the risk, and
the clinical use of MR imaging in RT. the detectability. Risk mitigation strategies sup-
port the generation of a standardized workflow,
Considering the novelty and peculiarity of this clearly defined protocols, and the definition
new therapeutical approach, it is important that checklist as well as standard operating proce-
all the team members are familiar with every step dures (SOPs) for a safe and efficient implementa-
of the entire workflow in case of unforeseen tion of MRgRT procedures. This adopted model
events and to rapidly make decisions [13, 14]. As should be tested on volunteers during dry run
experience develops over time, team roles and training sessions. After clinical implementation
responsibilities might be modified according to this model might be improved, as the experience
necessity whereby crossing the traditional bound- of the team increases [19].
aries of the respective roles, to enable a prompter An exhaustive QA guideline for medical phys-
and less resource-intensive treatment delivery. icists was proposed by ESTRO and ASTRO to
Furthermore, due the potential adaptive nature of help commissioning [20].
every treatment session, it is essential that critical
online decisions are made swiftly and consis-
tently within the team. To ensure this, close mul- MRI-Specific Training
tidisciplinary teamwork is required to define and and Equipment
optimize treatment workflow and thresholds for
decision-making and action levels. Before starting with MRgRT, dedicated SOPs for
The dedicated training must cover all the MRI safety should be written and the staff must
aspects of the workflow to enable new treatment be trained according to this, in addition to dedi-
plans to be created, based on the daily anatomy of cated training as described above. The following
target and OARs. Also, dedicated training in MR aspects should be addressed:
anatomy should be deemed essential. This new
treatment modality has important implication in • Potential hazards and risks (magnetic field
the typical organization of a radiation oncology risk, radio frequency field risk, cryogen risk,
department, because of the need of a dedicated biological effects due to magnetic field, pro-
team, rapid availability of radiation oncologists, jectile incident risk).
and medical physicists for each treatment frac- • Acoustic noise (patient compliance, hearing
tion, that must be imbued into the usual day loss).
activity. In contrast with the typical treatment • Magnet quench (emergency procedure).
• Access restriction (fringe field, safety zones, ensure the safety of any material that is imple-
MR unsafe/conditional/safe, screening of mented into the MR-Linac. It is recommended to
the patient [implants, foreign body metals, differentiate between (1) ferromagnetic safety,
pacemaker, or ICD implants], authorized (2) imaging artefact assessment, (3) proper coil
and safety-certified staff). (Additional infor- handling and usage, and (4) degree of dose
mation can be found in Chap. 3 of this attenuation.
book.) For ferromagnetic safety, vendors should
• Radio frequency (RF) and gradient fields (RF always provide a written confirmation regarding
absorption, looped conductors, patient contact MR-environment compatibility. Nevertheless, as
with bore or wires). part of risk assessment and QA management
• Emergencies (guidelines for the staff to ensure strategy, all equipment must be tested on site and
a safe evacuation of the patient). any ferromagnetic properties should be excluded.
Phantom measurements should be performed to
Internally approved guidelines for emergency assess both susceptibility and distortion artifacts
management must be created. These guidelines and to quantify their impact on image quality and
must be readily available to all involved staff spatial accuracy and to evaluate possible dose
(i.e., the emergency response team, including attenuation. Finally, proper coil placement needs
physicians and nurses). These guidelines must be to be tested to ensure optimal distance and loca-
consistent with other institutional guidelines on tion to the area of interest and to avoid imaging
MRI safety as well as national/international deterioration due to insufficient signal generation
guidelines. and detection.
Patients must be screened for MRI safety The knowledge of acquisition, processing,
before MR simulation (metal screening, implant and interpretation of MR images and the
screening, pacemaker, ICD screening) and sequences is a prerequisite for MRgRT [22–24].
informed about the procedure to familiarize with Considering the variety in the educational pro-
the equipment (i.e., hearing protection, squeeze grams for RTTs, radiation oncologists, medical
bulb alarm) and instructed on how the staff will physicists, and dosimetrists across different
communicate with them during treatment. All the countries, it is difficult to define a complete core
screening procedures for MR safety are essential of required skills and training pathways.
to avoid any harm to patient and personnel; there- Nevertheless, some guidance can be derived from
fore, the following indications must be learned the initial experience that is currently being made
by the dedicated team, including also periodic with the MR-Linac implementation. During the
training to refresh them. clinical implementation phase, the dedicated
All items that are at risk within the MR envi- team to MRgRT should include RTTs, radiation
ronment must be removed by the patient before oncologists, and medical physicists with a spe-
entering the treatment and simulation room. cialized training in MR technology. This core
Ferromagnetic detection systems (i.e., entry con- team should have received MR training as part of
trol systems, handheld metal detectors) can be their education course or national certification
used additionally to enhance safety, but only program or visited specific courses that provided
after the conventional screening procedure the basic knowledge of MR imaging. A refresher
abovementioned and not as a substitute for them. course or an internship in a radiology department
Moreover, for each subsequent fraction, the staff can be useful for all members of the team. In
should be informed about any procedure the addition, courses provided by vendors, universi-
patient may have had between the last and the ties (i.e., Utrecht), ESTRO, the Institute of
current fraction [21]. Cancer Research, or American College of
Regarding positioning equipment, not all the Radiology (ACR) or by site visits and peer-to-
available patient setup devices have been tested peer training at experienced MR-Linac depart-
or designed for MR use. Therefore, it is critical to ments provide additional opportunities to
improve skills and experience with MR-Linac vs pancreatic tumors [32–35]. It is recommended
technology and adaptive workflow [25–27]. to define internally the minimum requirements
Also, a RT-trained radiographer might be for treatment adaptation or which threshold val-
involved in the treatment process and not only in ues, or action levels, should be considered. These
positioning and imaging aspects. With dedicated considerations depend on the tumor site and the
training is it possible to achieve contouring con- clinical relevance of the daily anatomy change,
sistency comparable to that of clinicians. This for example, the variation of the tumor and
can help reduce the workload of the team and to OARs, the type of dose distribution (homoge-
acquire mutual skills to speed up the learning neous, heterogeneous, SBRT treatments), total
curve of the MR team [28]. dose, and fractionation. As a general rule, the
higher the dose per fraction and the proximity to
OARs, the smaller the therapeutic window for
raining and Education During
T anatomical variation. These minimum require-
the Delivery Phase ments can be implemented in different ways into
the clinical practice [34, 36]. One approach is to
The MR-Linac provides the fundamental ability predefine the type of plan adaptation for each
to adapt treatment plans on a daily basis. Is it treatment. It is helpful to implement clear dose
known that anatomy changes, moment by constraints and target coverage goals to define if
moment due to breathing motion, peristalsis, or a plan is acceptable for delivery (green light),
patient movement, as well tumor change over the within a range of accepted variability (yellow
course of treatment. The use of an MR-Linac light), or is an unacceptable parameter (red light)
may prevent this problem, as OAR motion and with different levels of action outlined for correc-
morphological changes in the target can be seen tion and decision-making. Furthermore, it might
in real time. This advantage may indicate the use be useful to set a time frame for online contour-
of real-time biological targeting based on bio- ing and a maximum number of optimization
marker or differential response, resistance, and cycles to avoid long adaptation times.
progression or even hypoxia and introduces the Future deformable dose accumulation in com-
possibility to translate the radiomics concepts bination with daily adaptation can be expected to
into the clinical practice in radiation oncology refine the current knowledge on toxicity parame-
[29–31]. ters, because dosimetry will be based on actual
The typical treatment workflow on a MR-Linac delivered, rather than planned, radiation dose.
is constituted by patient positioning, preadaptive Also, the possibility to obtain synthetic CTs from
imaging acquisition, recontouring of OARs and/ MR images might increase the accuracy of
or target or superimposition of previous contour, parameter calculations. The two available
treatment plan adaptation or shifting, plan quality MRgRT systems present differences, but both
assurance, pre-treatment imaging verification, allow the assessment of the anatomy on MRI
treatment delivery with the possibility of online prior to shifting or replanning. This assessment
delivery verification (i.e., cine-MRI or 3D MRI) can be with respect to reference plan dosimetry
or motion management (i.e., gating), and even- (MRIdian) or reference anatomy (Unity). When
tual post-treatment imaging. changes in OARs or targets appear, it is manda-
The goal of daily plan adaptation can be signi- tory to replan, especially for ultra hypofraction-
fied by improved target coverage, OAR sparing, ated regimens, with minimal PTV margins or
or both, depending on whether the tumor cover- when steep dose gradients are used.
age or the OAR sparing requires primary consid- The type of adaptive workflow depends on the
eration during the specific treatment session. The complexity and variation of the daily imaging
clinical benefit of daily adaptation greatly compared to the reference and can range from a
depends on the tumor site and can be of different simple shift in MLC position to a full reoptimiza-
importance, for example, head and neck tumors tion, based on patients’ anatomy [37, 38].
Similarly, OARs and target contours can be either Presently, the Unity system is implementing a
rigidly translated from the reference plan or fully gating system to account for target movement
recontoured, based on daily anatomical during the delivery phase. Moreover, this system
differences. provides different MRI sequences that can be
Another important issue is interobserver con- acquired during beam on phase [45]. These
touring variation. To account for this, each center options enable the reconstruction of the delivered
should assess the accuracy of its recontouring dose to be compared with the intended dose
process against a reference gold standard and delivery for more precise treatment evaluation
implement a QA program for online contouring and/or dose response assessment. This system
procedures. In order to make the recontouring can be used for offline adaptation, but in a next
process faster, consistent, and compatible with phase can be used for intra-fraction dose-guided
the daily clinical routine, several centers have adaptation [46, 47].
adopted a workflow with a partial recontouring of
OAR only in the proximity of PTV (i.e., the
2–3 cm surrounding area), since the recontouring raining and Education Procedures
T
is performed with the patient waiting in the treat- for MR-Specific Patient Selection
ment position [37, 39]. In this case, OAR con- and Indications for MR-Linac
straints have to be adapted to this concept (use of
absolute volume in cc) for a fast and reliable plan Patient selection is a crucial point in MRgRT
evaluation. which is provided in detail in Chap. 3. Selection
Some pre-treatment procedures can be time criteria and training are a fundamental part to
consuming, exceeding 15 min. In this case a pre- correctly homogeneously address patients to
delivery, or verification, MRI can be acquired to MR-Linac treatments. Among the most impor-
ensure the correct positioning of the planned iso- tant criteria are patient characteristics, target vol-
center. This image can be used to further opti- ume characteristics, and type. Since this is still a
mize patient positioning with small couch or novel technology, it is not feasible to address all
plan shifts, or conduct a full reoptimization in the possible diseases amenable for treatment on
the case of major and unacceptable target/OAR MR-Linac; however, the inclusion of patients
changes [40]. within a prospective protocol is suggested. The
Each center should define the level of action evaluation of patient’s allocation should be based
and threshold in the case of target movement out- according to the principles of health technology
side of the high-dose area, which is dependent assessment for cost-benefit analysis and the prin-
on, among the other factors, the used PTV mar- ciples of evidence-based medicine. There are
gins. The MRIdian system allows for respiratory- several ongoing clinical trials on MRgRT, as well
gated treatments using breath-hold techniques as published data. However, the acquisition of
with automated real-time anatomy structure strong evidence through prospective multicenter
tracking for lesions in the abdomen and thorax. cohorts (i.e., MOMENTUM study) is of outmost
Several visual feedback modalities for patients importance.
have been implemented in the clinical practice All patients should be screened for MRI com-
(i.e., MRI-compatible monitors in combination patibility, according to the adopted guidelines
with an adjustable mirror in the gantry or prism [48]. Patients should be classified as physically
glasses) [41–44]. These modalities facilitate vol- incompatible (i.e., non-MRI-conditional pace-
untary breath-hold delivery at the appropriate maker carriers, severe obesity), clinically incom-
respiratory phase. Preliminary experience shows patible (i.e., severe psychiatric disorder, severe
that this treatment modality is well tolerated or claustrophobia, inability to understand the
appreciated by patients and can reduce the duty instructions), borderline compatible (i.e., mild
cycle between 67% and 87% in adrenal and pan- claustrophobia), or fully compatible for
creatic tumor-gated MRgRT [41]. MRgRT. Patients unfit for MRgRT should be
referred to standard RT, while supportive inter- Also, moving targets might benefit from
vention (i.e., psychological intervention, anesthe- MRgRT, especially when they are close to OARs.
sia, pharmacological or supportive technique In fact with the daily adaptation it is possible to
such as music or aromatherapy) could be account not only for OAR position, but also fol-
implemented for borderline compatible patients low the tumor using motion management tools
[48–51]. and, when available, gating systems [58–60].
Other than physical compatibility, physicians This can be of particular interest in lung and pan-
should evaluate the general clinical status of the creatic tumors, liver and kidney lesions, head and
patient, especially regarding the balance between neck tumors, oligometastases, lymph nodes, and
the compliance degree with MRgRT and the reirradiation. Moreover, MRgRT is indicated in
expected benefit from it. In fact, treatment slots all the cases where MRI can provide additional
on a MR-Linac are generally longer than that on information (i.e., tumor shrinkage, radiomics,
a conventional RT in a range between 20 and early toxicity onset).
60 min [39, 52]. For this reason, elderly or frailty
does not represent an automatic exclusion crite-
rion, but a scoring system can be used to support Backup Solutions
the most suitable treatment modality [53].
Patients with severe obesity or cachexia (body Considering the elevated cost of an MR-Linac, it
mass index >40 or weight <40 kg) should be is unlikely that a single department might have
evaluated on a case-by-case basis, especially more than one system. Therefore, since a direct
considering the possibility of having artifacts, transfer from an MR-Linac to a conventional
patient heating, and bore size issue (gantry diam- Linac is not easy to plan, it is important to iden-
eter <70 cm) [6, 54]. tify alternative strategies to avoid unnecessary
The second selection criterion is the target treatment interruptions. The best solution might
characteristics and their relation with the sur- be represented by a collaboration with nearby
rounding normal tissues. The MR-Linac offers centers harboring the same technology with the
higher soft tissue contrast visibility, which may aim to ensure treatment continuation with the
allow safe and accurate dose administration in same standards, especially in those situations
scenarios with OARs in close proximity to the when long treatment interruption might affect the
target. Published clinical data is showing that clinical outcome (i.e., head and neck tumors).
SBRT might be safely administered to prostate, Another solution is represented by planning an
lymph node, and pancreas sites. In particular, alternative plan on a conventional Linac, using
early comparative studies in prostate settings are the same patient setup and the simulation CT
reporting a lower incidence of side effects, while used for electron density, or a new simulation
in the case of pancreas SBRT the rate of grade 3 CT. This last solution requires new plan summa-
or higher side effects is very limited using both tion and dose accumulation modalities that
available technologies [55]. restrict the adoption of this modality as routine.
Considering the evidence, the MR-Linac
might be best exploited in those situations where
CT density is homogeneous and the ability to dis- References
tinguish the tissue is reduced (i.e., head and neck,
upper abdomen) and the higher soft tissue con- 1. Landry G, Corradini S, Belka C. Magnetic resonance-
guided radiation therapy: the beginning of a new era.
trast provided by MRI can overcome this prob- Radiat Oncol. 2020;15:163.
lem. A typical case is represented by prostate 2. Corradini S, Alongi F, Andratschke N, et al.
cancer, where MRI might reduce the daily uncer- MR-guidance in clinical reality: current treatment
tainties, compared to cone-beam CT, by a more challenges and future perspectives. Radiat Oncol.
2019;14:92.
accurate definition of the prostate-rectum inter- 3. Mutic S, Dempsey JF. The ViewRay system: magnetic
face and the correct identification of the prostate resonance-guided and controlled radiotherapy. Semin
apex and penile bulb [56, 57]. Radiat Oncol. 2014;24:196–9.
4. Raaymakers BW, Jürgenliemk-Schulz IM, Bol GH, 19. Botman R, Tetar SU, Palacios MA, et al. The clini-
et al. First patients treated with a 1.5 T MRI-Linac: cal introduction of MR-guided radiation therapy
clinical proof of concept of a high-precision, high- from a RTT perspective. Clin Transl Radiat Oncol.
field MRI guided radiotherapy treatment. Phys Med 2019;18:140–5.
Biol. 2017;62(23):L41–50. Published 2017 Nov 14. 20. Tanadini-Lang S, Budgell G, Bohoudi O, et al. An
5. Acharya S, Fischer-Valuck BW, Kashani R, et al. ESTRO-ACROP guideline on quality assurance
Online magnetic resonance image guided adaptive and medical physics commissioning of online MRI
radiation therapy: first clinical applications [published guided radiotherapy systems based on a consensus
correction appears in Int J Radiat Oncol biol Phys. expert opinion. Radiother Oncol. 2023;181:109504.
2016 Sep 1;96(1):243]. Int J Radiat Oncol Biol Phys. 21. Green O, Henke LE, Parikh P, et al. Practical impli-
2016;94(2):394–403. cations of ferromagnetic artifacts in low-field MRI-
6. Klüter S. Technical design and concept of a 0.35 T guided radiotherapy. Cureus. 2018;10:e2359.
MR-Linac. Clin Transl Radiat Oncol. 2019;18:98–101. 22. van der Heide UA, Frantzen-Steneker M, Astreinidou
7. Lagendijk JJW, Raaymakers BW, Raaijmakers E, et al. MRI basics for radiation oncologists. Clin
AJE, et al. MRI/linac integration. Radiother Oncol. Transl Radiat Oncol. 2019;18:74–9.
2008;86:25–9. 23. Jonsson J, Nyholm T, Söderkvist K. The rationale for
8. Raaymakers BW, Lagendijk JJ, Overweg J, et al. MR-only treatment planning for external radiother-
Integrating a 1.5 T MRI scanner with a 6 MV apy. Clin Transl Radiat Oncol. 2019;18:60–5.
accelerator: proof of concept. Phys Med Biol. 24. Kurz C, Buizza G, Landry G, et al. Medical physics
2009;54(12):N229–37. challenges in clinical MR-guided radiotherapy. Radiat
9. Raaijmakers AJE, Hårdemark B, Raaymakers BW, Oncol. 2020;15:1–16.
et al. Dose optimization for the MRI-accelerator: 25. ESTRO. In-room MRI-guided RT. n.d. https://www.
IMRT in the presence of a magnetic field. Phys Med estro.org/Courses/In-room-MRI-guided-RT.
Biol. 2007;52:7045–54. 26. Utrecht U. MRI in radiotherapy course. n.d.
10. Woodings SJ, Bluemink JJ, de Vries JHW, et al. Beam h t t p s : / / w w w. u m c u t r e c h t . n l / n l / z i e k e n h u i s /
characterisation of the 1.5 T MRI-linac. Phys Med mri-in-radiotherapy-course.
Biol. 2018;63:85015. 27. Research TI of C. Magnetic resonance image guided
11. Park JM, Park S-Y, Wu H-G, Kim J. Commissioning radiotherapy (MRIgRT). n.d. http://www.icr.ac.uk/
experience of tri-Cobalt-60 MRI-guided radiation studying-and-training/opportunities-for-clinicians/
therapy system. Prog Med Phys. 2015;26:193. radiotherapy-a nd-i maging-t raining-c ourses/
12. Winkel D, Bol GH, Kroon PS, van Asselen B, et al. magnetic-r esonance-i mageguided-r adiotherapy-
Adaptive radiotherapy: the Elekta Unity MR-Linac (mrigrt).
concept. Clin Transl Radiat Oncol. 2019;18:54–9. 28. Adair Smith G, Dunlop A, Alexander SE, Barnes H,
13. McNair HA, Wiseman T, Joyce E, et al. International Casey F, Chick J, Gunapala R, Herbert T, Lawes R,
survey; current practice in on-line adaptive radiother- Mason SA, Mitchell A, Mohajer J, Murray J, Nill
apy (ART) delivered using magnetic resonance image S, Patel P, Pathmanathan A, Sritharan K, Sundahl
(MRI) guidance. Tech Innov Patient Support Radiat N, Tree AC, Westley R, Williams B, McNair
Oncol. 2020;16:1–9. HA. Evaluation of therapeutic radiographer con-
14. McGee KP, Tyagi N, Bayouth JE, et al. Findings of touring for magnetic resonance image guided online
the AAPM Ad Hoc committee on magnetic resonance adaptive prostate radiotherapy. Radiother Oncol.
imaging in radiation therapy: unmet needs, opportu- 2023;180:109457.
nities, and recommendations [published correction 29. Boldrini L, Cusumano D, Chiloiro G, et al. Delta
appears in Med Phys. 2023 Mar;50(3):1943]. Med radiomics for rectal cancer response prediction with
Phys. 2021;48(8):4523–31. hybrid 0.35 T magnetic resonance-guided radiother-
15. Klüter S, Schrenk O, Renkamp CK, et al. A practical apy (MRgRT): a hypothesis-generating study for an
implementation of risk management for the clinical innovative personalized medicine approach. Radiol
introduction of online adaptive Magnetic Resonance- Med. 2019;124:145.
guided radiotherapy. Phys Imaging Radiat Oncol. 30. Cusumano D, Boldrini L, Yadav P, et al. External
2021;17:53–7. validation of early regression index (ERI(TCP))
16. Younge KC, Wang Y, Thompson J, et al. Practical as predictor of pathologic complete response in
implementation of failure mode and effects analysis rectal cancer using magnetic resonance-guided
for safety and efficiency in stereotactic radiosurgery. radiation therapy. Int J Radiat Oncol Biol Phys.
Int J Radiat Oncol Biol Phys. 2015;91:1003–8. 2020;108:1347.
17. Huq MS, Fraass BA, Dunscombe PB, et al. A 31. Bibault J-E, Xing L, Giraud P, et al. Radiomics: a
method for evaluating quality assurance needs in primer for the radiation oncologist. Cancer Radiother.
radiation therapy. Int J Radiat Oncol Biol Phys. 2020;24:403–10.
2008;71:S170–3. 32. Finazzi T, Haasbeek CJA, Spoelstra FOB, et al.
18. Noel CE, Santanam L, Parikh PJ, et al. Process-based Clinical outcomes of stereotactic MR-guided adap-
quality management for clinical implementation of tive radiation therapy for high-risk lung tumors. Int J
adaptive radiotherapy. Med Phys. 2014;41:81717. Radiat Oncol Biol Phys. 2020;107:270–8.
33. Finazzi T, Palacios MA, Spoelstra FOB, et al. Role 47. Kontaxis C, de Muinck Keizer DM, et al. Delivered
of on-table plan adaptation in MR-guided ablative dose quantification in prostate radiotherapy using
radiation therapy for central lung tumors. Int J Radiat online 3D cine imaging and treatment log files on a
Oncol Biol Phys. 2019;104:933–41. combined 1.5T magnetic resonance imaging and lin-
34. Finazzi T, Palacios MA, Haasbeek CJA, et al. ear accelerator system. Phys Imaging Radiat Oncol.
Stereotactic MR-guided adaptive radiation ther- 2020;15:23–9.
apy for peripheral lung tumors. Radiother Oncol. 48. Sammet S. Magnetic resonance safety. Abdom
2020;144:46–52. Radiol. 2016;41:444–51.
35. Boldrini L, Cusumano D, Cellini F, et al. Online 49. Chadderdon AL, Carns DR, Pudalov LR, et al.
adaptive magnetic resonance guided radiotherapy for Underlying mechanisms of psychological interven-
pancreatic cancer: state of the art, pearls and pitfalls. tions in magnetic resonance imaging and image-
Radiat Oncol. 2019;14:71. guided radiology procedures. Top Magn Reson
36. Winkel D, Bol GH, Werensteijn-Honingh AM, et al. Imaging. 2020;29:157–63.
Evaluation of plan adaptation strategies for stereotac- 50. Walworth DD. Effect of live music therapy for
tic radiotherapy of lymph node oligometastases using patients undergoing magnetic resonance imaging. J
online magnetic resonance image guidance. Phys Music Ther. 2010;47:335–50.
Imaging Radiat Oncol. 2019;9:58–64. 51. Schellhammer F, Ostermann T, Krüger G, et al. Good
37. Bohoudi O, Bruynzeel AME, Senan S, et al. Fast scent in MRI: can scent management optimize patient
and robust online adaptive planning in stereotactic tolerance? Acta Radiol. 2013;54:795–9.
MR-guided adaptive radiation therapy (SMART) for 52. Werensteijn-Honingh AM, Kroon PS, Winkel D,
pancreatic cancer. Radiother Oncol. 2017;125:439–44. et al. Feasibility of stereotactic radiotherapy using
38. Kim J, Park JM, Choi CH, et al. Retrospective study a 1.5 T MR-Linac: multifraction treatment of pel-
comparing MR-guided radiation therapy (MRgRT) vic lymph node oligometastases. Radiother Oncol.
setup strategies for prostate treatment: repositioning 2019;134:50–4.
vs. replanning. Radiat Oncol. 2019;14:139. 53. Boldrini L, Colloca GF, Villani E, et al. Magnetic
39. Tetar SU, Bruynzeel AME, Lagerwaard FJ, et al. resonance-guided radiotherapy feasibility in elderly
Clinical implementation of magnetic resonance imag- cancer patients: proposal of the MASTER scoring
ing guided adaptive radiotherapy for localized prostate system. Tumori. 2021;107(1):26–31.
cancer. Phys Imaging Radiat Oncol. 2019;9:69–76. 54. Graves SA, Snyder JE, Boczkowski A, et al.
40. van Timmeren JE, Chamberlain M, Krayenbuehl J, Commissioning and performance evaluation of
et al. Treatment plan quality during online adaptive RadCalc for the Elekta unity MRI-linac. J Appl Clin
re-planning. Radiat Oncol. 2020;15:1–11. Med Phys. 2019;20:54–62.
41. van Sörnsen de Koste JR, Palacios MA, Bruynzeel 55. Corradini S, Alongi F, Andratschke N, et al.
AME, et al. MR-guided gated stereotactic radiation MRguidance in clinical reality: current treatment
therapy delivery for lung, adrenal, and pancreatic challenges and future perspectives. Radiat Oncol.
tumors: a geometric analysis. Int J Radiat Oncol Biol 2019;14:1–12.
Phys. 2018;102:858–66. 56. Alongi F, Rigo M, Figlia V, et al. 1.5 T MR guided
42. Green OL, Rankine LJ, Cai B, et al. First clini- and daily adapted SBRT for prostate cancer: feasibil-
cal implementation of real-time, real anatomy ity, preliminary clinical tolerability, quality of life and
tracking and radiation beam control. Med Phys. patient-reported outcomes during treatment. Radiat
2018;45:3728–40. Oncol. 2020;15:69.
43. Tetar S, Bruynzeel A, Bakker R, et al. Patient reported 57. Mazzola R, Figlia V, Rigo M, et al. Feasibility
outcome measurements on the tolerance of magnetic and safety of 1.5 T MR-guided and daily adapted
resonance imaging-guided radiation therapy. Cureus. abdominal-pelvic SBRT for elderly cancer patients:
2018;10:e2236. geriatric assessment tools and preliminary patien-
44. Klüter S, Katayama S, Spindeldreier CK, et al. First treported outcomes. J Cancer Res Clin Oncol.
prospective clinical evaluation of feasibility and 2020;146:2379.
patient acceptance of magnetic resonance-guided 58. Park JM, Wu H-G, Kim HJ, et al. Comparison of
radiotherapy in Germany. Strahlenther Onkol. treatment plans between IMRT with MR-Linac and
2020;196:691–8. VMAT for lung SABR. Radiat Oncol. 2019;14:105.
45. Jackson S, Glitzner M, Tijssen RHN, Raaymakers 59. Henke L, Kashani R, Robinson C, et al. Phase
BW. MRI B 0 homogeneity and geometric dis- I trial of stereotactic MR-guided online adap-
tortion with continuous linac gantry rotation tive radiation therapy (SMART) for the treat-
on an Elekta Unity MR-Linac. Phys Med Biol. ment of oligometastatic or unresectable primary
2019;64(12):12NT01. malignancies of the abdomen. Radiother Oncol.
46. Menten MJ, Mohajer JK, Nilawar R, et al. Automatic 2018;126:519–26.
reconstruction of the delivered dose of the day using 60. Witt JS, Rosenberg SA, Bassetti MF. MRI-guided
MR-Linac treatment log files and online MR imaging. adaptive radiotherapy for liver tumours: visualising
Radiother Oncol. 2020;145:88–94. the future. Lancet Oncol. 2020;21:e74–82.
An Introduction to MR Sequences
Used in Oncology with Application
5
to MR-Linac
Hersh Sagreiya, Karthik Sundaram,

Jooyoung (James) Sohn, and Poonam Yadav
Introduction sequences can highlight areas of fluid or edema,

fat saturation images can establish the presence
Magnetic resonance imaging (MRI) is a powerful of fat in tissues, and contrast-enhanced sequences
imaging modality that has several advantages can aid in identifying tumors or abscesses. MRI
over CT. Unlike CT, MRI does not require the use is commonly used as a problem-solving modal-
of ionizing radiation. Radiation exposure is of ity, and it is frequently recommended as a follow-
particular concern for children and young adults. up when there are indeterminate findings on CT
They are more sensitive to the stochastic effects or ultrasound. In fact, the use of multiple
of ionizing radiation due to rapid cell division sequences can help to narrow the differential
related to growth and development, their long diagnosis to a specific entity. In addition, MRI is
remaining lifespan allows more time to experi- able to provide superior soft tissue contrast. In
ence the deleterious effects of radiation, and chil- this example, sagittal T1, T2, and T1 postcontrast
dren could potentially receive a higher effective images of the uterus show excellent soft tissue
radiation dose if optimized pediatric protocols differentiation between the uterus and endome-
are not applied [1]. In addition, cumulative radia- trium (Fig. 5.1a–c). This highlights structures
tion dose also becomes more of a concern if serial within the uterus, such as Nabothian cysts, the
imaging is required, particularly in otherwise surrounding fat, and nearby structures (bladder,
healthy patients. bowel, spine, and musculature). Moreover, dif-
MRI also provides the ability to image using ferent sequences highlight different aspects of the
various sequences, which highlight different anatomy, such as the endometrium. This stands in
aspects of anatomy. For instance, T2-weighted comparison to CT, which only shows one view of
the anatomy (Fig. 5.1d). While ultrasound can
H. Sagreiya (*) · K. Sundaram also show similar anatomy, the relationship
Hospital of the University of Pennsylvania, between structures is not as clearly demarcated,
Department of Radiology, Division of Abdominal and there is a potential issue of operator depen-
Imaging, Philadelphia, PA, USA
dence, with image quality related to the level of
[email protected] training of the sonographer (Fig. 5.1e). The
improved soft tissue contrast of MRI can poten-
J. (J) Sohn · P. Yadav
Northwestern University Feinberg School of tially be important in certain scenarios. For
Medicine, Department of Radiation Oncology, instance, pancreatic ductal adenocarcinoma can
Chicago, IL, USA occasionally appear as isodense (similar in den-
sity) to the pancreatic parenchyma on CT, with
[email protected]

46 H. Sagreiya et al.
a b c
d e
Fig. 5.1 Differences in soft tissue differentiation between sequences. Ultrasound highlights different aspects of the
MRI, CT, and ultrasound for the same patient. MRI shows image. (a) T1 pre-contrast. (b) T2. (c) T1 postcontrast. (d)
better soft tissue differentiation than CT and can highlight CT pelvis with contrast. (e) Ultrasound pelvis
specific soft tissue characteristics through different
the only indication being a dilated pancreatic products from calcifications [6]. Moreover, CT is
duct if there is resultant obstruction [2]. However, typically preferred over MRI for generalized
the use of multiple MRI sequences (T1, T2, diffu- oncologic staging, with MRI typically reserved
sion, contrast enhanced) will typically readily for further characterization of suspicious find-
show the tumor in equivocal CT cases. ings. However, MRI is particularly useful for
MRI does have a few potential disadvantages. local staging for certain cancers, such as prostate,
First of all, MRI is typically more expensive, rectal, and gynecologic cancers, due to its
time-consuming, and less readily available than superior soft tissue contrast. Whole body MRI is
CT [3]. However, there has been the development currently used for cancer screening in patients
of abbreviated MR protocols, often targeted to a with syndromes that predispose them to develop-
particular indication, which eliminate unneces- ing cancer, including Li-Fraumeni syndrome,
sary sequences that provide redundant informa- constitutional mismatch repair deficiency, and
tion, and are hence able to decrease imaging hereditary paraganglioma-pheochromocytoma
time, interpretation time, and ultimately imaging syndromes [7]. These patients are at high risk for
cost [4]. In addition, although certain findings are developing tumors, and a major advantage of
more obvious on CT, such as calcifications, calci- MRI in this context is the lack of ionizing radia-
fication can be appreciated on MRI [5]. For tion, which facilitates serial imaging. Along the
instance, phase information from susceptibility- same lines, while whole body MRI screening of
weighted imaging can help to discriminate blood asymptomatic, low-risk individuals has been pos-
5 An Introduction to MR Sequences Used in Oncology with Application to MR-Linac 47
tulated, there are potential concerns regarding tion of this dipolar magnet from quantum
discovering indeterminate or false-positive find- mechanics is analogous to the description of
ings, with potentially deleterious psychological spinning objects from classical mechanics
effects for the patient, as well as expensive fol- [10]. Without a magnetic field, there is a ran-
low-up [8]. Finally, MRI is also used for screen- dom alignment of the magnetic poles of pro-
ing in specific scenarios, such as for the tons, and there is no net magnetization vector
development of hepatocellular carcinoma in (Fig. 5.2). However, in the presence of a mag-
patients with cirrhosis, for which it may have bet- netic field (B0), the protons align both parallel
ter performance than ultrasound [9]. and antiparallel to the magnetic field. As more
protons align parallel to rather than antiparallel
to the magnetic field, there is a net magnetiza-
Basics of MR Physics tion vector.
In MRI, the strength of the magnetic field is
MRI takes advantage of magnetically active measured in Tesla (T). While 1.5T and 3T are
nuclei. In order for nuclei to be magnetically commonly used in clinical scenarios, research
active, they must have a net charge. Magnetic indications can use even higher strength magnets
resonance imaging typically takes advantage (7T and beyond). Greater magnetic field strength
of the hydrogen ion, which is widely available increases the signal-to-noise ratio. In the pres-
throughout the human body as a fundamental ence of a magnetic field, protons precess or spin
molecular building block. These nuclei have an along the direction of magnetization (the axis of
intrinsic property, conceptually thought of as B0). The angular momentum of the precession of
spin, conceived as the nucleus spinning around protons in the presence of a magnetic field is
its central axis; although the nucleus does not related to the strength of the magnetic field by the
spin in the classical sense, it does have a mag- Larmor equation:
netic moment, creating a local magnetic field ω = γ B0
with a north and south pole, and the descrip-
Fig. 5.2 (Left) Random alignment of protons without a more protons are aligned parallel (yellow) rather than
magnetic field. (Right) Parallel and antiparallel alignment antiparallel to the magnetic field (green), there is a net
of protons within an external magnetic field (B0). Because magnetization vector (NMV)
In this equation, ω is the angular momentum In order to localize a particular point in 3D

(which is related to the frequency of precession), space, magnetic gradients must be applied in the
γ is the gyromagnetic ratio, and B0 is the strength x, y, and z directions. This requires gradient coils
of the magnetic field. in each orthogonal direction that gradually vary
A current running through a solenoid or a the magnetic field. As a result, for any given gra-
large coil generates a magnetic field [11]. In dient, the strength of the magnetic field is a func-
order to generate an MR image, a transmit coil tion of location along that axis. Since the
must first transmit a radiofrequency (RF) excita- frequency of precession is directly proportional
tion pulse. Protons that are precessing at the fre- to the strength of the magnetic field, the set of
quency of the RF pulse will become excited and frequencies of the transmitted RF pulse will
go from a low to a high energy state. In doing so, excite protons in a particular location in three-
the net magnetization vector tilts away from the dimensional space, and the subsequent relaxation
axis of the magnetic field and into the transverse of these protons will emit an RF pulse that even-
plane (Fig. 5.3). As the protons relax and return tually leads to the MR image.
to their previous state, they will emit an RF pulse. The first gradient that is applied along the
This is what is captured as the MR image. During z-axis, or the axis along B0, is the slice select gra-
this process, the net magnetization vector returns dient. This gradient selects a particular axial slice
to the direction of B0. in 3D space. Subsequent gradients localize data
Fig. 5.3 The net magnetization vector is initially in the direction of B0. After an RF pulse, net magnetization flips into
the transverse plane
along the x-axis and the y-axis. First, the gradient filling of k-space. For instance, compressed sens-
along the x-axis, known as the frequency- ing can accelerate MR image acquisition times
encoding gradient, is applied, encoding informa- by undersampling k-space and hence obtaining
tion along the x-axis. This includes two parts, a less data [15]. Even newer techniques have lever-
phasing and a dephasing lobe. Next, the gradient aged deep learning for compressed-sensing-
along the y-axis, known as the phase-encoding based acceleration of MR image acquisition [16].
gradient, is applied, encoding information along As previously mentioned, in order to generate an
the y-axis. The specific pattern in which the RF interpretable image, an inverse Fourier transform
pulse and various gradients are applied results in is performed on the K-space data.
different MR sequences, which are discussed The amount of time that it takes to create an
later in this chapter (Fig. 5.6). The subsequently MR image depends on three factors—the repeti-
emitted RF pulse is detected by a receive coil. tion time (TR), the number of excitations for
Phased-array coils consist of multiple coils that phase-encoding (NEX), and the number of
receive the MR signal. Parallel imaging is a tech- signals averaged (NSA) [17]. Since the vast

nique that can take advantage of phased-array majority of patients are narrower in the anterior-
radiofrequency coils by undersampling the data to-poster dimension, it is typically the phase-
(i.e., reducing the number of sampled k-space encoding direction for the sake of minimizing
lines, further described below) and hence reduc- examination time.
ing acquisition times [12]. The data from the
multiple coils is combined to create the image.
Next, the signal is amplified and then converted T1-Weighted Imaging, T2-Weighted
from analog to digital data. This digital data is Imaging, and Pulse Sequences
subsequently used to create the image.
The Fourier transformation relies on the con- The ability to differentiate within or between two
cept that a complex signal can be rewritten as the tissues (contrast) allows a radiological image to
sum of a series of waves, and the Fourier trans- be of diagnostic value for clinical, research, or
form decomposes this signal into the frequencies therapeutic purposes. Tissue contrast in modali-
and amplitudes of its constituent waves; in MRI, ties such as US and CT can be considered “fixed,”
the signal in question is the MR echo containing as differences between tissues can typically be
frequency and phase-encoded spatial information manipulated after acquiring the raw data or
needed to construct the image, and the Fourier altered by using exogenous contrast. Magnetic
transform applied to this complex signal repre- resonance imaging (MRI) allows for user altera-
sents it in k-space, a 2D Fourier space organizing tion of contrast before image acquisition through
information about spatial frequency and ampli- the choice of pulse sequence and parameters.
tude [13]. Going from k-space to the final image All MR images are affected by three constants
is performed using a 2D inverse Fourier trans- that affect tissue contrast: a longitudinal relax-
form. K-space contains frequency and phase ation constant (Fig. 5.4), a transverse relaxation
data, unlike the typical data one sees in images. constant (Fig. 5.4), and proton density—the den-
The center of k-space typically contains low spa- sity of protons in a voxel. The basic spin-echo
tial frequency image information about image sequence best describes the relationship between
shape and contrast, consisting of the majority of these parameters and signal intensity (SI) from
the signal, whereas the periphery of k-space has an image [18]:
high spatial frequency information about fine
 − TR
 − TE
details (edges, interface between tissues), con- SI ( TE,TR ) = ρ 1 − e T1  ∗ e T 2
sisting of a relatively weaker signal [14]. The  
amount of time that the patient is in the scanner where ρ represents a scaling factor proportional to
and the type of sequence obtained are both related the proton density, T1 represents the longitudinal
to the amount of time that it takes to fill k-space. relaxation, and T2 represents the transverse relax-
There are newer techniques that allow for faster ation. Manipulation of parameters such as echo
Fig. 5.4 T1 is related to relaxation of longitudinal magnetization, and T2 is related to relaxation of transverse
magnetization
time (TE), the time between excitation and the tion of the T1 time is the time that it takes for
middle of the acquired echo, and repetition time approximately 63% of the z-axis magnetization
(TR), the time between each successive excitation, to return to its equilibrium [20].
can fundamentally change the type of image. By In T1-weighted sequences, tissues, specifi-
varying TE and TR, the SI can be made “weighted” cally the hydrogen protons in tissues, with inher-
to differences in the T1 values of tissues (TR ~ T1, ently short T1 relaxation time demonstrate high
TE ≪ T2), T2 values (TR ≫ T1, TE ~ T2), or a signal intensity compared to tissues with long
proton density (TR ≫ T1, TE ≪ T2). relaxation times. In the case of spin-echo
There are variations from the spin-echo sequence acquisition, this can be achieved by set-
sequence now used in clinical MRI scanners, ting TE and TR to low values.
including gradient-echo sequences, spoiled gra- In general, T1-weighted imaging can provide
dient recalled echo, and inversion recovery anatomical details, especially useful for imaging
sequences, among others [19]. the head and neck, as well as the bones. Tissues
that appear bright include fat, blood products
(methemoglobin), slow-moving blood, radiation
T1-Weighted Sequences changes, melanin, and paramagnetic contrast
agents (i.e., gadolinium) (Fig. 5.5a and b) [21,
The magnetization vector realigns with the axis 22]. Tissues that appear dark include air, bones
of B0 through the process of T1 recovery, during (noting that marrow in bones will appear bright),
which the longitudinal magnetization increases and fast-flowing blood. T1 values are affected by
in magnitude or recovers (Fig. 5.4). The defini- magnetic field strength, but T2 values are not.
a b
c d
Fig. 5.5 WHO Grade IV glioblastoma. (a) T1. (b) T1 postcontrast. (c) T2. (d) T2 FLAIR
T2-Weighted Sequences magnetic fields of spinning nuclei—and T2*

decay—due to interaction with local magnetic
Transverse relaxation, or spin-spin relaxation, field inhomogeneities. In the spin-echo sequence
actually represents a decay of the transverse mag- (often used for T2-weighted images), the 180°
netization (loss of signal) through both T2 pulse (Fig. 5.6) refocuses field inhomogeneities,
decay—random interaction of the individual and hence T2* decay is negligible. This is not the
Fig. 5.6 Spin-echo sequence diagram. This includes the amount of time between the RF pulse and the MR signal.
RF pulse, the application of the slice selection, phase The repetition time (TR) is the amount of time between
encoding, and frequency encoding gradients, and the successive MR pulses
receipt of the MR signal. The echo time (TE) is the
case in gradient-echo recall sequences, which are Single-shot fast spin echo (SSFSE) consists of an
often used for T1-weighted images. T2 time is excitation by the RF pulse, followed by a rapid
defined as the time at which approximately 63% succession of 180-degree refocusing pulses,
of the nuclei spinning perpendicular to the main which results in all the echoes needed to fill
field have lost their phase coherence (Fig. 5.4) k-space for a given slice. The more rapid acquisi-
[19]. Hence, tissues with long T2 relaxation times tion of FSE and particularly SSFSE reduces the
(fluids) appear brighter than those with short T2 potential for motion artifact. The rapid succession
relaxation times. T2-weighted sequences are of refocusing pulses helps to mitigate susceptibil-
acquired at long TE and long TR [22]. ity artifact. That is because the refocusing pulses
in spin echo minimize T2* effects related to mag-
netic field inhomogeneities. Like spin echo, gradi-
ariations in Spin Echo
V ent-echo sequences used in modern scanners can
and Gradient Echo also leverage multiple echoes to decrease imaging
time. For instance, the EPI sequence, used in dif-
Fast spin echo (FSE) consists of an excitation by fusion-weighted imaging, obtains all necessary
the RF pulse, followed by multiple 180-degree k-space data after a single excitation by the RF
refocusing pulses occurring before the next exci- pulse. Flip angle is an important concept in gradi-
tation pulse, with resultant echoes more rapidly ent echo. While spin echo typically uses a
filling k-space. The echo train length (ETL) repre- 90-degree flip angle, gradient echo can use lower
sents the number of resultant echoes per RF pulse. flip angles, which affects the recovery of longitu-
dinal magnetization and ultimately T1-weighting. Diffusion-Weighted Imaging

In gradient echo, increasing the flip angle
increases T1-weighting. Diffusion-weighted imaging (DWI) is based on
In steady-state images, the TR value is short. As initial observations that spin-echo sequences are
a result, there is only partial recovery of longitudi- sensitive to the diffusion of water molecules [18,
nal magnetization and decay of transverse magne- 25]. This provides a noninvasive means of detect-
tization before the application of the next RF ing changes in the random motion of water mole-
excitation pulse, once again encouraging a higher cules due to thermal energy (Brownian motion)
energy state. The interplay between magnetic by MRI. Molecule movement can be described by
relaxation and RF excitation eventually reaches a a Gaussian function where the mean-squared dis-
steady state. In these kinds of rapidly acquired placement of water depends on the diffusion coef-
sequences, fluid appears hyperintense, and they ficient of the molecule (water) and time. The free
are useful for imaging vascular structures. movement of water becomes more restricted in
tissues due to complex factors including cellular-
ity, membrane permeability, and extracellular vol-
Clinical Applications ume fraction. Hence, the diffusion coefficient of
water is lower in tissues that restrict movement.
These sequences are often implemented for ana- Acquisition of images and quantification of an
tomic imaging and treatment planning. However, apparent diffusion coefficient (ADC) are based
they are also useful in diagnostic imaging to on the application of diffusion gradients to pulse
identify pathology, as fluid (cysts, bile, cerebro- sequences to make images sensitive to these
spinal fluid) and edema appear bright or hyperin- molecular changes. An initial T2-weighted image
tense on T2-weighted imaging. Moreover, with fat suppression is acquired with low gradi-
inflammation (from injury), infections, and radia- ents (b = 0–50 s/mm2). Next, one or more addi-
tion changes appear hyperintense on T2-weighted tional images are acquired by increasing the
imaging (Fig. 5.5c). For abdomen and pelvis, diffusion weighting (i.e., b = 100, 500, 1000 s/
T2-weighted imaging is the “workhorse” mm2, etc.), which leads to decreasing signal
sequence for evaluation of anatomy, pathology, intensities from tissues that allow more water dif-
and treatment planning. In a recent study, fusion, while those that “restrict” water move-
T2-weighted images were the preferred weight- ment demonstrate a higher signal. This allows the
ing for radiation treatment in most sites, except quantification of an apparent diffusion coefficient
the chest wall/breast and the lung/esophagus (ADC) and generation of an ADC map image.
where T1-weighting was preferred [23]. Hence, tumors often demonstrate high signal
intensity on high b-value diffusion-weighted
images while demonstrating low signal on gener-
Proton Density ated ADC maps. Lesions with intrinsic T2 hyper-
intensity that demonstrate high signal on both the
In proton density (PD) weighted images, tissues DWI and ADC maps exhibit T2 shine through
with a higher concentration of protons produce and do not represent true restricted diffusion.
the strongest signal and appear the brightest. This In oncological imaging, DWI-MR has been
is achieved by minimizing the impact of T1 and tested in almost all types of tumors in order to
T2 differences in tissues by acquiring at long TR differentiate benign from malignant disease, to
and short TE. The signal intensities of tissues are grade malignancies, to predict and assess treat-
similar to T2-weighted sequences where air and ment responses, and to identify residual or recur-
bones are dark, while fluid and fat are bright. PD rent tumors on follow-up exam [26]. An example
sequences are useful in the brain to differentiate case is shown for osteosarcoma of the humerus
gray and white matter and are useful in musculo- (Fig. 5.7). Sagittal STIR and sagittal T1 postcon-
skeletal/extremity imaging [24]. trast images demonstrate an irregular lesion cen-
a b
c d
Fig. 5.7 Osteosarcoma of the humerus with periosteal reaction and extraosseous component. (a) Sagittal T1 STIR. (b)
Sagittal T1 postcontrast. (c) Diffusion-weighted imaging. (d) Apparent diffusion coefficient
tered on the humerus, with associated weighted imaging and hypointense or dark on the
enhancement and extension into the soft tissues apparent diffusion coefficient map (c and d), a
(a and b). The tumor is hyperintense on diffusion- pattern that is typical for tumors.
at Reduction Techniques (Chemical

F a loss of signal on out-of-phase images compared
Shift, Fat Saturation, STIR) to in-phase images due to the presence of micro-
scopic or intracellular fat (Fig. 5.8a and b).
Fat demonstrates high signal intensity on T1- and Hepatic iron deposition is the opposite of this,
T2-weighted images due to the inherent short T1 with a loss of signal on in-phase images. This
time and prolonged T2 time, respectively. occurs because of the longer TE of the in-phase
However, there are situations in which “nulling” gradient-echo sequence. In an analogous fashion,
the fat signal becomes advantageous to visualize metallic clips and coils, in addition to gas as seen
non-fatty tissue to better visualize pathology. in the bowel, will demonstrate a larger region of
Multiple methods exist to “null” fat from even more pronounced signal loss on in-phase
sequences, including chemical shift, fat satura- images, a phenomenon referred to as “blooming.”
tion using a spoiler gradient, and inversion recov- In the example of angiomyolipoma, out-of-phase
ery techniques [27]. imaging emphasizes an “India ink” artifact, with
Hydrogen within either water or fat will have an emphasis of the black outline of the lesion due
different precessional frequencies, related to the to the interface between tissues (Fig. 5.8c and d).
Larmor equation previously discussed. As a This name comes from an analogous appearance
result, the precessional frequencies for hydrogen to a staining technique from pathology. However,
within fat and water will sometimes be in-phase the loss of signal of the lesion with fat saturation
and sometimes out-of-phase. In chemical shift clearly shows that it contains macroscopic or
imaging, in-phase and out-of-phase imaging con- extracellular fat (Fig. 5.8e). Fat saturation can be
sists of paired gradient echoes with different TE especially useful for T1 postcontrast sequences,
values based on whether hydrogen within fat as the suppression of fat can help to emphasize
molecules is either in-phase or out-of-phase with enhancing tissues.
hydrogen within water molecules. Fat and water Inversion recovery pulses are techniques for
within the same voxel will cancel each other out removing signals from tissues such as fat and
on out-of-phase imaging, and both add to the sig- water. In these sequences, before the initial
nal on in-phase imaging. The timing of the in- 90-degree RF excitation pulse, there is a
phase image (in milliseconds) is always double 180-degree inversion pulse that is specifically
that of the out-of-phase image. As a result, mag- timed to cancel signals from particular m olecules.
netic susceptibility artifacts are more pronounced The time between these pulses is known as the
on the in-phase images. Of note, the timing of inversion time (TI). For STIR (short tau inver-
in-phase and out-of-phase varies by the magnetic sion recovery), the TI is timed to occur when pro-
field strength, with times decreasing with increas- tons within fat are in the 90-degree position. By
ing magnetic field strength. its namesake, the TI for the STIR sequence is
In fat saturation, an RF pulse tuned to the fat relatively short. Since this inversion recovery
resonance frequency is used together with a pulse nulls the signal from materials that demon-
spoiler gradient to saturate and de-phase fat pro- strate short T1 relaxation, it can technically sup-
tons, which eliminates the signal from fat. This press the signal from tissues that have similar T1
causes tissues containing fat to appear darker on values to fat, which can include tissues with gad-
the image. Sometimes this process fails, such as olinium or proteinaceous material. STIR is often
in the presence of magnetic field distortions, used in extremity imaging. For example, T1
resulting in a failure of fat saturation. The Dixon STIR sequence of the humerus highlights areas
technique involves obtaining water-only images of abnormality associated with osteosarcoma
and fat-only images in addition to in-phase and (Fig. 5.7a). On the other hand, the TI can also be
out-of-phase images. timed to remove signal from protons within
Different fat reduction techniques can be use- water, using a much longer TI. This is put to good
ful depending upon the pathology that is being use in FLAIR (fluid attenuation inversion recov-
imaged. For instance, in hepatic steatosis, there is ery). This is often used in neuroimaging. For
a b
c d e
Fig. 5.8 In-/out-of-phase imaging versus fat saturation. Angiomyolipoma shows an “India ink” artifact outlining
Figure (a) and (b) demonstrate a patient with fatty liver the lesion on out-of-phase imaging, as well as a loss of
disease, which shows a loss of signal on out-of-phase signal within the lesion with fat saturation. (c) In-phase
imaging. (a) In-phase image. (b) Out-of-phase image. image. (d) Out-of-phase image. (e) Fat saturation
Figure (c–e) shows a patient with an angiomyolipoma.
instance, a T2 FLAIR image of the brain high- ity. After the injection of the contrast agent into
lights areas of edema within the brain (Fig. 5.5d).the bloodstream, there are different timings that
are used for imaging the relevant anatomy that
vary by the MR exam protocol. For instance,
Contrast Enhancement abdominal imaging protocols can include an arte-
rial phase, a portal venous phase, and a late or
MRI typically takes advantage of gadolinium- delayed phase (Fig. 5.9). In the figure, contrast-
based contrast agents. Since gadolinium leads to enhanced CT shows a lesion in hepatic segment
T1-shortening, causing tissues containing the 4A which is technically indeterminate. On
agent to appear bright, contrast-enhanced T2-weighted imaging and diffusion-weighted
sequences are naturally T1-weighted. In addition, imaging, it is hyperintense but still not fully char-
fat saturation can also increase lesion conspicu- acterized. However, on contrast-enhanced imag-
ing, there is peripheral nodular discontinuous contrast material [29]. The temporal resolution
enhancement on arterial phase imaging, with varies by machine, vendor, and clinical applica-
centripetal fill-in on subsequent contrast- tion. This allows for more precise characteriza-
enhanced sequences. This firmly establishes the tion of enhancement patterns, including
diagnosis of a benign hemangioma. It is impor- enhancement curves demonstrating the degree of
tant to note that different abdominal lesions have enhancement as a function of time, allowing for
characteristic enhancement patterns, which are the calculation of specific perfusion parameters.
critical for diagnosis. For instance, hepatocellular There are several clinical applications of dynamic
carcinoma will classically demonstrate non-rim contrast-enhanced MRI, including neuroimag-
arterial phase hyperenhancement with subse- ing, breast, cardiac, and prostate.
quent washout within a cirrhotic liver [28]. Much research has been done on the safety of
Dynamic contrast-enhanced MRI involves administering gadolinium-based contrast agents.
obtaining serial MR images before, during, and Nephrogenic systemic fibrosis (NSF) is a rare
after the administration of a bolus of gadolinium and life-threatening condition that was linked to
a b
c d
Fig. 5.9 (a) Non-contrast CT shows a lesion with inter- the arterial phase (e), portal venous phase (f), late phase
nal enhancement within liver segment 4A. While this (g), and delayed phase (h) demonstrate peripheral nodular
could be a benign hemangioma, it is indeterminate on discontinuous enhancement on the arterial phase with pro-
single-phase CT. Diffusion-weighted imaging (b) and gressive centripetal filling on each subsequent postcon-
T2-weighted imaging (c) demonstrate a hyperintense trast phase, firmly establishing the diagnosis of
lesion in this area, although it is not yet fully character- hemangioma
ized. T1-weighted imaging (d) and postcontrast series in
e f
g h
Fig. 5.9 (continued)
gadolinium administration in patients with acute linium administration, as long as there is a true
kidney injury or stage 4–5 chronic kidney dis- clinical indication. A recent consensus statement
ease, which represents an estimated glomerular from the ACR and the National Kidney
filtration rate of <30 mL/min/1.73 m2 [30]. Foundation states that the risk of NSF from
However, the risk of NSF has not been found to Group II GBCAs is felt to be very low, and no
be the same for all forms of gadolinium-based unconfounded cases of NSF were reported for
contrast agents (GBCAs). The American College the only available type of group III GBCAs—
of Radiology (ACR) Manual on Contrast Media gadoxetate disodium; hence, depending upon the
has divided GBCAs into three categories: group I clinical scenario, the harms of delaying or with-
(agents associated with the greatest number of holding GBCAs in patients with acute kidney
NSF cases), group II (agents associated with few, injury should be balanced against and could very
if any, unconfounded cases of NSF), and group well outweigh the risk of NSF [32].
III (agents for which data remains limited regard- Transient reactions to contrast such as flush-
ing NSF risk, but for which few, if any uncon- ing or nausea are typically self-limited. Allergic
founded cases of NSF have been reported) [31]. reactions are uncommon, and life-threatening
In particular, the manual considers the risk of reactions such as anaphylaxis are rare.
NSF among patients given standard doses of Gadolinium-based contrast agents are typically
group II contrast agents to be sufficiently low that avoided in pregnancy unless the benefits substan-
assessment of renal function with a questionnaire tially outweigh the risks, as it is an FDA Class C
or laboratory testing is optional prior to IV gado- drug.
Imaging Protocols Sequences in MR-Linac
Imaging protocols include several imaging The clinically used MR scanner strength is typi-
sequences. They can either be general (standard cally 3T or 1.5T in the radiology department.
MRI of the abdomen) or tailored to a particular However, 0.35T or 1.5T MR scanners are used in
indication (MR elastography, MR pancreatic cyst radiation oncology in combination with Linac.
protocol). Table 5.1 summarizes the different The basic physics behind all MR scanners is fun-
imaging sequences previously discussed, includ- damentally the same.
ing the concept behind them and their utility/
indications.
equences Used in Elekta Unity
S
MR-Linac
Table 5.1 Summary of MRI sequences
Utility/select The Elekta Unity MR-Linac (Elekta, Stockholm,
Sequence Concept indications Sweden) includes a 1.5T Philips big-bore MRI
T1 Relaxation of Fat, methemoglobin, (Philips Healthcare, Amsterdam, Netherlands)
longitudinal melanin, and contrast
and a 7 MV linear accelerator [33]. As this sys-
magnetization are hyperintense
T2 Decay of Fluid, edema, tem uses standard 1.5T imaging, sequences will
transverse inflammation, appear similar to others presented in this chapter.
magnetization infection, and radiation The sequences used depend upon the clinical
change are application for the precise delineation of the
hyperintense
Proton Concentration Musculoskeletal
organs at risk and the target; for instance, brain
density of protons imaging, gray-white applications may include T2 FLAIR and T1 post-
differentiation contrast images, as well as DWI and ADC maps
DWI/ADC Brownian Tumor, stroke, for patients receiving MR-guided radiotherapy
motion of inflammation, active [34]. ADC values can change with radiotherapy,
molecules demyelination, specific
clinical entities and a recent study sought to provide Unity-
(epidermoid cyst) specific recommendations for measuring ADC to
SSFP Steady-state Cine images showing increase its consistency, potentially facilitating
free precession movement, useful for its use for treatment monitoring as a quantitative
vascular structures
such as the heart
imaging biomarker [35]. The role of MR-Linac
In−/ Paired gradient Microscopic fat, iron in different clinical indications is an active area
out-of- echo with (hepatic steatosis, of research [36, 37].
phase different TE hemochromatosis,
values adrenal adenoma)
Fat Suppress signal Macroscopic fat
saturation from fat (myelolipoma), often Sequences Used in ViewRay MRidian
used postcontrast to
highlight enhancing The ViewRay MRIdian MR-Linac (ViewRay
tissue Inc., Oakwood, USA) includes a 0.35T split
STIR Short tau Suppress signal within
inversion fat, often used in
superconducting magnet and a 6 MV linear
recovery extremity imaging accelerator [38]. The MRIdian system can be
FLAIR Fluid Suppress signal within operated in either the clinical or MRI mode. In
attenuation fluid, often used in the clinical mode, the Linac and MRI work
inversion neuroimaging
simultaneously, and in the MR-only mode, the
recovery
T1 Gadolinium Tumor, abscess. system behaves as an MR simulator. In the clini-
postcontrast within tissues Dynamic technique can cal mode, the pulse sequence used for 3D volu-
highlight the contrast metric imaging and 2D cine MRI is a true fast
timing of lesions/tissue imaging with steady-state precession (TRUFI)
sequence, which is a type of balanced steady- Navigator

state free precession (bSSFP) sequence, yielding
a T2/T1-weighted contrast. Apart from a planning MRI and Linac integration allows an automatic
TRUFI sequence, T1-weighted, T2-weighted, mode for selecting trigger position and accep-
and DWI sequences are also available to users. tance window, which can be used to improve
respiratory motion artifact in the thoracic and
abdominal regions. This navigator-triggered
Volumetric TRUFI Sequence imaging technique can be implemented prior to
most standard spin-echo and gradient-echo-based
A low-resolution volumetric planning image can acquisitions. The automatic mode of navigator-
be acquired in 15 s for verification of patient triggered acquisition is divided into two phases,
position. 25-s high-resolution imaging protocols including a learning scout phase followed by a
can be used for 3D volumetric images with suf- standard imaging phase. The aim of the learning
ficient contrast for contouring and planning. The scout phase is to estimate the patient’s respiratory
system transfers the reconstructed images in the pattern and position based on the first five respi-
DICOM format to the system database. ratory cycles by using a series of navigator
T1: There are two different T1 sequences that echoes. The targeted diaphragm position can be
are integrated. The first is a 3D gradient-echo determined by averaging the acquired respiratory
(GRE) sequence for the head and neck, which curve during this period. With an appropriate set-
offers heavy T1 contrast, superior signal-to-noise ting of an acceptance window, a standard MR
ratio, and high spatial resolution; this sequence is acquisition starts during the imaging phase when
slightly more sensitive to body motions such as the diaphragm motion is within a targeted
respiratory breathing. For the thorax and abdomi- position.
nopelvic imaging, where patient breathing com- DWI: Diffusion-weighted imaging can be
monly creates motion artifact, the Turbo FLASH used in the MR-only mode. This sequence can be
(TFL) sequence, in conjunction with the used to measure intensity based on the diffusion
navigator-triggered prospective acquisition cor- of water molecules in and out of each voxel and
rection (PACE) technique, significantly reduces can provide functional imaging to quickly differ-
respiratory motion artifact while achieving a rea- entiate between tumor and edema.
sonable scan time. 2D cine: The TRUFI sequence is used for real-
T2: The system provides fast T2 imaging for all time tracking and automated beam control or
regions of the body. A rapid HASTE technique gating.
with navigator-echo-gating is recommended to Example Images: Examples of the ViewRay
reduce motion artifact from the patient’s respira- MRIdian MR-Linac are shown for cardiac,
tory motion. For imaging in regions without respi- pancreatic, and prostatic cases (Figs. 5.10, 5.11,
ratory motion, navigator-echo gating can be turned and 5.12).
off to reduce acquisition time.
Fig. 5.10 Cardiac example for MR-Linac. The top row shows ViewRay MRIdian TRUFI imaging with free-breathing
(172 s). The bottom row presents the image from CT simulation
Fig. 5.11 Pancreatic example for MR-Linac. The top row shows ViewRay MRIdian TRUFI imaging with breath-hold
(25 s). The bottom row presents images from CT simulation
Fig. 5.12 Prostate example for MR-Linac. The top row shows ViewRay MRIdian TRUFI imaging with free-breathing
(128 s). The bottom row presents the image from CT simulation
MRI Artifacts strates a decreased MR signal in the center of the

image (Fig. 5.13c) and is frequently encountered
This section reviews common MR imaging arti- when there is a large body habitus or a large vol-
facts. Due to the nature of MR signal acquisition, ume of ascites [41]. However, artifacts can also
several artifacts can be created that are worth aid in medical diagnosis. For instance, cerebral
knowing about. These artifacts can make medical cavernous malformations (or cavernomas) dem-
diagnosis more challenging. Aliasing artifact is onstrate susceptibility artifact on MRI
also known as wrap-around artifact (Fig. 5.13a). (Fig. 5.13d), related to the presence of hemosid-
This occurs when a body part is outside the field erin which will cause blooming on gradient-echo
of view, causing it to wrap around the opposite sequences [42]. Susceptibility artifacts are more
side of the image [39]. This is more of a problem pronounced at higher field strengths. Another
in the phase-encoding direction, and one solution example of an artifact being useful for diagnosis
is to increase the field of view in that direction. is the chemical shift artifact that was discussed
Moiré fringes, also known as the zebra artifact, earlier (Fig. 5.8).
are an interference pattern that is related to mag- Gibbs or truncation artifact occurs in areas of
netic field inhomogeneity (Fig. 5.13b); these are significant intensity changes in MR images, such
more frequently seen with gradient-echo as the boundary between the skull and the brain,
sequences [40]. The dielectric effect demon- and it appears as alternating lines extending from
a b
c d
Fig. 5.13 MR artifacts. (a) Proton density imaging of the ening at the center of the image. (d) Gradient echo shows
foot demonstrates aliasing or wrap-around artifact. (b) foci of susceptibility artifact with areas of blooming in the
MR breast shows an image with Moiré fringes, related to cerebral hemispheres, reflecting multiple cavernous mal-
signal interference. (c) T2-weighted imaging in a patient formations. This is an example in which an MR artifact
with ascites demonstrates the dielectric effect, with dark- can actually help with diagnosis
these tissue interfaces; this artifact is related to and results in dark bands across the image [44]. It
the truncation of k-space during MR data acqui- is thusly named because it was originally felt to
sition [43]. Gibbs artifact can be reduced by mimic the appearance of herringbone or cordu-
increasing the imaging matrix size. Herringbone roy fabric. Zipper artifact is related to unwanted
artifact (also known as spike or corduroy artifact) radiofrequency signals corrupting the MR images
is related to abnormal data point(s) in k-space and can either arise from an external source or
internally from the MRI system [45]. This arti- tant consideration is MRI device compatibility.
fact highlights the need for keeping the MR mag- While modern devices are increasingly being
net in an RF-shielded enclosure and results in made to be MR-compatible, it is important to be
white and black bands across the image. mindful of their MRI safety profile [51]. Metallic
Imaging artifacts due to metal implants are shrapnel and other foreign bodies should also be
commonly encountered. Near metal, the mag- carefully considered, particularly if they are near
netic field can vary widely, and the magnetization anatomically critical structures.
in a single voxel will precess at fluctuating rates; Magnetic gradients are essential for localiza-
this leads to dephasing and signal loss, as evi- tion in MRI, but they can create magnetic fields
denced by a black area on the image [46]. This is that vary with time. As a result, they can induce a
much more of a problem for gradient echo rather current, particularly in conductive materials, cre-
than spin echo. Motion artifact is another com- ating the potential for a burn risk. In particular,
mon problem in MRI, with examples including looping structures should be avoided. RF pulses
patient movement, cardiac movement, respira- can also deposit energy, and this is measured by
tory motion, bowel peristalsis, vessel pulsation, the specific absorption rate (SAR). The FDA
blood flow, and CSF flow [47]. There are several maintains a limit of a 1 °C increase of whole body
strategies for mitigating motion artifact. These temperature, and this should be rigorously mea-
include fast imaging sequences, reducing patient sured by the MR machine. Quenching is a rare,
discomfort within the MRI scanner, sedation in albeit life-threatening scenario, in which heating
the right clinical scenarios, imaging at specified of cryogenic material converts it from liquid to
time points during the cardiac/respiratory cycle, gas. As a result, helium replaces oxygen gas,
and post-examination motion correction algo- increasing the risk of suffocation, also resulting in
rithms [48]. For instance, PROPELLER (periodi- increased air pressure within the room.
cally rotated overlapping parallel lines with There are also patient-specific factors associ-
enhanced reconstruction) is a technique for fill- ated with MRI. Some patients find placement
ing k-space that is more resistant to motion [49]. within the MRI scanner to be uncomfortable, and
K-space is sampled in a rotating fashion analo- other patients can experience symptoms of claus-
gous to the blades of a propeller, particularly trophobia. Rapid gradient switching may create a
oversampling the central aspect of k-space. The lot of noise, and patients are typically given ear-
trade-off is that PROPELLER sequences can take plugs to mitigate this. Finally some patient popu-
slightly longer. lations, such as infants, small children, and adults
with altered mental status, may not be coopera-
tive with the examination, resulting in motion
General MRI Safety Considerations artifact and nondiagnostic images.
Robust screening precautions are essential in the

vicinity of an MRI machine. Ferromagnetic sub- Conclusion
stances near the vicinity of the scanner can
become dangerous projectiles, and this effect This section provides a brief overview regarding
increases exponentially as one gets closer to the the clinical indications of MRI. First, it covered
magnet. Shielding is a strategy that is often the basics of MR physics required for the genera-
employed in the vicinity of the scanner, as its tion of the image. Next, it covered the fundamen-
magnetic field can extend through walls and tals of T1- and T2-weighted imaging. There was
floors. In particular, access to the five Gauss line a discussion of MR sequences, including spin
is strictly monitored. The American College of echo and gradient echo, as well as different fat
Radiology has designated four MRI safety zones, reduction techniques. This chapter discussed
with particularly restricted access in Zone III and MRI contrast and contrast safety. Sequences
the magnet itself in Zone IV [50]. Another impor- were discussed in the context of the Elekta Unity
and ViewRay MRIdian MR-Linac systems. There 13. Gallagher TA, Nemeth AJ, Hacein-Bey L. An intro-
was a discussion of MRI artifacts, which both duction to the Fourier transform: relationship to
MRI. AJR Am J Roentgenol. 2008;190(5):1396–405.
hinder and aid image interpretation. Finally, gen- 14. Paschal CB, Morris HD. K-space in the clinic. J Magn
eral MRI safety concerns were discussed. While Reson Imaging. 2004;19(2):145–59.
it ultimately serves as an introductory blueprint 15. Jaspan ON, Fleysher R, Lipton ML. Compressed
for further study, these fundamental principles sensing MRI: a review of the clinical literature. Br J
Radiol. 2015;88(1056):20150487.
are essential for the successful operation of 16. Chen Y, Schönlieb CB, Liò P, Leiner T, Dragotti
MR-Linac in the clinic. PL, Wang G, et al. AI-based reconstruction for fast
MRI—a systematic review and meta-analysis. Proc
IEEE. 2022;110(2):224–45.
17. Petersein J, Saini S. Fast MR imaging: technical strat-
References egies. AJR Am J Roentgenol. 1995;165(5):1105–9.
18. Hahn EL. Spin echoes. Phys Rev. 1950;80(4):
1. Nagayama Y, Oda S, Nakaura T, Tsuji A, Urata J, 580–94.
Furusawa M, et al. Radiation dose reduction at pedi- 19. Brown RW, Cheng Y-CN, Haacke M, Thompson MR,
atric CT: use of low tube voltage and iterative recon- Venkatesan R. Magnetic resonance imaging: physical
struction. Radiographics. 2018;38(5):1421–40. principles and sequence design. 2nd ed. Hoboken, NJ:
2. Scialpi M, Reginelli A, D’Andrea A, Gravante S, Wiley-Blackwell; 2014.
Falcone G, Baccari P, et al. Pancreatic tumors imag- 20. Bloch F. The principle of nuclear induction. Science.
ing: an update. Int J Surg. 2016;28(Suppl 1):S142–55. 1953;118(3068):425–30.
3. Glover M, Whorms D, Singh R, Almeida RR, 21. Lu H, Clingman C, Golay X, van Zijl PC. Determining
Prabhakar AM, Saini S, et al. A radiology-focused the longitudinal relaxation time (T1) of blood at 3.0
analysis of transparency and usability of top Tesla. Magn Reson Med. 2004;52(3):679–82.
U.S. Hospitals’ Chargemasters. Acad Radiol. 22. Stanisz GJ, Odrobina EE, Pun J, Escaravage M,
2020;27(11):1603–7. Graham SJ, Bronskill MJ, et al. T1, T2 relaxation and
4. Canellas R, Rosenkrantz AB, Taouli B, Sala E, Saini magnetization transfer in tissue at 3T. Magn Reson
S, Pedrosa I, et al. Abbreviated MRI protocols for the Med. 2005;54(3):507–12.
abdomen. Radiographics. 2019;39(3):744–58. 23. Eccles CL, Adair Smith G, Bower L, Hafeez S,
5. Wang Y, Osborne MT, Tung B, Li M, Li Y. Imaging Herbert T, Hunt A, et al. Magnetic resonance imag-
cardiovascular calcification. J Am Heart Assoc. ing sequence evaluation of an MR Linac system;
2018;7(13):e008564. early clinical experience. Tech Innov Patient Support
6. Haller S, Haacke EM, Thurnher MM, Barkhof Radiat Oncol. 2019;12:56–63.
F. Susceptibility-weighted imaging: technical essen- 24. Westbrook C, Talbot J. MRI in practice. 5th ed.
tials and clinical neurologic applications. Radiology. Hoboken, NJ: Wiley-Blackwell; 2018.
2021;299(1):3–26. 25. Carr HY, Purcell EM. Effects of diffusion on free pre-
7. Petralia G, Zugni F, Summers PE, Colombo A, Pricolo cession in nuclear magnetic resonance experiments.
P, Grazioli L, et al. Whole-body magnetic resonance Phys Rev. 1954;94(3):630–8.
imaging (WB-MRI) for cancer screening: recommen- 26. Yankeelov TE, Pickens DR, Price RR. Quantitative
dations for use. Radiol Med. 2021;126(11):1434–50. MRI in cancer. 1st ed. Boca Raton, FL: CRC Press;
8. Busacchio D, Mazzocco K, Gandini S, Pricolo 2012.
P, Masiero M, Summers PE, et al. Preliminary 27. Bitar R, Leung G, Perng R, Tadros S, Moody AR,
observations regarding the expectations, accept- Sarrazin J, et al. MR pulse sequences: what every
ability and satisfaction of whole-body MRI in radiologist wants to know but is afraid to ask.
self-referring asymptomatic subjects. Br J Radiol. Radiographics. 2006;26(2):513–37.
2021;94(1118):20191031. 28. Chernyak V, Fowler KJ, Kamaya A, Kielar AZ,
9. Kim SY, An J, Lim YS, Han S, Lee JY, Byun JH, et al. Elsayes KM, Bashir MR, et al. Liver imaging report-
MRI with liver-specific contrast for surveillance of ing and data system (LI-RADS) version 2018: imag-
patients with cirrhosis at high risk of hepatocellular ing of hepatocellular carcinoma in at-risk patients.
carcinoma. JAMA Oncol. 2017;3(4):456–63. Radiology. 2018;289(3):816–30.
10. Grover VP, Tognarelli JM, Crossey MM, Cox IJ, 29. Verma S, Turkbey B, Muradyan N, Rajesh A,
Taylor-Robinson SD, McPhail MJ. Magnetic reso- Cornud F, Haider MA, et al. Overview of dynamic
nance imaging: principles and techniques: lessons for contrast-enhanced MRI in prostate cancer diag-
clinicians. J Clin Exp Hepatol. 2015;5(3):246–55. nosis and management. AJR Am J Roentgenol.
11. Roth C, Deshmukh S. Fundamentals of body MRI. 2012;198(6):1277–88.
2nd ed. Philadelphia, PA: Elsevier; 2017. 30. Woolen SA, Shankar PR, Gagnier JJ, MacEachern
12. Glockner JF, Hu HH, Stanley DW, Angelos L, MP, Singer L, Davenport MS. Risk of nephrogenic
King K. Parallel MR imaging: a user’s guide. systemic fibrosis in patients with stage 4 or 5 chronic
Radiographics. 2005;25(5):1279–97. kidney disease receiving a group II gadolinium-based
contrast agent: a systematic review and meta-analysis. 40. Alfudhili K, Masci PG, Delacoste J, Ledoux JB,
JAMA Intern Med. 2020;180(2):223–30. Berchier G, Dunet V, et al. Current artefacts in car-
31. ACR. ACR manual on contrast media. Reston, VA: diac and chest magnetic resonance imaging: tips and
American College of Radiology; 2022. tricks. Br J Radiol. 2016;89(1062):20150987.
32. Weinreb JC, Rodby RA, Yee J, Wang CL, Fine D, 41. Cornfeld D, Weinreb J. Simple changes to 1.5-T
McDonald RJ, et al. Use of intravenous gadolinium- MRI abdomen and pelvis protocols to optimize
based contrast media in patients with kidney disease: results at 3 T. AJR Am J Roentgenol. 2008;190(2):
consensus statements from the American College W140–50.
of Radiology and the National Kidney Foundation. 42. Campbell PG, Jabbour P, Yadla S, Awad IA. Emerging
Radiology. 2021;298(1):28–35. clinical imaging techniques for cerebral cavern-
33. Snyder JE, St-Aubin J, Yaddanapudi S, Boczkowski A, ous malformations: a systematic review. Neurosurg
Dunkerley DAP, Graves SA, et al. Commissioning of a Focus. 2010;29(3):E6.
1.5T Elekta Unity MR-linac: a single institution expe- 43. Kellner E, Dhital B, Kiselev VG, Reisert M. Gibbs-
rience. J Appl Clin Med Phys. 2020;21(7):160–72. ringing artifact removal based on local subvoxel-
34. Dunkerley DAP, Hyer DE, Snyder JE, St-Aubin JJ, shifts. Magn Reson Med. 2016;76(5):1574–81.
Anderson CM, Caster JM, et al. Clinical imple- 44. Zhuo J, Gullapalli RP. AAPM/RSNA physics tutorial
mentational and site-specific workflows for a 1.5T for residents: MR artifacts, safety, and quality control.
MR-Linac. J Clin Med. 2022;11(6):1662. Radiographics. 2006;26(1):275–97.
35. Kooreman ES, van Houdt PJ, Keesman R, Pos FJ, van 45. Graves MJ, Mitchell DG. Body MRI artifacts in clini-
Pelt VWJ, Nowee ME, et al. ADC measurements on cal practice: a physicist’s and radiologist’s perspec-
the Unity MR-linac - a recommendation on behalf tive. J Magn Reson Imaging. 2013;38(2):269–87.
of the Elekta Unity MR-linac consortium. Radiother 46. Hargreaves BA, Worters PW, Pauly KB, Pauly JM,
Oncol. 2020;153:106–13. Koch KM, Gold GE. Metal-induced artifacts in
36. Civil YA, Oei AL, Duvivier KM, Bijker N, Meijnen MRI. AJR Am J Roentgenol. 2011;197(3):547–55.
P, Donkers L, et al. Prediction of pathologic complete 47. Zaitsev M, Maclaren J, Herbst M. Motion artifacts in
response after single-dose MR-guided partial breast MRI: a complex problem with many partial solutions.
irradiation in low-risk breast cancer patients: the J Magn Reson Imaging. 2015;42(4):887–901.
ABLATIVE-2 trial-a study protocol. BMC Cancer. 48. Havsteen I, Ohlhues A, Madsen KH, Nybing JD,
2023;23(1):419. Christensen H, Christensen A. Are movement arti-
37. Regnery S, de Colle C, Eze C, Corradini S, Thieke facts in magnetic resonance imaging a real problem?-
C, Sedlaczek O, et al. Pulmonary magnetic a narrative review. Front Neurol. 2017;8:232.
resonance-guided online adaptive radiotherapy of 49. Forbes KP, Pipe JG, Bird CR, Heiserman
locally advanced: the PUMA trial. Radiat Oncol. JE. PROPELLER MRI: clinical testing of a novel
2023;18(1):74. technique for quantification and compensation of head
38. Kluter S. Technical design and concept of a 0.35 T motion. J Magn Reson Imaging. 2001;14(3):215–22.
MR-Linac. Clin Transl Radiat Oncol. 2019;18:98–101. 50. ACR. ACR manual on MR safety. Reston, VA:
39. Stadler A, Schima W, Ba-Ssalamah A, Kettenbach J, American College of Radiology; 2020.
Eisenhuber E. Artifacts in body MR imaging: their 51. Shellock FG, Crues JV. MRI bioeffects, safety, and
appearance and how to eliminate them. Eur Radiol. patient management. 2nd ed. Los Angeles, CA:
2007;17(5):1242–55. Biomedical Research Publishing Group; 2022.
Motion Management and Tracking
6
Elizabeth Huynh, Eenas A. Omari, Sara N. Lim,
and Raymond H. Mak
Introduction linear accelerators (MR-Linacs) has provided cli-

nicians with the ability to reduce uncertainty and
Motion management is a critical component of target margins. This has expanded the ability to
modern image-guided radiation therapy, and dose escalate with steeper dose gradients and/or
accurate modeling of motion during simulation, hypofractionated treatments while reducing dose
planning, and treatment delivery over the past exposure to surrounding organs.
two decades has enabled the advent of high- Motion of both targets and organs-at-risk
precision radiation therapy techniques such as (OARs) can occur due to a variety of physiologi-
stereotactic body radiation therapy (SBRT). cal processes, including breathing, cardiac activ-
Historically, motion management (particularly ity, digestion, swallowing, and bladder filling/
for respiratory motion) relied on adding large voiding, and can lead to significant intra- and
planning margins to account for this motion. The inter-fraction variation in the position of the
development of multiple motion management tumor and surrounding normal tissues. Failure to
approaches with X-ray imaging-based linear account for this variation during treatment plan-
accelerators and now more recently MRI-guided ning and delivery can result in decreased treat-
ment accuracy and increased toxicity to
E. Huynh (*) surrounding normal tissues.
Department of Physics and Engineering, London
Regional Cancer Program, London Health Sciences
Centre, London, ON, Canada
e-mail: [email protected] Types of Motion
E. A. Omari
Department of Radiation Oncology, Medical College Respiratory Motion
of Wisconsin, Milwaukee, WI, USA
e-mail: [email protected] Respiratory motion is typically periodic (on aver-
S. N. Lim age ~ 5 s per breath or 12 breaths per minute) and
Department of Radiation Oncology, Northwestern more predictably modeled, but there can be sig-
Memorial Hospital, Northwestern University
nificant variation in amplitude, period, and regu-
e-mail: [email protected] larity with potential significant patient-to-patient,
R. H. Mak
intra-fraction, and inter-fraction variability.
Department of Radiation Oncology, Brigham and Respiratory motion is particularly relevant for
Women’s Hospital, Dana-Farber Cancer Institute, thoracic (e.g., breast, lung, intracardiac) and
Boston, MA, USA upper abdominal cancers (e.g., gastric, liver, pan-

70 E. Huynh et al.
creas, esophageal). There is substantial variation timing of ingestion. The process of digestion also
in the amplitude of respiratory motion, and there results in production of gas that can result in sig-
is a need for motion management techniques for nificant intra-fraction and inter-fraction move-
lung tumors, with the greatest motion typically ment of the stomach, duodenum, small and large
encountered for lower lobe tumors or organs bowel, and rectum. Traditional physiological
adjacent to the diaphragm [1, 2]. In general, management techniques for radiation therapy for
respiratory motion is most substantial in the abdominal and pelvic sites include pre-specified
superior-inferior direction, ranging up to 2–5 cm fasting intervals (e.g., nothing by mouth for
in reported series, but lateral and anterior- 4–8 hours before simulation and treatment) to
posterior motion can also be substantial in a sub- decompress the stomach and duodenum and min-
set of patients (ranging up to 1–2 cm in selected imize peristaltic movement, dietary restrictions
patients) [2, 3]. There are various respiratory (e.g., avoiding gas forming foods), and bowel
motion management techniques, discussed in preparation regimens (e.g., daily medications for
section “X-Rray- and CT-Based Techniques for gas-reduction such as simethicone, laxatives,
Respiratory Motion Management”. and/or enemas to decompress the bowels) to opti-
mize reproducibility of gastrointestinal organ
location and filling, which are applicable for both
Cardiac Motion X-ray-based and MRI-guided approaches.
Similarly, bladder filling can result in substan-
Cardiac motion has an inherent period (normally tial intra-fraction and inter-faction variability in
1–2 beats per second) similar to respiratory bladder volume and shape which impacts a vari-
motion; however, the motion induced by cardiac ety of pelvic tumor sites (e.g., prostate, bladder,
contraction and relaxation during systole and gynecological, rectal). In particular, X-ray-based
diastole, respectively, can be volumetrically more assessments of bladder volume can be particu-
complex. While cardiac motion has not been his- larly susceptible to intra-fraction filling and can
torically accounted for in the treatment of tho- result in substantial target shifts as reported in
racic and upper abdominal tumors due to prostate cancer [6]. Conventional methods to
limitations in the frequency of X-ray imaging mitigate bladder motion issues include bladder
technologies relative to the cardiac cycle, the regimens with pre-specified timing of bladder
ability for MRI to perform continuous cine imag- voiding and hydration instructions prior to treat-
ing allows for real-time imaging of cardiac ment and implantation of fiducials for prostate
motion. More recent applications for SBRT to cancer for intra-fraction monitoring.
treat intracardiac tumors [4] and for the ablation Lastly, swallowing and phonation can impact
of cardiac arrhythmias such as ventricular tachy- the location of oral, oropharyngeal, pharyngeal,
cardia [5] provide new indications and impetus to and hypopharyngeal organs and tumor targets for
develop novel cardiac motion management head and neck cancers. The impact of this motion
techniques. is greatest for laryngeal tumor locations with up
to 2 cm of laryngeal movement with swallowing
[7], and more recent investigations with MRI-
Physiological Motion based dynamic imaging approaches have quanti-
of Gastrointestinal, Urinary, fied the expected intra-fraction motion [8–10].
and Head and Neck Organs These physiological motions can result in sig-
nificant variability in delivered dose either
Physiological motion of the gastrointestinal through displacement of the tumor target or shift
organs can be due to distention of the organs and in location of the organ itself and can be critical
peristalsis of liquids, solid food, and stool con- in clinical scenarios with high doses of radiation
tent, which can be dependent on the quantity and delivered with steep gradients between tumors
6 Motion Management and Tracking 71
and OARs (e.g., pancreatic cancer adjacent to Respiratory Gating

duodenum, stomach, and/or bowel). The continu-
ous imaging and tracking capabilities of MRI- Respiratory-gated radiation therapy with X-ray
guided linear accelerators provide advantages in approaches typically utilizes an external respira-
real-time monitoring of intra-fraction physiologi- tory surrogate to gate the treatment beam on and
cal movement. off during the patient’s breathing cycle, so that
the radiation is delivered only when the tumor is
in specific phases of the breathing cycle or during
X-Ray- and CT-Based Techniques a breath-hold. These active motion management
for Respiratory Motion approaches in general allow for reduction of the
Management target volume.
While external surrogates can be used for gat-
Historically, motion management with X-ray- ing, internal surrogates such as implantation of
and CT-based approaches in radiation therapy fiducials in and/or around the tumor may be
planning has focused on respiratory motion. required for internal verification of the gating
Approaches for respiratory management include process. Fiducials (implanted metal markers) are
motion-encompassing methods (e.g., 4D-CT), required due to the inherent limitations of cone
respiratory gating, breath-hold, forced shallow beam CT or intra-fraction triggered kV imaging
breathing with abdominal compression, and real- in resolving soft tissue or small tumors in the tho-
time tracking (often requiring fiducial place- rax and upper abdomen.
ment). These techniques can be further divided
into passive and active approaches: motion-
encompassing methods and forced shallow Breath-Hold
breathing are passive approaches, while respira-
tory gating, breath-hold, and respiratory- Breath-hold treatments are similar to respiratory
synchronization are active. Active approaches gated treatments, but with treatment delivery
account for variations in breathing patterns in occurring only when a patient is holding their
real-time, while passive approaches do not. breath at a pre-specified depth. For X-ray-based
approaches, external surrogates and/or implanted
fiducials are required as described above. Breath-
Motion Encompassing Methods hold approaches have some advantages to gating
approaches including improved duty cycle if the
4D-CT involves acquiring multiple CT scans of patient can consistently reproduce the breath-
the patient during the breathing cycle and creat- hold and the ability to maneuver critical organs
ing a 4D image set that allows the clinician to away from the target zone (e.g., deep inspiration
define an internal target volume (ITV) for the breath-hold for left-sided breast cancer to posi-
tumor that encompasses the range of tumor location the high-dose region further away from the
tion throughout the breathing cycle and also iden- heart [11]).
tify normal tissue location during the breathing
cycle. Some potential limitations of the 4D-CT
and ITV approach is that ITV can still be prohibi- Forced Shallow Breathing
tively large with significant respiratory motion with Abdominal Compression
and expose surrounding organs, such as the lungs
for lung tumors, to excess dose. There are also Compression devices can also be utilized to sup-
concerns regarding dose blurring and interplay press respiratory motion and is most effective for
effects [2]. lower lobe lung tumors. Drawbacks of this
72 E. Huynh et al.
approach include increased inter-fraction vari- 4D-MRI

ability and the potentially increased motion in a
small proportion of cases [12]. Respiratory correlated 4D-MRI can be acquired
to generate motion averaged or mid-position
images for daily treatment planning and daily
Tracking motion assessment [13]. This is of importance
when treating abdominal and thoracic tumors
Motion management with fiducials for X-ray- with a non-gated treatment. The average or mid-
guided radiation therapy involves the implanta- position images generated from the respiratory
tion of small, visible (i.e., radio-opaque) markers binned 4D-MRI can be used for treatment plan-
that are placed within or near the target area. ning, daily image registration/fusion, and daily
These markers are then tracked during treatment re-planning (i.e., contouring and plan reoptimiza-
to monitor and compensate for any movement of tion). The determination between the motion-
the target. averaged and mid-position images is based on
image quality and the presence of artifacts. For
example, Fig. 6.1 shows the reconstructed
Limitations of X-Ray-Based Motion MR-Linac 4D-MRI for a liver patient where the
Management mid-position image would be the better option
due to the blurring artifacts caused by the large
Key limitations of X-ray-based respiratory respiratory motion as shown on the motion aver-
motion management include higher normal tissue aged image. The mid-position image takes longer
dose exposure with motion encompassing (e.g., to reconstruct but shows sharper boundaries
4D-CT) approaches, the dependency on external despite the presence of large respiratory motion.
surrogates, and/or implantation of fiducial mark- If similar quality is observed between the motion
ers to enable solutions that can reduce target vol- averaged and the mid-position images, the motion
ume. The technologies are highly dependent on averaged image may be the better option due to
accurate modeling and reconstructions and are the faster reconstruction time.
not easily applied to non-periodic motion (e.g., 4D-MRI can also be used to determine the
physiological motion), and there is a limitation maximum extent of the motion with respect to
on intra-fraction frequency of imaging due to the original planning target volume (PTV) for
cumulative imaging X-ray exposure concerns. non-gated treatments [15]. If the extent of the
motion lies outside of the planned volume, a new
PTV can be generated based on the daily motion
MRI-Guided Linear Accelerator to avoid target underdosing.
Techniques for Motion
Management
Breath-Hold MRI
The combination of a MRI with a linear accelera-
tor allows for new motion management strate- Similar to X-ray-based breath-hold techniques,
gies. MRI-based approaches for motion MRI-based breath-hold techniques require
management in radiation therapy planning patient cooperation to reproducibly hold their
include both respiratory-gated MRI and breath- breath during simulation and treatment delivery.
hold MRI, while motion management during In contrast to X-ray-based breath-hold tech-
treatment delivery can be enhanced with continu- niques, MRI-based breath-hold treatments are
ous, cine MRI. not dependent on external surrogates or place-
Fig. 6.1 Elekta Unity MR-Linac 4D-MR images of a liver patient showing the motion averaged (top), mid-position
(middle), and respiratory binned (bottom) images [14]
ment of fiducial markers for verification. and creating a cine image set that shows the
Furthermore, multiple image acquisitions can be tumor and surrounding normal tissues in differ-
obtained until a favorable and reproducible ent positions during the breathing cycle. Since
breath-hold is obtained without additional radia- cine MRI does not expose patients to ionizing
tion dose. The ability to select breath-holds that radiation, the cine MRI can be acquired continu-
result in favorable geometric relationships is par- ously during radiation delivery for intra-fraction
ticularly helpful for cases with tumors abutting monitoring and gating. While MRI acquisition
central organs (e.g., central lung tumor, Fig. 6.2). times were previously a barrier for real-time
Clinically, the patient can be coached to either imaging, selection and optimization of MR
perform the breath-hold at the end of exhale or sequences for radiation therapy monitoring allow
end of inhale. Some factors determining the for rapid image acquisition (e.g., 4–8 frames per
breath-hold phase selection are reproducibility second) in two-dimensional planes that can be
and the duty cycle for which the patient is capa- selected to intersect with the tumor target or criti-
ble of holding their breath. For MR-Linacs with cal OAR for continuous visualization and
gating capability, such as the ViewRay MRIdian, tracking.
the beam is held automatically when the target or Furthermore, continuous MRI with tumor
surrogate is outside of the breath-hold window. tracking enables monitoring for non-periodic
For MR-Linacs without gating capability, such as motion, such as the detection of bowel gas move-
the early version of the Elekta Unity, the beam is ment causing a shift in tumor location during
held manually when the target or surrogate lies treatment delivery (Fig. 6.3).
outside the breath-hold window.
imitations of MRI-Guided Motion

L
ine MRI with Breath-Hold Technique
C Management
and Target Tracking
Despite the advantages of MRI-guided motion
Cine MRI involves acquiring multiple MR management, there are key limitations associated
images of the patient during the breathing cycle with the approach. MRI bores are relatively nar-
74 E. Huynh et al.
Fig. 6.2 Comparison of 4D-CT (left) versus MR-guided MR-guided breath-hold technique and adjustment of the
breath-hold (right) approaches for SBRT for a large right patient’s breath-hold depth allowed for positioning of the
lower lobe lung tumor with significant respiratory motion. tumor away from the esophagus and safe delivery of
Left: 4D-CT-based ITV generation with 5 mm PTV 12 Gy × 5 fractions (57 Gy isodose line in color wash,
results in high-dose overlap with the esophagus (57 Gy esophagus in yellow)
isodose line in color wash, esophagus in yellow). Right:
Fig. 6.3 Intra-fraction physiological motion: interest (yellow). Left: Decompressed rectum that matches
Representative frames from continuous sagittal MR cine rectal contours from initial simulation image and MRI of
imaging on the ViewRay MRIdian during beam-on for a the day; Right: Intra-fraction shift of gas from bowel into
prostate cancer received MRI-guided SBRT with the pros- the rectum causing expansion of the rectum and shift of
tate target tracking structure (red) and gating region of the prostate target during beam delivery
row and more enclosed compared to CT bores, accurately. Often in these cases, a surrogate will
and therefore, the patient must be able to tolerate be used for tracking instead. The accuracy of the
being in the MRI for an extended period of time. tracking software can also be compromised by
Furthermore, MRI is susceptible to motion arti- the presence of nearby MR artifacts.
facts as acquiring 3D volume images require In summary, X-ray-based versus MRI-based
more time than CT scans, and the motion over the motion management approaches both have
MRI acquisition time will cause blurring of the unique advantages and disadvantages for radia-
image. Tracking of small lesions is challenging tion therapy planning and delivery. The selection
with current MR-Linac systems as the tracking of the technology for a given patient and clinical
software has difficulty tracking small lesions indication needs to be personalized.
Simulation during treatment. Furthermore, patient set-up

with immobilization devices on MR-Linacs are
Patients treated on an MR-Linac will typically limited by the size of the MR bore and also need
undergo CT simulation, simulation on the to include the MR radiofrequency (RF) receive
MR-Linac, and, at times, an additional simula- coils. Depending on the MR-Linac configuration,
tion on a diagnostic MRI scanner or MR simula- the requirements for incorporation of the receive
tor that may have imaging capabilities that the coils in the immobilization strategy will differ.
MR-Linac does not possess. The CT scan is used The Elekta Unity system has posterior RF receive
for dose calculation and MR-Linac scan for treat- coils integrated in the couch top and an anterior
ment planning and contouring, and the additional coil mounted on a bridge, whereas the ViewRay
MR scan is usually acquired on a higher field MRIdian has flexible surface RF receive coils
strength MRI scanner, equipped with a flat top that need to be positioned anteriorly and posteri-
table and laser alignment system, with superior orly to the patient [16, 17]. Additional
image quality and more available sequences MR-specific equipment such as hearing protec-
compared with the MR-Linac for guiding delin- tion (headphones or earplugs) and emergency
eation of the treatment target. While the order in squeeze bulb also need to be considered and
which the simulations are performed may vary incorporated into the patient set-up with the
across institutions, motion management strate- immobilization device. Immobilization devices
gies for inter-fraction and intra-fraction motion for the MR-Linac must be MR conditional or MR
are considered at the first simulation and executed safe. The immobilization devices must also fit
in each successive simulation such that the through the MR bore with enough clearance to
patient’s treatment position is reproducible and account for variation in patient geometry, primar-
representative of their internal anatomy and ily patient width, as the diameter of the bore is
external position at treatment. the limiting factor. For MR-Linacs that allow for
lateral couch shifts, such as the ViewRay
MRIdian, enough clearance between the immo-
Positioning and Immobilization bilization device and the bore must also be con-
sidered to allow for lateral shifts at treatment,
For conventional X-ray-based radiation therapy, such as 1–2 cm.
rigid immobilization devices are heavily relied For treatment sites in the thorax, abdomen, or
on for motion management in order to reproduce pelvis, rigid immobilization devices such as vac-
the patient’s position from simulation to treat- uum cushion-based immobilization may be used
ment and minimize motion during treatment. (e.g., Vac-Loc Cushions (CIVCO) or VacQfix
This is particularly important for conventional Vacuum Cushions (Qfix)) to aid in patient posi-
Linac treatments because soft tissue is often not tioning reproducibility. For MR-Linacs requiring
well visualized on kV radiographs or cone beam surface RF receive coils, the receive coil must fit
CT, and imaging internal anatomy during treat- on top or under the vacuum cushion, and the vac-
ment is limited to kV images that only provide uum cushion must be formed to fit through the
good visualization of high-density structures bore with enough clearance for lateral shifts, if
such as bone or implanted radio-opaque fiducial applicable. Some institutions choose not to use
markers. any rigid immobilization devices, but instead opt
Considerations for immobilization devices for for a non-rigid immobilization such as only the
treatments on a MR-Linac are different than for surface RF receive coils. The RF receive coils on
conventional Linac treatments because soft tissue the ViewRay MRIdian are flexible coils that are
is better visualized on MRI, treatment plans may placed on the anterior and posterior of the patient,
be adapted each fraction for variations in internal secured with velcro straps, centered over the
anatomy and external position, and internal anat- treatment area. Many groups find that this is
omy may be visualized and tracked for gating enough immobilization to maintain patient com-
76 E. Huynh et al.
fort and ensure that the patient remains as still as For online adaptive treatments, day-to-day
possible for the duration of the treatment [18]. reproducibility of the exact patient position is not
Having the patient secured in only the RF receive absolutely critical as the treatment plan will be
coils also reduces the probability of clearance adapted for changes in the patient’s internal anat-
issues with the bore since there is no additional omy and external position. Treatment times,
width added to the patient from a rigid immobili- especially online adaptive treatments, on an
zation device. On the Elekta Unity system, there MR-Linac are longer than conventional Linac
are less issues with RF coil interference since the treatments. Currently, online adaptive treatments
posterior RF receive coils are integrated in the on an MR-Linac can take on average 45 min [19],
couch and an anterior coil mounted on a bridge although software and hardware improvements
which can slide superiorly-inferiorly and can be are constantly being made by vendors to decrease
adjusted in height to avoid patient and immobili- MR-Linac treatment times. Given the longer
zation collisions. treatment times, patient comfort is very impor-
For treatment sites in the brain and head and tant for safe and effective treatment [20, 21]; if a
neck, patients are immobilized with a thermo- patient is comfortable, they are less likely to
plastic mask. Headphones are typically used for move during treatment. Patients may be treated
hearing protection and communication with ther- with their arms above their head as in conven-
apists. When the thermoplastic mask is used for tional Linac treatments or treated with their arms
immobilization, hearing protection and commu- down by their sides. For online adaptive treat-
nication must also be accommodated through the ments, many groups choose to treat patients with
mask. In this situation, ear plugs can be used in their arms down by their sides as this is more
place of headphones that directly connect to the comfortable for the long treatment times that are
audio system and fit more easily through the needed for online adaptive treatments. For treat-
mask. When flexible surface RF receive coils are ments that use the MR for image guidance and
used, the posterior coil may lie flat on the couch are not online adaptive treatments, many groups
top or on top of the headrest beneath the patient’s choose to treat patients with arms above their
head. The anterior coil is then applied and posi- head as these treatments are shorter than online
tioned to avoid touching the patient’s head, such adaptive treatments and this position may have
as through the use of a holder or grip ring. treatment planning advantages, allowing acces-
Patient positioning should be reproduced at sibility of more gantry angles in the treatment
each simulation, CT, MR-Linac, and, if applica- plan.
ble, additional MRI simulation, including the
rigid immobilization devices. The MR coils
should also be positioned as they would at treat- Respiratory Motion
ment for each simulation. Some MR-Linacs,
such as the ViewRay MRIdian, provide dummy Acquisition of MRIs requires tens of seconds to
coils which have the same geometry and materi- several minutes depending on the sequence
als as the RF receive coils, without the electron- parameters, field of view, and resolution. As a
ics in the coil. This ensures that the patient’s result, respiratory motion, especially for treat-
position at simulation is most similar to treatment ment areas in the thorax and abdomen, can cause
to minimize registration error between the significant motion artifacts when the MRI is
images. This reduces the amount of deformation acquired while the patient is free breathing. MRIs
needed when performing a deformable registra- acquired over these time periods encompass mul-
tion with the CT to the MR-Linac MRI and a tiple breathing cycles and result in motion arti-
more accurate rigid registration between the facts that appear as blurring of the image. Tumor
MR-Linac and additional MRI simulation targets in the pelvis may also be subject to the
images. impact of respiratory motion when the patient
breathes with their belly as opposed to chest, candidacy would be assessed by acquiring MRIs
resulting in blurring of the MRI. An MRI with and evaluating the amount of respiratory motion
these motion artifacts for treatment planning cre- artifact present. The tumor target and OARs are
ates uncertainty in contouring of the tumor tar- contoured on the breath-hold MRI, and the treat-
gets and OARs and dosimetry. ment plan is subsequently created on this MRI.
On gated MR-Linac systems, such as the During treatment, gated MR-Linac systems
ViewRay MRIdian, respiratory motion can be are designed to track a target that moves with
managed with a breath-hold technique, where the respiratory motion and deliver radiation only
patient is instructed to hold their breath at a con- when the target that is being tracked is in the pre-
sistent capacity for tens of seconds during image defined breath-hold position. Thus, patients that
acquisition and during treatment. At simulation, use the breath-hold technique will also require an
the patient must first be assessed for breath-hold internal target that moves with respiratory motion
candidacy. A good candidate for the breath-hold and can be tracked. Ideally, the target that is
technique is a patient that is able to hold their tracked is the treatment target itself, that is, the
breath for at least the minimum duration of a 3D tumor (Fig. 6.4), but may also be a nearby organ
image acquisition on the MR-Linac (e.g., as a surrogate for motion. The motion of the
20–30 s) reproducibly at a consistent capacity selected tracking target can be evaluated at simu-
(i.e., not releasing their breath throughout the lation on the MR-Linac, and multiple tracking
breath hold). This can be assessed at CT simula- target candidates for tracking during treatment
tion using systems that are also applied for con- may be identified. For treatments in the pelvis,
ventional Linac simulations and treatments for identifying a target to track for breath-hold can
respiratory motion, such as the Varian respiratory be difficult as often the treatment target (e.g., dis-
gating for scanners device, that characterize and eased lymph nodes) may not move with breath-
monitor the patient’s respiratory pattern. If the ing; however, OARs such as the bowel may be
patient is a good candidate to be treated with a impacted by respiratory motion. However, bowel
breath-hold technique, all simulation images (CT is not a good candidate for a tracking target as its
and MRIs) should be acquired in the treatment shape, position, and/or track of movement can
position with breath-hold. If simulation on the vary in a short period of time depending on the
MR-Linac precedes CT simulation, breath-hold presence of gas and/or its content. In these types
Fig. 6.4 Representative frames from the sagittal cine treatment breath-hold position (left) and when the tumor
during MR-Linac simulation on the ViewRay MRIdian is in the correct breath-hold position for treatment (right)
outlining the tumor to be treated (red) when not in the
78 E. Huynh et al.
Fig. 6.5 Motion averaged images on the Elekta Unity for tory binned images. Left, axial view; middle, coronal
a liver cancer patient showing the planned ITV (blue) vs. view; right, sagittal view
daily ITV (pink) drawn from the daily 4D-MRI respira-
of scenarios, the tracking target may not be the treatment adaptive planning. The respiratory
tumor itself, but a surrogate for respiratory binned images can be used to visualize the extent
motion such as a kidney. of the daily motion with respect to the daily
If a patient is not a good candidate for breath- PTV. PTV adjustments can be made based on the
hold or there are no sufficient internal targets to extent of motion. Figure 6.5 shows an example of
track for breath-hold, the patient may be consid- an ITV adjustment made based on the daily respi-
ered for simulation and treatment free breathing. ratory binned images for a patient with metastatic
The patient should only be treated free breathing melanoma to the liver. In this case, the extent of
if the target and critical OARs are sufficiently the motion was larger than the initial planned
visible despite motion artifacts on the MRI, and motion.
additional planning considerations should be Breath-hold techniques are not widely adopted
incorporated for the uncertainty of tumor and on the Elekta Unity system (non-gating version
OAR position, such as larger planning target vol- of the system). However, they can be feasible by
ume (PTV) margins for the tumor or the use of coaching the patient and manually holding the
planning organ-at-risk volumes (PRVs) for beam when the target is outside of the breath-
OARs. hold window. This technique is more reliant on
For non-gated treatments on the MR-Linac, user operation and will need to be carefully
for example, the Elekta Unity system with no gat- implemented. A recently released upgraded ver-
ing capability, motion can be managed through sion of the Elekta Unity system allows for breath-
more conventional techniques. Patient selection hold treatments (deep inspiration breath-hold,
should account for the feasibility of treatment on inhale, exhale), where the patient can be imaged
the system depending on the target’s motion and treated in a voluntary breath-hold position.
extent. If the motion assessed during 4D-CT sim-
ulation does not exceed the clinical threshold for
treatment without gating (e.g., 1.5 cm), then the Tracking and Gating
patient may be a good candidate, given that no
other contraindications are present. The extent of Cine MR images are single plane images that are
the motion to allow for non-gated treatments may taken with very short repetition and echo times,
vary between clinics. To reduce respiratory allowing multiple images to be taken per second.
motion, a compression belt can be used. The This allows close to real-time imaging of a
measured motion is then accounted for in the patient’s internal anatomy. More importantly, this
planning process by determining the ITV which allows for both visualization of tumors and OARs
is used to generate the PTV. during treatment and the development of algo-
Prior to treatment, daily motion-average or rithms to track them from cine frame to cine
mid-position images generated from the acquired frame [22]. Motion management in MR-guided
4D-MRI is used for image registration and pre- radiotherapy involves the use of 1 to 3 planes of
these cine MR images, a predefined tracking tar- dently seen through the motion of the diaphragm
get structure and a boundary that defines the (superior-inferior motion) and chest (anterior-
range of motion allowable for the target. The tar- posterior motion). However, lateral movement is
get structure is defined during the initial planninggenerally not visible on the sagittal plane.
process and often redrawn during the adaptive Consequently, lateral tumor motion out of the
workflow accounting for the difference in its treatment region or lateral OAR motion into the
daily position. During treatment, this structure is treatment region may not also be easily seen and
deformed first from the adaptive re-planning accounted for. Single plane tracking in other
MRI to the first cine, then from each preceding planes such as the coronal plane provides infor-
cine onto the next succeeding cine image. mation on the lateral motion of the tumor and
Because of this, care must be taken when defin- OARs as well as in the superior-inferior direc-
ing the tracking structure. tions, but not in the anterior-posterior direction.
Currently, there are generally two different Single plane tracking in the axial plane provides
types of motion management, namely, tracking lateral motion and anterior-posterior information
with manual gating, where users have to actively but not superior-inferior motion. This remains a
determine when to turn the beam off manually, considerable drawback in machines capable of
and tracking with automatic gating, where the only single plane cine, as information on motion
Linac is designed to automatically beam off when in at least one direction is lacking.
certain predefined conditions are met. These con- Tracking in multiple planes provides informa-
ditions are user defined and ideally involve track- tion on tumor and OAR motion in all three direc-
ing the tumor itself, though this is not always tions, superior-inferior, anterior-posterior, and
possible depending on the tumor size and relative lateral. A recent upgrade to the ViewRay
intensity difference from its environment. In MRIdian, the A3i system, has multiplane track-
automatic gating, these conditions would gener- ing capability. This enables tracking in all three
ally involve constraints on the motion of the planes, although most often the tumor or OAR is
tumor or OAR and the confidence in the defor- tracked in the sagittal and coronal planes during
mation of the structure being tracked. Acquisition treatment. Figure 6.6 provides an example of
speed of cine images is on the order of a few hun- multiplane tracking for a pancreatic tumor on the
dred milliseconds per image (four or eight frames ViewRay MRIdian A3i system showing the
per second for Viewray MRIdian and five frames tumor tracked in the sagittal and the coronal
per second for Elekta Unity), allowing users to planes.
track motion multiple times during a single respi- The Elekta Unity system allows for three-
ratory cycle. plane tracking for non-gated treatments.
Figure 6.7 shows a frame with three-plane track-
ing from continuous MR cine imaging during
Single Plane Vs. Multiplane Tracking treatment for a patient with metastatic renal cell
carcinoma treating the intra-abdominal lymph
Due to computational limitations, for now, real- nodes.
time 4D-MRI cannot be generated on MR-Linacs The upgraded version of the Elekta system
for motion management; instead, single or mul- allows for automatic gating in two planes, the
tiple 2D cine planes are utilized. Early versions sagittal and coronal. The two planes cover the
of the ViewRay MRIdian enable single plane motion extents in all three directions: laterally,
tracking and gating during treatment in the sagit- anterior-posterior, and superior-inferior. Gating
tal plane containing a tracking structure used to strategies on the system are divided into two
gate the radiation beam. The sagittal view is suf- categories: non-respiratory and respiratory. The
ficient to visualize the superior-inferior and non-respiratory “exception gating” allows the
anterior-posterior motion, thus allowing the user beam to be held when the target moves outside
to account for respiratory motion as most evi- of a defined tolerance. This technique can be
80 E. Huynh et al.
Fig. 6.6 Representative frames from continuous MR with a pancreatic tumor tracking structure (red) and gating
cine imaging during beam-on for a pancreatic tumor region of interest (yellow). Left, sagittal cine view; right,
receiving MRI-guided SBRT on the ViewRay MRIdian coronal cine view
Fig. 6.7 Representative frames from continuous MR Unity system. The green contour is the “Mask” used for
cine imaging during treatment for a patient with meta- tracking. Top left, coronal cine view; bottom left, axial
static renal cell carcinoma, treating the intra-abdominal cine view; right, sagittal cine view
lymph nodes, receiving MRI-guided SBRT on the Elekta
used for pelvis and head and neck cases. For ing structures in the tracking region. The
respiratory gating, three strategies can be used: deformation algorithm is intensity based, so
free breathing (time averaged position), full the difference in intensity should be stable
exhalation, and breath-hold (deep inspiration regardless of differences in B0 homogeneity
breath-hold, inhale, exhale). For the free breath- due to gantry position. Figure 6.8 shows an
ing strategy, the patient can breathe freely, and example of an ideal, contrast-enhanced liver
the beam will be held if the target moves out- lesion that can be used for tracking. This
side of the defined tolerance. The full exhala- lesion was successfully tracked because of the
tion strategy allows the patient to breathe freely, contrast difference between the lesion,
and the beam will only be enabled at the end of abdominal fat, and the surrounding liver.
the exhale position. With the breath-hold strat- Contrast agents, such as gadoxetic acid, may
egy, the patient can be imaged and treated in a be used to help increase intensity differences
voluntary breath-hold position. These respira- between the tracking region and its surround-
tory-based techniques are most suitable for ings when the difference is not as distinct.
patients with treatment sites in the abdomen When tracking contrast-enhanced lesions,
and thorax. the amount of time between injection of con-
trast and start of treatment should be taken
into account during the adaptive workflow to
Considerations for Tracking minimize the contrast from washing out at
and Gating Structures treatment. For gadoxetic acid, since contrast
is enhanced for 20 min to 120 min post injec-
There are several important considerations for tion, this translates to an approximately
determining the appropriate structure to track on 100 min time limit from start of imaging to
the real-time cine that will gate the radiation the end of treatment before contrast begins to
beam on and off. decrease. In addition, while blood vessels/
ducts may seem like ideal candidates for
1. The most important consideration is the rela- tracking on the 3D MR image, the fluid flow
tive intensity difference between the region in them could cause their brightness to be
being tracked and its surroundings. There unstable relative to their surroundings in sin-
must be a stable, distinct, and well-defined gle plane cine MRI and could result in subop-
difference in image intensity from surround- timal tracking in real time.
a b
Fig. 6.8 (a) Sagittal view of an MR acquired on the Gd-based contrast, Eovist, administered to differentiate
ViewRay MRIdian of a liver lesion that did not require the liver lesion (red arrow) from the healthy liver that was
Gd-contrast since the liver lesion had good contrast with successfully used for tracking (right)
the rest of the liver (left) (b) and another patient that had
82 E. Huynh et al.
2. Imaging speed considerations: of particular might be necessary to track the whole liver or
interest for this are lung and liver lesions. a surrogate that can be assumed to have cor-
Because the pixel spacing for cine images are related motion. In these cases, however, a
relatively large, it may not be possible to reli- reconsideration of tolerances such as limits on
ably track small lesions. For example, in the overall motion of the surrogate might be
lungs, while the intensity difference between necessary.
lesions and the surrounding lung is large and 4. Distinctness of the tracking region: ideally,
stable, some lesions are small enough and the tracking region would be the only region
undergo enough motion that a lower acquisi- with differing densities. For instance, in
tion speed may be required to allow for greater Fig. 6.7, there are multiple dark regions all
temporal averaging. This would, in turn, allow over the liver that are similar in size and inten-
for better tumor visualization. However, sity to the gross tumor volume. During beam
because slower acquisition does result in delivery, the patient’s breathing motion can
greater averaging, it can wash out the inten- cause deformation of the liver, resulting in the
sity difference between the tracking region software and users being unable to determine
and its surroundings. While usually not an which dark region is the actual lesion. In this
issue for lung lesions due to the large differ- instance, the decision was made to track a
ences in voxel intensity between lung tissue whole segment of the liver with tighter allow-
and tumor, this may decrease the intensity dif- able motion.
ference in other cases (e.g., between liver
lesions and healthy liver).
3. In relation to consideration #2, the size of the Multi-Target Tracking
tracking region cannot be too small (less than
~0.5 cm minimum distance). If the lesion is Current MR-guided linacs can image and track in
only a few voxels large, then blurring and multiple cine planes. In cases where tracking is
motion during cine will make it difficult not limited to a single plane, motion of a target can
only for the machine to reliably identify the only reflect motion in two out of three directions
lesion but also for users to verify the accuracy (e.g., superior-inferior and anterior-posterior in a
of the region being tracked. Figure 6.9 shows sagittal cine MRI). As a consequence of this,
an example of a lesion that was not a good tracking multiple targets with only one plane
candidate for tracking due to both its size and involves either (1) having the targets in the same
predominance of similar sized dark spots in image plane; (2) tracking only one lesion after
the surrounding liver. In such instances, it determining the correlation in motion between
the two targets and adjusting the PTV margin to
account for not visualizing the second target dur-
ing treatment; or (3) having multiple isocenters to
treat each target separately.
Some systems such as the ViewRay MRIdian
A3i system has the ability to track multiple struc-
tures (e.g., target and an adjacent OAR) in differ-
ent planes and to simultaneously gate on each
structure, which allows for enhanced verification
of proper positioning for critical structures. For
example, diseased abdominal lymph nodes may
Fig. 6.9 Sagittal view of a liver lesion treated on the not move significantly with respiration; however,
ViewRay MRIdian that was a poor choice for tracking due adjacent critical OARs such as the bowel may
to the very small area occupied by the hypointense region
and the presence of multiple, similarly sized hypointense
move. When these patients are treated with the
regions. The gross tumor volume is shown in blue breath-hold technique, tracking multiple struc-
tures ensures that the patient is in the correct Both structures need to be within the predefined
breath-hold position during radiation delivery. As gating region of interest in order for the beam to
seen in Fig. 6.10, the kidney can be tracked as a gate on. If either of the structures are outside of
surrogate for respiratory motion in the sagittal the gating region of interest, the beam will not
plane, and the abdominal lymph node can be turn on. A similar concept may be applied to
tracked in the coronal plane, ensuring that the patients where multiple tumors are being treated
treatment target has not moved during treatment. simultaneously.
Fig. 6.10 Representative frames from continuous MR node was tracked on the coronal plane (right) to verify
cine imaging during beam-on for a diseased abdominal that the treatment target was not moving. Top row, both
lymph node receiving MRI-guided SBRT on the ViewRay the kidney and abdominal node tracking structures (red)
MRIdian. Multiplane tracking with multiple tracking tar- are within the gating region of interest (yellow); the beam
gets was applied. A portion of the kidney was tracked on is gated on. Bottom row, the kidney tracking structure is
the sagittal plane (left) as a surrogate for respiratory outside of the gating region of interest, but the abdominal
motion and consequently OAR motion, and the abdominal node remains inside, the beam is gated off
84 E. Huynh et al.
Planning Considerations Given the features available with MR-guided

linacs, enhanced soft tissue image-guided set-up,
Planning considerations for patient immobiliza- accounting for inter-fraction motion through
tion are similar to conventional Linac treatment online adaptive planning, the size of margin
plans. As recommended by AAPM TG-176, required to maintain appropriate target coverage
attenuation through immobilization devices for MR-Linacs may be reduced compared with
should be accounted for in the dose calculation conventional Linac treatments. For example, a
[23]; this should also include attenuation through prospective study with rectal cancer patients
the RF receive coils. treated on an MR-Linac initially used conven-
For patient positioning, the decision between tional Linac PTV margins and then reduced the
treating patients with their arms up versus arms PTV margins after evaluating motion during
down is dependent on patient comfort but also treatment. The PTV margins were reduced for
has implications for treatment planning strate- both the mesorectum and elective nodal regions
gies. When treating a patient with their arms from 1 cm to 4–6 mm for the mesorectum and
positioned down, gantry angles that would allow 8 mm to 4–6 mm for the elective nodal regions
the beam to traverse through the arms are elimi- [25]. Without the use of real-time tumor imaging,
nated due to the uncertainty in arm position. the PTV margin must account for some uncer-
Furthermore, additional gantry angles may be tainty when surrogates are used for motion such
avoided clockwise and counterclockwise from as fiducials [26]. Under MR guidance with real-
each arm (e.g., within 2 cm in each direction time cine imaging and gating during treatment,
from the arm) since the exact arm position at the tumor itself is tracked in real-time enabling a
simulation may not be reproduced at each treat- reduction in the PTV margin. For breath-hold
ment day, and this reduces the probability that an treatments on the MR-Linac for lung, pancreas,
arm may be in the beam path. Although this fur- and adrenal tumors, van Sornsen de Koste et al.
ther limits the potential gantry angles that can be found that a PTV margin of 3 mm was sufficient
used in the plan, of note is that some of these gan- to account for almost all the uncertainties in the
try angles that are avoided due to the arms are study, however, recommended that a larger PTV
already avoided in an arms up scenario due to the margin be considered for free breathing scenarios
couch corners/rails that some groups already [27]. PTV margins of 3 mm are typically used on
avoid for the high attenuation through the outer the ViewRay MRIdian with gated breath-hold
edges of the couch. The arms could also purpose- treatments.
fully be placed a bit farther from the body to When treating patients on the MR-Linac with
reduce the number of avoidance angles. Arm a non-gated approach, the total treatment volume
clearance will need to be accessed at each treat- will be larger since an ITV must be generated,
ment on MRI. Van den Wollenberg et al. com- such as on the Elekta Unity MR-Linac. The
pared plan quality metrics for patients with selection of PTV margin is site specific, but typi-
oligometastatic liver disease treated on a conven- cally ranges from 3 mm to 1 cm from the
tional Linac with arms up versus an MR-Linac ITV. Using a gating approach could reduce the
with a 15 IMRT beam plan with arms down and PTV margin used and overall treatment volume
avoiding the arms. They found that target cover- since the PTV would be expanded from the GTV
age was comparable between the plans. Dose or CTV, instead of an ITV.
conformity and dose to organs-at-risk were Patients treated with the breath-hold technique
slightly worse for MR-Linac plans due to the have a significant reduction in tumor motion
restriction of gantry angles. However, the compared to being treated free breathing. Ehrbar
MR-Linac plans still met all the clinical target et al. evaluated the residual motion during gating
objectives and OAR constraints [24]. for lung and liver lesions and found that it was
small and comparable for all patients (mean 2D work for all patients, and some patients will still
root mean square error of 2.4 mm). Furthermore, need coaching from a therapist.
compared to the static planned dose, with gating,
the dose to the CTV was slightly reduced, with a
difference of -0.6% for the mean dose and -1.7% Patient Education and Coaching
for the dose to 95% of the CTV. This study indi-
cates that gating based on internal anatomy mini- For gated treatments, a patient can perform either
mizes target motion during radiation delivery and deep inspiration, mid-inspiration, or end expira-
safe and robust delivery of high radiation doses to tion breath-hold, according to what is clinically
these tumors [28]. determined most suitable for the treatment. For
instance, a mid-inspiration breath could allow the
patient to adjust the amount of air in the lungs
Patient Coaching more easily, alternating inhaling or exhaling
more air as needed [29]. On the other hand, end
One major advantage of motion management expiration or deep inspiration may position the
with MR-Linacs is that radiation delivery can be tumor farther away from OARs and/or be easier
triggered by the motion of the lesion itself, for patients to maintain [30]. Ultimately, the cli-
instead of using a surrogate such as in surface- nician will need to assess both the breath-hold
guided radiation therapy. Alongside these bene- capabilities of the patient and the resulting loca-
fits, however, the importance of proper patient tion of the tumor relative to OAR before deciding
education cannot be overstated. Generally, on which breath-hold techniques would be the
patients will need to be coached on their breath- most ideal before simulation. This assessment
ing patterns and, especially in gating, be trained can be done at the initial patient consultation
to take reproducible breaths. where the physician can evaluate the breath-hold
capabilities of the candidate patient and educate
the patient on the breath-hold expectations at
Feedback Mechanism simulation and treatment.
The best opportunity for active patient coach-
In some MR-Linacs, the patient can view the cine ing is during the simulation. First, a patient would
images used to track tumor motion. The cine is be instructed to perform a deep or mid-inspiration
projected in real time onto a screen located at the or end expiration breath-hold, and a 3D MRI is
wall close to the opposite end of the MRI bore. acquired. A temporary tracking structure is then
By using mirrors, patients are able to view the drawn on the treatment region as verified by the
screen throughout the treatment and watch as physician, and a cine is performed. While watch-
their tumor moves according to how much air ing the cine, the patient can be coached to attempt
they inhale/exhale. The ViewRay MRIdian A3i to reproduce the breath-hold performed during
system has the capability to display both the cine the acquisition of the 3D MRI. After the simula-
and tracking structure or only the tracking struction, the patient should be reminded to practice
ture for patient viewing. This feedback mecha- breath-hold prior to treatment.
nism can empower patients to self-coach,
deciding when they will breath-hold and when to
recover their breath, while the therapist super- Conclusion
vises the treatment, as opposed to following
instructions dictated by a therapist throughout the MR-Linacs offer a new era for motion manage-
duration of treatment. While the feedback mech- ment strategies with visualization of soft tissue
anism works well for some patients, it does not targets and the ability to gate the radiation beam
86 E. Huynh et al.
by tracking the actual tumor itself. These 10. Gurney-Champion OJ, McQuaid D, Dunlop A, et al.
improved motion management strategies over MRI-based assessment of 3D intrafractional motion
of head and neck cancer for radiation therapy. Int J
X-ray-based image-guided radiation therapy Radiat Oncol Biol Phys. 2018;100:306–16.
offer the opportunity to reduce target margins, 11. Sixel KE, Aznar MC, Ung YC. Deep inspiration
thereby reducing toxicity to normal tissues and breath hold to reduce irradiated heart volume in
yet maintaining accuracy in treating tumors. breast cancer patients. Int J Radiat Oncol Biol Phys.
2001;49:199–204.
While there are still several limitations with 12. Mampuya WA, Nakamura M, Matsuo Y, et al.
MR-guided motion management strategies, it is Interfraction variation in lung tumor position with
anticipated that the continuous development of abdominal compression during stereotactic body
MR-Linacs, both hardware and software, will radiotherapy. Med Phys. 2013;40:091718.
13. Stemkens B, Paulson ES, Tijssen RHN. Nuts and
overcome these limitations of current MR-Linacs. bolts of 4D-MRI for radiotherapy. Phys Med Biol.
2018;63:21TR01.
Acknowledgments We thank Zhaohui Han and Jennifer 14. Paulson ES, Ahunbay E, Chen X, et al. 4D-MRI driven
Campbell for their assistance in creating ViewRay MR-guided online adaptive radiotherapy for abdomi-
MRIdian figures. nal stereotactic body radiation therapy on a high field
MR-Linac: implementation and initial clinical experi-
ence. Clin Transl Radiat Oncol. 2020;23:72–9.
15. Omari E, Jassar H, Straza M, et al. Daily 4DMRI
References assessed motion and its impact on ITV for non-gated
MRI-guided adaptive radiation therapy of liver can-
1. Brandner ED, Wu A, Chen H, et al. Abdominal organ cer. Hoboken, NJ: Wiley; 2022.
motion measured using 4D CT. Int J Radiat Oncol 16. Winkel D, Bol GH, Kroon PS, et al. Adaptive radio-
Biol Phys. 2006;65:554–60. therapy: the Elekta Unity MR-linac concept. Clin
2. Keall PJ, Mageras GS, Balter JM, et al. The man- Transl Radiat Oncol. 2019;18:54–9.
agement of respiratory motion in radiation oncol- 17. Mutic S, Dempsey JF. The ViewRay system: magnetic
ogy report of AAPM Task Group 76. Med Phys. resonance-guided and controlled radiotherapy. Semin
2006;33:3874–900. Radiat Oncol. 2014;24:196–9.
3. Liu HH, Balter P, Tutt T, et al. Assessing respiration- 18. Cuccia F, Alongi F, Belka C, et al. Patient positioning
induced tumor motion and internal target volume and immobilization procedures for hybrid MR-Linac
using four-dimensional computed tomography for systems. Radiat Oncol. 2021;16:183.
radiotherapy of lung cancer. Int J Radiat Oncol Biol 19. Güngör G, Serbez İ, Temur B, et al. Time analysis of
Phys. 2007;68:531–40. online adaptive magnetic resonance-guided radiation
4. Sim AJ, Palm RF, DeLozier KB, et al. MR-guided therapy workflow according to anatomical sites. Pract
stereotactic body radiation therapy for intracardiac Radiat Oncol. 2021;11:e11–21.
and pericardial metastases. Clin Transl Radiat Oncol. 20. Raaymakers BW, Jürgenliemk-Schulz IM, Bol GH,
2020;25:102–6. et al. First patients treated with a 1.5 T MRI-Linac:
5. Bernstein HM, Leon W, Daly ME, et al. Noninvasive clinical proof of concept of a high-precision, high-
stereotactic radiation for refractory ventricular tachy- field MRI guided radiotherapy treatment. Phys Med
cardia after failure of cardiac sympathetic denerva- Biol. 2017;62:L41–50.
tion. JACC Case Rep. 2022;4:1189–94. 21. Kurz C, Buizza G, Landry G, et al. Medical physics
6. Dees-Ribbers HM, Betgen A, Pos FJ, et al. challenges in clinical MR-guided radiotherapy. Radiat
Inter- and intra-fractional bladder motion dur- Oncol. 2020;15:93.
ing radiotherapy for bladder cancer: a compari- 22. Menten MJ, Fast MF, Wetscherek A, et al. The
son of full and empty bladders. Radiother Oncol. impact of 2D cine MR imaging parameters on auto-
2014;113:254–9. mated tumor and organ localization for MR-guided
7. Hamlet S, Ezzell G, Aref A. Larynx motion associated real-time adaptive radiotherapy. Phys Med Biol.
with swallowing during radiation therapy. Int J Radiat 2018;63:235005.
Oncol Biol Phys. 1994;28:467–70. 23. Olch AJ, Gerig L, Li H, et al. Dosimetric effects
8. Li H, Chen H-C, Dolly S, et al. An integrated model- caused by couch tops and immobilization devices:
driven method for in-treatment upper airway motion report of AAPM Task Group 176. Med Phys.
tracking using cine MRI in head and neck radiation 2014;41:061501.
therapy. Med Phys. 2016;43:4700. 24. van den Wollenberg W, de Ruiter P, Nowee ME, et al.
9. Bruijnen T, Stemkens B, Terhaard CHJ, et al. Investigating the impact of patient arm position in an
Intrafraction motion quantification and planning tar- MR-linac on liver SBRT treatment plans. Med Phys.
get volume margin determination of head-and-neck 2019;46:5144–51.
tumors using cine magnetic resonance imaging. 25. Intven MPW, de Mol van Otterloo SR, Mook S,
Radiother Oncol. 2019;130:82–8. et al. Online adaptive MR-guided radiotherapy for
rectal cancer; feasibility of the workflow on a 1.5T efficiency and dose reconstruction for stereotactic
MR-linac: clinical implementation and initial experi- treatments of the liver and lung. Radiother Oncol.
ence. Radiother Oncol. 2021;154:172–8. 2022;174:101–8.
26. Ge J, Santanam L, Yang D, et al. Accuracy and 29. Lens E, Gurney-Champion OJ, Tekelenburg DR,
consistency of respiratory gating in abdomi- et al. Abdominal organ motion during inhalation
nal cancer patients. Int J Radiat Oncol Biol Phys. and exhalation breath-holds: pancreatic motion at
2013;85:854–61. different lung volumes compared. Radiother Oncol.
27. van Sörnsen de Koste JR, Palacios MA, Bruynzeel 2016;121:268–75.
AME, et al. MR-guided gated stereotactic radiation 30. Rodriguez LL, Kotecha R, Tom MC, et al. Impact of
therapy delivery for lung, adrenal, and pancreatic breath-hold MR-guided radiotherapy (MRgRTBH)
tumors: a geometric analysis. Int J Radiat Oncol Biol vs. free-breathing CT image-guided radiotherapy
Phys. 2018;102:858–66. (CT-IGRTFB) on gastrointestinal sparing and dose
28. Ehrbar S, Braga Käser S, Chamberlain M, et al. conformality in adrenal SBRT. Int J Radiat Oncol Biol
MR-guided beam gating: residual motion, gating Phys. 2021;111:e527.
sCT and Dose Calculation
7
Emilie Alvarez-Andres, Fernanda Villegas,
Anaïs Barateau, and Charlotte Robert
Introduction photon beam RT since it conditions cross sec-

tions of photon-matter interactions. In addition to
rucial Role of the CT Scanner
C its usefulness for dose calculation, the planning
in Radiatiotherapy CT also allows the patient to be repositioned at
each treatment session using in-room kV X-rays
Conventional 3D radiation therapy (RT) plan- or cone beam CT (CBCT) imaging, which are
ning, as implemented since the introduction of essential for the implementation of image-guided
the CT scanner in the 1990s in RT departments, radiation therapy (IGRT). However, CT suffers
uses Hounsfield units (HU) to qualify tissue from very poor contrast when it comes to identi-
properties in terms of electron densities (ED). fying structures in volumes containing soft tis-
This correlation is obtained through a calibration sue, hence the use of combined CT/MRI imaging
curve for each set of acquisition/reconstruction for volume delineation, particularly for brain and
CT scan parameters, which allows the corre- pelvis [1] (Fig. 7.1, upper part). The recent com-
sponding HUs to be derived from inserts with mercialisation of hybrid machines combining an
known ED. This property is essential in external MR device and a linear accelerator (MR-Linac)
E. Alvarez-Andres A. Barateau
Department of Radiotherapy and Radiation Univ Rennes, CLCC Eugène Marquis, Inserm,
Oncology, Faculty of Medicine and University LTSI - UMR 1099, F-35000 Rennes, France
Hospital Carl Gustav Carus of Technische Universität e-mail: [email protected]
Dresden, Dresden, Germany
C. Robert (*)
OncoRay – National Center for Radiation Research UMR 1030 Radiothérapie Moléculaire et Innovation
in Oncology, Faculty of Medicine and University Thérapeutique, ImmunoRadAI, Université Paris-
Hospital Carl Gustav Carus, Technische Universität Saclay, Institut Gustave Roussy, Inserm,
Dresden; Helmholtz-Zentrum Dresden-Rossendorf, Villejuif, France
Dresden, Germany
Department of Radiation Oncology, Gustave Roussy,
Villejuif, France
F. Villegas e-mail: [email protected]
Radiotherapy Physics and Engineering, Medical
Radiation Physics and Nuclear Medicine, Karolinska
University Hospital, Stockholm, Sweden
Department of Oncology-Pathology, Karolinska
Institute, Stockholm, Sweden
e-mail:
[email protected]

90 E. Alvarez-Andres et al.
Fig. 7.1 Comparison of the two possible workflows on electronic density assignments in segmented regions
implemented in RT in 2023. In the first case (top), treat- (BDA, bulk density assignment). By contrast, the MRI-
ment planning is based on CT imaging to which MRI can only workflow (bottom) eliminates the planning CT and
be added to facilitate the delineation of target structures, therefore uses a sCT as early as at the planning stage,
where arising spatial registration uncertainties between required for dose calculation. In this workflow, geometric
the two modalities are addressed by increased margins. In fidelity over large fields of view becomes important to
this workflow, the main challenges rely in the production correctly restore all volumes and especially the external
of high-fidelity, artefact-free MR images, and the ability contour. The implemented method for generating sCT
to generate dosimetrically accurate sCT at each treatment must be generalisable to any new patient anatomy. Patient-
session for workflows that include MR-Linac treatments. specific quality controls must be implemented to ensure
However, this process can take advantage of the informa- that there are no dose-impacting mistakes in the genera-
tion provided by the CT scan acquired during the planning tion of the sCT. As CT imaging is still required to set up
stage for the implementation of adaptive radiation therapy the patient in the X-ray-based image-guided RT workflow,
(ART), i.e. based on the anatomy of the day. Two methods it is ultimately necessary to ensure that the generated sCT
are proposed by vendors for this purpose: methods based can perform this task
on deformable image registration (DIR) or methods based
7 sCT and Dose Calculation 91
has brought breakthrough technological solu- uation correction. In the early 2010s, methods to
tions to access the patient’s anatomy of the day assign uniform electron densities in previously
that, together with the subsequent adaptation of segmented regions and multi-atlas strategies
treatment plans, and makes real the possibility to were already proposed as solutions to this issue
“see what you treat”. The latter, when applied to [9, 10].
an online re-optimisation of treatment plans Recently, several groups have evaluated the
workflow, has however raised questions, hence cost-effectiveness of MR-only workflows, par-
bringing the need to generate synthetic CT scans ticularly in the management of prostate cancer
(sCT), also called in the literature “virtual CT”, [11, 12]. Based on ten patients, Persson et al. [11]
“substitute CT”, or “pseudo CT”. showed that the elimination of the CT scan and
the acceleration of the target delineation stage
were the main cost-saving factors, leading to a
Other Applications of sCT moderate cost reduction of about 200 EUR/
patient. A Finnish team tackled a similar prob-
The concept of generating sCTs is however not lem, but also considering notions of investment
new and has already been proposed for the imple- in the purchase of heavy equipment [12]. For a
mentation of MRI-only workflows, i.e. without period of 10 years and 300 patients annually
requiring the acquisition of a planning CT, for managed for prostate cancer, a 2% cost reduction
brain, head and neck, and pelvis applications was evaluated between CT + MRI workflow and
(Fig. 7.1, lower part). The goals of such MR-only MRI-only workflow. These studies are, however,
workflows are manifold. First, it allows to limit merely indicative as the prices of each procedure
the imaging departments’ burden and patient vary from one country to another and are based
irradiation by eliminating an X-ray-based exami- on local experiences and workflows. In both stud-
nation. Second, it removes anatomy differences ies, costs linked to the step of generating syn-
deriving from the non-simultaneous CT and MRI thetic scans were however not considered.
examinations, especially for pelvic tumour sites
where bladder and rectal fillings may differ.
Finally, it reduces the uncertainty underlying Challenges in MRI-Based Workflows
multimodal registration – evaluated to be about
2 mm for the brain and the pelvis and up to 5 mm The implementation of MRI-based workflows is
in the abdomen [2–6] – and so the derived mar- not straightforward in many respects, as illus-
gins between the clinical target volume (CTV) trated in Fig. 7.1, particularly since it requires the
and the planning target volume (PTV). MRI- development of specific quality assurance (QA)
based planning is also widely used in brachyther- programmes. These challenges are detailed in the
apy since the introduction of the concept of following lines which serve as an introduction to
image-guided adaptive brachytherapy, which has all the sections that will follow.
demonstrated benefits in terms of local control First, when using MR images for treatment
in locally advanced cervical cancer [7]. However, planning purposes, their geometric fidelity must
in brachytherapy, dose calculations are still cur- be ensured. This is of particular concern for MRI-
rently based on the TG43 formalism and can be only workflows, requiring large fields of
performed on MRI without methodological dif- view (FOV). Indeed, distortions can affect the
ficulties, as the patient is assumed to be only external contours, which are of major importance
composed of water [8], thus not requiring sCTs. for dose calculation, but also the definition of tar-
In nuclear medicine, the need to use sCT already get structures, their accuracy being crucial espe-
emerged at the time of the introduction of the cially for the implementation of stereotactic
PET-MRI bi-modality, for which the replacement treatments. Magnetic susceptibility artefacts or
of the CT modality by MRI raised methodologi- chemical shifts can also affect the images, lead-
cal questions for the implementation of the atten- ing to errors in the generation of sCT images, in
addition to difficulties/errors in volumes regions [14]. Then, atlas-based and voxel-based

delineation. approaches were both first introduced in 2011.
Second, for MRI-only workflows, robust Most recent voxel-based methods of sCT genera-
methods for generating sCTs need to be imple- tion from MRI are now based on machine learn-
mented, i.e. that are reliable regardless of the ing and particularly deep learning (DL). Table 7.1
patient anatomy and the quality of the MR presents the global characteristics of these meth-
images used as input data. In addition to routine ods and their respective advantages and draw-
QA, patient-specific quality checks must there- backs. Some of the sCT generation results
fore be conducted with the aim of verifying the produced by research teams will be presented in
trustworthiness of the result. Several solutions this section, the performance metric being
are possible for this purpose, including visual described in Table 7.2 and in section “Metrics for
checks, comparison with a reference CT, the use sCT Evaluation”.
of a second algorithm for sCT generation, or the
generation of uncertainty maps as second output
of the tool [13]. For treatments where patient Bulk Density Assignment
repositioning is based on X-rays, it is also impor-
tant to verify the ability of the generated sCT to The bulk density assignment method requires the
perform this task accurately and most impor- segmentation of the MRI into sub-volumes before
tantly without systematic errors. Finally, the use assigning them with pre-defined ED depending
of MR-Linac requires both rapid generation and on the tissue categories (such as air, bone, and
quality control of sCTs, with the patient lying on soft tissue).
the table. For MRI/CT-based workflows, the According to Johnstone et al. [14], the number
availability of the planning CT scanner is a of tissue classes considered and the assigned den-
strong advantage. sity values differ between studies. Several sCT
In this chapter, the first part will present the
studies, for prostate and brain RT, have consid-
methods proposed in the literature to generate ered one single water class [15–17]. The reported
sCTs, from the basic methods, such as bulk den- success of this simple approach is, however, lim-
sity, to more recent methods based on deep learn- ited with dose errors potentially exceeding 2%
ing. The second part will discuss the metrics used [18–22]. More complex scenarios were imple-
to evaluate the quality of these sCTs in the con- mented with the aim of identifying the number of
text of RT. The third part will summarise the key tissue categories necessary to achieve acceptable
points of MRI quality and recent recommenda- dosimetric errors.
tions for QA programmes for the implementation Based on four prostate patients, Lee et al. [23]
of safe MRI-based or even MRI-only workflow. generated a first water-only sCT (sCTW) by
A review of implemented academic and indus- assigning the whole body with the water ED
trial solutions in the clinic will conclude the derived from ICRU Report 46 [24]. A second
chapter. Throughout the chapter, the focus will be sCT (sCTWB), considering both water and bone
on methods integrated into commercially avail- classes, was computed with the additional bone
able MR-Linacs. assignment obtained averaging the bone CT
intensities over all the patients. Based on a 64 Gy
PTV prescription dose, the maximum root mean
sCT Generation from MRI square differences (RMSD) inside the 70% CT
for Radiotherapy Dose Calculation: and sCTs isodoses were equal to 1.33 Gy and
State of the art 0.88 Gy for sCTW and sCTWB, respectively, prov-
ing the crucial need of a minimum of a two class-
The first reported way for RT dose calculation in density assignment approach.
the context of an MRI-only workflow was to per- This observation was further confirmed in
form a bulk density assignment on MRI sub- 2008 by Eilertsen et al. [20], with ten prostate
Table 7.1 Presentation of the main sCT generation methods with their prerequisites, advantages, and drawbacks
Bulk density Voxel-based approaches
assignment Atlas-based approaches Statistics Machine learning
Definition Assignation of Performing deformable Use of voxel intensity information for sCT
homogeneous registration of the given prediction
densities to MRI MRI to one or several
segmented classes co-registered MRI-CT
atlas(es)
Pre-requisite Segmentations on Atlas database (MRI-CT Statistic model Training step to establish
MRI + density values pairs) definition the MRI to CT
per structure translation model
Advantages • Simple • Continuous values of • Continuous values • Continuous values of
• Fast HU of HU if HU
regression • Fast
• Does not require access
to matched data for the
latest deep-learning
methods
Inconvenients • Need of specific • Dependent on • Potential need of • Potential need of
MR sequence(s) for deformable registration specific MR specific MR
bone delineation accuracy sequence(s) sequence(s) enhancing
• Homogeneous areas • Calculation time enhancing bone bone visibility
in the sCT (depending on the visibility • Not robust to atypical
number of patients • Model patient anatomies (if
considered in the performance not present in training
database) dependency data)
• Not robust to atypical • Learning step
patient anatomies computationally costly
patients. The axial T2-weighted MRI were con- The bulk density assignment method is simple
verted into water-equivalent sCT by assigning to implement but faces one major challenge,
1.02 g/cm3 to the body (sCTWO) and two water- being the distinction between bone tissue and air-
bone sCTs with the additional bone segmentation filled cavities, since both types of tissue show a
and densities assignment of 1.3 g/cm3 (sCTWB1) similar signal on routine anatomical MR sequences.
and 2.1 g/cm3 (sCTWB2). After applying a col- Several sequences have been developed to better
lapsed cone dose algorithm, the dose volume his- visualise tissues with very short T2 relaxation
togram (DVH)-based differences between mean times, in particular bones, tendons, and ligaments.
doses to the CTV on CT and sCT were equal to These sequences, called “UTE” for ultrashort echo
−0.9%, 0.2%, and 4.2% for sCTWO, sCTWB1, and time or “ZTE” for zero echo time, however, greatly
sCTWB2, respectively, highlighting the importance extend the examination times and thus the risks of
of the adequate bone density assignment. patients movements. The bulk density assignment
The same conclusions were obtained in case method, if manually implemented, can be also
of head-and-neck tumours for seven patients with tedious and biased by high inter-operator variabil-
a 2D axial T1-weighted MRI [25]. In this study, ity. Automatic segmentation solutions were devel-
three sCTs were generated, namely, a water oped very early on, particularly in the context of
equivalent sCT (sCTW), a water/bone sCT nuclear medicine, based either on clustering meth-
(sCTWB), and a water/bone/air sCT (sCTWBA), ods or on neural networks [26, 27], and are now
based on bone and air densities equal to 1.53 g/ commonly used in RT [28, 29].
cm3 and 0 g/cm3, respectively. For the PTV Several studies have evaluated the bulk den-
receiving 70 Gy, V100% [minimum; maximum] sity assignment method in the context of
were [6.5%; 65.5%], [−0.2%; 5.7%], and [1.5%; MR-Linac treatments, demonstrating its clinical
7.4%], respectively. applicability. For breast cancer, Koerkamp et al.
Table 7.2 Metrics used in the literature for the evaluation of sCT scans generated from
94
MR images. This table is inspired by (Boulanger et al., 2021) [32] and (Vandewinckele
et al., 2020) [13]
Type of
metrics Metric Definition Ideal value + Advantages/- Disadvantages.
Image Intensity- ME: Mean error 1 0 HU + Is usually reported
evaluation based ME = ∑iN=1 ( sCTi − CTi ) + Easy to be calculated
metrics N + Can be calculated within different regions of interest
MAE: Mean 1 0 HU - Does not show the spread in the voxel-wise differences since mean/
absolute error MAE = ∑iN=1 sCTi − CTi standard deviation are calculated
N
- Hide patient-wise dispersion as values at the cohort scale are usually
MSE: Mean 1 2
0 HU reported
squared error MSE = ∑iN=1 ( sCTi − CTi )
- No information about the exact location of the error (only organ-wise)
N
RMSE: Root 0 HU - Difficult to compare between studies
1 N 2 - Non-trivial association with dosimetric impact
mean squared RMSE = ∑i=1 ( sCTi − CTi )
error N - Gives information of absolute error and not of the relative error
PSNR: Peak Maximum + Easy to be calculated
 Q2 
signal-to-noise PSNR =   value of dB + Gives some information about the relative error
ratio  MSE  - No information about the location of the error
- Not representative of the human vision
SSIM: Structural 1 + Carries information about inter-dependencies between pixels
similarity metric SSIM =
( 2 µCT µsCT + C1 ) ( 2σ CTsCT + C2 ) + More accurate representation of the human vision compared to PSNR
( µCT 2 + µsCT 2 + C1 ) (σ CT 2 + σ sCT 2 + C2 ) - More difficult to calculate
NCC: 1 + Evaluates the level of correlation between two signals
1 ( CTi − µCT ) ( sCTi − µsCT )
Normalised NCC = ∑i + Less sensitive to linear variations in intensities in the two images
cross-correlation N σ CTσ sCT compared to conventional cross-validation
- Unintuitive interpretation
Geometric DSC: Dice 2 (VCT ∩ VsCT ) 1 + Relevant for dose calculation/dose reporting
fidelity similarity DSC = + Relevant for repositioning
metrics coefficient VCT + VpCT - DSC is dependent on voxel size
- DSC is dependent on the volume of the structures and thresholds/
segmentation quality to assess it
- Gives no information on actual HU values
HD: Hausdorff H(sCT, CT) = max (h(sCT, CT), h(CT, sCT)) 0 mm + Focus on the boundary of the contours = > relevant for repositioning
distance - Sensitive to outliers = > 95th percentile is often preferred
- Non-trivial association with dosimetric impact
MASD: Mean dave ( SCT ,SsCT ) + dave ( SCT ,SsCT ) 0 mm + Focus on the boundary of the contours = > relevant for repositioning
absolute surface MASD( A,R ) = - Non-trivial association with dosimetric impact
distance 2
E. Alvarez-Andres et al.
7
Type of
metrics Metric Definition Ideal value + Advantages/- Disadvantages.
Dose Dose Voxel-to-voxel Difference between the dose distribution 0 Gy or 0% + Clinically relevant
evaluation difference dose differences computed on the reference CT and on the sCT + Spatial evaluation possible with subtraction of dose distributions
metrics - Average can hide potential deviations
DVH difference Dose differences on DVH specific points, for a 0 Gy or 0% + Easy to calculate if both DVHs are available
given structure - No spatial information
- Care should be taken in transferring/warping structures sets
- A difference in DVH or DVH parameter can be caused by an “error” in
the sCT or in the contour or by a difference in the anatomy
sCT and Dose Calculation
Gamma Mean gamma Value of the mean gamma 0 + Used to give an overall representation of error usability of the sCT
analysis Gamma pass rate Percentage of pixels/voxels with a gamma 100% + Analysis can be tweaked to preference (global/local, 2D/3D, dose
metrics value lower than 1 difference, DTA value and threshold)
+ Spatial evaluation possible with gamma maps
- Dependency to target volumes, dose algorithm, treatment technique
- Difficult to compare between studies
Patient repositioning Matching Compare translational and rotational 0 mm, 0° + Vital information about ability of the sCT for patient repositioning
accuracy differences in case of: - True matching value is unknown: No gold standard available
- CBCT and CT/sCT registration - No standard on the procedure, e.g. handling of residual misregistration
- kV imaging and CT/sCT-derived DRR
registration
Abbreviations: N number of voxels, sCTi intensity of the ith voxel of the sCT, CTi intensity of the ith voxel of the CT, Q range of voxel values of sCT and reference CT, μCT mean value
of CT, μsCT mean value of sCT, σCT and σsCT standard deviation of the reference CT and sCT, C1 and C2 are expressed as (k1Q)2 and (k2Q)2, V volume of CT and sCT, dave absolute
Euclidean distance, h(X, Y) is the largest distance from a point in X to the nearest point in Y, SCT surface of the segmentation of a volume of interest in the CT, SsCT surface of the seg-
mentation of a volume of interest in the sCT, DRR digitally reconstructed radiograph, DTA distance to agreement
95
[30] collected ten patients from the ABLATIVE Atlas-Based Approaches

trial and assigned mean ED from the CT seg-
mented volumes for the sCT generation. Three The atlas (multi-atlas) approach consists of
scenarios were investigated: lungs-gross tumour forming one (several) atlas(es), corresponding to
volume (GTV)-remaining body (sCTLT), lungs- a registered pair(s) of CT and MRI, prior to the
GTV- chest wall-ipsilateral breast-remaining sCT generation. The MRI of the atlas(es) is (are)
body (sCTLTCB), and lungs-GTV-chest wall- deformably registered to the incoming patient
ipsilateral breast-ribs-remaining body (sCTLTCBR). MRI, before applying the obtained deformation
Intensity modulated radiation therapy (IMRT) field(s) to the corresponding CT atlas(es). In
treatment plans were directly optimised on the case of the multi-atlas approach, the multiple
sCT before being re-calculated on CT with a generated warped CT atlases are finally com-
Monte Carlo algorithm simulating a 1.5T bined to form a final sCT. This method has sev-
MR-Linac. The obtained medians for the body eral advantages [14, 32]. Firstly, it is based on
mean absolute error (MAE) were equal to the use of conventional MRI anatomical
106HU, 104HU, and 104HU for sCTLT, sCTLTCB, sequences and, thus, does not involve any
and sCTLTCBR, respectively. For these same increase in acquisition time. Furthermore, this
images, median gamma pass rates of 95.8%, method has been shown to be robust to artefacts
96.3%, and 95.9% were obtained for the 3D that could affect MRI [33]. Finally, the same
2%/0 mm criterion combined with a 10% dose method can be applied to the deformation of
threshold applied to 20 Gy. Thus, it suggested organs-at-risk (OAR), fulfilling two clinical
that segmenting the lungs, GTV, chest wall, and tasks in a single operation. The major difficul-
ipsilateral breast, and assigning them specific ties, however, relate to the ability of such meth-
ED, led to the best compromise between image ods to accurately recover the external contour,
processing complexity and performance. highly impacting the dose computation [34], and
However, this study raises the issue of the need its application to atypical anatomies, in particu-
for an alternative workflow in case the scanner is lar for brain tumours where a surgical procedure
unavailable. may have led to the removal of part of the skull
Similarly, Ding et al. [31] investigated the [35]. The calculations are also very time inten-
optimal tissue classes in the case of five patients sive, with, for instance, 10 min for a 15 patient
presenting cervix carcinoma treated with the large atlas computed on a NVIDIA Titan X
Elekta Unity (Elekta AB, Stockholm, Sweden). graphical processing unit (GPU) [36].
In the first scenario, the sacrum, ilium, ischium, According to the review of Johnstone et al.
pubis, and coccyx structures were set as water- [14], straightforward approaches, involving the
equivalent, whereas the rest of the body, includ- use of a single average atlas for pelvis sCT gen-
ing femoral heads, was assigned with the averages eration, have first been introduced in the litera-
of relative ED of the corresponding CT structures ture [34, 37]. In 2015, Sjölund et al. [38] collected
(sCTFB). For the second scenario, these CTs and T1-weighted MRIs of ten head tumour
CT-derived relative ED averages were assigned patients to generate the multi-atlas database. The
to all the volumes (sCTEB). Dose calculation on Morphon deformable registration resulted in a
sCTEB provided better results than on sCTFB, at series of warped CT atlases that were either com-
least in terms of gamma analysis, which proved bined with mean or median operators. The head
the superiority of complete bone segmentation MAE resulted in differences of 123.2HU ±
over the conventional approach only including 18.6HU and 113.4HU ± 17.6HU for the mean
femoral heads. Yet, this higher number of required and median strategies, respectively, the latter
volumes may result in a time-consuming seg- being of superior accuracy to combine multiple
mentation task in clinics. warped CT atlases.
A more elaborated approach was proposed by clinical translation owing to the potential need of
Farjam et al. [39] for prostate patients. It involved exotic MRI sequences enhancing bone visibility,
23 pairs of CT and dual fast field echo mDixon such as UTE or ZTE [43].
MRI. During the fusion of the warped CT atlases, First introduced in 1974 [44], the Gaussian
the generalised registration error [40], a metric mixture models (GMM) belong to the unsuper-
evaluating the registration inaccuracy, was com- vised machine learning category and are often
puted between the deformed MRI atlases and the used for data clustering. In the context of sCT
processed MRI of the incoming patient. The final generation, tissues from MRI and CT are assumed
sCTs were finally obtained through a weighted to have Gaussian distributions, enabling the elab-
mean, with weights equal to the metric inverse. oration of a regression between these two inten-
For the 13 patients of the testing set, the mean sity types [45]. The main studies relying on this
MAE were equal to 47HU, 116HU, 36HU, and approach were reviewed by Johnstone et al. [14].
47HU for the body, bone, fat, and muscle regions, In 2011, Johansson et al. collected data from five
respectively. The initial treatment plan recalcula- patients presenting brain CT, UTE, and
tion resulted in mean and maximum DVH- T2-weighted MRI sequences [46]. The GMM,
extracted doses to the PTV equal to 0.8% ± 0.2% which was elaborated based on head voxels only,
and 0.9% ± 0.3%, respectively. resulted in a MAE of 137HU. The potential of
In this same perspective of similarity-based these sCTs for dosimetry and patient setup accu-
sCT process optimisation, Uh et al. [41] per- racy was evaluated by Jonsson et al. [47]. The
formed a pioneering study in 2014, involving 26 dose calculation on CT and sCT of five glioblas-
brain tumour paediatric patients. The fusion of toma patients resulted in maximum mean DVH
warped CT atlases, following a pattern recogni- differences of 0.8% in target volumes and
tion approach, was performed with the extraction 3%/3 mm gamma pass rates without dose thresh-
of 9x9 voxels size patches. The final sCT value old ranging from 99.4% and 100%. Regarding
for a given voxel V was a weighted mean of all the patient setup accuracy, the volume delinea-
corresponding encompassed V-based patches tion on sCT-derived digitally reconstructed radio-
voxels. A similarity evaluation involving the graphs (DRR) and then transposition to
incoming patient MRI and the corresponding CT-derived DRR resulted in an absence of visual
multi-atlas MRI patches was used to assess the discrepancies. An improvement of this approach
ponderation factors. To evaluate the role of the was proposed in 2013 by Johansson et al. [48].
multi-atlas set size, either 6 or all 12 patients The spatial location details of the voxels were
were used to generate the sCT. The body RMSD added to the model via the consideration of the
resulted in mean performance of 219 HU and 207 distance voxels/external segmentation and voxels
HU for the 6- and 12-patient-based sCT approach, 3D coordinates. The database was composed of
respectively (p < 0.01). The chi index, a gamma nine patients, all presenting brain CT and UTE
index-inspired metric [42], with a 2%/2 mm cri- MRI. The model without and with the latter mod-
terion, led to rates equal to 98.3% ± 1.5% and ifications resulted in mean absolute prediction
98.5% ± 1.3% for the 6- and 12-patient-based deviations of 140HU ± 22HU and 130HU ±
sCT generation, respectively, illustrating in this 18HU, respectively, proving the superiority of
case the limited added value of the multi-atlas set the updated approach. Based on a non-AI regres-
enlargement. sion method, Pandey et al. [49] designed in 2019
a hybrid model relying on a linear function to
predict HU involving MRI intensities and coordi-
Voxel-Based Approaches nates, as well as a patient-specific random contri-
bution. The database was composed of 12 head
The voxel-based methods, first developed in tumour patients, with CT and T1-weighted
2011, rely on MRI voxel information to predict MRI. The head MAE was equal to 71.1HU ±
sCT. They present the disadvantage of limited 5.5HU.
The fuzzy c-means (FCM) algorithms have regression. However, they are inaccurate when
also been widely used in fields such as segmenta- dealing with non-bijective functions such as
tion of medical images [50], astronomy [51], and translation between MRI and CT [57]. Based on
economy [52]. In the context of sCT generation, five patients with brain CT and 3D T1-weighted
they are used to obtain the probabilities for vox- MRI, Yang et al. [58] reported a peak signal-to-
els to belong to given tissue classes before assign- noise ratio (PSNR) of 27.27 dB ± 1.00 dB based
ing corresponding HU intensities. In 2020, Hsu on an approach where features, namely, the local
et al. [53] collected five cases presenting binary pattern and the Gabor wavelet, were first
abdominal 3D mDixon sequences that were first computed in all MRI patches, before disregard-
pre-processed and then used as FCM algorithm ing the less suited for sCT generation. In Wang
input for classification into low-intensity repre- et al. [59], an auto-context approach was adopted,
senting high-density tissue, mid-intensity involv- implying that, except for the first sCT training set
ing air/bone marrow/lung/solid bone, and case computation, the model included sCT-
high-intensity for fat. Assigned intensities varied derived features. This model evaluation relied on
between -1000HU for the air and 354HU for the a cohort of 14 brain lesion patients with CT and
spine. These HU were multiplicated by the prob- T1-weighted MRI. The treatment plans, dynamic
ability matrices and then summed to obtain the conformal arc (DCA) or VMAT, were re-
final sCT. Minimum Dice similarity coefficients calculated with AAA and resulted in averages
(DSC) of 0.24, 0.84, and 0.94 were achieved in inferior or equal to 0.171Gy and 0.019Gy for
air, fat, and lung, respectively. Based on a 4-tissue PTV and OAR DVH differences, respectively.
classification with FCM, namely, low-density, With similar performance, Lei et al. [57] pro-
mid-density, bone, and marrow, Bredfeldt et al. posed to develop a model for MRI tissue classifi-
[54] enrolled 16 hepatocellular carcinoma cation into bone, air, and water with a first random
patients with CT and 3D Dixon MRI. Median forest algorithm, before using the probability
MAE were all below 60HU. After VMAT plans map-derived features to build a second random
calculation with the anisotropic analytical algo- forest model, as additional input to MRI-derived
rithm (AAA), median PTV DVH differences features. Based on 14 brain CT and T1-weighted
were between ±0.17Gy. To enhance the accuracy MRI, body MAE and PSNR were respectively
of a FCM approach aiming at classifying six tho- equal to 57.5HU ± 8.5HU and 28.33 dB ±
racic tissues, Su et al. [55] proved that the acqui- 1.68 dB. Average DVH differences were not
sition of UTE-mDixon and the association of superior to 0.123Gy and 0.020Gy for PTV and
reconstructed water-only Dixon MRI, fat-only OAR, respectively. A comparison of multi-atlas,
Dixon MRI, and the water-only divided by the random forests without auto-context and with
sum of water-only and fat-only Dixon images auto-context sCT approaches was performed by
should be used as algorithm input. Body MAE Nie et al. [60] for 22 pelvis cases. It respectively
inferior to 50HU were reported. A similar order resulted in MAE of 66.1HU ± 6.9HU, 51.2HU ±
of magnitude was obtained by Liu et al. [56] with 3.8HU, and 48.1HU ± 4.6HU, proving the supe-
a MAE of 49.1HU ± 17.8HU in pelvic soft tis- riority of the voxel-based method over
sues and 274.4HU ± 26.9HU for bones. In this multi-atlas.
study, to prevent the classification into bone of DL concept was first introduced in the mid-
non-bone mask belonging voxels, the five-class- 1980s and is now widespread in many medical
based FCM algorithm was regularised. The and non-medical fields. DL relies on a series of
dosimetry analysis, involving VMAT plans, non-linear functions organised into layers,
resulted in maximum mean DVH differences of enabling the access to highly deep features [61].
0.5% for PTV. DL networks are trained using an optimisation
An additional approach is random forests, i.e. process that requires a loss function to calculate
the combination of decision nodes organised into the model error. The choice of the loss function(s)
trees that can be both used for classification and has an impact on sCT accuracy, as well as the
training data and the tuning of network hyperpa- 2021 [32, 43]. They classified the popular DL
rameters [62]. The DL model can be trained with architectures into two categories: generator-only
2D, 2.5D, or 3D images. 2D images are typically and generative adversarial network (GAN) with
axial slices. A 2.5D model corresponds to three its variants. Several architectures were proposed
different 2D trainings in axial, sagittal, and coro- in each category. Here, we summarise the most
nal planes. For a given MRI, the final sCT is used DL architectures for sCT generation from
hence generated by averaging (or computing the MRI, with their advantages and drawbacks.
median value) of the three sCTs obtained from Among the generator-only models, the most
each training. A 3D DL model allows to reduce famous architecture is the U-Net, displayed in
inter-slice artefact, but it is more computationally Fig. 7.2 and which was initially designed for
costly than 2D or 2.5D. microscopy images delineation [63]. It relies on
Two reviews about the sCT generation with the combination of an encoder and a decoder,
DL methods in radiotherapy were published in respectively, aiming at computing abstract features
Fig. 7.2 Original 2D U-Net architecture. The encoder, detailed images computation, the equivalent encoder fea-
corresponding to the left branch of the U-shape, is com- ture maps and the feature maps of the current layer are
posed of the same block repeated four times. This block is concatenated. The series of operations CONV 3, ReLU,
first composed of a convolution with a 3 × 3 kernel size CONV 3, and ReLU is then applied. At the very end, a
(CONV 3) combined with a rectified linear unit (ReLU) convolution with a 1 × 1 kernel size is performed, before
activation function which has the advantage to avoid gradi- generating the output image, corresponding in this context
ents vanishing. After repeating this step twice, a max pool- to a sCT. Reproduced from Ronneberger O, Fischer P,
ing with a 2 × 2 kernel size is applied resulting in features Brox T, U-Net: Convolutional Networks for Biomedical
map size reduction by a factor 2. Two CONV3, both fol- Image Segmentation, Medical Image Computing and
lowed by ReLU, are then applied, before the data reaches Computer-Assisted Intervention – MICCAI 2015,
the decoder, i.e. the right symmetrical branch of the archi- MICCAI 2015, Lecture Notes in Computer Science, vol
tecture. It involves four main blocks, which are first com- 9351, 234–241, 2015, https://doi.org/10.1007/978-3-319-
posed of a transposed convolution. To further enhance 24574-4_28, with permission from Springer Nature
maps and reconstructing the output images from loss (with U-Net or GAN architecture) offered
these features maps. However, this network archi- higher sCT accuracy.
tecture suffers from certain limitations. First, the The conditional GAN (cGAN) has also been
convolutions composing the U-Net are not a global widely used for sCT generation since its first
operation, hampering the consideration of distant introduction in 2014 by Mirza et al. [72]. The
spatial context [64]. Chen et al. [65], therefore, cGAN particularity is to offer the operator the
proposed the TransU-Net, a U-shape-based neural opportunity to influence the generator output by
network whose encoder is a combination between adding extra data to the generator and discrimina-
a convolutional neural network and a transformer, tor. Despite their higher likelihood to experience
i.e. an attention only- based architecture [66]. mode-collapse in case of small datasets [73],
Based on the MICCAI 2015 Multi-Atlas Abdomen cGAN proved great success in various medical
Labelling Challenge cohort, the segmentation of fields. In 2019, Kazemifar et al. [74] proposed a
abdominal CT by Trans-U-Net and the original brain sCT study with a 2D cGAN, composed of a
U-Net resulted in mean DSC equal to 77.48% and U-Net as generator, and a combination of six con-
74.68% respectively, proving the potential interest volutional layers as well as five dense layers as
of this new architecture, which however has not discriminator. Regarding the cohort, 77 patients
yet been tested on the task of sCT generation. It presenting a couple of a brain CT and a 2D
was also proved to lead to brain sCT of 65.6 HU, contrast- enhanced T1-weighted MRI sequence
surpassing the U-Net performance of 80.8 HU were enrolled. The MAE, air and bone DSC were
[67]. Second, the optimal number of total layers respectively equal to 47.2HU ± 11.0HU, 0.70 ±
without considering the input, also known as 0.07, and 0.80 ± 0.06. After the re-computation
depth, is not straightforward to evaluate since it of the VMAT treatment plan on the sCT, mean
depends on the network goal [68]. To overcome differences between Dmean, Dmax, D95%, and D5%
this issue, Zhou et al. [69] recently introduced the metrics were all below 1% for both PTV and
U-Net ++, a deeply supervised segmentation net- OARs.
work combining U-Nets of different depths. For Still in the GAN category, cycleGANs, intro-
the BraTS 2013 challenge focused on brain tumour duced by Zhu et al. [75], have been massively
delineation based on MRI, intersection-over-union adopted to compute sCT since they present the
score of 91.21% ± 0.68% was reached for U-Net opportunity to use unpaired data. This architec-
++, while a rate of 89.21% ± 1.55% was reported ture consists of an image transformation task
for the original U-Net. through the combination of two GANs. The gen-
The GAN architecture for sCT generation erators have the role to predict Image 1 (CT)
from MRI was first proposed by Goodfellow from Image 2 (MRI) and Image 2 (MRI) from
et al. in 2014 [70]. It consists of the combination Image 1 (CT) in case of GAN 1 and 2, respec-
of a generator, having the role to compute, from tively. Regarding the discriminators, original
random noise, highly realistic synthetic images and synthetic Images 1 as well as original and
to fool the discriminator, and a discriminator synthetic Images 2 are classified by GAN 1 and
aiming at accurately classifying ground truth 2, respectively. One-to-one mappings are
images and synthetic images as real and false, assumed by cycleGAN, which may not always
respectively. Largent et al. compared a machine be an accurate representation of the clinical real-
learning method (patch-based) to a U-Net and a ity, in the case, for instance, of the generation of
GAN architecture, for prostate sCT generation one sCT from multiple MRI [76]. Peng et al.
from T2 3T MRI [71]. The best image quality [77] studied sCT generation with cGAN and
and dose results were obtained for the GAN DL cycleGAN for 173 nasopharyngeal carcinoma
method. In this study, they also compared perfor- patient data. The generators of the cGAN and the
mance from different loss functions: L2, single- cycleGAN were respectively a modified U-Net
scale perceptual loss, multiscale perceptual loss, and a combination of U-Net with residual neural
and weighted multiscale perceptual loss. The L2 networks. All discriminators corresponded to a
series of six convolutional layers. Matched CT computation with Monte Carlo, the 3D 3%/3 mm
and MRI, after deformable registration, and gamma index with 10% dose threshold were
unmatched data were used as cGAN and cycle- respectively equal to 99.3% ± 0.71%, 99.9% ±
GAN input, respectively. A better image accu- 0.10%, and 99.9% ± 0.21% for pelvis, thorax,
racy was obtained for cGAN than for cycleGAN, and abdomen. This dosimetry performance
for equivalent dose results in terms of gamma proved the high potential of the perceptual loss-
analysis. derived sCT.
The recent studies on sCT for MR-Linacs Similarly, Lapaeva et al. [80] investigated
have mostly focused on thorax, abdomen, and abdominal sCT accuracy from a cycleGAN, and
pelvis body regions [78–86]. Indeed, in 2022, collected 69, 54, 30, 20, 11, and 2 patients,
Lenkowicz et al. [78] collected 60 patients treated respectively treated for liver, nodes, adrenal
with the MRIdian (ViewRay, California, USA) glands, pancreas, kidney, and spleen. It led to a
for lung tumours. All of them underwent a CT total of 144 patients presenting 186 pairs of CT
without contrast agent and a true fast imaging and MRIdian-available true-FISP MRI sequence.
with steady-state precession (FISP) MRI The selected network architecture generator and
sequence. To generate the DL model of the sCT discriminator were, respectively, a HighResNet
(sCTDL), axial slices were used as inputs for the and a PatchGAN. This architecture offered the
2D cGAN, composed of a U-Net-based network opportunity to perform the training based on non-
as generator and a 70 × 70 PatchGAN as discrim- matching images. The external contour-based
inator [87]. Testing sets 1 and 2 were respectively MAE and SSIM were respectively equal to
composed of in-house and outer data. The body 70.10HU ± 18.97HU and 0.981 ± 0.009. The
MAE were equal to 56.8HU ± 12.4HU and 53HU Monte Carlo-based dose re-computation on sCT
± 8.5HU for testing sets 1 and 2, respectively. For resulted in 3D 2%/2 mm local gamma indices
the dosimetry analysis, the GTV ED was forced with, respectively, 50% and 90% dose thresholds
to 1, leading to a second sCT (sCTDL+Bulk). When equal to 99.4% ± 0.6% and 99.5% ± 0.6%. The
evaluating the sCTDL, the global 3D 3%/3 mm differences in DVH for the target volume and
with 10% dose threshold gamma index led to OAR did not exceed ±2%, proving the high sCT
99.1 ± 0.9% and 97.3 ± 5.7% for testing sets 1 accuracy.
and 2, respectively. Similarly, the sCTDL+Bulk anal-
ysis resulted in pass rates 99.4% ± 0.7% and
97.6% ± 5.5% for testing sets 1 and 2, Metrics for sCT Evaluation
respectively.
Kang et al. [79] outperformed the dosimetry Several metrics have been introduced in the lit-
performance reported by Lenkowicz et al. [78], erature to evaluate sCT generation. They are
potentially due to the implementation of a per- gathered in Table 7.2. These metrics can be clas-
ceptual loss to ensure highly detailed final sCT sified into four broad categories: (1) intensity-
[71]. The total cohort was composed of abdomi- based metrics, which aim at quantifying
nal, breast, and prostate tumour sites, each differences between sCT and CT intensities at the
including 30 patients. The TrueFISP MR voxel scale or, in a more complex way, their tex-
sequence was acquired for each patient on the tural differences; (2) metrics to evaluate the geo-
MRIdian. The generator and discriminator of metric fidelity; and metrics of greater clinical
the 2.5D cycleGAN used in this study respec- interest, either (3) based on the comparison of
tively corresponded to a U-Net and a ResNet. dose maps or DVH produced by CT and sCT or
The body MAE achieved for the prostate, breast, (4) on the ability of the sCT to provide accurate
and abdomen patients were equal to 55.3HU ± images to deal with patients setup. The following
5.5HU, 63.7HU ± 3.8HU, and 58.8HU ± 4.4HU, paragraphs introduce these metrics by illustrating
respectively. The total mean structural similar- some typical values reported in recent studies,
ity metric (SSIM) was equal to 0.90. After dose mostly published in 2021, or after.
Intensity-Based Metrics Wang 2014 et al. [102], these terms respectively

rely on signals mean, signals standard deviation,
First, the mean absolute error (MAE) is by far the and standardised signals analyses. The target
most reported metric, mainly due to its computa- value is 1, and the range of possible values ranges
tion ease. It corresponds to the average of abso- between −1 and 1. The literature reported mean/
lute voxel-wise differences between a reference median performance varying between 0.70 and
CT and sCT. Despite its frequent occurrence, 0.98 for brain [88, 96–98, 103], pelvis [85, 94,
MAE should be manipulated with caution, since 101, 104], and head-and-neck [105] sCT,
it is heavily impacted by CT/sCT or CT/MRI reg- respectively.
istration errors. This metric can be calculated in
whole body contour and in volumes of interest
(depending on the anatomical localisation). The etrics to Characterise Geometric
M
MAEs in the literature range from 47 HU [74] to Fidelity
83 HU [88] for the brain and from 24 HU [89] to
59 HU [90] for the pelvis. The Dice similarity coefficient (DSC), computed
The mean error (ME), representing the average as twice the cardinality of two volumes intersec-
of voxel-wise HU differences, has been moder- tion divided by the sum of the two volumes cardi-
ately reported in the literature. It indicates a sys- nalities, quantifies volumes overlap. As specified
tematic under/over HU predictions trend of neural by Reinke et al. in 2021 [106], this index presents
networks when used as loss function. It has the multiple limitations, including underestimated
major disadvantage to potentially mask intensities small volume scores or a dependence on voxel
differences due to the absence of the absolute size. It also depends on a large set of parameters,
value operator, and additionally shares some limi- including the involved images (sCT against CT/
tations with the MAE. The reported mean/median MRI), the investigated sub-volumes (mostly air
ME values in the different studies are between and bone), and the method used to obtain these
−11 HU and 28 HU [30, 36, 78, 91–94], regard- sub-volumes (threshold or segmentation). In the
less of the anatomical localisation and the method literature, air mean indices between 0.65 and
of sCT generation. DL and multi-atlas-based sCT 0.92 were reported for brain, pelvis, and abdo-
were also reported to present ME closer to zero men sCT [43, 80, 94, 97, 107]. Mean bone per-
compared to bulk density assignment. formance were ranging from 0.80 to 0.93 for
A third metric evaluating intensities restora- brain and pelvis-femur studies [43, 74, 90, 94,
tion accuracy is the peak signal-to-noise ratio 97, 107–109].
(PSNR), expressed in dB, which is proportional
to the logarithm of the ratio between the maxi-
mum intensity of the reference image (here, the etrics Assessing the Resulting
M
CT) squared and the mean squared error. The Dosimetric Impact
higher the value of the metric, the better the abil-
ity of the algorithm to accurately restore the Regarding dosimetry quality, gamma indices
HU. Although it presents the limitation to not [110], quantifying two dose maps similarity in
accurately represent the human vision [95], it has terms of Euclidean distance and dose discrep-
been used for brain [88, 96–100], pelvis [85, 94, ancy, have been frequently reported, partly owing
101], and abdomen [80, 86] studies with mean/ to their wide availability in clinics. This metric
median ranges comprised between 21.422 dB has the advantage of adding a distance tolerance
and 39.02 dB. which may be meaningful due to residual misreg-
First introduced in 2004, structural similarity istrations. However, this metric suffers from
metric (SSIM) aims at evaluating the structural some limitations, such as highly disregarding or
similarity between two images, i.e. the lumi- penalising errors in small dose regions for the
nance, contrast, and similarity. As reported by global and local approaches, respectively [111].
Each study uses its own gamma criteria: dose dif- and CT/sCT (or CT/sCT-derived DRR) and eval-
ference value (in %), distance to agreement value uating the translational and rotational differences.
(in mm), dose threshold, local or global analysis, Maximum means of −0.4 mm and (±)0.1° (when
2D or 3D. For pelvis and brain, mean 1%/1 mm reported) were obtained for pelvis, brain, and
and 2%/2 mm with 10% dose threshold were, head-and-neck sCT when registering CT/sCT-
respectively, reported to range from 97.3% to derived DRR and kV imaging [112, 116, 117].
99.0% [85, 112] and from 97.95% to 99.9% [85, For CT/sCT and CBCT fusion, the maximum
96, 112, 113]. For the 3D 2%/2 mm criterion with means were equal to (±)0.7 mm and 0.5° (when
10% dose threshold, the dosimetric performance reported) [89, 112, 117–119].
of bulk density assignment, multi-atlas, and
cGAN were compared in 2022 by O’Connor
et al. [93] in the context of definitive pelvic radio- ew Metrics in the Context of MRI-
N
therapy and resulted in a same mean pass rate of Only Workflow
99.7%.
A second metric assessed for sCT dosimetry The metrics presented above require a reference
quality evaluation is the relative or absolute CT to assess the reliability of the sCT scan, and
Dose Volume Histogram-derived criteria differ- are thus found in the literature as reference met-
ences between CT and sCT, such as the dose rics. When implementing a MRI-only RT work-
received by X% of a given volume denoted DX%. flow, the CT scan will no longer be acquired, so
However, no knowledge on the spatial location of these metrics will not be available for every sCT
the errors is accessible with this metric. Target scan generated. However, in these configurations,
and OAR maximum means/medians dose errors the implementation of a sCT quality assurance
were equal to −0.8% and 1.6%, respectively, for process will still be necessary. Several strategies
the brain and pelvis [74, 85, 90, 114]. Values as have been proposed for this purpose, including
high as ±30% and ±75% in target and OAR, the comparison to reference values obtained from
respectively, were reported for some bulk assign- templates to check the plausibility of the pre-
ment strategies [115]. dicted HUs [92, 120]. The use of a secondary
As already mentioned, in the aim of dose cal- independent sCT generation algorithm is another
culation from MRI, a dose difference in a voxel is path for a more secure implementation of such a
not necessarily linked to a HU error in this voxel workflow [13].
but can be attributed to HU errors throughout the
beam pathway. In this context, a metric reflecting
the dosimetric utility of sCT has recently been MRI-QA for Distortion/Artefact-Free
proposed and consists of assessing the water Synthetic CT
equivalent depth from the skin surface to the
treatment plane isocentre at 36 equidistant gantry Generalities
angles. This metric demonstrated a better correla-
tion with the dose difference at the isocentre Geometric fidelity is a key feature in MRI-guided
compared to conventional metrics such as MAE RT as it directly affects the restoration of external
or ME in prostate treatments [104]. contours and the accuracy of OAR and target vol-
ume delineation. Inaccurate reconstruction of
external contours can have a significant impact
Accuracy of Patient Repositioning on dose evaluation due to errors in X-ray path
length and source-to-skin distance estimates.
The sCT ability to set up the patient as precisely Let’s note that coil bridges should be used in CT-
as in the CT-based workflow should be verified free treatment planning to avoid distorting the
[13]. It has been assessed in the literature by reg- body contours due to the coils weight, which has
istering in-room imaging such as CBCT (or kV) been reported to lead to displacements higher
than 1 cm [121]. Inaccuracies in the definition of field inhomogeneity or offset (e.g. chemical
target volumes due to image distortion also affect shift), and Δf represents the frequency variation
dose coverage, especially when considering or offset. This field strength independent quantity
small volumes and stereotactic irradiation condi- is often provided in a spherical volume of interest
tions [122, 123], with errors up to 32% reported described by its diameter (DSV for diameter
when comparing the minimum dose received by spherical volume). In 2023, various vendor
99% of the target volume in multiple brain metas- reports guaranteed values in the order of 1.0–
tases radiosurgery [122]. Obviously, the dosimet- 2.0 ppm VRMS (volume root mean square) for
ric impact is reduced when large target volumes 1.5 and 3.0T magnetic fields for a 45 cm DSV for
are considered, with deviations of less than 1% RT dedicated MR devices. A detailed review of
observed in prostate and head and neck cancers the properties of current machines in terms of B0
[124, 125]. field homogeneity in the context of radiotherapy,
Several causes affect the geometric fidelity of including MR-Linac, has been performed by
MR images, including B0 inhomogeneities, resid- Gach et al. [127]. The shimming process aims to
ual gradient non-linearities (GNL), eddy cur- homogenise the B0 field as much as possible but
rents, and associated trajectory errors in k-space cannot deal with patient-related B0 inhomogene-
[126, 127]. The following paragraphs briefly ities caused by the variations in tissue magnetic
review their definitions and the methods pro- susceptibility χ (in particular, differences in sus-
posed to quantify and mitigate them. Several ceptibility between air and biological tissue)
working groups have already defined consensus which has been shown to be the main cause of
QA guidelines for MRI, in the context of RT residual B0 inhomogeneities today, even with the
treatment planning incorporating MRI or for application of an active shimming [127].
MRI-guided online radiotherapy systems [128, Spin echo-based pulse sequences are less
129]. The key points of these reports are described affected than gradient-recalled echo sequences
here. The main reconstruction artefacts, e.g. mag- by B0 field inhomogeneities due to their intrinsic
netic susceptibility artefacts, which can largely principle, which incorporates an additional 180°
affect as well the quality of the segmentation pro- refocusing pulse. Increasing the readout band-
duced by the radiation oncologist and the perfor- width or receiver bandwidth, i.e. the frequency
mance of the sCT generation tools, are also range used when receiving RF pulses, can reduce
briefly discussed here. the impact of these field inhomogeneities on the
reconstructed image, however resulting in a
higher amount of noise encompassed and thus a
B0 Inhomogeneities lower signal-to-noise ratio. The phase difference
mapping technique is recommended to map, at a
The principle of MRI is based on the application voxel scale, these field inhomogeneities [130].
of a static polarising magnetic field that is
assumed to be spatially invariant. Any deviation
from this principle affects, among other things, esidual Gradient Non-linearity-
R
the uniformity of the image, its spatial integrity, Induced Distortions
and blurring. The homogeneity of the magnetic
field is conventionally expressed in parts per mil- Spatial encoding in MRI is based on the succes-
lion (ppm) of the static magnetic field value. sive application of magnetic field gradients.
Conventional reconstruction methods based on
B f
=
ppm = direct fast Fourier transform assume that these
B0 fL gradients are perfectly linear throughout the
entire imaging volume. Due to various reasons,
where B0 is the magnetic field strength, fL is the including engineering limitations, physiological
Larmor frequency, ΔB represents the magnetic constraints, or intentional sacrifices to maximise
system performance, gradient fields are not per- tions resulting from GNL have been shown to be
fectly linear and always contain higher-order greater than 1 mm at radii >10 cm from the iso-
non-linear components [131], directly impacting centre of the magnet and to reach maximum val-
geometric fidelity. A priori knowledge of GNL ues of 2–3 mm [129].
fields can be managed retrospectively by apply- The use of the reverse gradient technique,
ing post-processing in the image domain or pro- which takes advantage of the application of oppo-
spectively during image reconstruction by MRI site readout gradients for GNL isolation, is the
vendors to reduce the impact of these distortions recommended method for the characterisation of
in the final images. However, despite the appli- GNL-induced 3D distortion maps [129].
cation of these algorithms, residual distortions Evaluations should be performed by applying the
remain. In 2019, Gach et al. illustrated effects of 3D corrections provided by the vendor and using
GNL for two ViewRay MRI-60Co and MRI- a large phantom (>80% of the usable FOV) with
Linac systems (Fig. 7.3) for 2D slice excitations fully characterised landmarks.
[132]. This figure illustrates the fact that distor-
tions increase when the distance from the iso-
centre of the magnet increases and the major Susceptibility Artefacts
impact of this effect on the restoration of exter-
nal contours and organs located at the periphery Magnetic susceptibility, denoted χ, quantifies the
of the FOV. This is even more crucial in radio- magnetisation properties of a material when sub-
therapy where large diameter scanners (≥70 cm) jected to a magnetic field. The quantity χ corre-
are used in order to manage the clutter of immo- sponds to a proportionality constant between the
bilisation devices and to correctly capture the macroscopic magnetisation M of the material and
full anatomy of the patient. Geometric distor- the local magnetic field B0 [133]. The value of χ
a b
Fig. 7.3 Illustration of geometric distortions generated H, Curcuru AN, Wittland EJ, Maraghechi B, Cai B, Mutic
by residual gradient non-linearities, highlighting the S, Green OL, MRI quality control for low-field MR-IGRT
greater impact at large distances from the isocentre. These systems: Lessons learned, Journal of Applied Clinical
distortions are illustrated here for the MRI-60Co (a) and Medical Physics, 20(10), 53-66, 2019, https://doi.
the MR-Linac (b) [132]. Reproduced from Michael Gach org/10.1002/acm2.12713, CC.BY 4.0
can be positive if the magnetisation is aligned ecent Guidelines from Scientific

R
with the field (paramagnetism) or negative, in the Societies on the Frequency of Quality
case where the two vectors are opposite (diamag- Controls
netism). Ferromagnetism characterises materials
for which residual magnetism is retained when Table 7.3 summarises the consensus of the
the external magnetic field is removed. Iron and AAPM Working Group 284 established in 2021
nickel, for example, are ferromagnetic materials. regarding monthly quality checks, emphasising
As already mentioned, tissue susceptibility the importance of monthly geometric integrity of
can affect both the amplitude and frequency of the system verifications. A daily check is even
echo signals, with the total B field becoming a recommended during the commissioning stage.
combination of the applied static field B0 and the For this test, the use of a large diameter phantom
field offset ΔB created by the magnetisation M of (>30 cm), which can be a homemade, vendor-
the object. Frequency shifts in the echo signal can supplied, or commercial phantom, is preferred.
thus be observed as the applied field strength B0 The acceptability criterion is not strict, with a tol-
increases, even though variations in χ are gener- erance equal to 2 mm across a 25 cm FOV. The
ally small (of the order of 0.1 ppm). Susceptibility- same group recommends an independent assess-
weighted sequences, which consist of the ment of GNL during commissioning and after
acquisition of one or more gradient echo signals, any major hardware repair, as well as after hard-
from which amplitude and frequency are ware and software upgrades.
extracted, are used to derive susceptibility maps
[134, 135]. The ESTRO-ACROP working group focused
Magnetic susceptibility artefacts are thus on the commissioning of online MRI-guided
caused by local gradients created at the interfaces radiotherapy systems and made proposals based
of tissues with very different magnetic suscepti- on consensus expert opinion [128]. Their recom-
bilities (air/tissue, air/bone, metallic implant, cal- mendations, published in 2023, are in favour of
cium deposits). They are responsible for two either a weekly or monthly check of B0 field inho-
main effects: image distortions and hyposignal mogeneities using vendor-provided 3D phantoms
areas. The presence of ferromagnetic materials in and analysis software. The use of third-party
or near the FOV also causes distortions of the B0 solutions and in-house developed methods is
magnetic field lines, leading to areas of intense mentioned here as an alternative, used for inde-
hyposignal in MR images. The use of a low mag- pendent tests. A particularity of the MR-Linac is
netic field or the preferential choice of spin echo the presence of a linear accelerator which can
sequences can reduce the impact of these arte- interfere with the magnetic field of the MRI. The
facts. The selection of specific phase and fre- working group therefore recommends the evalua-
quency encoding directions, the use of high RF tion of the level of MR distortions as a function
pulse bandwidths, the use of slice encoding for of the gantry position on an annual basis.
metal artefact correction (SEMAC), or multi-
acquisition with variable-resonance image com-
binations are other ways to achieve this goal verall Distortion Values Reported
O
[136]. Recent methods to correct susceptibility on MR-Linacs
artefacts have been proposed in the image domain
to directly inpaint the artefact regions [136]. Several recent studies have quantified spatial
Tools to simulate the distortion maps generated distortions on MR-Linacs [138–142]. For the
by the patient herself/himself by taking the sus- ViewRay system, Marasini et al. [138] reported
ceptibility values in the different tissues as inputs geometric distortions of 3.2 mm ± 0.02 mm,
are also available and could be interesting solu- 2.9 mm ± 0.02 mm, and 1.8 mm ± 0.01 mm in
tions to generate distortion-free MR images right-left, anterior-posterior, and superior-
[137]. inferior directions using the MRID3D phantom
Table 7.3 Consensus of the AAPM Working Group 284 establishing the monthly quality controls to be implemented
[129]. Reproduced from Glide-Hurst CK, Paulson ES, McGee K, Tyagi N, Hu Y, Balter J, Bayouth J, Task group 284
report: magnetic resonance imaging simulation in radiotherapy: considerations for clinical implementation, optimiza-
tion, and quality assurance, Medical Physics, 48(7), 2021, https://doi.org/10.1002/mp.14695, with permission from John
Wiley and Sons
Test Suggested equipment Protocol/parameter/tolerance
System
Room temperature Visual inspection of digital readout Functional
and humidity
Cold head operation Audible pump sound Functional
Cryogen level Digital readout from MRI console Manufacturer specified (no change from
indicator baseline)
Mechanical
Table movement Ruler ±1.0 mm from set distances
smoothness and
accuracy
Laser alignment with Ruler and phantoms with internal ±2.0 mm from expected distance offsets
imaging isocenter landmarks
Laser movement Ruler ±2.0 mm from set distances
smoothness and
accuracy
Patient marking
Laser marking Patient marking software ±2.0 mm
accuracy
Image qualitya
Central frequency Digital readout from MRI console Manufacturer specified
after imaging reference phantom (i.e.,
ACR large MRI phantom or
homogeneous phantom)
Transmitter gain ± 5% from baseline
Flexible RF coil Homogeneous phantom Individual elements, exceeds minimum
testing vendor-provided threshold
Geometric accuracy Manufacturer supplied, in-house, or Verify ≤2 mm across 25 cm FOV
commercially available phantom
>30 cm in diameter/width
High contrast spatial ACR large MRI phantom with ≤1.0 mm
resolution procedures/tolerances defined in Ref.
[83]
Low contrast detectability Total number of discernible spokes (for four
slices) for fields <3 T should range from 21
(0.3 T) to 36 (1.5 T), and 40 for 3 T.
Artifact evaluation No obvious image artifacts
Percent image uniformity Head coil:
≥ 87.5% for <3 T
≥ 82.0% for 3 T
Percent signal ghosting Ghosting ratio ≤ 2.5%
a
May skip measurement if weekly ACR MRI QA results performed by Diagnostic Radiology Department staff are
made available for the QMP to review. Baseline values shall he established following the ACR 2015 QC Manual (collect
data for 10 days and determine baseline value). Action limits will be magnet and system-specific as determined by the
QMP and manufacturer recommendations
with the vendor-provided distortion correction 0.25 mm ± 0.03 mm, 0.30 mm ± 0.01 mm, and
turned off. The onboard Siemens distortion cor- 0.27 mm ± 0.01 mm for the 90° gantry angle. In
rection function highly improved the results, 2019, based on a home-made phantom, a mean
with the mean values falling to distortion over the whole phantom of
0.60 ± 0.28 mm was reported by Dorsch et al. In this section, a distinction between MR-only
[141], with 99.8% of the control points present- workflows with conventional Linac or MR-Linac
ing distortions below 1.5 mm. In 2022, is made because there are some key differences in
Damyanovich et al. evaluated the temporal sta- the sCT generation process as well as in its role in
bility of the Elekta Unity system based on 76 the workflow (see Fig. 7.1). For example, an
weekly measurements [142]. GNL and 3D dis- essential aspect to evaluate when implementing a
torsions were considered in this analysis, which sCT product in a conventional Linac workflow is
involved a large FOV 3D-phantom containing its high patient setup reproducibility. By contrast,
1932 markers embedded in seven parallel plates, in an MR-Linac workflow, this is not an issue as
spaced 25 mm × 25 mm in- and 55 mm through- patient positioning is verified by MRI. Rather,
plane. Time-averaged absolute distortions were one of the biggest challenges the sCT faces when
reported to be in the range of 0.30–0.49 mm, implementing an MR-only MR-Linac workflow
0.53–0.80 mm, 1.0–1.4 mm, and 2.28–2.37 mm will be setting up a fast and robust online QA that
for DSV respectively, equal to 200, 300, 400, will keep up with quick adaptive radiotherapy
and 500 mm and considering the 98th percentile. that these hybrid machines offer.
The statistical process control they implemented
proved that the distortions were stable over time.
Significant changes in GNL were however MR-Only Workflows
observed following servicing at the 1-year mark. with Conventional Linac
In an MR-only RT workflow with conventional

s CT Implementation in MR-only Linac, the sCT represents the substitute of a
Radiotherapy Workflows planning CT. Therefore, the straightforward
commissioning process should compare the per-
Introduction formance of the sCT against the gold standard
CT. To assess the ability to correctly recover
Since synthetic CTs are directly used for dose intensities, metrics such as MAE (Table 7.2) are
calculation, it is expected that they should per- amongst the most popular, yet the relevance of
form as good as a CT to be clinically imple- measurements in the whole body or just in a par-
mented. This section will give an overview on the ticular region of interest are still under discus-
academic and industrial sCT solutions, both AI sions. Susceptibility artefacts, from bone tissue
and non-AI-based, that have been commissioned and air cavities to dental prosthesis or postopera-
and implemented into clinical radiotherapy tive metal clips, are another point of interest in
workflows. the commissioning evaluation. The magnitude of
The lack of a consensus QA programme for the impact on dose calculation from these arte-
these images seems to be the main issue that has facts may vary depending on the approach used
hindered their wider use in the clinic. Indeed, to generate the sCT. Dosimetric evaluation, on
sCT solutions are not yet generalisable to all situ- the other hand, seems to focus on the analysis of
ations, limiting their use to one specific patient DVH parameters and gamma pass rates
group or diagnosis. Hospitals that have success- (Table 7.2). However, the exact criteria to use are
fully implemented a sCT in the clinic had to tai- also debatable as parameters such as dose calcu-
lor the commissioning process according to their lation algorithms and dose delivery techniques
needs. Yet, two aspects are common to all com- may play a role.
missioning processes reported in the literature: Patient positioning verification is central
geometric fidelity and dosimetric performance. for the commissioning of sCT in the case of
Each clinic might have decided to focus more on MR-only RT with conventional Linacs.
some aspects than others, choosing different met- Correct generation of DRRs from sCT is nec-
rics to evaluate. essary for 2D kV verification and therefore
should be included in the QA procedure. Gold assessment of the individual MR device perfor-
fiducial identification for prostate cancer mance and its stability over time. A maximum
patients may also become central to commis- distortion of 3 mm in the outer edges of volumes
sioning procedures. The available sCT solu- of pelvic examinations was reported [145–147].
tions have tackled this issue in different ways In particular, Kemppanien et al. [146] found that
forcing quality controls to be accordingly such a distortion influenced up to 0.1% the dose
tailored. to the PTV. On the other hand, each group
Finally, the importance of risk analysis should reported different methodologies for dose distri-
be highlighted because it allows the clinical team bution comparison between sCT and CT, not
to understand and address the major obstacles in facilitating inter-clinic comparison. Among the
the implementation process. By quantifying the factors that influenced dose distribution analysis
risk involved in each step, appropriate safeguards were the image modality selected as base for the
can be put in place and monitored throughout the dose planning (CT or sCT), the type of registra-
implementation process. tion used to fuse structures and dose plans
between image modalities, the treatment tech-
linical sCT Generation with Non-DL
C nique, and the calculation algorithm of the treat-
Solutions ment planning system. Despite these differences
The review paper by Edmund and Nyholm [143] in methodology, all groups reported mean dose
on the status of sCT solutions for RT and PET/ differences to the target below 1%. Finally,
MR until November 2015 mentioned two groups patient position verification was an essential
that performed an early MR-only radiotherapy part of the clinical feasibility for an MR-only
workflow implementation for prostate and brain workflow. Gold fiducials are not visible in the
cancers. The Helsinki group used a voxel-based MRCAT prostate sCT; hence they need to be
dual regression approach to treat prostate cases, correctly identified on the MRI and thereafter
while the Michigan group chose a voxel-based delineated on the sCT. Some groups have intro-
probabilistic approach for their brain cases. duced automatic localisation of fiducials, but
Another review [14] that included data until manual observation remains the most reliable
March 2017 reported centres in New York and method [147].
Turku as the first to use the CE-marked com- By 2018, PHILIPS released a second product,
mercial sCT solution developed by PHILIPS MRCAT pelvis, still based on a 3D mDIXON
(Koniklijke Philips N.V., Eindhoven, The fast field echo sequence but in which the assign-
Netherlands) for the treatment of prostate cases. ment of HU is continuous (PHILIPS brochure).
The PHILIPS product, MR for Calculating The resulting sCT can be used for other pelvic
Attenuation (MRCAT) prostate, is generated cancers such as rectal/anal as well as gynaeco-
from a 3D mDIXON fast field echo sequence logical. Yu et al. [148] conducted a dosimetric
and then uses bulk density assignment for five performance evaluation of the MRCAT pelvis
tissues: air, adipose, water, trabecular/spongy generated by three centres with different field
bone, and compact/cortical bone. In a review strengths and treatment planning systems. They
published in 2019 [144], two more centres included 77 prostate, 43 rectum, and 27 gynaeco-
(Utrecht, Netherlands and Odense, Denmark) logical cases. As expected, they reported that
were identified to have successfully imple- MRCAT pelvis better reproduces the CT when
mented the MRCAT prostate into clinics. They compared to the bulk density MRCAT prostate
evaluated different aspects of an MR-only RT based on HU profile comparisons. The MAE
workflow, and three common factors were of between CT and MRCAT pelvis sCT was equal
interest: (1) geometric fidelity of MR images, to 120 ± 15 HU for bone and 33 ± 4 HU for soft
(2) dosimetric performance of the sCT com- tissue. They reported visualisation of gold fidu-
pared to CT, and (3) patient positioning verifica- cials and rectal spacer on the MRCAT pelvis sCT,
tion through gold fiducials. The first included an thus being superior to MRCAT prostate. However,
they also observed a slight overestimation (but flows for prostate cancer can be safely imple-
still clinically acceptable) of PTV doses in some mented in clinics. Nonetheless, they highlighted
rectal cases, which they attributed to the fact that the lack of guidelines for QA hinders the wide-
this sCT solution classifies air within the body as spread implementation of MR-only RT
soft tissue. workflows.
An additional commercially available sCT Noteworthy is the patient specific QA proce-
generator (CE-marked) is the MRIPlanner dure proposed by Palmer et al. (2018) [152],
software by Spectronic Medical AB (Helsingborg, based on CBCT imaging. To enable this proce-
Sweden). This solution requires a T2-weighted dure, two criteria need to be evaluated, namely,
MR image, and thereafter a statistical decompo- the CBCT HU stability over time and the com-
sition algorithm is used to assign HU [149]. The parison of the dose distributions of the CBCT
MR-OPERA study [150] was designed to vali- with the reference CT. Periodic phantom mea-
date this product for prostate cancer patients surements over a period of 8 weeks presented a
across four Swedish institutions using MR and deviation of less than 40 HU for materials rela-
CT devices from different vendors. A total of 170 tive ED ranging from 0.2 to 1.7. Ten prostate
patients were recruited with the only exclusion patients from the MR-PROTECT study were
criterion being patients with hip prostheses. No analysed in this work. Mean dose deviations were
specific MR scan protocol was required other less than 0.9% for either CT-sCT or CBCT-sCT
than a T2-weighted with a large FOV. All centres for target DVHs. Finally, the CBCT QA tool was
were given access to the MRIPlanner installed on tested by introducing known errors in the
a cloud-based service where the sCT were gener- sCT. Only a water-equivalent sCT failed to be
ated from the uploaded MR images. Dose distri- detected by the QA. In conclusion, the proposed
butions comparison between the CT and sCT patient-specific QA proved to be clinically feasi-
showed a difference below 0.5% for all OARs ble enabling a CT free QA. However, this QA
and targets. The mean gamma pass rate for the cannot be performed before the first CBCT
2%/2 mm criterion was found to be 99.12%. acquisition, preventing this solution to be appli-
These results demonstrated that the sCTs pro- cable in all dose regimes such as
duced with MRIPlanner were dosimetrically hypofractionation.
accurate and robust across a variety of vendors, In the review by Bird et al. [144], the Helsinki
field strengths, and treatment techniques. and Michigan groups were mentioned for their
The MR-OPERA study was followed by the early clinical results after implementing their in-
MR-PROTECT study which explored the feasi- house sCT solutions. The Helsinki group [153],
bility of implementing an MR-only radiotherapy who treated prostate cancer with MR-only work-
protocol for prostate [151]. A total of 40 patients flow since 2012, was the first to report similar
were included between March 2017 and May clinical outcomes between patients treated either
2018. All patients had gold markers implanted as with MR-only or CT-based workflows by moni-
part of the setup verification protocol. Patients toring the early prostate-specific antigen and the
were planned in an MR-only workflow; however, radiation therapy oncology group grades. They
a CT was acquired directly after the MRI session pointed out that the main limitations of using sCT
but only used for QA purposes and retrospective in a clinical environment were the identification
analysis. The retrospective DVH analysis and the of the gold markers, large artefacts due to hip
gamma pass rates were in line with the results implants and obese patients. The Michigan
obtained in the MR-OPERA study, after the out- group, on the other hand, developed a voxel-
liers were corrected for rectal air. They also noted based weighted summation method for their sCT
that experience and training of the MR staff miti- generation of the pelvic region. Kim et al. [154]
gated risks associated with MRI acquisition evaluated the MR-only workflow implementation
which inevitably impact sCT generation. Their through a failure mode and effects analysis that
overall results demonstrated that MR-only work- helps to develop a robust quality management
programme. Particularly for sCT, bone segmenta- lesions and focusing on gliomas and metastases
tion and air pocket identification led to improve- [159]. Of special interest was data from 14
ments in QA and quality control checklists, as patients who presented missing bone skull due to
well as limited risks for gross anatomic errors surgery performed before RT. They found a MAE
leading to incorrect HU assignment and impact- of 177 ± 19 HU for bone in a volume of interest
ing dose calculation. (VOI) around the resected bone which was of the
The HIPSTER (high precision prostate substi- same order as the MAE for the entire cohort of
tute CT-based external beam radiotherapy) trial is patients. The average DSC for the whole skull
a sCT prospective clinical study consisting of 30 was 0.92 compared to 0.90 in the
prostate patients recruited between 2017 and VOI. Unfortunately, no specific data on the
2019 from two Australian institutions. The impact on dose distribution was reported. The
authors used an in-house developed sCT using a overall positive results from the two validation
hybrid atlas-voxel method [155]. The acceptance studies encouraged the Lund group to conduct
criteria for the dose calculation were isocentre the first prospective clinical study for the imple-
dose within 2% and gamma pass rate larger than mentation of an MR-only workflow for brain
90% for the 2%/2 mm criterion. The study dem- radiotherapy [119]. A total of 21 patients with
onstrated that the specially tailored QA pro- grades III and IV gliomas were treated between
gramme was robust for MR-only RT workflow March 2021 and March 2022. A QA programme
implementation as 90% of the cases showed a was specifically developed to ensure the safe use
mean isocentre dose difference below 1.5% and of the sCT. The prospective part of the QA con-
mean gamma pass rate of 99.7% [156]. sisted of four control steps. The first was per-
formed by a script that checked that the MRI
linical sCT Generation with DL
C acquisition parameters were those required for
Solutions sCT generation. Thereafter, the sCT was visually
DL approaches to generate sCT are an expanding inspected for artefacts and misalignments against
area of research. Their popularity stems from MRI. Then, a HU comparison to the CT was per-
their ability to bypass some of the intrinsic limi- formed. The areas where bone resection was
tations of non-AI sCT methods. For example, observed in the sCT were also controlled based
the fast sCT generation, potentially from a single on the CT. Lastly, calculated plans were accepted
MR sequence, reduces scan time or the superior- if they fulfilled either one of these two criteria:
ity it has in reproducing uncommon anatomies (1) dose distribution differences between sCT
compared to atlas-based algorithms. and CT should be below 1% for the target and
Palmer et al. [157] were the first to validate the OAR, and (2) OAR absolute dose should be 10%
commercial AI-based sCT product released by below clinical tolerance. The retrospective part of
Spectronic Medical AB for head and neck radio- the QA consisted of verifying the rigid registra-
therapy. A convolutional neural network (CNN)- tion equivalence between CBCT to sCT and
based transfer function estimation algorithm CBCT to CT. Overall, the success of this study
[158] was used to automatically generate sCT relied on the establishment of acceptance criteria
from a T1-weighted Dixon Vibe MR sequence. tailored to local clinical routines and this solution
They reported a MAE of 67 ± 14 HU for the over- limitations evaluation.
all body and 195 ± 27 HU for bone particularly. An additional commercially available sCT is
As for dose distributions comparison, the differ- provided by Siemens Healthineers (Erlangen,
ence in absorbed dose mean ranged from −0.3% Germany) as part of syngo.via RT Image Suite.
to 0.02% for the tested DVH parameters. An This sCT is generated by a multi-layer neural net-
additional validation study performed by the work (CNN for segmentation and a cGAN for
Lund group reported similar promising results image reconstruction) with a T1-weighted Dixon
for the sCT generated with MRIPlanner Vibe MR sequence as input [160]. In 2022,
(Spectronic Medical AB) but applied to brain Masitho et al. [114] performed the validation of
this product for brain MR-only workflow. Apart modes and the lack of a general consensus in QA
from evaluation of geometry fidelity and dose criteria [162]. An increase in prospective studies
calculation performance, they also focused on the with larger cohorts is expected to happen in the
ability to produce DRR from sCTs that can be next few years as more hospitals access these
used for 2D/2D non-coplanar kV X-ray patient solutions.
setup verification. Results showed a MAE of
136 ± 13 HU for the brain. After adding extra
fixation to the mask to prevent variation in the MR-Only Workflows with MR-Linac
rotational lateral direction when acquiring the
MR sequence, they concluded that patient setup Unlike the MR-only workflow with conventional
with sCT DRR as reference was equivalent to CT Linac, the MR-only workflow with an MR-Linac
DRR-based setup. The clinical implementation hybrid system opens the possibility for fast treat-
evaluation for MR-only RT brain workflow is ment adaptation, both intrafraction and interfrac-
planned for the near future. tion, as anatomy changes can be observed in real
PHILIPS (Koniklijke Philips N.V.) has also time. In this context, the role of the sCT is to pro-
launched two AI-based products, MRCAT brain vide fast and accurate 3D ED maps for interfrac-
and MRCAT head and neck. Both use as input a tion dose calculation based on the MRI of the
3D T1-weighted mDIXON sequence and gener- day. On the other hand, the sCT impacts neither
ate a sCT with continuous HU directly on the MR the target and OAR delineation nor the patient
console. As part of a clinical implementation of position verification, as this is entirely handled
the MRCAT brain, Emin et al. [161] conducted a by the MRI. Thus, the major challenge of the
two-phase process consisting of a retrospective implementation process of sCT is to develop a
study (CT-based RT workflow with sCT on the fast (on-the-fly) QA for evaluation of image
background) and a prospective study (MR-only fidelity.
RT workflow with CT in the background). They In the 2020 review paper by Kurtz et al.
reported a mean dose difference below 1% for [163], six MR-Linac approaches were men-
PTV D2%, PTV D98%, and PTV Dmean, based tioned, and only two were commercialised at
on 60 glioblastoma patients (30 patients per that time, namely, the MRIdian by ViewRay
phase). As the MRCAT brain has a fixed voxel Technologies, Inc. (California, USA) and the
size (1.1 mm × 1.1 mm × 1 mm), the importance Unity by Elekta AB (Stockholm, Sweden). The
of re-gridding the images to matching voxel sizes workflows of these machines are different
before performing dose comparisons was empha- because of the differences in their designs. Yet,
sised due to smaller structures (e.g. cochleae and none of them operate in a true MR-only work-
chiasma) being particularly sensitive to the rigid flow as there is always a CT image acquired
registration process between CT and sCT. Indeed, before the start of the treatment. The dose plan
a clear reduction in the spread of the DVH differ- optimised on this CT image is referred to as the
ence distributions was reported when comparing base plan. With the MRIdian system, one may
pre- and post re-gridded data for the OAR choose between modifying the table position
studied. and delivering the base plan or performing a full
Academic AI-based sCT solutions for image re-optimisation [163]. On the other hand, the
reconstruction of other more difficult sites such Unity system does not allow for table correc-
as breast, liver and abdomen can be found in the tion. So the base plan must be adapted in one of
literature [32]. MRI-only clinical implementation the two ways: (1) adapt-to-position (ATP) where
using such solutions is still in its infancy, even a rigid translational registration and limited
though the geometric and dosimetric evaluations optimiser is used to reproduce DVH from the
seem to be favourable. The main drawbacks for base plan or (2) adapt-to-shape (ATS) where
these sCT solutions can be reduced to two rea- recontouring and full re-optimisation are possi-
sons: the lack of knowledge about the failure ble [164].
Plans re-optimisation is currently performed anatomical and clinical scenarios. Thus, sCT is
on a bulk density sCT generated by different still a future development for a use in MR-Linac
methods depending on the system. The MRIdian workflows. The advent of AI has increased this
system method consists of using DIR to fuse the area of research, and of interest are the DL sCT
MRI and the planning CT thereby creating a solutions generated from MRI acquired in low
deformed version of the CT. Discrepancies magnetic fields such as the MRIdian system.
between image modalities such as gas or body Apart from the advantage of offering a lower
shape are compensated by dedicated structures electron return effect, the images produced by
where ED is overridden. The second method, this system are particularly convenient for lung
used in the ATS workflow of the Unity system, cases [168]. Lenkowicz et al. [78], previously
starts by extracting the contours along with their introduced, were the first to produce an in-house
average ED from the planning CT. These values sCT based on cGAN architecture trained with 32
are then propagated to the daily MR image by patients presenting lung lesions. They concluded
performing a DIR. ATS workflow allows for that their approach was suitable for palliative
manual recontouring before the average ED is cases; however improvements such as using a
assigned to each contour [165]. Tang et al. [165] larger training cohort and more complex neural
added a step to the workflow where the reference networks needed to be performed to obtain a bet-
plan (on CT) and the re-calculated plan (on sCT) ter dosimetry performance which is required for
were compared on-the-fly. If major discrepancies curative purposes.
were found, then the sCT was modified, and the In 2022, Therapanacea (Paris, France) intro-
plan re-calculated in an iterative fashion until the duced the MR-Box option in their ART-Plan™
dose difference would be within the acceptable toolbox. The MR-Box is an AI-powered,
clinical tolerance. For 54 out of 61 patients with CE-marked, and FDA-cleared software provid-
various treatment sites, the gamma pass rate was ing sCT images and automated OAR delineation
99.7% with the criteria 3%/3 mm and a DVH dif- from MRI. GAN architecture is used to generate
ference of less than 1.5% when comparing high-fidelity sCT images that match the underly-
against the first generated sCT. Major differences ing distribution of the training CT scans while
were found in two lung cases, where the ED gra- maintaining a voxel level anatomical correspon-
dient within the target and the adjoining tissue dence with the input MRI. Particularly, a 3D
could not be reproduced by bulk density method, cGAN architecture, similar to the Pix2Pix archi-
and thus, these patients were treated with ATP tecture, is used for brain imaging, while for pel-
workflow instead. Their study demonstrated the vis, the underlying DL model is a cycleGAN. The
importance of extra dosimetric checks on the sCT solution can be used in MRI-only workflows or
to enable a quick identification and correction of workflows based on an MR-Linac. In 2022, a
the sources of errors that otherwise would remain multi-centric evaluation was conducted consider-
undetected in pre-treatment check or plan adapta- ing pelvic anatomies and 20 test patients from 8
tion phase. A prospective risk study was con- different institutions using a 0.35 T MR [169].
ducted by the Memorial Sloan Kettering Cancer Reconstruction performance was assessed using
Center department for the ATP workflow, high- the OARs delineated for the treatment planning.
lighting the need to train planners/therapists to The results showed that the model was able to
review ED in all three planes of the entire image reliably restore the intensities of the femoral
[166]. heads, while those of the rectum and the sigmoid
In 2023, more sophisticated sCT generation were more difficult to reconstruct. In 2022, a
methods are not yet clinically implemented in clinical trial was launched at Gustave Roussy
any commercially available MR-Linac. As Cancer Campus (Villejuif, France) to evaluate the
pointed out by Ng et al. [167], the real challenge performance of the sCT in patients being treated
is the development of a generalisable and reliable for glial tumours. Results are expected by the end
sCT solution that would be applicable to various of 2023.
DL-generated sCT promise to shorten the re- ratised today, at least for workflows based on
planning phase in an MR-Linac workflow while MRI-only. Like in numerous fields, the arrival of
increasing accuracy compared to atlas-based or DL, and its associated robustness, has largely
bulk density methods. To be compatible with contributed to this change in practices. However,
online adaptive workflows, the time between workflows based on MR-Linac still have reduced
generation and quality control of these images benefits from these advances, since no DL solu-
cannot exceed a few minutes [170]. Indeed, some tion was yet commercially integrated. A major
studies have reported that generation of sCT can shortcoming widely identified by several authors
be achieved in less than 10 s if GPU-based net- is the lack of recommendations concerning the
works are used [171]. However, time for MRI commissioning stage and QA procedures. At
acquisition must also be accounted for, and cur- least the absence of high magnitude distortions
rently the dedicated MR sequences that are used and artefacts should be checked in this context.
for sCT generation are time-consuming and, The literature review also highlighted the need
thus, not yet applicable for online adaptive for new profiles of physicists with mixed knowl-
MRgRT [172]. edge in RT physics, MRI physics, and AI for the
Once these hurdles are overcome, the last implementation of safe workflows.
remaining challenge for the adoption of AI-based
sCT in MR-linac workflows is the development
of a robust image QA that can be carried out References
online within the span of a few minutes. A good
starting point for relevant QA criteria and toler- 1. Owrangi AM, Greer PB, Glide-Hurst CK. MRI-only
treatment planning: benefits and challenges. Phys
ance levels can be drawn from the QA pro-
Med Biol. 2018;63(5):05TR01.
grammes already developed for conventional 2. van Herk M, Kooy HM. Automatic three-
Linac MR-only workflows. Adjustments should dimensional correlation of CT-CT, CT-MRI, and
be done considering the hospitals needs and CT-SPECT using chamfer matching. Med Phys.
1994;21(7):1163–78.
assess the limitations of the MR-Linac devices.
3. Ulin K, Urie MM, Cherlow JM. Results of a multi-
As already mentioned, several strategies can also institutional benchmark test for cranial CT/MR
be implemented for this purpose, such as com- image registration. Int J Radiat Oncol Biol Phys.
parison with reference values obtained from 2010;77(5):1584–9.
4. Dean CJ, Sykes JR, Cooper RA, Hatfield P, Carey
models to check the plausibility of the predicted
B, Swift S, u. a. An evaluation of four CT-MRI co-
HUs or the use of a second independent algo- registration techniques for radiotherapy treatment
rithm similar to the second monitor unit calcula- planning of prone rectal cancer patients. Br J Radiol
tion algorithms. As long as a consensus in QA is 2012;85(1009):61–8.
5. Korsager AS, Carl J, Østergaard LR. Comparison of
not achieved, a planning CT will still be needed
manual and automatic MR-CT registration for radio-
for patient-specific QA purposes for the foresee- therapy of prostate cancer. J Appl Clin Med Phys.
able future. A novel idea, however, has been put 2016;17(3):294–303.
forward by Goodburn et al. [173] where a poten- 6. Brock KK. Deformable registration accuracy con-
sortium. Results of a multi-institution deformable
tial solution to save time would consist of remov- registration accuracy study (MIDRAS). Int J Radiat
ing the sCT visualisation step and directly use the Oncol Biol Phys. 2010;76(2):583–96.
k-space data as treatment planning software 7. Pötter R, Tanderup K, Kirisits C, de Leeuw A,
input. Kirchheiner K, Nout R, u. a. The EMBRACE II
study: the outcome and prospect of two decades
of evolution within the GEC-ESTRO GYN work-
ing group and the EMBRACE studies. Clin Transl
Conclusion Radiat Oncol. 2018;9:48–60.
8. Rivard MJ, Coursey BM, DeWerd LA, Hanson
WF, Huq MS, Ibbott GS, u. a. Update of AAPM
In conclusion, the implementation of sCT in clin- Task Group No. 43 report: a revised AAPM proto-
ics, which was rather experimental and anecdotal col for brachytherapy dose calculations. Med Phys.
a few years ago, is in the process of being democ- 2004;31(3):633–74.
9. Burgos N, Cardoso MJ, Modat M, Pedemonte based electron density treatment planning. J Appl
S, Dickson J, Barnes A, u. a. Attenuation correc- Clin Med Phys. 2011;12(4):3522.
tion synthesis for hybrid PET-MR scanners. Med 22. Lambert J, Greer PB, Menk F, Patterson J, Parker
Image Comput Comput Assist Interv. 2013;16(Pt J, Dahl K, u. a. MRI-guided prostate radiation
1):147–54. therapy planning: investigation of dosimetric accu-
10. Martinez-Möller A, Souvatzoglou M, Delso G, racy of MRI-based dose planning. Radiother Oncol.
Bundschuh RA, Chefd’hotel C, Ziegler SI, u. 2011;98(3):330–4.
a. Tissue classification as a potential approach 23. Lee YK, Bollet M, Charles-Edwards G, Flower
for attenuation correction in whole-body P ET/ MA, Leach MO, McNair H, u. a. Radiotherapy
MRI: evaluation with PET/CT data. J Nucl Med. treatment planning of prostate cancer using mag-
2009;50(4):520–6. netic resonance imaging alone. Radiother Oncol.
11. Persson E, Svanberg N, Scherman J, Jamtheim 2003;66(2):203–16.
Gustafsson C, Fridhammar A, Hjalte F, u. a. MRI- 24. ICRU Report 46. Photon, electron, proton and neu-
only radiotherapy from an economic perspective: can tron interaction data for body tissues. 1992.
new techniques in prostate cancer treatment be cost 25. Chin AL, Lin A, Anamalayil S, Teo BKK. Feasibility
saving? Clin Transl Radiat Oncol. 2023;38:183–7. and limitations of bulk density assignment in MRI
12. Keyriläinen J, Sjöblom O, Turnbull-Smith S, for head and neck IMRT treatment planning. J Appl
Hovirinta T, Minn H. Clinical experience and cost Clin Med Phys. 2014;15(5):4851.
evaluation of magnetic resonance imaging -only 26. Zaidi H, Montandon ML, Slosman DO. Magnetic
workflow in radiation therapy planning of prostate resonance imaging-guided attenuation and scatter
cancer. Phys Imaging Radiat Oncol. 2021;19:66–71. corrections in three-dimensional brain positron emis-
13. Vandewinckele L, Claessens M, Dinkla A, Brouwer sion tomography. Med Phys. 2003;30(5):937–48.
C, Crijns W, Verellen D, u. a. Overview of artificial 27. Wagenknecht G, Kops ER, Tellmann L, Herzog
intelligence-based applications in radiotherapy: H. Knowledge-based segmentation of attenuation-
recommendations for implementation and quality relevant regions of the head in T1-weighted MR
assurance. Radiother Oncol. 2020;153:55–66. images for attenuation correction in MR/PET
14. Johnstone E, Wyatt JJ, Henry AM, Short SC, Sebag- systems. In: 2009 IEEE nuclear science sympo-
Montefiore D, Murray L, u. a. Systematic review of sium conference record (NSS/MIC). Orlando,
synthetic computed tomography generation meth- FL: IEEE; 2009. [zitiert 27. März 2023]. S. 3338–
odologies for use in magnetic resonance imaging- 43. Verfügbar unter: http://ieeexplore.ieee.org/
only radiation therapy. Int J Radiat Oncol Biol Phys. document/5401751/.
2018;100(1):199–217. 28. Harrison K, Pullen H, Welsh C, Oktay O, Alvarez-
15. Prabhakar R, Julka PK, Ganesh T, Munshi A, Joshi Valle J, Jena R. Machine learning for auto-
RC, Rath GK. Feasibility of using MRI alone for 3D segmentation in radiotherapy planning. Clin Oncol.
radiation treatment planning in brain tumors. Jpn J 2022;34(2):74–88.
Clin Oncol. 2007;37(6):405–11. 29. Robert C, Munoz A, Moreau D, Mazurier J, Sidorski
16. Chen L, Price RA, Wang L, Li J, Qin L, McNeeley G, Gasnier A, u. a. Clinical implementation of deep-
S, u. a. MRI-based treatment planning for radiother- learning based auto-contouring tools-experience of
apy: dosimetric verification for prostate IMRT. Int J three French radiotherapy centers. Cancer Radiother.
Radiat Oncol Biol Phys. 2004;60(2):636–47. 2021;25(6–7):607–16.
17. Weber DC, Wang H, Albrecht S, Ozsahin M, 30. Groot Koerkamp ML, de Hond YJM, Maspero M,
Tkachuk E, Rouzaud M, u. a. Open low-field mag- Kontaxis C, Mandija S, Vasmel JE, u. a. Synthetic CT
netic resonance imaging for target definition, dose for single-fraction neoadjuvant partial breast irradia-
calculations and set-up verification during three- tion on an MRI-linac. Phys Med Biol. 2021;66(8).
dimensional CRT for glioblastoma multiforme. Clin 31. Ding S, Liu H, Li Y, Wang B, Li R, Liu B, u. a.
Oncol. 2008;20(2):157–67. Assessment of dose accuracy for online MR-guided
18. Ramsey CR, Oliver AL. Magnetic resonance imag- radiotherapy for cervical carcinoma. J Radiat Res
ing based digitally reconstructed radiographs, Appl Sci. 2021;14(1):159–70.
virtual simulation, and three-dimensional treat- 32. Boulanger M, Nunes JC, Chourak H, Largent A,
ment planning for brain neoplasms. Med Phys. Tahri S, Acosta O, u. a. Deep learning methods to
1998;25(10):1928–34. generate synthetic CT from MRI in radiotherapy: a
19. Wang C, Chao M, Lee L, Xing L. MRI-based literature review. Phys Med. 2021;89:265–81.
treatment planning with electron density informa- 33. Wyatt J, McCallum H. Applying a commercial atlas-
tion mapped from CT images: a preliminary study. based synthetic computed tomography algorithm to
Technol Cancer Res Treat. 2008;7(5):341–8. patients with hip prostheses for prostate magnetic
20. Eilertsen K, Vestad LNTA, Geier O, Skretting A. A resonance-only radiotherapy. Radiother Oncol.
simulation of MRI based dose calculations on the 2019;133:100–5.
basis of radiotherapy planning CT images. Acta 34. Dowling JA, Lambert J, Parker J, Salvado O, Fripp
Oncol. 2008;47(7):1294–302. J, Capp A, u. a. An atlas-based electron density
21. Karotki A, Mah K, Meijer G, Meltsner mapping method for magnetic resonance imaging
M. Comparison of bulk electron density and voxel- (MRI)-alone treatment planning and adaptive MRI-
based prostate radiation therapy. Int J Radiat Oncol (MR) data by incorporation of spatial information—
Biol Phys. 2012;83(1):e5-11. potential application for MR-only radiotherapy and
35. Demol B, Boydev C, Korhonen J, Reynaert attenuation correction in positron emission tomogra-
N. Dosimetric characterization of MRI-only phy. Acta Oncol. 2013;52(7):1369–73.
treatment planning for brain tumors in atlas- 49. Pandey A, Sa Y, Guo B, Zhang R. Feasibility of gen-
based pseudo- CT images generated from stan- erating synthetic CT from T1-weighted MRI using a
dard T1-weighted MR images. Med Phys. linear mixed-effects regression model. Biomed Phys
2016;43(12):6557. Eng Express. 2019;5(4):047004.
36. Han X. MR-based synthetic CT generation using 50. Ranjbarzadeh R, Saadi SB. Automated liver and
a deep convolutional neural network method. Med tumor segmentation based on concave and convex
Phys. 2017;44(4):1408–19. points using fuzzy c-means and mean shift cluster-
37. Greer PB, Dowling JA, Lambert JA, Fripp J, Parker ing. Measurement. 2020;150:107086.
J, Denham JW, u. a. A magnetic resonance imaging- 51. Bandyopadhyay S, Das S, Datta A. Detection
based workflow for planning radiation therapy for of coronal holes in solar disk image using fast
prostate cancer. Med J Aust. 2011;194(4):S24-27. fuzzy C-means algorithm. In: Das NR, Sarkar S,
38. Sjölund J, Forsberg D, Andersson M, Knutsson Herausgeber. Computers and devices for com-
H. Generating patient specific pseudo-CT of the munication [Internet]. Singapore: Springer
head from MR using atlas-based regression. Phys Singapore; 2021 [zitiert 27. März 2023]. S.
Med Biol. 2015;60(2):825–39. 29–35. (Lecture Notes in Networks and Systems;
39. Farjam R, Tyagi N, Deasy JO, Hunt MA. Dosimetric Bd. 147). Verfügbar unter: http://link.springer.
evaluation of an atlas-based synthetic CT genera- com/10.1007/978-981-15-8366-7_5.
tion approach for MR-only radiotherapy of pelvis 52. Jahangoshai Rezaee M, Jozmaleki M, Valipour
anatomy. J Appl Clin Med Phys. 2019;20(1):101–9. M. Integrating dynamic fuzzy C-means, data envel-
40. Farjam R, Tyagi N, Veeraraghavan H, Apte A, opment analysis and artificial neural network to
Zakian K, Hunt MA, u. a. Multiatlas approach with online prediction performance of companies in stock
local registration goodness weighting for MRI-based exchange. Phys A. 2018;489:78–93.
electron density mapping of head and neck anatomy. 53. Hsu SH, DuPre P, Peng Q, Tomé WA. A tech-
Med Phys. 2017;44(7):3706–17. nique to generate synthetic CT from MRI for
41. Uh J, Merchant TE, Li Y, Li X, Hua C. MRI- abdominal radiotherapy. J Appl Clin Med Phys.
based treatment planning with pseudo CT gen- 2020;21(2):136–43.
erated through atlas registration. Med Phys. 54. Bredfeldt JS, Liu L, Feng M, Cao Y, Balter
2014;41(5):051711. JM. Synthetic CT for MRI-based liver stereotactic
42. Bakai A, Alber M, Nüsslin F. A revision of body radiotherapy treatment planning. Phys Med
the gamma- evaluation concept for the com- Biol. 2017;62(8):2922–34.
parison of dose distributions. Phys Med Biol. 55. Su KH, Friel HT, Kuo JW, Al Helo R, Baydoun
2003;48(21):3543–53. A, Stehning C, u. a. UTE-mDixon-based tho-
43. Spadea MF, Pileggi G, Zaffino P, Salome P, Catana rax synthetic CT generation. Med Phys.
C, Izquierdo-Garcia D, u. a. Deep convolution neu- 2019;46(8):3520–31.
ral network (DCNN) multiplane approach to syn- 56. Liu C, Zhang X, Si W, Ni X. Multiview self-
thetic CT generation from MR images-application in supervised segmentation for OARs delineation in
brain proton therapy. Int J Radiat Oncol Biol Phys. radiotherapy. Evid Based Complement Alternat
2019;105(3):495–503. Med. 2021;2021:8894222.
44. Thompson M, Duda RO, Hart PE. Pattern classifica- 57. Lei Y, Harms J, Wang T, Tian S, Zhou J, Shu HK, u.
tion and scene analysis. Leonardo. 1974;7(4):370. a. MRI-based synthetic CT generation using seman-
45. Edmund JM, Kjer HM, Van Leemput K, Hansen tic random forest with iterative refinement. Phys
RH, Andersen JAL, Andreasen D. A voxel-based Med Biol. 2019;64(8):085001.
investigation for MRI-only radiotherapy of the 58. Yang X, Lei Y, Shu HK, Rossi P, Mao H, Shim H,
brain using ultra short echo times. Phys Med Biol. u. a. Pseudo CT estimation from MRI using patch-
2014;59(23):7501–19. based random forest. Proc SPIE Int Soc Opt Eng.
46. Johansson A, Karlsson M, Nyholm T. CT substitute 2017;10133:101332Q.
derived from MRI sequences with ultrashort echo 59. Wang T, Manohar N, Lei Y, Dhabaan A, Shu HK,
time. Med Phys. 2011;38(5):2708–14. Liu T, u. a. MRI-based treatment planning for brain
47. Jonsson JH, Akhtari MM, Karlsson MG, Johansson stereotactic radiosurgery: dosimetric validation of a
A, Asklund T, Nyholm T. Accuracy of inverse treat- learning-based pseudo-CT generation method. Med
ment planning on substitute CT images derived Dosim. 2019;44(3):199–204.
from MR data for brain lesions. Radiat Oncol. 60. Nie D, Cao X, Gao Y, Wang L, Shen D. Estimating
2015;10:13. CT image from MRI data using 3D fully convolu-
48. Johansson A, Garpebring A, Karlsson M, Asklund tional networks. Deep Learn Data Label Med Appl.
T, Nyholm T. Improved quality of computed tomog- 2016;2016:170–8.
raphy substitute derived from magnetic resonance
61. LeCun Y, Bengio Y, Hinton G. Deep learning. 75. Zhu JY, Park T, Isola P, Efros AA. Unpaired image-to-
Nature. 2015;521(7553):436–44. image translation using cycle-consistent adversarial
62. Tahri S, Barateau A, Cadin C, Chourak H, Ribault S, networks. In Proceedings of the IEEE international
Nozahic F, u. a. A high-performance method of deep conference on computer vision. 2017;2223–32.
learning for prostate MR-only radiotherapy planning 76. Brou Boni KND, Klein J, Gulyban A, Reynaert N,
using an optimized Pix2Pix architecture. Phys Med. Pasquier D. Improving generalization in MR-to-CT
2022;103:108–18. synthesis in radiotherapy by using an augmented
63. Ronneberger O, Fischer P, Brox T. U-Net: convolu- cycle generative adversarial network with unpaired
tional networks for biomedical image segmentation. data. Med Phys. 2021;48(6):3003–10.
2015 [zitiert 28. April 2023]; Verfügbar unter. 77. Peng Y, Chen S, Qin A, Chen M, Gao X, Liu Y, u.
https://arxiv.org/abs/1505.04597. a. Magnetic resonance-based synthetic computed
64. Liang J, Yang C, Zhong J, Ye X. BTSwin-Unet: tomography images generated using generative
3D U-shaped symmetrical swin transformer- adversarial networks for nasopharyngeal carcinoma
based network for brain tumor segmentation radiotherapy treatment planning. Radiother Oncol.
with self-supervised pre-training. Neural Process 2020;150:217–24.
Lett [Internet]. 2022 [zitiert 28. April 2023]; 78. Lenkowicz J, Votta C, Nardini M, Quaranta F,
Verfügbar unter. https://link.springer.com/10.1007/ Catucci F, Boldrini L, u. a. A deep learning approach
s11063-022-10919-1. to generate synthetic CT in low field MR-guided
65. Chen J, Lu Y, Yu Q, Luo X, Adeli E, Wang Y, u. radiotherapy for lung cases. Radiother Oncol.
a. TransUNet: transformers make strong encod- 2022;176:31–8.
ers for medical image segmentation. 2021 [zitiert 79. Kang SK, An HJ, Jin H, Jin K, Chie EK, Park JM, u.
28. April 2023]; Verfügbar unter: https://arxiv.org/ a. Synthetic CT generation from weakly paired MR
abs/2102.04306. images using cycle-consistent GAN for MR-guided
66. Vaswani A, Shazeer N, Parmar N, Uszkoreit J, radiotherapy. s. 2021;11(3):263–71.
Jones L, Gomez AN, u. a. Attention is all you need 80. Lapaeva M, La Greca Saint-Esteven A, Wallimann
[Internet]. arXiv; 2017 [zitiert 28. April 2023]. P, Günther M, Konukoglu E, Andratschke N, u. a.
Verfügbar unter: http://arxiv.org/abs/1706.03762. Synthetic computed tomography for low-field mag-
67. Weijie Chen, Seyed Iman Zare Estakhraji, Alan B netic resonance-guided radiotherapy in the abdo-
McMillan. Exploration of vision transformer models men. Phys Imaging Radiat Oncol. 2022;24:173–9.
in medical images synthesis. 81. Garcia Hernandez A, Fau P, Wojak J, Mailleux
68. Iqbal A, Sharif M, Khan MA, Nisar W, Alhaisoni H, Benkreira M, Rapacchi S, u. a. Synthetic com-
M. FF-UNet: a U-shaped deep convolutional neural puted tomography generation for abdominal
network for multimodal biomedical image segmen- adaptive radiotherapy using low-field magnetic
tation. Cognit Comput. 2022;14(4):1287–302. resonance imaging. Phys Imaging Radiat Oncol.
69. Zhou Z, Siddiquee MMR, Tajbakhsh N, Liang 2023;25:100425.
J. UNet++: redesigning skip connections to exploit 82. Olberg S, Chun J, Su Choi B, Park I, Kim H, Kim
multiscale features in image segmentation. IEEE T, u. a. Abdominal synthetic CT reconstruction with
Trans Med Imaging. 2020;39(6):1856–67. intensity projection prior for MRI-only adaptive
70. Goodfellow IJ, Pouget-Abadie J, Mirza M, Xu B, radiotherapy. Phys Med Biol. 2021;66(20).
Warde-Farley D, Ozair S, u. a. Generative adversar- 83. Nousiainen K, Santurio GV, Lundahl N, Cronholm
ial networks. 2014 [zitiert 1. Mai 2023]; Verfügbar R, Siversson C, Edmund JM. Evaluation of MRI-
unter: https://arxiv.org/abs/1406.2661. only based online adaptive radiotherapy of abdomi-
71. Largent A, Barateau A, Nunes JC, Mylona E, Castelli nal region on MR-linac. J Appl Clin Med Phys.
J, Lafond C, u. a. Comparison of deep learning-based 2022;24:e13838.
and patch-based methods for pseudo-CT generation 84. Farjam R, Nagar H, Kathy Zhou X, Ouellette D,
in MRI-based prostate dose planning. Int J Radiat Chiara Formenti S, DeWyngaert JK. Deep learning-
Oncol Biol Phys. 2019;105(5):1137–50. based synthetic CT generation for MR-only radio-
72. Mirza M, Osindero S. Conditional generative adver- therapy of prostate cancer patients with 0.35T
sarial nets. 2014 [zitiert 28. April 2023]; Verfügbar MRI linear accelerator. J Appl Clin Med Phys.
unter: https://arxiv.org/abs/1411.1784. 2021;22(8):93–104.
73. Shahbazi M, Danelljan M, Paudel DP, Van Gool 85. Hsu SH, Han Z, Leeman JE, Hu YH, Mak RH,
L. Collapse by conditioning: training class- Sudhyadhom A. Synthetic CT generation for MRI-
conditional GANs with limited data. 2022 [zitiert guided adaptive radiotherapy in prostate cancer.
28. April 2023]; Verfügbar unter: https://arxiv.org/ Front Oncol. 2022;12:969463.
abs/2201.06578. 86. Fu J, Singhrao K, Cao M, Yu V, Santhanam AP, Yang
74. Kazemifar S, McGuire S, Timmerman R, Wardak Y, u. a. Generation of abdominal synthetic CTs from
Z, Nguyen D, Park Y, u. a. MRI-only brain radio- 0.35T MR images using generative adversarial net-
therapy: assessing the dosimetric accuracy of syn- works for MR-only liver radiotherapy. Biomed Phys
thetic CT images generated using a deep learning Eng Express. 2020;6(1):015033.
approach. Radiother Oncol. 2019;136:56–63.
87. Isola P, Zhu JY, Zhou T, Efros AA. Image-to- 100. Jabbarpour A, Mahdavi SR, Vafaei Sadr A, Esmaili
image translation with conditional adversarial net- G, Shiri I, Zaidi H. Unsupervised pseudo CT gen-
works. ArXiv161107004 Cs [Internet]. 2018 [zitiert eration using heterogenous multicentric CT/MR
24. Juni 2021]; Verfügbar unter: http://arxiv.org/ images and CycleGAN: dosimetric assessment for
abs/1611.07004. 3D conformal radiotherapy. Comput Biol Med.
88. Sreeja S, Mubarak DMN. Pseudo computed tomog- 2022;143:105277.
raphy estimation from brain MRI using anatomic 101. Reaungamornrat S, Sari H, Catana C, Kamen
signature and joint dictionary learning. Proc Comput A. Multimodal image synthesis based on disentan-
Sci. 2023;218:1946–57. glement representations of anatomical and modality
89. Ahangari S, Hansen NL, Olin AB, Nøttrup TJ, specific features, learned using uncooperative rela-
Ryssel H, Berthelsen AK, u. a. Toward PET/MRI as tivistic GAN. Med Image Anal. 2022;80:102514.
one-stop shop for radiotherapy planning in cervical 102. Wang Z, Bovik AC, Sheikh HR, Simoncelli
cancer patients. Acta Oncol. s2021;60(8):1045–53. EP. Image quality assessment: from error visibility
90. O’Connor LM, Dowling JA, Choi JH, Martin J, to structural similarity. IEEE Trans Image Process.
Warren-Forward H, Richardson H, u. a. Validation of 2004;13(4):600–12.
an MRI-only planning workflow for definitive pelvic 103. Zimmermann L, Knäusl B, Stock M, Lütgendorf-
radiotherapy. Radiat Oncol. 2022;17(1):55. Caucig C, Georg D, Kuess P. An MRI sequence
91. Zhao S, Geng C, Guo C, Tian F, Tang X. SARU: independent convolutional neural network for syn-
a self-attention ResUNet to generate synthetic CT thetic head CT generation in proton therapy. Z Med
images for MR-only BNCT treatment planning. Med Phys. 2022;32(2):218–27.
Phys. 2023;50(1):117–27. 104. Hyuk Choi J, Asadi B, Simpson J, Dowling JA,
92. Chourak H, Barateau A, Tahri S, Cadin C, Lafond Chalup S, Welsh J, u. a. Investigation of a water equiv-
C, Nunes JC, u. a. Quality assurance for MRI-only alent depth method for dosimetric accuracy evalua-
radiation therapy: a voxel-wise population-based tion of synthetic CT. Phys Med. 2023;105:102507.
methodology for image and dose assessment of 105. Zhao Y, Wang H, Yu C, Court LE, Wang X, Wang
synthetic CT generation methods. Front Oncol. Q, u. a. Compensation cycle consistent generative
2022;12:968689. adversarial networks (Comp-GAN) for synthetic CT
93. O’Connor LM, Choi JH, Dowling JA, Warren- generation from MR scans with truncated anatomy.
Forward H, Martin J, Greer PB. Comparison Med Phys. 2023;50(7):4399.
of synthetic computed tomography generation 106. Reinke A, Tizabi MD, Sudre CH, Eisenmann M,
methods, incorporating male and female anatom- Rädsch T, Baumgartner M, Acion L, Antonelli
ical differences, for magnetic resonance imaging- M, Bakas S, Bankhead P, Benis A, Cardoso MJ,
only definitive pelvic radiotherapy. Front Oncol. Cheplygina V, Christodoulou E, Cimini B, Collins
2022;12:822687. GS, Farahani K, Ginneken BV, Glocker B, Maier-
94. Bahrami A, Karimian A, Arabi H. Comparison Hein L. Common limitations of image processing
of different deep learning architectures for syn- metrics: A Picture Story. ArXiv; 2021. https://doi.
thetic CT generation from MR images. Phys Med. org/10.48550/arXiv.2104.05642
2021;90:99–107. 107. Kaushik S, Bylund M, Cozzini C, Shanbhag D,
95. Mori S. A real-time, single-exposure, dual-energy Petit SF, Wyatt JJ, u. a. Region of interest focused
subtraction mask for markerless tumor track- MRI to Synthetic CT Translation using Regression
ing in radiotherapy: proof of concept. Phys Med. and Classification Multi-task Network. arXiv;
2019;63:63–9. 2022.
96. Wang J, Yan B, Wu X, Jiang X, Zuo Y, Yang 108. Lena B, Florkow MC, Ferrer CJ, van Stralen M,
Y. Development of an unsupervised cycle Seevinck PR, Vonken EJPA, u. a. Synthetic CT for
contrastive unpaired translation network for the planning of MR-HIFU treatment of bone metas-
MRI-to-CT synthesis. J Appl Clin Med Phys. tases in pelvic and femoral bones: a feasibility study.
2022;23(11):e13775. Eur Radiol. 2022;32(7):4537–46.
97. Wang C, Uh J, Merchant TE, Hua CH, Acharya 109. Eshraghi Boroojeni P, Chen Y, Commean PK, Eldeniz
S. Facilitating MR-guided adaptive proton therapy in C, Skolnick GB, Merrill C, u. a. Deep-learning syn-
children using deep learning-based synthetic CT. Int thesized pseudo-CT for MR high-resolution pediat-
J Part Ther. 2022;8(3):11–20. ric cranial bone imaging (MR-HiPCB). Magn Reson
98. Ranjan A, Lalwani D, Misra R. GAN for synthe- Med. 2022;88(5):2285–97.
sizing CT from T2-weighted MRI data towards 110. Low DA, Harms WB, Mutic S, Purdy JA. A tech-
MR-guided radiation treatment. MAGMA. nique for the quantitative evaluation of dose distribu-
2022;35(3):449–57. tions. Med Phys. 1998;25(5):656–61.
99. Wang CC, Wu PH, Lin G, Huang YL, Lin YC, Chang 111. Yu L, Baker A, Kairn T, Livingstone A, Trapp J,
YPE, u. a. Magnetic resonance-based synthetic com- Crowe S. A structure-based gamma evaluation
puted tomography using generative adversarial net- method for identifying clinically relevant dose
works for intracranial tumor radiotherapy treatment differences in organs at risk. Phys Eng Sci Med.
planning. J Pers Med. 2022;12(3):361. 2023;46(3):1033.
112. Liu X, Emami H, Nejad-Davarani SP, Morris E, 123. Pappas EP, Alshanqity M, Moutsatsos A, Lababidi
Schultz L, Dong M, u. a. Performance of deep learn- H, Alsafi K, Georgiou K, u. a. MRI-related geomet-
ing synthetic CTs for MR-only brain radiation ther- ric distortions in stereotactic radiotherapy treatment
apy. J Appl Clin Med Phys. 2021;22(1):308–17. planning: evaluation and dosimetric impact. Technol
113. Maspero M, Bentvelzen LG, Savenije MHF, Cancer Res Treat. 2017;16(6):1120–9.
Guerreiro F, Seravalli E, Janssens GO, u. a. 124. Adjeiwaah M, Bylund M, Lundman JA, Karlsson
Deep learning-based synthetic CT generation for CT, Jonsson JH, Nyholm T. Quantifying the effect
paediatric brain MR-only photon and proton radio- of 3T magnetic resonance imaging residual sys-
therapy. Radiother Oncol. 2020;153:197–204. tem distortions and patient-induced susceptibility
114. Masitho S, Szkitsak J, Grigo J, Fietkau R, Putz F, distortions on radiation therapy treatment planning
Bert C. Feasibility of artificial-intelligence-based for prostate cancer. Int J Radiat Oncol Biol Phys.
synthetic computed tomography in a magnetic 2018;100(2):317–24.
resonance-only radiotherapy workflow for brain 125. Adjeiwaah M, Bylund M, Lundman JA, Söderström
radiotherapy: two-way dose validation and 2D/2D K, Zackrisson B, Jonsson JH, u. a. Dosimetric
kV-image-based positioning. Phys Imaging Radiat impact of MRI distortions: a study on head and
Oncol. 2022;24:111–7. neck cancers. Int J Radiat Oncol Biol Phys.
115. Ding S, Liu H, Li Y, Wang B, Li R, Huang 2019;103(4):994–1003.
X. Dosimetric accuracy of MR-guided online 126. Curcuru AN, Lewis BC, Kim T, Yang D, Michael
adaptive planning for nasopharyngeal carcinoma GH. Effects of B0 eddy currents on imaging iso-
radiotherapy on 1.5 T MR-Linac. Front Oncol. center shifts in 0.35-T MRI-guided radiotherapy
2022;12:858076. (MR-IGRT) system. Med Phys. 2021;48(6):2929–38.
116. Kemppainen R, Suilamo S, Ranta I, Pesola M, 127. Gach HM, Curcuru AN, Mutic S, Kim T. B0 field
Halkola A, Eufemio A, u. a. Assessment of dosimet- homogeneity recommendations, specifications, and
ric and positioning accuracy of a magnetic resonance measurement units for MRI in radiation therapy.
imaging-only solution for external beam radiother- Med Phys 2020;47(9):4101–4114.
apy of pelvic anatomy. Phys Imaging Radiat Oncol. 128. Tanadini-Lang S, Budgell G, Bohoudi O, Corradini
2019;11:1–8. S, Cusumano D, Güngör G, u. a. An ESTRO-
117. Palmér E, Nordström F, Karlsson A, Petruson ACROP guideline on quality assurance and medical
K, Ljungberg M, Sohlin M. Head and neck can- physics commissioning of online MRI guided radio-
cer patient positioning using synthetic CT data in therapy systems based on a consensus expert opin-
MRI-only radiation therapy. J Appl Clin Med Phys. ion. Radiother Oncol. 2023;181:109504.
2022;23(4):e13525. 129. Glide-Hurst CK, Paulson ES, McGee K, Tyagi N,
118. Maspero M, Tyyger MD, Tijssen RHN, Seevinck Hu Y, Balter J, u. a. Task group 284 report: mag-
PR, Intven MPW, van den Berg CAT. Feasibility of netic resonance imaging simulation in radiother-
magnetic resonance imaging-only rectum radiotherapy: considerations for clinical implementation,
apy with a commercial synthetic computed tomog- optimization, and quality assurance. Med Phys.
raphy generation solution. Phys Imaging Radiat 2021;48(7):e636–70.
Oncol. 2018;7:58–64. 130. Jackson E, Bronskill M, Drost D, Och J, Pooley R,
119. Lerner M, Medin J, Jamtheim Gustafsson C, Alkner Sobol W, u. a. Acceptance testing and quality assur-
S, Olsson LE. Prospective clinical feasibility study ance procedures for magnetic resonance imaging
for MRI-only brain radiotherapy. Front Oncol. facilities [internet]. AAPM; 2010 Dez [zitiert 2. Mai
2022;11:812643. 2023]. Verfügbar unter: https://www.aapm.org/pubs/
120. Wang C, Uh J, Patni T, Merchant T, Li Y, Hua CH, reports/detail.asp?docid=101.
u. a. Toward MR-only proton therapy planning for 131. Tao S, Trzasko JD, Gunter JL, Weavers PT, Shu
pediatric brain tumors: synthesis of relative proton Y, Huston J, u. a. Gradient nonlinearity calibration
stopping power images with multiple sequence MRI and correction for a compact, asymmetric magnetic
and development of an online quality assurance tool. resonance imaging gradient system. Phys Med Biol.
Med Phys. 2022;49(3):1559–70. 2017;62(2):N18–31.
121. Buatti JS, Gallagher KJ, Bailey I, Griglock T, Heard 132. Michael Gach H, Curcuru AN, Wittland EJ,
M. An evaluation of quality assurance guidelines Maraghechi B, Cai B, Mutic S, u. a. MRI quality
comparing the American College of Radiology and control for low-field MR-IGRT systems: lessons
American Association of Physicists in Medicine task learned. J Appl Clin Med Phys. 2019;20(10):53–66.
group 284 for magnetic resonance simulation. J Appl 133. Duyn JH, Schenck J. Contributions to magnetic sus-
Clin Med Phys August 2022;23(8):e13730. ceptibility of brain tissue. NMR Biomed. 2017;30(4).
122. Karaiskos P, Moutsatsos A, Pappas E, Georgiou E, 134. Kames C, Doucette J, Rauscher A. Multi-echo
Roussakis A, Torrens M, u. a. A simple and efficient dipole inversion for magnetic susceptibility map-
methodology to improve geometric accuracy in ping. Magn Reson Med. 2023;89(6):2391–401.
gamma knife radiation surgery: implementation in 135. Tyagi N, Fontenla S, Zhang J, Cloutier M, Kadbi
multiple brain metastases. Int J Radiat Oncol Biol M, Mechalakos J, u. a. Dosimetric and workflow
Phys. 2014;90(5):1234–41. evaluation of first commercial synthetic CT soft-
ware for clinical use in pelvis. Phys Med Biol. CT software for radiotherapy. J Appl Clin Med Phys.
2017;62(8):2961–75. 2021;22:207.
136. Xie K, Gao L, Lu Z, Li C, Xi Q, Zhang F, u. a. 149. Siversson C, Nordström F, Nilsson T, Nyholm T,
Inpainting the metal artifact region in MRI images Jonsson J, Gunnlaugsson A, u. a. Technical note:
by using generative adversarial networks with gated MRI only prostate radiotherapy planning using the
convolution. Med Phys. 2022;49(10):6424–38. statistical decomposition algorithm. Med Phys.
137. Lundman JA, Bylund M, Garpebring A, Thellenberg 2015;42(10):6090–7.
Karlsson C, Nyholm T. Phys Imaging Radiat Oncol. 150. Persson E, Gustafsson C, Nordström F, Sohlin M,
2017;1:41–5. Gunnlaugsson A, Petruson K, u. a. MR-OPERA:
138. Marasini S, Quinn B, Cole M, Flores R, Kim a multicenter/multivendor validation of magnetic
T. System-dependent image distortion related to resonance imaging-only prostate treatment planning
gantry positions of a 0.35 T MRgRT: characteriza- using synthetic computed tomography images. Int J
tion and the corresponding correction. J Appl Clin Radiat Oncol Biol Phys. 2017;99(3):692–700.
Med Phys. 2023;24(1):e13826. 151. Persson E, Jamtheim Gustafsson C, Ambolt P,
139. Lewis BC, Shin J, Quinn B, Barberi E, Sievert D, Engelholm S, Ceberg S, Bäck S, u. a. MR-PROTECT:
Kim JS, u. a. First clinical experience of correcting clinical feasibility of a prostate MRI-only radiother-
phantom-based image distortion related to gantry apy treatment workflow and investigation of accep-
position on a 0.35T MR-Linac. J Appl Clin Med tance criteria. Radiat Oncol. 2020;15(1):77.
Phys. 2021;22(11):21–8. 152. Palmér E, Persson E, Ambolt P, Gustafsson C,
140. Lewis B, Guta A, Mackey S, Gach HM, Mutic S, Gunnlaugsson A, Olsson LE. Cone beam CT for QA
Green O, u. a. Evaluation of diffusion-weighted MRI of synthetic CT in MRI only for prostate patients. J
and geometric distortion on a 0.35T MR-LINAC Appl Clin Med Phys. 2018;19(6):44–52.
at multiple gantry angles. J Appl Clin Med Phys. 153. Tenhunen M, Korhonen J, Kapanen M, Seppälä T,
2021;22(2):118–25. Koivula L, Collan J, u. a. MRI-only based radiation
141. Dorsch S, Mann P, Elter A, Runz A, Spindeldreier therapy of prostate cancer: workflow and early clini-
CK, Klüter S, u. a. Measurement of isocen- cal experience. Acta Oncol. 2018;57(7):902–7.
ter alignment accuracy and image distortion of 154. Kim J, Miller B, Siddiqui MS, Movsas B, Glide-
an 0.35 T MR-Linac system. Phys Med Biol. Hurst C. FMEA of MR-only treatment planning in
2019;64(20):205011. the pelvis. Adv Radiat Oncol. 2019;4(1):168–76.
142. Damyanovich AZ, Tadic T, Foltz WD, Jelveh 155. Dowling JA, Sun J, Pichler P, Rivest-Hénault D,
S, Bissonnette JP. Time-course assessment of Ghose S, Richardson H, u. a. Automatic substitute
3D-image distortion on the 1.5 T Marlin/Elekta computed tomography generation and contour-
Unity MR-LINAC. Phys Med. 2022;100:90–8. ing for magnetic resonance imaging (MRI)-alone
143. Edmund JM, Nyholm T. A review of substitute CT external beam radiation therapy from standard
generation for MRI-only radiation therapy. Radiat MRI sequences. Int J Radiat Oncol Biol Phys.
Oncol. 2017;12(1):28. 2015;93(5):1144–53.
144. Bird D, Henry AM, Sebag-Montefiore D, Buckley 156. Greer P, Martin J, Sidhom M, Hunter P, Pichler P,
DL, Al-Qaisieh B, Speight R. A systematic review Choi JH, u. a. A multi-center prospective study for
of the clinical implementation of pelvic magnetic implementation of an MRI-only prostate treatment
resonance imaging-only planning for external beam planning workflow. Front Oncol. 2019;9:826.
radiation therapy1. Int J Radiat Oncol Biol Phys. 157. Palmér E, Karlsson A, Nordström F, Petruson K,
2019;105(3):479–92. Siversson C, Ljungberg M, u. a. Synthetic computed
145. Tyagi N, Fontenla S, Zelefsky M, Chong-Ton M, tomography data allows for accurate absorbed dose
Ostergren K, Shah N, u. a. Clinical workflow for calculations in a magnetic resonance imaging only
MR-only simulation and planning in prostate. Radiat workflow for head and neck radiotherapy. Phys
Oncol. 2017;12(1):119. Imaging Radiat Oncol. 2021;17:36–42.
146. Kemppainen R, Suilamo S, Tuokkola T, Lindholm P, 158. Cronholm RO, Karlsson A, Siversson C. Whitepaper:
Deppe MH, Keyriläinen J. Magnetic resonance-only MRI only radiotherapy planning using the transfer
simulation and dose calculation in external beam function estimation algorithm. 2020.
radiation therapy: a feasibility study for pelvic can- 159. Lerner M, Medin J, Jamtheim Gustafsson C, Alkner
cers. Acta Oncol. 2017;56(6):792–8. S, Siversson C, Olsson LE. Clinical validation of a
147. Kerkmeijer LGW, Maspero M, Meijer GJ, van der commercially available deep learning software for
Voort van Zyp JRN, de Boer HCJ, van den Berg synthetic CT generation for brain. Radiat Oncol.
CAT. Magnetic resonance imaging only workflow 2021;16(1):66.
for radiotherapy simulation and planning in prostate 160. Hoesl M, Escobar Corral, Mistry N. Whitepaper:
cancer. Clin Oncol. 2018;30(11):692–701. MR-based synthetic CT reimagined, an AI-based
148. Yu V, Keyrilainen J, Suilamo S, Beslimane I, Dresner algorithm for continuous Hounsfield units in the
A, Halkola A, u. a. A multi-institutional analysis of a pelvis and brain – with syngo.via RT Image Suite
general pelvis continuous Hounsfield unit synthetic (VB60). Siemens-Healthineers; 2022.
161. Emin S, Rossi E, Gagliardi G, Hedman M, Villegas 168. Das IJ, Yadav P, Mittal BB. Emergence of MR-Linac
F. Uncertainty-aware MR-based CT synthesis for in radiation oncology: successes and challenges
robust proton planning of skull-based tumour. of riding on the MRgRT bandwagon. J Clin Med.
ESTRO. 2023. 2022;11(17):5136.
162. Jonsson J, Nyholm T, Söderkvist K. The rationale 169. Güngör G, Azria D, Balermpas P, Boldrini L, Chuong
for MR-only treatment planning for external radio- MD, De Ridder M, u. a. A multi-centric evaluation
therapy. Clin Transl Radiat Oncol. 2019;18:60–5. of AI-driven synthetic CT generation form low field
163. Kurz C, Buizza G, Landry G, Kamp F, Rabe M, magnetic resonance imaging. Int J Radiat Oncol.
Paganelli C, u. a. Medical physics challenges in 2022;114(3):S163.
clinical MR-guided radiotherapy. Radiat Oncol. 170. Cusumano D, Boldrini L, Dhont J, Fiorino C, Green
2020;15(1):93. O, Güngör G, u. a. Artificial intelligence in magnetic
164. Dunkerley DAP, Hyer DE, Snyder JE, St-Aubin resonance guided radiotherapy: medical and physi-
JJ, Anderson CM, Caster JM, u. a. Clinical imple- cal considerations on state of art and future perspec-
mentational and site-specific workflows for a 1.5T tives. Phys Med. 2021;85:175–91.
MR-Linac. J Clin Med. 2022;11(6):1662. 171. Maspero M, Savenije MHF, Dinkla AM, Seevinck
165. Tang B, Liu M, Wang B, Diao P, Li J, Feng X, u. a. PR, Intven MPW, Jurgenliemk-Schulz IM, u. a.
Improving the clinical workflow of a MR-Linac by Dose evaluation of fast synthetic-CT generation
dosimetric evaluation of synthetic CT. Front Oncol. using a generative adversarial network for gen-
2022;12:920443. eral pelvis MR-only radiotherapy. Phys Med Biol.
166. Liang J, Scripes PG, Tyagi N, Subashi E, Wunner 2018;63(18):185001.
T, Cote N, u. a. Risk analysis of the Unity 1.5 T 172. Thorwarth D, Low DA. Technical challenges of real-
MR-Linac adapt-to-position workflow. J Appl time adaptive MR-guided radiotherapy. Front Oncol.
Clin Med Phys [Internet]. 2023 [zitiert 3. Mai 2021;11:634507.
2023];24(3). Verfügbar unter: https://onlinelibrary. 173. Goodburn RJ, Philippens MEP, Lefebvre TL,
wiley.com/doi/10.1002/acm2.13850. Khalifa A, Bruijnen T, Freedman JN, u. a. The future
167. Ng J, Gregucci F, Pennell RT, Nagar H, Golden of MRI in radiation therapy: challenges and oppor-
EB, Knisely JPS, u. a. MRI-LINAC: a transforma- tunities for the MR community. Magn Reson Med.
tive technology in radiation oncology. Front Oncol. 2022;88(6):2592–608.
2023;13:1117874.
Treatment Planning
Considerations for an MR-Linac
8
Justine M. Cunningham, Jennifer L. Dolan,
Kate Aldridge, and Ergys Subashi
Treatment Planning System algorithms that result in intensity modulated radi-

ation therapy (IMRT) deliveries paired with
MR-guided radiation therapy (MRgRT) deliv- accurate dose calculation algorithms (further dis-
ered with MR-Linacs is motivated by advanced cussed in sections “Planning Geometry and
simulation and on-board imaging. MRI-primary Optimization and Dose Calculation”). Additional
contouring, treatment planning, localization, and imaging (e.g., CT, PET) is often required for
motion management enhance the targeting accu- electron density (section “Electron Density”) or
racy of this treatment modality. The gain in tar- delineation purposes (section “Contouring”).
geting accuracy facilitates the shrinking of target Furthermore, the effective clinical application of
margins, sparing of organs at risk (OAR), online MRgRT relies on ultra-fast optimization and
adaptive treatment planning and delivery, and dose calculation to the point where such opera-
furthermore escalation or de-escalation of pre- tions can be performed while the patient waits on
scribed dose. Thus, MR-Linacs have largely the table for treatment.
become specialized treatment machines with uti-
lization focused on hypo-fractionation and ste-
reotactic style treatments [1]. Particularly, Hardware and Software Requirements
MR-Linacs have proven impactful in abdominal,
pelvic, and cranial treatment sites [2–4]. TPS hardware and software system requirements
Complex treatments, such as stereotactic body align with the safety and security necessitated by
radiation therapy (SBRT), demand advanced any commercial planning system. High-
treatment planning systems (TPS) to meet the performance computational abilities are required
intricate and diverse treatment planning needs. to enable Monte Carlo optimizations and dose
Such needs largely rely on inverse optimization calculations with computation times at less than
1 min. This requires powerful computers and
code which are optimized for CPUs and/or GPUs.
J. M. Cunningham (*) · J. L. Dolan · K. Aldridge An example of MR-Linac TPS has dual CPUs
Henry Ford Health, Department of Radiation
Oncology, Detroit, MI, USA with a minimum of 14 processor cores each,
e-mail: [email protected]; [email protected]; 128 GB RAM, 512 GB hard disk space, and
[email protected] 4 GB of video memory.
E. Subashi Additional emphasis should be placed on data
Memorial Sloan Kettering Cancer Center, transfer and interconnectivity quality assurance
Department of Medical Physics, New York, NY, USA (QA), as the MR-Linac TPS and treatment deliv-
124 J. M. Cunningham et al.
ery system (TDS) are likely to reside separately tional TPSs to assess the non-dosimetric and
from a clinic’s primary record and verify system dosimetric TPS commissioning and for ongo-
and other necessary subsystems [5, 6]. ing QA of the TPS [8]. Additional TPS testing
can include data acquisition and processing,
beam model iteration, and validation following
I ntegration with Treatment Delivery the American Association of Physicists in
Software Medicine (AAPM) Medical Physics Practice
Guidelines [9, 10]. Data acquisition and pro-
From a system architecture perspective, the nov- cessing includes CT calibration data and treat-
elty for MR-Linacs is the presence of a TPS in ment beam data following AAPM report of
parallel to the TDS. Since MRgRT TDSs are Task Group 106 [11]. Validation tests can
capable of on-table adaptive therapy (section include basic photon dose calculation, as well
“Adaptive Planning Considerations”), they are as dose calculations in heterogeneous media
responsible for more than just treatment delivery. and for intensity modulated radiation therapy
In a common adaptive work flow, the original CT (IMRT). IMRT commissioning tests can include
is deformed to the daily MRI for electron density, the tests outlined in AAPM Report of Task
and then the original plan is calculated onto the Group 119 [12]. Dose calculation for
current anatomy. Dose evaluation statistics are MR-Linacs is achieved utilizing Monte Carlo-
presented to the clinician to review and decide on based dose calculation algorithms; therefore
whether to adapt or not (section “Plan beam modeling and beam model commission-
Evaluation”). In the background, the system can ing can follow standards outlined in the AAPM
begin optimizing, generating a new plan, and Report of Task Group 157 and 105 [13, 14].
present a dose distribution by Monte Carlo with More sophisticated treatment techniques
excellent accuracy, all while the patient waits result in increased complexity of plans and more
inside the machine. The integration of the treat- advanced software. Therefore, there are testing
ment planning activities with the treatment deliv- requirements in addition to MRI-specific imag-
ery software presents unique considerations in ing QA requirements, to consider when imple-
the commissioning and QA of the planning sys- menting MR-guided radiation therapy.
tems [5, 7]. Specifically, the magnetic field may cause non-
negligible dosimetric effects, and the beam model
within the TPS can account for these effects,
Patient Database necessitating thorough validation. Current prac-
tices have included this testing within end-to-end
The TPS and TDS will require secure access to a verification tests [15] or specifically designed
patient and machine delivery database. Each sub- phantoms [16].
system in the integrated treatment planning and Online adaptive treatment techniques offer
treatment delivery system must accurately access additional TPS commissioning challenges. If
beam and patient data. Commissioning and rou- the treatment delivery system planning platform
tine QA should aim at the evaluation for the accu- is unique from the TPS, the same processes and
racy and security of such connectivity [5, 6]. verification of treatment planning must be per-
formed utilizing the online adaptive platform.
End-to-end verification tests isolating couch
Commissioning and Quality Assurance attenuation, electron density conversion, con-
tour generation, and optimization robustness are
Commissioning and QA of the treatment plan- some of the many tests that can be considered
ning systems (TPS) associated with MR-Linacs during adaptive treatment planning validation
can follow established guidelines for conven- [17–19].
8 Treatment Planning Considerations for an MR-Linac 125
Contouring reatment Planning Imaging

T
Datasets
The integration of an MRI scanner with a linear
accelerator offers unique advantages compared to A combination of treatment planning CTs and
conventional CT-based on-board imaging. The MRIs are used for OAR and target contouring.
soft tissue contrast achievable on MR-Linacs has MR-Linac platforms offer both T1-weighted and
been shown to provide better visualization of T2-weighted images. High-field MR-Linacs have
radiotherapy targets and OARs compared to on- been evaluated for the preference of MR-derived
board CT-based imaging techniques [20]. sequences including mDIXON and T1- and
Functional imaging during simulation aids in T2-weighted images for different anatomical
improving the precision of target delineation and regions [34, 35]. Low-field MR-Linacs require a
OARs [21]. Additionally, the incorporation of true fast imaging with steady-state precession
multi-planar MR-cine image acquisition intro- (TRUFI) sequence yielding a T2−/T1-weighted
duces new motion management techniques such contrast for planning [36]. The possibility of
as real-time motion monitoring and gating which short acquisition times for TRUFI image
may facilitate the use of reduced margins sequences allows for images to be acquired at
[22–24]. breath-hold, facilitating the ability to treat at
Standard contouring tools available across breath-hold using cine image real-time motion
various treatment planning platforms can be uti- monitoring and gating during treatment delivery.
lized for MR-Linac treatment planning. MRI- MRI sequences with longer acquisition times can
based target volume and OAR contouring use 4D-MRI to facilitate treatment planning on
recommendations exist for various locations motion averaged images [28]. These may be used
throughout the body [25–28]. Additionally, sev- with 4DCT (four-dimensional CT) to account for
eral auto-segmentation techniques are being breathing motion using an internal target volume.
studied not only for their use in initial segmenta- Additionally, motion encompassing methods
tion for planning but also to be utilized during with abdominal compression have been used to
online adaptive radiotherapy [29–32]. Auto- reduce the amplitude of breathing motion [37].
segmentation methods can be categorized into Recommended slice thicknesses of one to three
methods that do and do not use prior knowledge. millimeters are achievable for conventional and
Methods that do not use prior knowledge include stereotactic treatment planning. Although MRI is
intensity-based segmentation where voxels are the gold standard for soft tissue visualization,
classified into different tissue types based on contrast-enhanced CT-based imaging or
intensity levels. Methods that require prior F-fluorodeoxyglucose PET imaging may prove
knowledge include atlas based, statistical mod- useful in the contouring of targets in various
els, and models based on machine learning and regions throughout the body and can be regis-
artificial intelligence. Atlas-based segmentation tered to the treatment planning MRI to aid in tis-
uses deformable or rigid registration to transform sue delineation.
a structure of interest from a reference image to
test image. Segmentation using machine learning Contrast
is achieved through using a training dataset to To aid in the visualization and delineation of the
establish features for tissue segmentation and disease being treated, gadolinium-based contrast
applying it to a test image [33]. Auto-segmentation agents can be used. Spinal cord tumors, cranial
techniques are an invaluable tool for increased tumors, and liver metastasis are common ana-
planning efficiency, specifically in relation to tomical sites where contrast agents are used to
online adaptive treatment planning. better visualize the disease. These contrast agents
cannot only be used for target delineation but can facts [38]. Due to the high dependence of geo-
also enhance online gating capabilities. For metric accuracy in radiation delivery, the
example, when treating liver metastasis, contrast reduction of geometric distortions and artifacts
can be distributed, if kidney function allows, to through quantification and mitigation is vital for
the patient on the day of treatment, and the lesion the delineation of target volumes and OARs.
itself can be used for real time motion monitoring
and gating during treatment delivery. An example
of this can be seen in Fig. 8.1. Margins
rtifacts and Geometric Distortions

A Planning target volume margins are applied to
Two primary categories of geometric distortions account for uncertainties in target volume delin-
exist in MR imaging including system level, eation and positioning, organ motion, variations
resulting from gradient nonlinearity in the spatial in patient positioning, and uncertainties in dose
encoding gradients and magnetic field inhomoge- delivery and target coverage. The required mar-
neities, as well as patient-level distortions due to gin size is impacted by image-guided technique
chemical shift artifacts and susceptibility arti- and frequency of use. In MRgRT direct visualiza-
Fig. 8.1 Top images, liver lesion with the CTV outlined the cine MRI with 3.5 mm × 3.5 mm resolution and four
in red and PTV outlined in pink in the axial sagittal and frames per second used for gating. The smoothed CTV
coronal view of a 14 s TRUFI image taken at end exhale volume, shown in red falls within the gating boundary
with an image resolution of 1.5 mm × 1.5 mm × 3 mm. (yellow), which was generated using a 3 mm isotropic
The bottom image includes a sagittal and coronal view of margin from the smoothed CTV volume
tion of target volumes and OAR in conjunction anatomical structure is outlined on each plane
with functional imaging during simulation and with a desired margin generating a gating bound-
treatment delivery may facilitate the reduction in ary. A defined percentage of the structure of
margins due to the reduced uncertainties in the interest allowed outside of the gating boundary is
delineation of target volume and patient localiza- used to gate the beam off. For each frame of cine
tion at time of treatment. Additionally, planar MRI’s taken, the gating structure is deformably
cine MR images are acquired continuously dur- registered to a selected key frame outlining the
ing treatment for delivery verification. These cine desired treatment position. If the defined param-
images can be used for motion monitoring with eters are met, the beam is turned on. This tech-
manual gating or real-time motion management nique can be used for gated breath-hold, gated
with continuous feedback. Frequently, real-time free-breathing treatments, and gated treatments
motion management is performed utilizing a con- that are unaffected by respiratory motion but may
tour of interest that can be outlined as the target be affected by peristalsis in the gastrointestinal
volume in the MR cine image; however one may track, changes in the genitourinary track, or
also choose to monitor an organ at risk to ensure patient movement [36]. The delivery technique is
it does not enter the radiation field or a surrogate crucial to determining the size of the margins
contour if the target volume is not visible. For required for various clinical sites. Depending on
mobile targets, a large PTV margin was previ- the contour of interest chosen for motion moni-
ously used to provide sufficient coverage of the toring, the gating boundary may correspond to
moving lesion. However, continual monitoring of the planning target volume margins.
the precise location of the target makes it feasible
to reduce PTV treatment margins, therefore
reducing the dose to OARs. These motion man- Online Adaptive Considerations
agement techniques may further facilitate the
reduction in margins due to reducing the uncer- During the online adaptive treatment planning
tainties in dose delivery [22, 24, 39, 40]. Margins workflow, efficiency in re-contouring and re-
will be specific to the anatomical area being planning is vital to reduce treatment times.
treated, immobilization devices used, and the Limiting the number of organs that require recon-
motion management technique selected for treat- touring to the immediate regions adjacent to the
ment delivery. target volume can facilitate this. Additionally,
defining those structures that can be strictly
deformed and reviewed from the planning image
otion Management with Cine
M to the daily image will aid in reducing the num-
Imaging ber of contours that will be manually recontoured
online.
During motion monitoring, two-dimensional cine Gross tumor volumes and clinical tumor vol-
MRIs are acquired continuously in up to three umes are frequently redrawn daily. Planning
planes. During delivery, the contour of interest is rules imbedded into the treatment plan can be
displayed on the cine MRI allowing the treatment used to automatically generate the resulting plan-
delivery team to assess any changes in anatomy ning target volume based on pre-established mar-
and intervene when necessary. This method relies gin expansions. These rules can include the
on manual monitoring of the changing anatomy, generation of target volumes using pre-
without automatic beam gating. During real-time established margins, optimization structures that
motion management with continuous feedback, are frequently used in planning, and planning at
the system can gate the beam automatically. An risk volumes. The predefinition of margin expan-
Fig. 8.2 Adaptive pancreas TRUFI MRI shown in the (green) which will be redrawn within the normal tissue
axial, sagittal, and coronal orientations. Structures out- contouring ring (light blue) daily during the adaptive pro-
lined for the adaptive procedure include the liver (pale cedure. Finally, the CTV OPT (color wash green) and
green), right kidney (dark blue), left kidney (dark green), PTV OPT (color wash purple) are automatically regener-
and spinal canal (dark green) which will be deformed onto ated for daily adaptive treatments by cropping the rigidly
the MRI of the day and reviewed and edited, as necessary. copied PTV (red) and CTV (orange) 3 mm away from the
The gastrointestinal OAR include the colon (pink), stom- gastrointestinal structures
ach (yellow), duodenum (orange), and small bowel
sions within the plan not only improves efficiency propagated onto the MRI via bulk density over-
in the generation of these structures but also rides, deformable registration of a CT simulation
reduces the opportunity for error associated with to the MR, and synthetic CT (sCT) generation.
the manual generation of these structures. Due to
their impact on the adaptive workflow, it is
imperative that these structures and rules are out- CT Primary
lined during the initial planning process. An
example of a normal tissue contouring ring and At the initiation of the treatment planning pro-
contours generated through treatment planning cess, the primary imaging dataset is chosen. In
rules is shown in Fig. 8.2. some MRgRT treatment planning software ver-
sions, the user can choose the patient’s CT data-
set to be used as the primary image dataset.
Electron Density Supplemental datasets, such as the MRI simula-
tion, can be fused to the CT dataset for contour-
MR imaging is not sufficient for radiotherapy ing purposes. However, this workflow may lead
treatment planning due to its intrinsic lack of to image registration errors as significant as
electron density (ED) information. 1-2 mm, where the magnitude of these errors
Conventionally, radiation dose is calculated using depends on the co-registration technique used as
CT images where voxel Hounsfield units are con- well as the anatomical site [41]. These image reg-
verted to ED using a CT to ED conversion curve. istration errors are systematic and will be propa-
In MRgRT, there are distinct options available to gated throughout the treatment planning and
overcome this limitation. Plans can be developed delivery process, which could result in a geomet-
on a patient CT as opposed to an MRI, allowing ric miss. This method allows the user to calculate
for the dose calculation process to be performed dose in the conventional fashion, but this will
on the gold standard dataset for dose calculation. also result in MRI to CT co-registration errors
For MR only options, ED information can be during daily image guidance.
MR Primary sCT. Multi-atlas methods can take into consider-

ation the variation in size and shape of individual
The MR image can also be chosen as the primary patient anatomy, where the most similar CT or
imaging dataset. In this scenario, electron density CT-MRI based on registration metrics, to the cur-
must be propagated to the treatment planning rent anatomy, will be deformably registered to
image for dose calculation. This can be done the treatment planning MRI and used to generate
through bulk density overrides or a deformable the sCT. The generation of sCT through machine
registration of the patient’s CT simulation or sCT learning is an ongoing research initiative. Various
to the treatment planning MRI. In each scenario, neural networks have been developed utilizing
the electron density map should be reviewed by a both paired MRI/CT patient images and unpaired
physicist prior to the final dose calculation. images, where co-registration errors are not prop-
MR-only treatment planning can be performed agated to the synthetic CT generation with prom-
using sCT datasets generated from MRI and used ising results. An example of a sCT generated
for dose calculation. A significant review of the using MRCAT (MR for calculation attenuation),
production of sCT has been done by Owrangi an FDA-approved commercial sCT software, is
et al. [38] and Chap. 7. Voxel, atlas-based, and shown in Fig. 8.3.
deep learning methods have been under develop- The ED map can be generated from the CT
ment in recent years. Voxel-based methods use simulation image dataset for dose calculation
classification of tissues on a voxel-by-voxel using a deformable image registration. The CT
basis. Categories of tissue include water, air, adi- simulation and MR simulation are typically taken
pose, trabecular bone, and cortical bone. For each sequentially with consistent immobilization
classification of tissue, a bulk HU value is devices to ensure similar anatomical properties
assigned. The simplest scenario involves between the two imaging datasets. In abdominal
assigning a water-equivalent HU value to the
and pelvic anatomy, there are frequently differ-
entirety of the patient volume; however, in the ences in the air present in each of the imaging
presence of heterogeneous media including bone datasets. One method that has been introduced to
and air, the uncertainty in dose calculation will overcome this issue has been to implement con-
increase. Therefore, various MRI sequences can tours that can be used for bulk density overrides
be used and combined to segment air, bone, and of the tissues present in each of the images to air
soft tissue in different anatomical regions of the or water. Specifically, the air in the CT image can
body and assign the corresponding HU value. be contoured and overridden to water, while the
Both single and multi-atlas-based techniques air in the MRI can be contoured and overridden
have been developed. In single atlas-based tech- to air. An example of this is shown in Fig. 8.4.
niques, the average patient anatomy in the form This method can also be used to account for any
of a CT or CT-MRI pair is deformably registered differences in air-tissue interfaces between the
to the treatment planning MRI to produce the synthetic CT and treatment planning MRI.
Fig. 8.3 MRCAT sCT shown on the top, with the associated electron density heat map (middle) and treatment planning
MRI (bottom)
Fig. 8.4 Axial and coronal views of the deformed CT structure in light blue is used to bulk density override the
(top), electron density map (middle), and treatment plan- visible air in MRI to an electron density of air. The over-
ning MRI (bottom). The override to water structure in ride to air structure is set to a higher priority than that of
dark blue is used to bulk density override the visible air in the override to water structure
CT to an electron density of water. The override to air
Planning Geometry MR-Linac treatment platforms. As mentioned

and Optimization above, the isocenter location is sometimes placed
outside of the treatment volume, which results in
Isocenter Placement frequent off-axis treatments to lateral targets
where beam angles must be chosen carefully, tak-
In conventional radiation therapy, the isocenter ing into consideration the MLC position limita-
location is optimally placed within the center of tions and ensuring the target volume is adequately
the target volume or the radiation field. However, encompassed in the beams eye view (BEV). An
MR-Linacs have a limited bore size resulting in example of beam geometry is shown in Fig. 8.5,
either limited or prohibited table movement in including the off-axis beam eye view showing the
the anterior/posterior and lateral directions. As a limitations of the MLC movement and beam
result, the isocenter location is typically placed in angle choice.
the center of the target or radiation field in the Conventional treatment delivery times range
superior-inferior direction but is centrally located from 15 to 30 min, depending on the number of
in the axial plane. Depending on the laterality of beams angles and complexity (e.g., number of
the tumor location, this may result in large off- segments) of the treatment delivery. This is due to
axis segment openings. To reduce the distance slow gantry rotation and MLC movement.
from the isocenter to the target volume center, the However, the length of the adaptive treatment
patient is positioned with the treatment site as process can be much longer and range from 45 to
close to midline as possible. 90 min due to the need to generate and review
new contours and treatment plans. Patient posi-
tioning for adaptive treatments is often modified
Beam Geometry to increase patient comfort, including simulating
patients immobilized with their arms down,
Current delivery platforms offer 3D conformal resulting in the requirement of choosing beam
and static IMRT-based planning methods. angles avoiding entry through the patient arms.
Conventional beams geometries used in CT-based Additional adaptive treatment considerations
planning methods are acceptable for MRI-based include maintaining suboptimal beam angles in
planning while keeping in mind additional limi- the base plan, as they may be required to main-
tations due a decreased size in the field-of-view tain plan quality due to inter-fraction motion. A
(FOV), limited beam angles, and lack of collima- pseudo arc technique, with equally spaced static
tor rotation, off-axis treatments, and longer treat- IMRT beams, is frequently adopted for treatment
ment times. The largest FOV achievable for MR planning. Future developments in MR-Linac
simulation images is smaller than the FOV treatment delivery platforms may lead to arc-
achievable on CT simulation machines. This may based treatment options, where these limitations
result in edges of the patient body contour being will still need to be considered during the beam
cut off in simulation images, causing limited selection process.
choices of beam angles due to missing electron
density information. Additionally, due to manu-
facturer designs, some MR-Linacs have an asso- Treatment Delivery Techniques
ciated dead zone where a range of beam angles
cannot be used for treatment planning. It is also 3D Conformal
recommended to avoid treatment through the Three-dimensional conformal treatment planning
edges of the couch due to uncertainties a ssociated on an MR-Linac follows a similar concept to that
with patient position and variation in attenuation of CT-based planning techniques. A set beam
through the edges of the couch. MLC motion is geometry is chosen following the limitations out-
limited to a single plane (e.g., left-right or supe- lined in section “Beam Geometry”, MLCs can
rior-inferior) as collimator rotation is disabled on either be manually positioned or automatically fit
Fig. 8.5 Left breast APBI off-axis beam geometry with indicate the off-axis nature of the beams required due to
the isocenter shown as a red cross, the CTVsb color the limited couch movement. These beams are also shown
washed red, and PTVsb color washed purple. Two BEVs in the bottom left of the figure
of the beams at gantry angle 140° (top) and 312° (bottom)
to a target structure with a specified margin. The ments, re-optimized and re-calculated to consider
dose is calculated to a normalization or weight- the physical limitations of the MLC. During the
ing point, and fixed beams can be weighted to optimization, various parameters can be chosen
create the desired dose distribution. that will impact the dose calculation including
the bixel size, turning the magnetic field on/off
IMRT for dose calculation, editing the grid resolution
Current delivery platforms are limited to step- and selection of MLC sequencing options. Bixel
and-shoot IMRT delivery techniques, allowing size is limited in one dimension by the width of
for highly conformal dose distributions improv- the MLC. A smaller bixel size will result in a
ing target coverage while reducing dose to finer resolution of the fluence; however it may
OAR. IMRT plans are generated by dividing the result in increased modulation. A larger bixel size
beams into bixels, then optimizing and calculat- will reduce the fluence resolution, decreasing the
ing the bixel fluence based on user defined objec- modulation of the plan; however, it may result in
tives and constraints as shown by Das et al. [42]. sub-optimal plans. Once the desired bixel fluence
Fluences are then converted into deliverable seg- is achieved, MLC sequencing options can be cho-
sen to put an emphasis on plan quality by creat- Adaptive treatment planning often includes an
ing a best matched deliverable fluence to the isotoxic approach. Instead of adaptation based on
desired bixel fluence, degree of modulation avoiding under dosing the target or overdosing
through discretization of the fluence’s in the bixel OARs, an isotoxic approach involves prescribing
map, or simply limiting the number of segments. to a predefined dose-volume constraint to adja-
However, step-and-shoot when compared to slid- cent OARs while escalating the dose to treatment
ing window and arc-based treatment delivery targets as much as technically possible to obtain
techniques is less efficient and often results in a favorable therapeutic ratio. This is frequently
longer treatment times. Due to the inherent lon- adopted in the treatment of pancreatic cancers;
ger treatment times, the planning process often however, it is being adopted to other anatomical
includes treatment delivery time reduction tech- areas including in stereotactic liver radiation
niques, such as limiting the number of MLC seg- therapy and tumor regression in the treatment of
ments while maintaining satisfactory plan quality. lung cancer [43–45].
However, choosing to use too few segments will
result in a suboptimal dose distribution.
Optimization strategies, such as allowing the user Dose Calculation
to generate a series of plans for a single cost func-
tion with varying number of segments, may be Integrated MR-Linac TPSs have the unique
useful in balancing plan complexity with plan requirement and challenge of calculating radia-
quality. tion dose in magnetic environments. Magnetic
fields influence the trajectory of electrons liber-
ated by megavoltage-level photons and thus
Optimization impact the spatial distribution of their energy
deposition and furthermore radiation therapy
The goal of the optimization process is to reduce dose. This influence, known as the Lorentz force,
the cost function to near zero (minimum). can manifest as alterations in beam characteris-
However, this is typically not achievable, and an tics (e.g., depth versus dose, penumbra) and clini-
acceptable solution is found by reducing the cost cally is of concern at interfaces of significant
function until negligible improvements can be electron density change (e.g., air/skin or lung/tis-
made and the optimizer converges. The cost sue) resulting in the electron return effect (ERE).
function is used to define the difference between The magnitude of this force is related to the mag-
the desired dose distribution and the obtained netic field strength. An example of the dose depo-
one. Each treatment planning platform may sition changes for a 0.35T MR-Linac with and
define their cost function differently; however, in without the consideration of the magnetic field in
each scenario the cost function parameters are the dose calculation for a left breast treatment is
defined by the user and may depend on dose- shown in Fig. 8.6, where prominent changes in
based criteria or biological criteria. A physical the surface dose are observed. The effect of the
cost function is mainly used and defined using magnetic field on dose deposition is dependent
objectives and constraints based off dose volume on field strength, patient anatomy, and beam
histogram criteria. Once the problem is defined geometry, necessitating a review of common
within the cost function, the optimizer then approaches used in conventional treatment plan-
begins to search for a minimum in the cost func- ning [46]. The application of magnetic field
tion to generate the desired dose distribution. effects on the mechanics of radiation transport
Inherently, static IMRT results in low-dose and dose deposition is best handled in Monte
streaking in the beam path. This streaking can be Carlo dose calculation algorithms where each
minimized through avoidance and ring optimiza- particle history can be appropriately altered by
tion structures. the field’s presence. Broadly used Monte Carlo
Fig. 8.6 An example of the dose deposition changes for a treatment. Notable changes in 6Gy isodose line illustrat-
0.35T MR-Linac with and without the consideration of ing the surface dose can be seen anteriorly and
the magnetic field in the dose calculation for a left breast posteriorly
codes including GEANT4, PENELOPE, MCNP, patient-specific collimated beam is then trans-
and EGS have demonstrated the utility of Monte ported to and through the patient anatomy. The
Carlo dose calculation techniques in the mag- electron density map described in section
netic field environment. Therefore, MR-Linacs “Electron Density” provides the data needed to
rely on such algorithms [14, 47]. query radiation interaction cross sections that
In addition, MR-Linac planning systems dictate the radiation transport and dose
require an enhanced calculation speed due to the deposition.
time-limited nature of on-table adaptive radiation However, a balance must be obtained between
therapy treatment planning. To transition from efficiency and accuracy. Dose grid size and the
the general-purpose Monte Carlo codes to algo- number of particle histories become important
rithms that are suitable for this application, vari- parameters to statistically converge the dose cal-
ance reduction techniques and efficiency culation to a resolution level that is meaningful
enhancing methods must be implemented [14]. for treatment planning and evaluation.
These well-established Monte Carlo engines
have since been used to validate commercial
packages and their chosen balance between accu- Electron Return Effect
racy and efficiency [48–51].
To improve efficiency, the Monte Carlo dose As mentioned above, a strong motivator behind
calculation algorithms typically use a machine- the use of Monte Carlo dose calculation algo-
specific phase space that acts as the source or rithms in the presence of a magnetic field is the
starting point to each history in the calculation. influence of the B-field on the trajectory of liber-
The output of the forward or inverse optimization ated electrons. The magnitude of this effect
process described in 8.5.3 and 8.5.4 is a series of increases with increasing magnetic field strength,
MLC segments at user-defined gantry angles. along with its clinical impact, and the need to
These segments are then used as the first patient- take these effects into consideration during dose
specific step in the final dose calculation. The calculation [52]. The impact of the magnetic field
includes a reduction in the depth of maximum the magnetic field, cryostat, and modified scatter-
dose, an asymmetric dose kernel (due to a ing on the dose distribution that have been vali-
deflected electron path), and changes in the beam dated in treatment planning studies [60–62]. This
penumbra. It also introduces the electron return algorithm utilizes graphical processing units to
effect at in-field tissue-air interfaces, which is increase calculation speeds through paralleliza-
due to electrons being forced into a circular path tion of the process and has shown within 1%
in air and re-entering the tissue interface leading agreement in a homogenous phantom along the
to an increase in local dose [53–55]. The ERE beam central axis and within 2% agreement in a
results in an increase dose at the proximal end of heterogeneous phantom [50, 63].
an air cavity and a decreased dose at the distal
end; this can directly impact patient treatments in
the nasal and oral cavities and thorax and abdom- Adaptive Planning Considerations
inal regions with tissue-lung interfaces or gas
pockets. These effects are considered through the Treatment planning relies on a detailed model of
simulation of the dose perturbation from the elec- patient anatomy, including all clinical assess-
tron return effect within the TPS dose calculation ments needed to characterize the disease and all
algorithms. However, there are additional mitiga- physical measurements needed to characterize
tion strategies that can be considered during plan- the relevant properties affecting the distribution
ning, including opposing beams or increasing the of radiation dose. An accurate patient model must
number of beams to balance or reduce hot and account for spatiotemporal changes in any clini-
cold spots, introducing additional OAR contours cal findings or physical properties, and the pro-
such as a skin contour to evaluate and limit dose cess of designing a radiotherapy treatment plan
to these areas, as well incorporating bolus to must allow for revisions that adapt to such
reduce exit skin dose [56–58]. changes. In this context, adaptive planning is the
process of adjusting the radiation treatment plan
to account for changes in clinical findings and
Heterogeneity Corrections patient physical properties. Currently, the selec-
tion of a set of clinical criteria (or triggers) for
Most modern treatment planning systems utilize determining the need for adaptation is still work
deterministic model-based dose calculation algo- in progress, with initial research providing prom-
rithms that rely on the scaling of dose kernels to ising results [64–66].
account for the impact of tissue heterogeneities
on secondary electron dose clouds. Monte Carlo
dose calculation algorithms are based on proba- Reference Plans
bilistic models and are inherently more accurate
in heterogeneous environments, as they directly While it is recognized that a robust treatment
account for the impact of heterogeneous medi- plan must account for spatiotemporal changes in
ums on electron transport [14]. Theoretically, the patient model, current methods and measure-
their superior handling of inhomogeneity is clear; ments for fully characterizing this model are site-
additionally, significant experimental verification specific and will improve with standardization.
efforts have been undertaken to prove their valid- Therefore, an initial treatment plan is typically
ity in clinical environments [49, 50, 59]. devised based on a snapshot of the patient’s anat-
A low field MR-Linac currently operating omy from the day of simulation. Depending on
clinically has the option to take a static magnetic treatment site, the measurements may include an
field into account utilizing fast Monte Carlo dose estimate of the magnitude of changes due to
calculation algorithm [36]. A clinical high field physiological motion (e.g., cardiac, respiratory,
MR-Linac similarly employs a Monte Carlo dose or gastrointestinal motility) or processes (e.g.,
calculation algorithm that models the effects of bladder, rectum, and stomach filling/emptying).
The inclusion of functional imaging during simu- and OARs. These changes are often observed (1)
lation aids in improving the precision of target after the delivery of several fractions, (2) at each
delineation and OARs [21]. The reference plan is fraction, and (3) during the time of irradiation.
the treatment plan created based on data available These timescales correspond, respectively, to
at the time of simulation. Generally, the planning changes over several days, daily, and over a few
procedures for creating a reference plan are the minutes or seconds. The planning methods cor-
same as, or similar, the institutional procedures responding to each category are commonly
established for that treatment site. This also known as offline (or inter-fraction), online (or on-
applies to safety checks and QA procedures. table), and real-time (or dynamic) adaptive radio-
therapy [7, 67, 68]. Figure 8.7 shows a schematic
diagram presenting the relevant steps for each
Adaptation Timescales adaptation method.
Adaptive planning strategies have traditionally ffline Adaptive Planning

O
focused on spatiotemporal changes categorized Offline (or inter-fraction) adaptive planning mon-
over three timescales and assessed by longitudi- itors change after the delivery of one to several
nal anatomic and functional images of the target fractions and accordingly revise the treatment
Fig. 8.7 Schematic diagram of workflow steps for offline, online, and real-time adaptive planning for each treatment
session
plan to optimize radiation dose. While safety necessitates verification and motion monitoring
margins are often employed when preparing the images in a workflow that includes closed-loop
reference plan, additional modifications may be feedback [86, 87]. This in turn prolongs patient
needed to account for the magnitude of variabil- on-table time and requires automation at each
ity in target and OAR position, shape, and vol- step of adaptation. Online re-planning remains a
ume [69–74]. The requirements and constraints resource-intensive method as it routinely requires
for offline plan adaptation and QA are typically the presence of the entire treatment team (dosime-
the same as for reference plans. Offline adapta- trists, physicists, physicians) at the machine dur-
tion targets slow, systematic, or abrupt changes in ing the process of adaptation.
the patient model, can use the imaging modality
of choice for the given treatment site and is easy eal-Time Adaptive Planning
R
to implement. Real-time (or dynamic) adaptive planning moni-
tors changes during the delivery of radiotherapy
Online Adaptive Planning and revises the delivered plan accordingly to
The most widely implemented adaptation strat- optimize radiation dose. This adaptation strategy
egy in MR-Linacs is online (or on-table) adaptive monitors changes in patient anatomy that happen
planning. Online adaptive planning accounts for in short timescales (seconds to minutes) and for
daily changes in the patient model and accord- which a predictive model can be generated [88–
ingly revises the treatment plan to optimize the 90]. Common predictive models for real-time
delivered radiation dose. This adaptation method adaptation include those for baseline drifts and
is employed in treatment sites where daily vari- physiological motion due to cardiac, respiratory,
ability is expected to often be beyond the safety and GI tract motility [91, 92]. Recent work has
margins used in the reference plan. The online demonstrated that clinical implementation of
adapted plan uses an image acquired immedi- real-time adaptive radiotherapy can improve
ately after the patient is positioned on the delivery [93]. Real-time adaptation requires the
treatment couch and is commonly generated with system latencies, i.e., the time from detection to
semi-automatic methods that rely on the refer- prediction and execution, to be smaller than the
ence plan as the starting point. In this approach, time for the anatomical change to occur [94].
the images used for daily planning are acquired Therefore, the temporal requirements for imag-
in the MR-Linac and do not necessarily match ing and data acquisition are dependent on the
the image contrast from the reference plan. treatment site and the magnitude of changes in
Adaptation is achieved in two ways, broadly patient anatomy. Adaptation is typically imple-
based on user workflows that allow for a physical mented through dynamic modifications of MLC
or virtual couch shift and no contour editing or patterns to account for predicted changes in
full re-planning with daily re-contouring and patient model. Currently, the predictive models
optimization [23, 75, 76]. A common, semi- are mainly based on changes in target position
automatic technique for online adaptation is and shape, instead of OAR or functional changes.
based on the generation of a library of reference While still primarily a research method, the feasi-
plans that are optimized for anticipated changes bility of real-time adaptation in MR-Linacs has
in patient anatomy (e.g., bladder, rectum filling/ been successfully demonstrated in multiple sites
emptying) [77–80]. Planning templates that [90, 91, 95, 96].
attempt to provide robustness against changes in
patient anatomy, known as class solution tem-
plates, have also shown initial feasibility [81–83]. Adaptation to Anatomical Changes
The generation of safety margins typically fol-
lows the same rules as in reference plans [84, 85]. The adaptive planning methods described above
The expectation that anatomical changes may are predominantly driven by target and OAR ana-
occur during and after online plan adaptation tomical variability. The variability is due to
changes in position, size, and/or shape and can MRI. While the definition of functional MRI is
happen at various timescales. Therefore, based dependent on treatment site and will improve
on the timescales for these changes, all methods with standardization, the most often used
of plan adaptation (offline, online, real-time) can sequences for functional MRI in radiation ther-
be clinically used. The most common triggers for apy include DWI-MRI, perfusion, and DCE-
adaptation to anatomical variability involve non- MRI. These methods, respectively, estimate the
rigid deformations of the target/OARs at a mag- local rate of water diffusion, blood flow and oxy-
nitude beyond the safety margins used in the genation, and vessel permeability into the extra-
reference plan. The availability of MR images vascular/extracellular space. Recent work has
with various contrast mechanisms allows for demonstrated the feasibility of functional imag-
improved delineations and can enable methods ing in MR-Linacs and suggested consensus pro-
for adaptation triggers that combine contouring, tocols for future clinical use [106–112].
optimization, and dosimetric evaluation [64, 65, Functional imaging provides complementary
84, 85]. However, robust and fully automated cri- information to anatomical imaging and may help
teria for triggering adaptation are still needed, improve the precision of target/OAR delineation
with initial work providing promising results and offline or online adaptive planning [21, 84,
[64–66]. 85, 113–115]. Figure 8.8 presents an example of
combined anatomical and functional imaging
used in online adaptive radiotherapy to the pros-
Adaptation to Functional Changes tate and dominant intraprostatic lesion [85, 116].
Additionally, targeting based on functional imag-
Multiparametric (mp) MRI is increasingly being ing is feasible in a multi-institutional setting, and
used for diagnosis, staging, and treatment treatments can provide improvements in out-
response assessment in several disease sites [97– comes [117]. It is worth noting that other non-
105]. Imaging sequences within mpMRI proto- MRI-derived functional biomarkers can be used
cols commonly include T1/T2-weighted, for offline adaptation and could be combined to
diffusion-weighted (DW), perfusion, dynamic increase the precision of online or real-time
contrast-enhanced (DCE), and spectroscopic adaptation.
a b
c d e
Fig. 8.8 Use of functional biomarkers for treatment plan- ical T2-weighted MRI (c). T1 relaxometry map (d). T2
ning (a). Dose overlay on anatomical T2-weighted image relaxometry map (e). Apparent diffusion coefficient
(b). Contours for CTV (red, prostate gland) and GTV (ADC) map. Arrows point to the location of the dominant
(blue, dominant intraprostatic lesion) overlaid on anatom- intraprostatic lesion
Plan Evaluation organs in which max dose is a clinically impor-

tant measure for tissue response (e.g., spinal
Dose Calculation Accuracy cord). DVH (dose volume histogram) metrics and
dose indices that model treatment response (e.g.,
The point-spread kernel for secondary electrons TCP/NTCP) have been found to be less sensitive
involved in dose deposition is significantly to statistical noise, and calculations with uncer-
affected by the presence of the magnetic field. tainty less than 2% are sufficiently accurate for
The impact on dose distribution is a decreased reporting and plan evaluation [14, 127]. The
depth for build-up, a shifted and asymmetric concept of the uncertainty volume histogram

beam profile and the ERE [53, 55, 118]. The (UVH) is an equally valuable tool in the evalua-
magnitude of these effects increases with mag- tion of plan quality as it allows physicist and phy-
netic field strength and requires accurate model- sician to assess whether the dose uncertainty in
ing for dose calculations [52]. Currently, clinical targets/OARs is clinically acceptable [127].
treatment planning systems for any MR-Linac While it is recognized that dose accumulation for
must employ a dose calculation engine based on time-varying anatomy will provide critical infor-
Monte Carlo (MC) computation methods [63, mation for treatment response assessment, the
119–121]. Analytical or model-based dose calcu- implementation of image registration algorithms
lation engines with sufficient accuracy have yet for routine clinical use remains challenging [128,
to be implemented for routine use [122]. 129].
Guidelines and recommendations for the clinical
evaluation of the accuracy in MC-based algo-
rithms, including acceptance testing and com- Plan Summation
missioning, are described by AAPM TG 105 and
TG 157 [13, 14]. Particular attention is devoted to Accumulated dose has been shown to be a better
the differences with conventional TPS algo- predictor of treatment response in stereotactic
rithms, especially as it relates to statistical uncer- body radiotherapy [130]. Dose accumulation
tainty, as well as the conversion from requires summation of the dose from the adapted
dose-to-water and dose-to-medium [123]. plans delivered at each fraction, which in turn
Additional validation may be necessary at tissue necessitates implementation of rigid or deform-
interfaces with high/low density and at specific able registration [128]. The accuracy and utility
treatment sites where the ERE may be of concern of plan summations are critically dependent on
[16, 46, 49, 50, 58, 59, 61, 124–126]. Of note, the accuracy of the deformation vector fields esti-
prescription dose in MC-based TPS systems mated with image registration, on the assump-
should be volume-based (e.g., to the PTV) as tions of outcome modeling for time-varying dose
opposed to point-based (e.g., Dmax) to avoid a sys- delivery, and on the specific changes in patient
tematic bias that can arise from statistical uncer- anatomy (e.g., tumor response) [131–135]. Note
tainty in point doses [13, 14]. that each step in this process can be achieved
using multiple methods with varying degrees of
precision [136]. Despite the variability in method
Evaluation Metrics implementation, dose summation in MR-Linac
treatments has been successfully reported in mul-
The clinical evaluation metrics used in assessing tiple sites [87, 91, 137–142]. These findings
the quality of conventional plans remain some of highlight the potential for substantial improve-
the most used metrics for also assessing the qual- ments in response/toxicity assessment and adap-
ity of MR-Linac plans. Given the statistical tation strategies. Consensus protocols for dose
uncertainty associated with MC-based calcula- accumulation in MR-guided adaptive radiother-
tion engines, point-based evaluation metrics must apy will provide valuable help to the community
be used with caution, particularly for serial [143].
Quality Assurance algorithm requirements in MR-Linac treatment

planning. This was followed by a detailed discus-
The implementation of adaptive radiotherapy, sion of adaptive planning considerations to
particularly for online and real-time adaptive account for functional and anatomical changes
planning, requires efficient automatic and semi- and concluded with practical considerations for
automatic tools at each step of the process. plan evaluation techniques. The techniques
Given the time constraints with the patient on described in this chapter are applicable to low-
the table, QA for online and real-time planning and high-field MR-Linacs, but there are certain
presents challenges not encountered in conven- benefits and limitations associated with each. For
tional treatment planning methods [7]. As example, high-field MR-Linacs provide a better
shown in Fig. 8.7, while QA for offline adapta- signal-to-noise ratio compared to low-field
tion is identical to conventional planning, QA MR-Linacs which may be beneficial in some
and safety checks for online and real-time anatomical locations for contouring and adaptive
adaptation are integrated and performed in par- treatment planning. However, low-field
allel with the planning process [144, 145]. MR-Linacs have compensated through image
Specific to real-time adaptation, QA and safety reconstruction software to enhance image qual-
checks need to be incorporated in the predictive ity. Similar optimization and delivery techniques
model for robust testing of tracking accuracy, are implemented on both available units, allow-
target/OAR detection, and delivery. ing for comparable plan quality. However, there
Furthermore, IMRT plan verification is tradi- is an increased electron return effect in high-field
tionally based on measurements prior to treat- MRI which requires more considerations when
ment delivery which is practically infeasible for developing dose calculation algorithms and plan-
online and real-time planning. For adaptive ning techniques. Both platforms offer adaptive
planning, calculation-based tools that combine planning capabilities with different workflows,
integrity and independent dose checks become allowing for each individual clinic to choose
the primary verification methods, followed by which system is more suitable to their needs. The
ad hoc measurement-based methods [122, 146– benefits of MRgRT are numerous including
150]. Additional QA and safety checks for the enhanced soft tissue visualization, incorporating
semi-automated steps during plan adaptation functional imaging, adaptive planning capabili-
may be feasible [151, 152]. Failure mode and ties, and real-time motion monitoring utilizing
effects analysis (FMEA) will enable the clinical 2D cine MRI.
team to identify all critical and impactful steps
in the treatment process and has been shown to
provide improvements in safety and efficiency References
[144, 145, 153, 154].
1. Corradini S, Alongi F, Andratschke N, Azria D,
Bohoudi O, Boldrini L, et al. ESTRO-ACROP rec-
ommendations on the clinical implementation of
Summary hybrid MR-linac systems in radiation oncology.
Radiother Oncol. 2021;159:146–54.
A general overview of the requirements of an 2. Kishan AU, Ma TM, Lamb JM, Casado M, Wilhalme
H, Low DA, et al. Magnetic resonance imaging-
MR-Linac treatment planning system, as well as guided vs computed tomography-guided stereo-
the commissioning and QA practices, was tactic body radiotherapy for prostate cancer: the
reviewed. Followed by current contouring initia- MIRAGE randomized clinical trial. JAMA Oncol.
tives and practices regarding offline and online 2023;9(3):365.
3. Rudra S, Jiang N, Rosenberg SA, Olsen JR, Roach
adaptive planning. A practical guide to methods MC, Wan L, et al. Using adaptive magnetic reso-
for estimating electron density was included for nance image-guided radiation therapy for treat-
MR-based planning with an overview of treatment of inoperable pancreatic cancer. Cancer Med.
ment plan optimization, dose calculation, and 2019;8(5):2123–32.
4. Tan H, Stewart J, Ruschin M, Wang MH, Myrehaug 15. Chen X, Paulson ES, Ahunbay E, Sanli A,
S, Tseng CL, et al. Inter-fraction dynamics dur- Klawikowski S, Li XA. Measurement validation of
ing post-operative 5 fraction cavity hypofraction- treatment planning for a MR-Linac. J Appl Clin Med
ated stereotactic radiotherapy with a MR LINAC: Phys. 2019;20(7):28–38.
a prospective serial imaging study. J Neuro-Oncol. 16. Shortall J, Vasquez Osorio E, Aitkenhead A,
2022;156(3):569–77. Berresford J, Agnew J, Budgell G, et al. Experimental
5. Fraass B, Doppke K, Hunt M, Kutcher G, verification the electron return effect around spheri-
Starkschall G, Stern R, et al. American Association cal air cavities for the MR-Linac using Monte Carlo
of Physicists in Medicine Radiation Therapy calculation. Med Phys. 2020;47(6):2506–15.
Committee Task Group 53: quality assurance for 17. Elter A, Rippke C, Johnen W, Mann P, Hellwich E,
clinical radiotherapy treatment planning. Med Phys. Schwahofer A, et al. End-to-end test for fractionated
1998;25(10):1773–829. online adaptive MR-guided radiotherapy using a
6. Siochi RA, Balter P, Bloch CD, Santanam L, Blodgett deformable anthropomorphic pelvis phantom. Phys
K, Curran BH, et al. Report of Task Group 201 of Med Biol. 2021;66(24).
the American Association of Physicists in Medicine: 18. Hoffmans D, Niebuhr N, Bohoudi O, Pfaffenberger
quality management of external beam therapy data A, Palacios M. An end-to-end test for MR-guided
transfer. Med Phys. 2021;48(6):e86–e114. online adaptive radiotherapy. Phys Med Biol.
7. Glide-Hurst CK, Lee P, Yock AD, Olsen JR, Cao M, 2020;65(12):125012.
Siddiqui F, et al. Adaptive radiation therapy (ART) 19. Kisling K, Keiper TD, Branco D, Kim GG, Moore
strategies and technical considerations: a state of the KL, Ray X. Clinical commissioning of an adaptive
ART review from NRG oncology. Int J Radiat Oncol radiotherapy platform: results and recommenda-
Biol Phys. 2021;109(4):1054–75. tions. J Appl Clin Med Phys. 2022;23(12):e13801.
8. IAEA TRS-430, Commissioning and quality assur- 20. Noel CE, Parikh PJ, Spencer CR, Green OL, Hu
ance of computerized planning systems for radiation Y, Mutic S, et al. Comparison of onboard low-field
treatment of cancer. Vienna: International Atomic magnetic resonance imaging versus onboard com-
Energy Agency; 2004. puted tomography for anatomy visualization in
9. Smilowitz JB, Das IJ, Feygelman V, Fraass BA, radiotherapy. Acta Oncol. 2015;54(9):1474–82.
Kry SF, Marshall IR, et al. AAPM medical physics 21. Bauman G, Haider M, Van der Heide UA, Ménard
practice guideline 5.a.: commissioning and QA of C. Boosting imaging defined dominant prostatic
treatment planning dose calculations - megavoltage tumors: a systematic review. Radiother Oncol.
photon and electron beams. J Appl Clin Med Phys. 2013;107(3):274–81.
2015;16(5):14–34. 22. Dunlop A, Mitchell A, Tree A, Barnes H, Bower
10. Geurts MW, Jacqmin DJ, Jones LE, Kry SF, L, Chick J, et al. Daily adaptive radiotherapy for
Mihailidis DN, Ohrt JD, et al. AAPM medical phys- patients with prostate cancer using a high field
ics practice guideline 5.b: commissioning and QA of MR-linac: initial clinical experiences and assess-
treatment planning dose calculations—megavoltage ment of delivered doses compared to a C-arm linac.
photon and electron beams. J Appl Clin Med Phys. Clin Transl Radiat Oncol. 2020;23:35–42.
2022;23(9):e13641. 23. Winkel D, Bol GH, Kroon PS, van Asselen B,
11. Das IJ, Cheng C-W, Watts RJ, Ahnesjö A, Gibbons J, Hackett SS, Werensteijn-Honingh AM, et al.
Li XA, et al. Accelerator beam data commissioning Adaptive radiotherapy: the Elekta Unity MR-linac
equipment and procedures: report of the TG-106 of concept. Clin Transl Radiat Oncol. 2019;18:54–9.
the Therapy Physics Committee of the AAPM. Med 24. Winkel D, Bol GH, Werensteijn-Honingh AM,
Phys. 2008;35(9):4186–215. Intven MPW, Eppinga WSC, Hes J, et al. Target
12. Ezzell GA, Burmeister JW, Dogan N, LoSasso TJ, coverage and dose criteria based evaluation of the
Mechalakos JG, Mihailidis D, et al. IMRT commis- first clinical 1.5T MR-linac SBRT treatments of
sioning: multiple institution planning and dosimetry lymph node oligometastases compared with con-
comparisons, a report from AAPM Task Group 119. ventional CBCT-linac treatment. Radiother Oncol.
Med Phys. 2009;36(11):5359–73. 2020;146:118–25.
13. Ma CMC, Chetty IJ, Deng J, Faddegon B, Jiang SB, 25. Hall WA, Heerkens HD, Paulson ES, Meijer GJ,
Li J, et al. Beam modeling and beam model commis- Kotte AN, Knechtges P, et al. Pancreatic gross tumor
sioning for Monte Carlo dose calculation-based radi- volume contouring on computed tomography (CT)
ation therapy treatment planning: report of AAPM compared with magnetic resonance imaging (MRI):
Task Group 157. Med Phys. 2020;47(1):e1–e18. results of an international contouring conference.
14. Chetty IJ, Curran B, Cygler JE, DeMarco JJ, Ezzell Pract Radiat Oncol. 2018;8(2):107–15.
G, Faddegon BA, et al. Report of the AAPM Task 26. Heerkens HD, Hall WA, Li XA, Knechtges P, Dalah
Group No. 105: issues associated with clinical E, Paulson ES, et al. Recommendations for MRI-
implementation of Monte Carlo-based photon and based contouring of gross tumor volume and organs
electron external beam treatment planning. Med at risk for radiation therapy of pancreatic cancer.
Phys. 2007;34(12):4818–53. Pract Radiat Oncol. 2017;7(2):126–36.
27. Lukovic J, Henke L, Gani C, Kim TK, Stanescu T, reality: current treatment challenges and future per-
Hosni A, et al. MRI-based upper abdominal organs- spectives. Radiat Oncol. 2019;14(1):92.
at-
risk atlas for radiation oncology. Int J Radiat 41. Devic S. MRI simulation for radiotherapy treatment
Oncol Biol Phys. 2020;106(4):743–53. planning. Med Phys. 2012;39(11):6701–11.
28. Vasmel JE, Groot Koerkamp ML, Kirby AM, Russell 42. Das IJ, Sanfilippo NJ, Fogliata A, Cozzi L, Physics
NS, Shaitelman SF, Vesprini D, et al. Consensus on Io. Intensity modulated radiation therapy: a clinical
contouring primary breast tumors on MRI in the set- overview. IOP Publishing; 2020.
ting of neoadjuvant partial breast irradiation in trials. 43. Hawranko R, Sohn JJ, Neiderer K, Bump E, Harris
Pract Radiat Oncol. 2020;10(6):e466–e74. T, Fields EC, et al. Investigation of isotoxic dose
29. Chen X, Ma X, Yan X, Luo F, Yang S, Wang Z, escalation and plan quality with TDABC analysis
et al. Personalized auto-segmentation for magnetic on a 0.35 T MR-Linac (MRL) system in ablative
resonance imaging-guided adaptive radiotherapy of 5-fraction stereotactic magnetic resonance-guided
prostate cancer. Med Phys. 2022;49(8):4971–9. radiation therapy (MRgRT) for primary pancreatic
30. Ding J, Zhang Y, Amjad A, Xu J, Thill D, Li cancer. J Clin Med. 2022;11(9):2584.
XA. Automatic contour refinement for deep learn- 44. Hansen AT, Poulsen PR, Høyer M, Worm
ing auto-segmentation of complex organs in MRI- ES. Isotoxic dose prescription level strategies for
guided adaptive radiation therapy. Adv Radiat stereotactic liver radiotherapy: the price of dose uni-
Oncol. 2022;7(5):100968. formity. Acta Oncol. 2020;59(5):558–64.
31. Kawula M, Hadi I, Nierer L, Vagni M, Cusumano 45. Warren M, Webster G, Ryder D, Rowbottom C,
D, Boldrini L, et al. Patient-specific transfer learning Faivre-Finn C. An isotoxic planning comparison
for auto-segmentation in adaptive 0.35 T MRgRT of study for stage II-III non-small cell lung cancer: is
prostate cancer: a bi-centric evaluation. Med Phys. intensity-modulated radiotherapy the answer? Clin
2023;50:1573. Oncol (R Coll Radiol). 2014;26(8):461–7.
32. Liang Y, Schott D, Zhang Y, Wang Z, Nasief H, 46. Kim A, Lim-Reinders S, McCann C, Ahmad SB,
Paulson E, et al. Auto-segmentation of pancre- Sahgal A, Lee J, et al. Magnetic field dose effects
atic tumor in multi-parametric MRI using deep on different radiation beam geometries for hypofrac-
convolutional neural networks. Radiother Oncol. tionated partial breast irradiation. J Appl Clin Med
2020;145:193–200. Phys. 2017;18(6):62–70.
33. Sharp G, Fritscher KD, Pekar V, Peroni M, 47. Kirkby C, Stanescu T, Fallone BG. Magnetic field
Shusharina N, Veeraraghavan H, et al. Vision 20/20: effects on the energy deposition spectra of MV pho-
perspectives on automated image segmentation for ton radiation. Phys Med Biol. 2009;54(2):243–57.
radiotherapy. Med Phys. 2014;41(5):050902. 48. Wang Y, Mazur TR, Green O, Hu Y, Li H, Rodriguez
34. Eccles CL, Adair Smith G, Bower L, Hafeez S, V, et al. A GPU-accelerated Monte Carlo dose calcu-
Herbert T, Hunt A, et al. Magnetic resonance imag- lation platform and its application toward validating
ing sequence evaluation of an MR Linac system; an MRI-guided radiation therapy beam model. Med
early clinical experience. Tech Innov Patient Support Phys. 2016;43(7):4040.
Radiat Oncol. 2019;12:56–63. 49. Khan AU, Simiele EA, Lotey R, DeWerd LA, Yadav
35. Whiteside L, McDaid L, Hales RB, Rodgers J, Dubec P. Development and evaluation of a GEANT4-based
M, Huddart RA, et al. To see or not to see: evaluation Monte Carlo Model of a 0.35 T MR-guided radia-
of magnetic resonance imaging sequences for use tion therapy (MRgRT) linear accelerator. Med Phys.
in MR Linac-based radiotherapy treatment. J Med 2021;48(4):1967–82.
Imaging Radiat Sci. 2022;53(3):362–73. 50. Ahmad SB, Sarfehnia A, Paudel MR, Kim A,
36. Klüter S. Technical design and concept of a Hissoiny S, Sahgal A, et al. Evaluation of a com-
0.35 T MR-Linac. Clin Transl Radiat Oncol. mercial MRI Linac based Monte Carlo dose cal-
2019;18:98–101. culation algorithm with GEANT4. Med Phys.
37. Heerkens HD, Reerink O, Intven MPW, Hiensch 2016;43(2):894–907.
RR, van den Berg CAT, Crijns SPM, et al. Pancreatic 51. Li Y, Ding S, Wang B, Liu H, Huang X, Song
tumor motion reduction by use of a custom abdomi- T. Extension and validation of a GPU-Monte
nal corset. Phys Imaging Radiat Oncol. 2017;2:7–10. Carlo dose engine gDPM for 1.5 T MR-LINAC
38. Owrangi AM, Greer PB, Glide-Hurst CK. MRI-only online independent dose verification. Med Phys.
treatment planning: benefits and challenges. Phys 2021;48(10):6174–83.
Med Biol. 2018;63(5):05tr1. 52. Raaijmakers AJ, Raaymakers BW, Lagendijk
39. Yang B, Yuan J, Poon DMC, Geng H, Lam WW, JJ. Magnetic-field-induced dose effects in
Cheung KY, et al. Assessment of planning target MR-guided radiotherapy systems: dependence
volume margins in 1.5 T magnetic resonance- on the magnetic field strength. Phys Med Biol.
guided stereotactic body radiation therapy for 2008;53(4):909.
localized prostate cancer. Precision. Radiat Oncol. 53. Raaijmakers AJ, Raaymakers BW, van der Meer
2022;6(2):127–35. S, Lagendijk JJ. Integrating a MRI scanner with a
40. Corradini S, Alongi F, Andratschke N, Belka C, 6 MV radiotherapy accelerator: impact of the sur-
Boldrini L, Cellini F, et al. MR-guidance in clinical face orientation on the entrance and exit dose due
to the transverse magnetic field. Phys Med Biol. 66. Tyran M, Jiang N, Cao M, Raldow A, Lamb JM, Low
2007;52(4):929–39. D, et al. Retrospective evaluation of decision-making
54. Kurz C, Buizza G, Landry G, Kamp F, Rabe M, for pancreatic stereotactic MR-guided adaptive
Paganelli C, et al. Medical physics challenges in radiotherapy. Radiother Oncol. 2018;129(2):319–25.
clinical MR-guided radiotherapy. Radiat Oncol. 67. Green OL, Henke LE, Hugo GD. Practical clinical
2020;15(1):93. workflows for online and offline adaptive radiation
55. Raaijmakers AJ, Raaymakers BW, Lagendijk therapy. Semin Radiat Oncol. 2019;29(3):219–27.
JJ. Integrating a MRI scanner with a 6 MV radiother- 68. Hunt A, Hansen VN, Oelfke U, Nill S, Hafeez
apy accelerator: dose increase at tissue-air interfaces S. Adaptive radiotherapy enabled by MRI guidance.
in a lateral magnetic field due to returning electrons. Clin Oncol. 2018;30(11):711–9.
Phys Med Biol. 2005;50(7):1363–76. 69. Langen KM, Jones DT. Organ motion and its
56. Huang C-Y, Yang B, Lam WW, Geng H, Cheung management. Int J Radiat Oncol Biol Phys.
KY, Yu SK. Magnetic field induced dose effects 2001;50(1):265–78.
in radiation therapy using MR-linacs. Med Phys. 70. Yan D, Lockman D. Organ/patient geometric varia-
2023;50:3623. tion in external beam radiotherapy and its effects.
57. Oborn BM, Metcalfe PE, Butson MJ, Rosenfeld Med Phys. 2001;28(4):593–602.
AB. Monte Carlo characterization of skin doses in 71. Van Herk M, Witte M, Van Der Geer J, Schneider
6 MV transverse field MRI-linac systems: effect of C, Lebesque JV. Biologic and physical fractionation
field size, surface orientation, magnetic field strength, effects of random geometric errors. Int J Radiat
and exit bolus. Med Phys. 2010;37(10):5208–17. Oncol Biol Phys. 2003;57(5):1460–71.
58. Godoy Scripes P, Subashi E, Burleson S, Liang J, 72. Bortfeld T, Jiang SB, Rietzel E. Effects of motion
Romesser P, Crane C, et al. Impact of varying air on the total dose distribution. Semin Radiat Oncol.
cavity on planning dosimetry for rectum patients 2004;14(1):41–51.
treated on a 1.5 T hybrid MR-linac system. J Appl 73. Purdy JA. Current ICRU definitions of volumes: lim-
Clin Med Phys. 2020;21(7):144–52. itations and future directions. Semin Radiat Oncol.
59. Paudel MR, Kim A, Sarfehnia A, Ahmad SB, 2004;14(1):27–40.
Beachey DJ, Sahgal A, et al. Experimental evalua- 74. Schaly B, Kempe JA, Bauman GS, Battista JJ, Van
tion of a GPU-based Monte Carlo dose calculation Dyk J. Tracking the dose distribution in radiation
algorithm in the Monaco treatment planning system. therapy by accounting for variable anatomy. Phys
J Appl Clin Med Phys. 2016;17(6):230–41. Med Biol. 2004;49(5):791–805.
60. Uilkema S, van der Heide U, Sonke JJ, Moreau M, 75. Mutic S, Dempsey JF. The ViewRay system: mag-
van Triest B, Nijkamp J. A 1.5 T transverse magnetic netic resonance–guided and controlled radiotherapy.
field in radiotherapy of rectal cancer: impact on the Semin Radiat Oncol. 2014;24(3):196–9.
dose distribution. Med Phys. 2015;42(12):7182–9. 76. Gupta A, Dunlop A, Mitchell A, McQuaid D, Nill
61. Menten MJ, Fast MF, Nill S, Kamerling CP, S, Barnes H, et al. Online adaptive radiotherapy
McDonald F, Oelfke U. Lung stereotactic body for head and neck cancers on the MR linear accel-
radiotherapy with an MR-linac–quantifying the erator: introducing a novel modified adapt-to-shape
impact of the magnetic field and real-time tumor approach. Clin Transl Radiat Oncol. 2022;32:48–51.
tracking. Radiother Oncol. 2016;119(3):461–6. 77. Heijkoop ST, Langerak TR, Quint S, Bondar L,
62. van Heijst TC, den Hartogh MD, Lagendijk JJ, Mens JWM, Heijmen BJM, et al. Clinical imple-
van den Bongard HJ, van Asselen B. MR-guided mentation of an online adaptive plan-of-the-day
breast radiotherapy: feasibility and magnetic- protocol for nonrigid motion management in locally
field impact on skin dose. Phys Med Biol. advanced cervical cancer IMRT. Int J Radiat Oncol
2013;58(17):5917–30. Biol Phys. 2014;90(3):673–9.
63. Bol G, Hissoiny S, Lagendijk J, Raaymakers 78. Meijer GJ, van der Toorn P-P, Bal M, Schuring D,
B. Fast online Monte Carlo-based IMRT planning Weterings J, de Wildt M. High precision bladder
for the MRI linear accelerator. Phys Med Biol. cancer irradiation by integrating a library planning
2012;57(5):1375. procedure of 6 prospectively generated SIB IMRT
64. McDonald BA, Vedam S, Yang J, Wang J, Castillo plans with image guidance using lipiodol markers.
P, Lee B, et al. Initial feasibility and clinical imple- Radiother Oncol. 2012;105(2):174–9.
mentation of daily MR-guided adaptive head and 79. Vestergaard A, Muren LP, Søndergaard J, Elstrøm
neck cancer radiation therapy on a 1.5 T MR-linac UV, Høyer M, Petersen JB. Adaptive plan selec-
system: prospective R-ideal 2a/2b systematic clini- tion vs. re-optimisation in radiotherapy for bladder
cal evaluation of technical innovation. Int J Radiat cancer: a dose accumulation comparison. Radiother
Oncol Biol Phys. 2021;109(5):1606–18. Oncol. 2013;109(3):457–62.
65. McCulloch MM, Lee C, Rosen BS, Kamp JD, 80. Beekman C, van Triest B, van Beek S, Sonke JJ,
Lockhart CM, Lee JY, et al. Predictive models to Remeijer P. Margin and PTV volume reduction
determine clinically relevant deviations in delivered using a population based library of plans strat-
dose for head and neck cancer. Pract Radiat Oncol. egy for rectal cancer radiotherapy. Med Phys.
2019;9(4):422–31. 2018;45(10):4345–54.
81. Korsholm ME, Waring LW, Edmund JM. A criterion 93. Keall P, Nguyen DT, O’Brien R, Hewson E, Ball H,
for the reliable use of MRI-only radiotherapy. Radiat Poulsen P, et al. Real-time image guided ablative
Oncol. 2014;9(1):16. prostate cancer radiation therapy: results from the
82. Ruggieri R, Rigo M, Naccarato S, Gurrera D, TROG 15.01 SPARK trial. Int J Radiat Oncol Biol
Figlia V, Mazzola R, et al. Adaptive SBRT by 1.5 Phys. 2020;107(3):530–8.
T MR-linac for prostate cancer: on the accuracy of 94. Keall PJ, Sawant A, Berbeco RI, Booth JT, Cho B,
dose delivery in view of the prolonged session time. Cerviño LI, et al. AAPM Task Group 264: the safe
Phys Med. 2020;80:34–41. clinical implementation of MLC tracking in radio-
83. Naccarato S, Rigo M, Pellegrini R, Voet P, Akhiat therapy. Med Phys. 2021;48(5):e44–64.
H, Gurrera D, et al. Automated planning for pros- 95. Uijtewaal P, Borman PT, Woodhead PL, Kontaxis
tate stereotactic body radiation therapy on the 1.5 T C, Hackett SL, Verhoeff J, et al. First experimental
MR-Linac. Adv Radiat Oncol. 2022;7(3):100865. demonstration of VMAT combined with MLC track-
84. Tyagi N, Liang J, Burleson S, Subashi E, Godoy ing for single and multi fraction lung SBRT on an
Scripes P, Tringale KR, et al. Feasibility of ablative MR-linac. Radiother Oncol. 2022;174:149–57.
stereotactic body radiation therapy of pancreas can- 96. NCT03621644. Stereotactic MRI-guided On-table
cer patients on a 1.5 Tesla magnetic resonance-linac Adaptive Radiation Therapy (SMART) for locally
system using abdominal compression. Phys Imaging advanced pancreatic cancer. Stereotactic MRI-
Radiat Oncol. 2021;19:53–9. guided On-table Adaptive Radiation Therapy
85. Brennan VS, Burleson S, Kostrzewa C, Godoy (SMART) for Locally Advanced Pancreatic Cancer.
Scripes P, Subashi E, Zhang Z, et al. SBRT focal 97. Marino MA, Helbich T, Baltzer P, Pinker-Domenig
dose intensification using an MR-Linac adap- K. Multiparametric MRI of the breast: a review. J
tive planning for intermediate-risk prostate can- Magn Reson Imaging. 2018;47(2):301–15.
cer: an analysis of the dosimetric impact of 98. Hoetker AM, Garcia-Aguilar J, Gollub
intra-fractional organ changes. Radiother Oncol. MJ. Multiparametric MRI of rectal cancer in the
2023;179:109441. assessment of response to therapy: a systematic
86. Paulson ES, Ahunbay E, Chen X, Mickevicius review. Dis Colon Rectum. 2014;57(6):790–9.
NJ, Chen G-P, Schultz C, et al. 4D-MRI driven 99. Gürses B, Böge M, Altınmakas E, Balık
MR-guided online adaptive radiotherapy for E. Multiparametric MRI in rectal cancer. Diagn
abdominal stereotactic body radiation therapy on Interv Radiol. 2019;25(3):175.
a high field MR-Linac: implementation and ini- 100. Mahajan A, Engineer R, Chopra S, Mahanshetty
tial clinical experience. Clin Transl Radiat Oncol. U, Juvekar S, Shrivastava S, et al. Role of 3T
2020;23:72–9. multiparametric-MRI with BOLD hypoxia imaging
87. Alam S, Veeraraghavan H, Tringale K, Amoateng for diagnosis and post therapy response evaluation
E, Subashi E, Wu AJ, et al. Inter- and intrafraction of postoperative recurrent cervical cancers. Eur J
motion assessment and accumulated dose quantifi- Radiol Open. 2016;3:22–30.
cation of upper gastrointestinal organs during mag- 101. Clemente EJI, Navarro OM, Navallas M, Ladera
netic resonance-guided ablative radiation therapy E, Torner F, Sunol M, et al. Multiparametric MRI
of pancreas patients. Phys Imaging Radiat Oncol. evaluation of bone sarcomas in children. Insights
2022;21:54–61. Imaging. 2022;13(1):1–18.
88. Fast MF, Nill S, Bedford JL, Oelfke U. Dynamic 102. Necchi A, Bandini M, Calareso G, Raggi D,
tumor tracking using the Elekta Agility MLC. Med Pederzoli F, Farè E, et al. Multiparametric magnetic
Phys. 2014;41(11):111719. resonance imaging as a noninvasive assessment of
89. Glitzner M, Woodhead P, Borman P, Lagendijk tumor response to neoadjuvant pembrolizumab
J, Raaymakers B. MLC-tracking performance in muscle- invasive bladder cancer: preliminary
on the Elekta unity MRI-linac. Phys Med Biol. findings from the PURE-01 study. Eur Urol.
2019;64(15):15NT02. 2020;77(5):636–43.
90. Feldman AM, Modh A, Glide-Hurst C, Chetty IJ, 103. Rossi SH, Prezzi D, Kelly-Morland C, Goh
Movsas B. Real-time magnetic resonance-guided V. Imaging for the diagnosis and response assessment
liver stereotactic body radiation therapy: an institu- of renal tumours. World J Urol. 2018;36:1927–42.
tional report using a magnetic resonance-linac sys- 104. Haider M, Yao X, Loblaw A, Finelli
tem. Cureus. 2019;11(9):e5774. A. Multiparametric magnetic resonance imaging in
91. Akdag O, Borman PTS, Woodhead P, Uijtewaal P, the diagnosis of prostate cancer: a systematic review.
Mandija S, Van Asselen B, et al. First experimental Clin Oncol. 2016;28(9):550–67.
exploration of real-time cardiorespiratory motion 105. Padhani AR, Miles KA. Multiparametric imag-
management for future stereotactic arrhythmia ing of tumor response to therapy. Radiology.
radioablation treatments on the MR-linac. Phys Med 2010;256(2):348–64.
Biol. 2022;67(6). 106. Kooreman ES, van Houdt PJ, Nowee ME, van
92. Zhang Y, Cao Y, Kashani R, Lawrence TS, Balter Pelt VWJ, Tijssen RHN, Paulson ES, et al.
JM. Real-time prediction of stomach motions based Feasibility and accuracy of quantitative imaging
upon gastric contraction and breathing models. Phys on a 1.5 T MR-linear accelerator. Radiother Oncol.
Med Biol. 2022;68(1):015001. 2019;133:156–62.
107. Kooreman ES, van Houdt PJ, Keesman R, Pos FJ, 118. Raaymakers BW, Raaijmakers AJ, Kotte AN, Jette
van Pelt VWJ, Nowee ME, et al. ADC measure- D, Lagendijk JJ. Integrating a MRI scanner with
ments on the Unity MR-linac - a recommendation a 6 MV radiotherapy accelerator: dose deposi-
on behalf of the Elekta Unity MR-linac consortium. tion in a transverse magnetic field. Phys Med Biol.
Radiother Oncol. 2020;153:106–13. 2004;49(17):4109–18.
108. van Houdt PJ, Saeed H, Thorwarth D, Fuller CD, 119. Hissoiny S, Ozell B, Bouchard H, Després
Hall WA, McDonald BA, et al. Integration of quan- P. GPUMCD: a new GPU-oriented Monte Carlo dose
titative imaging biomarkers in clinical trials for calculation platform. Med Phys. 2011;38(2):754–64.
MR-guided radiotherapy: conceptual guidance for 120. Hissoiny S, Raaijmakers AJ, Ozell B, Després
multicentre studies from the MR-Linac Consortium P, Raaymakers BW. Fast dose calculation in
Imaging Biomarker Working Group. Eur J Cancer. magnetic fields with GPUMCD. Phys Med Biol.

1990;2021(153):64–71. 2011;56(16):5119–29.
109. Kooreman ES, van Houdt PJ, Keesman R, van Pelt 121. Wooten HO, Green O, Yang M, DeWees T, Kashani
VWJ, Nowee ME, Pos F, et al. Daily intravoxel inco- R, Olsen J, et al. Quality of intensity modulated radi-
herent motion (IVIM) in prostate cancer patients ation therapy treatment plans using a 60Co magnetic
during MR-guided radiotherapy-a multicenter study. resonance image guidance radiation therapy system.
Front Oncol. 2021;11:705964. Int J Radiat Oncol Biol Phys. 2015;92(4):771–8.
110. Kooreman ES, Tanaka M, Ter Beek LC, Peters FP, 122. Li Y, Wang B, Ding S, Liu H, Liu B, Xia Y, et al.
Marijnen CAM, van der Heide UA, et al. T(1ρ) Feasibility of using a commercial collapsed cone
for radiotherapy treatment response monitoring dose engine for 1.5T MR-LINAC online indepen-
in rectal cancer patients: a pilot study. J Clin Med. dent dose verification. Phys Med. 2020;80:288–96.
2022;11(7):1998. 123. Kry SF, Feygelman V, Balter P, Knöös T, Charlie
111. Kooreman ES, van Pelt V, Nowee ME, Pos F, van Ma C-M, Snyder M, et al. AAPM Task Group
der Heide UA, van Houdt PJ. Longitudinal correla- 329: reference dose specification for dose calcula-
tions between intravoxel incoherent motion (IVIM) tions: dose-to-water or dose-to-muscle? Med Phys.
and dynamic contrast-enhanced (DCE) MRI dur- 2020;47(3):e52–64.
ing radiotherapy in prostate cancer patients. Front 124. Shortall J, Vasquez Osorio E, Chuter R, McWilliam
Oncol. 2022;12:897130. A, Choudhury A, Kirkby K, et al. Assessing
112. Mickevicius NJ, Kim JP, Zhao J, Morris ZS, Hurst localized dosimetric effects due to unplanned
NJ Jr, Glide-Hurst CK. Toward magnetic resonance gas cavities during pelvic MR-guided radiother-
fingerprinting for low-field MR-guided radiation apy using Monte Carlo simulations. Med Phys.
therapy. Med Phys. 2021;48(11):6930–40. 2019;46(12):5807–15.
113. Steenbergen P, Haustermans K, Lerut E, Oyen R, De 125. Xia W, Zhang K, Li M, Tian Y, Men K, Wang J,
Wever L, Van den Bergh L, et al. Prostate tumor delin- et al. Impact of magnetic field on dose distribution
eation using multiparametric magnetic resonance in MR-guided radiotherapy of head and neck cancer.
imaging: inter-observer variability and pathology Front Oncol. 2020;10:1739.
validation. Radiother Oncol. 2015;115(2):186–90. 126. Chen X, Prior P, Chen GP, Schultz CJ, Li XA. Dose
114. Dinh CV, Steenbergen P, Ghobadi G, van der Poel effects of 1.5 T transverse magnetic field on tissue
H, Heijmink SW, de Jong J, et al. Multicenter vali- interfaces in MRI-guided radiotherapy. Med Phys.
dation of prostate tumor localization using multi- 2016;43(8 Part 1):4797–802.
parametric MRI and prior knowledge. Med Phys. 127. Chetty IJ, Rosu M, Kessler ML, Fraass BA, Ten
2017;44(3):949–61. Haken RK, Kong FM, et al. Reporting and analyz-
115. van Schie MA, Steenbergen P, Dinh CV, Ghobadi G, ing statistical uncertainties in Monte Carlo-based
van Houdt PJ, Pos FJ, et al. Repeatability of dose treatment planning. Int J Radiat Oncol Biol Phys.
painting by numbers treatment planning in pros- 2006;65(4):1249–59.
tate cancer radiotherapy based on multiparamet- 128. Brock KK, Mutic S, McNutt TR, Li H, Kessler
ric magnetic resonance imaging. Phys Med Biol. ML. Use of image registration and fusion algorithms
2017;62(14):5575–88. and techniques in radiotherapy: report of the AAPM
116. Subashi E, LoCastro E, Apte A, Zelefsky MJ, Tyagi Radiation Therapy Committee Task Group No. 132.
N. Quantitative relaxometry for target localization Med Phys. 2017;44(7):e43–76.
and response assessment in ultra-hypofractionated 129. Rong Y, Rosu-Bubulac M, Benedict SH, Cui Y, Ruo
MR-guided radiotherapy to the prostate and R, Connell T, et al. Rigid and deformable image
DIL. Int J Radiat Oncol Biol Phys. 2022;114(3, registration for radiation therapy: a self-study evalu-
Supplement):S33. ation guide for NRG oncology clinical trial partici-
117. Kerkmeijer LGW, Groen VH, Pos FJ, Haustermans pation. Pract Radiat Oncol. 2021;11(4):282–98.
K, Monninkhof EM, Smeenk RJ, et al. Focal boost 130. Swaminath A, Massey C, Brierley JD, Dinniwell
to the intraprostatic tumor in external beam radio- R, Wong R, Kim JJ, et al. Accumulated delivered
therapy for patients with localized prostate cancer: dose response of stereotactic body radiation therapy
results from the FLAME randomized phase III trial. for liver metastases. Int J Radiat Oncol Biol Phys.
J Clin Oncol. 2021;39(7):787–96. 2015;93(3):639–48.
131. Jaffray DA, Lindsay PE, Brock KK, Deasy JO, 143. McDonald BA, Zachiu C, Christodouleas J, Naser
Tomé WA. Accurate accumulation of dose for MA, Ruschin M, Sonke JJ, et al. Dose accumulation
improved understanding of radiation effects in nor- for MR-guided adaptive radiotherapy: from practical
mal tissue. Int J Radiat Oncol Biol Phys. 2010;76(3 considerations to state-of-the-art clinical implemen-
Suppl):S135–9. tation. Front Oncol. 2022;12:1086258.
132. Sonke JJ, Aznar M, Rasch C. Adaptive radiother- 144. Noel CE, Santanam L, Parikh PJ, Mutic S. Process-
apy for anatomical changes. Semin Radiat Oncol. based quality management for clinical imple-
2019;29(3):245–57. mentation of adaptive radiotherapy. Med Phys.
133. Chetty IJ, Rosu-Bubulac M. Deformable registra- 2014;41(8Part1):081717.
tion for dose accumulation. Semin Radiat Oncol. 145. Cai B, Green OL, Kashani R, Rodriguez VL, Mutic
2019;29(3):198–208. S, Yang D. A practical implementation of physics
134. Veiga C, Lourenço AM, Mouinuddin S, van Herk M, quality assurance for photon adaptive radiotherapy.
Modat M, Ourselin S, et al. Toward adaptive radio- Z Med Phys. 2018;28(3):211–23.
therapy for head and neck patients: uncertainties in 146. Yang J, Zhang P, Tyagi N, Scripes PG, Subashi E,
dose warping due to the choice of deformable regis- Liang J, et al. Integration of an independent monitor
tration algorithm. Med Phys. 2015;42(2):760–9. unit check for high-magnetic-field MR-guided radia-
135. Heukelom J, Fuller CD. Head and neck cancer adaption therapy system. Front Oncol. 2022;12:747825.
tive radiation therapy (ART): conceptual consid- 147. Chen X, Ahunbay E, Paulson ES, Chen G, Li
erations for the informed clinician. Semin Radiat XA. A daily end-to-end quality assurance work-
Oncol. 2019;29(3):258–73. flow for MR-guided online adaptive radiation
136. Kadoya N, Nakajima Y, Saito M, Miyabe Y, Kurooka therapy on MR-Linac. J Appl Clin Med Phys.
M, Kito S, et al. Multi-institutional validation study 2020;21(1):205–12.
of commercially available deformable image reg- 148. Hackett S, van Asselen B, Feist G, Pencea S, Akhiat
istration software for thoracic images. Int J Radiat H, Wolthaus J, et al. Collapsed cone algorithm can
Oncol Biol Phys. 2016;96(2):422–31. be used for quality assurance for Monaco treatment
137. Zhang Y, Balter JM, Dow JS, Cao Y, Lawrence TS, plans for the MR-Linac. Med Phys. 2016;43(6 Part
Kashani R. Development of an abdominal dose 11):3441.
accumulation tool and assessments of accumulated 149. Lim SB, Godoy Scripes P, Napolitano M, Subashi E,
dose in gastrointestinal organs. Phys Med Biol. Tyagi N, Cervino Arriba L, et al. An investigation of
2023;68. using log-file analysis for automated patient-specific
138. Kontaxis C, Bol G, Lagendijk J, Raaymakers quality assurance in MRgRT. J Appl Clin Med Phys.
B. A new methodology for inter-and intrafraction 2021;22(9):183–8.
plan adaptation for the MR-linac. Phys Med Biol. 150. Peng C, Chen G, Ahunbay E, Wang D, Lawton C,
2015;60(19):7485. Li XA. Validation of an online replanning technique
139. Menten MJ, Mohajer JK, Nilawar R, Bertholet for prostate adaptive radiotherapy. Phys Med Biol.
J, Dunlop A, Pathmanathan AU, et al. Automatic 2011;56(12):3659.
reconstruction of the delivered dose of the day using 151. Altman MB, Kavanaugh JA, Wooten HO, Green OL,
MR-linac treatment log files and online MR imag- DeWees TA, Gay H, et al. A framework for auto-
ing. Radiother Oncol. 2020;145:88–94. mated contour quality assurance in radiation ther-
140. Fast M, van de Schoot A, van de Lindt T, Carbaat C, apy including adaptive techniques. Phys Med Biol.
van der Heide U, Sonke J-J. Tumor trailing for liver 2015;60(13):5199–209.
SBRT on the MR-Linac. Int J Radiat Oncol Biol 152. Chen HC, Tan J, Dolly S, Kavanaugh J, Anastasio
Phys. 2019;103(2):468–78. MA, Low DA, et al. Automated contouring error
141. Lim SY, Tran A, Tran ANK, Sobremonte A, Fuller detection based on supervised geometric attribute
CD, Simmons L, et al. Dose accumulation of daily distribution models for radiation therapy: a general
adaptive plans to decide optimal plan adaptation strategy. Med Phys. 2015;42(2):1048–59.
strategy for head-and-neck patients treated with 153. Huq MS, Fraass BA, Dunscombe PB, Gibbons JP Jr,
MR-Linac. Med Dosim. 2022;47(1):103–9. Ibbott GS, Mundt AJ, et al. The report of Task Group
142. Willigenburg T, van der Velden JM, Zachiu C, 100 of the AAPM: application of risk analysis meth-
Teunissen FR, Lagendijk JJ, Raaymakers BW, et al. ods to radiation therapy quality management. Med
Accumulated bladder wall dose is correlated with Phys. 2016;43(7):4209.
patient-reported acute urinary toxicity in prostate 154. Liang J, Scripes PG, Tyagi N, Subashi E, Wunner
cancer patients treated with stereotactic, daily adap- T, Cote N, et al. Risk analysis of the Unity 1.5 T
tive MR-guided radiotherapy. Radiother Oncol. MR-Linac adapt-to-position workflow. J Appl Clin
2022;171:182–8. Med Phys. 2022;24:e13850.
The First Low-Field MRI-Guided
Radiation Therapy Hybrid
9
Integrated System: MRIdian
Poonam Yadav, Jooyoung (James) Sohn,

and Ahtesham U. Khan
Introduction What Is an MR-Linac?
A significant difficulty in radiation therapy is to MR-Linacs are hybrid machines that combine an
provide high radiation doses to malignant tissue MRI scanner with a radiation delivery device
while reducing the damage to nearby healthy tis- such as a linear accelerator or a radioisotope irra-
sue. Real-time awareness of the tumor’s location diator. Including an imaging modality with the
is necessary to accomplish this, but current radiation delivery machine enables image-guided
X-ray-based imaging systems cannot consis- radiation therapy (IGRT) and adaptive radiother-
tently provide this information. Although mag- apy (ART). IGRT leads to routine visualization
netic resonance imaging (MRI) has the ability to of the patient’s anatomy and allows accurate
produce precise and thorough images of malig- patient positioning based on daily images. ART
nancies, combining MRI with radiation therapy aims to modify the treatment plan based on the
has not proven practical until recently. Several changes in the patient’s anatomy or tumor biol-
developments have resulted in the creation of ogy. Imaging modalities such as computed
MR-Linac systems, which allow for MRI-guided tomography (CT), positron emission tomography
radiation therapy in real time. Originally devel- (PET), or MRI can be employed to implement
oped and put into practice by Lagendijk et al., IGRT and ART. MRI has several advantages over
this groundbreaking technology has now been other imaging modalities that have been described
successfully used for Co-60 beams and linac in Chap. 2 of this book, but mainly including
beams [1, 2]. In this chapter, we will focus on the exquisite soft-tissue contrast, the ability to
ViewRay MRIdian system, which is a state-of- acquire both functional and anatomical informa-
the-art MR-Linac system that provides enhanced tion, and not employing ionizing radiation. Cine
precision and accuracy in radiation therapy MRI images can also be acquired over a long
delivery. period without any adverse health risks. The radi-
ation isocenter coincides with the MRI isocenter
for MR-Linacs and can be used to align the
patient on the couch using pretreatment volumet-
P. Yadav (*) · J. Sohn · A. U. Khan ric images. During the treatment delivery, cine
Department of Radiation Oncology, Northwestern images are usually acquired to monitor the tar-
Chicago, IL, USA
get’s motion especially due to respiratory motion.
e-mail: [email protected]; Beam gating can be employed to accurately
[email protected]; [email protected] deliver the radiation to the desired location.
150 P. Yadav et al.
While MR-Linacs provide many advantages over trons that are responsible for radiation dose) and
conventional treatment delivery systems, the causes them to move along a curved path with a
presence of a strong magnetic field in the vicinity radius of
of the beam delivery system creates unique chal- Etotal β
lenges related to treatment planning, delivery, r= (9.2)
qBc
dosimetry, training, and safety. There are many
MR-Linac systems that have been described in where Etotal is the total energy of the particle, β is
Chapters 10 and 11. This chapter focuses on the relativistic velocity, c is the speed of light, and
details related to the MRIdian 0.35T low-field r is the radius of curvature. The radius of the cur-
MR-Linac. vature is inversely proportional to the magnetic
field strength; therefore, higher magnetic fields
impact the incident particle trajectory more drasti-
Electron Return Effect (ERE) cally than lower fields. This chapter focuses on
MR-Linacs that treat with a photon beam and
High magnetic field strengths are desired in MRI deposit energy via the generation of secondary
to enhance the signal-to-noise ratio (SNR) of the electrons. It is of note that the path of the second-
images. However, any charged particles experi- ary electrons is subjected to the Lorentz force.
encing magnetic fields are also subjected to the The trajectory of the generated electrons is
Lorentz force (F): impacted greatly at high physical density gradient
   interfaces such as tissue-air interfaces. When
F = q v×B ( ) (9.1) crossing an interface with different material com-
where q is the charge, v is the velocity, and B is positions or densities, the circular motion of the
the magnetic field. The Lorentz force impacts the charged particles induces the electron return
trajectory of the charged particles (mainly elec- effect (ERE) illustrated in Fig. 9.1. An enhance-
Fig. 9.1 The impact of a magnetic field on the charged particle trajectories is illustrated
9 The First Low-Field MRI-Guided Radiation Therapy Hybrid Integrated System: MRIdian 151
ment or reduction of the radiation dose can occur, bolus is typically placed on the skin of the patient
relative to a zero magnetic field, depending on the during delivery to reduce the skin dose due to the
magnetic field strength and direction, density gra- ERE [5]. The impact of the magnetic field on
dient, electron energy, and angle of incidence. dose distribution can also be noted in dosimetric
Based on Eq. (9.2), the computed radius of an data such as percent depth dose (PDD) and lateral
electron beam is shown in Fig. 9.2 in terms of the beam profiles. For MR-Linacs with the magnetic
Lorentz force as a function of electron kinetic field orthogonal to the radiation beam, a smaller
energy. For media with lower stopping power, buildup region was observed [6]. The ERE curves
where the range of the secondary electrons is the secondary electrons toward the proximal
large relative to the radius of curvature, a large region of the medium creating a transient charged
impact on the dose distribution can be expected particle equilibrium (TCPE) at shallower depths.
due to the external magnetic field. Further, the An asymmetry was found in the lateral beam pro-
ERE can be significant for thoracic and abdomi- file orthogonal to the magnetic field direction
nal treatment sites, where high-density gradients leading to an increase in penumbra toward the
are often encountered by the radiation beam [3, direction of the Lorentz force [6]. The impact of
4]. The presence of the ERE may increase the the magnetic field on output factors has been
skin dose, especially for the skin near the region reported to be negligible for depths greater than
where the beam exits the patient. Therefore, a the depth of the maximum dose [7].
Fig. 9.2 The radius of curvature of electrons, due to the Lorentz force, as a function of kinetic energy
152 P. Yadav et al.
lanning and Dosimetry Challenges

P treatment MRI images can be co-registered with
for MR-Linacs the planning MRI images, and dose calculation
can be accomplished using the Hounsfield unit
Several challenges arise when utilizing (HU) information from the corresponding CT
MR-Linacs for radiation therapy treatment. images after a deformation map is applied based
Besides considering the conventional problems on the daily image set. This method circumvents
or workflow requirements that are encountered in the usage of daily CT imaging when an adapted
operating an MRI scanner or a linac separately, plan is created for the patient.
the presence of both machines in the same vault The presence of a magnetic field also creates
puts forth additional considerations and dosimetry challenges. The accuracy of the con-
challenges for clinicians. All equipment must be ventional dose calculation algorithms, such as
MR-safe, and the vault must be constructed con- convolution superposition or pencil beam, is
sidering both the impact of external radiofre- inadequate under magnetic fields for clinical use.
quency (RF) noise on image quality and the Therefore, Monte Carlo (MC) methods are
shielding requirements for high-energy photon employed to calculate absorbed dose in the pres-
beams. The personnel working with the machine ence of an external magnetic field. It is also chal-
must be trained on relevant aspects of MRI safety lenging to acquire 3D scans with a dosimeter
and radiation safety. Appropriate signs and warn- since the water tank and the associated motion
ings must be placed outside the vault to indicate stages must be MR-compatible and fit inside the
the use of a magnet and a high-energy radiation bore of the machine. Air-filled ionization cham-
beam. The linac components generate RF noise bers (ICs) are the preferred detectors when per-
that can induce artifacts in the MRI images. forming reference dosimetry in radiotherapy.
Although MRI scanners look similar to CT scan- Due to the low density of the air cavity inside the
ners, the scan acquisition procedure for these IC, the response of the chamber usually has a
modalities differs in various aspects such as dependence on the magnetic field strength and
couch motion, gantry rotation, scanning time, direction [8]. This dependence calls for addi-
and noisiness from the scanner. To achieve high tional correction factors for ICs when measuring
SNR for the acquired images, site-specific absorbed dose in MR-Linacs. Solid-state dosim-
receiver coils might be employed, which may eters typically have a much smaller dependence
intervene with patient comfort and attenuate the on the magnetic field but are usually not employed
incident radiation beam. Additionally, the total as reference detectors for absolute absorbed dose
treatment time on an MR-Linac can differ signifi- measurement [9].
cantly from conventional linacs, especially if
ART is used.
Pretreatment and peri-treatment imaging is System Design of MRIdian
performed using MRI, which typically has poor
spatial integrity due to B-field inhomogeneities The 2014 version of the MRIdian machine from
and nonlinear gradients when compared to X-ray- ViewRay housed three 60Co sources spaced 120°
based modalities. Therefore, monitoring and apart generating a dose rate of 550 cGy/min at
maintaining high spatial fidelity for MR images the isocenter [10]. Utilizing a radioactive mate-
is of paramount importance when utilizing rial that can be exploited by any malevolent par-
MR-Linacs. Since MRI is based on the nuclear ties poses safety concerns and requires repeated
spin properties of the imaged media, information replenishment of the source over time. The
related to electron density is not obtainable from smaller buildup region of this machine due to the
MRI images, which is required for dose calcula- lower energy spectrum of the 60Co source, rela-
tion. However, the initial treatment plan can be tive to megavoltage photon beams, also leads to
made using the CT image set from the CT simu- lower skin sparring. The collimation of the beam
lation data. If ART is implemented, the daily pre- requires three independent sets of multileaf col-
limators (MLCs), which adds complexity to direction (IEC-Y). The linac components such as
beam delivery. the accelerator, magnetron, pulse transformer,
ViewRay transitioned all of the 60Co-based port circulator, and gun driver are mounted on a
machines into linacs over the course of the past gantry ring placed between the two magnet
few years. The latest MR-Linac leverages the halves. The beam-generating components oper-
acceleration of electron beams using S-band ate under a magnetic field strength of <100
microwaves in a standing-wave waveguide. The Gauss, while the linac itself operates under a
linac is capable of producing a 6 MV flattening magnetic field strength of <2 Gauss. Magnetic
filter-free (FFF) photon beam with a propagation fields stronger than 2 Gauss can impact the tra-
direction orthogonal to the direction of both the jectory of the electron beam inside the wave-
MLC motion and the main magnetic field [11]. guides and can disrupt the acceleration process.
The beam is collimated by dual-stacked and dual- To mitigate the impact of the strong magnetic
focused MLC banks that move in an arc to follow field on the linac operation, the beam-generating
the photon beam divergence. The upper and components are placed inside shielding “buck-
lower MLC stacks are capable of full overtravel ets” comprised of five concentric cylindrical fer-
and comprise of leaves with a width of 0.83 cm romagnetic steel in order to create a magnetic
projected at the isocenter. All leaves have a height free zone. The impact of the RF noise, generated
of 5.5 cm and are composed of a 17.7 g/cc non- by the linac components, on MR image quality is
ferromagnetic tungsten alloy. The leaves are reduced by utilizing RF-absorbing carbon fiber
tilted to match the divergence of the photon beam and RF-reflecting copper shields.
in the IEC-Y direction. The upper and lower The bore of the MR-Linac is 70 cm in diame-
stacks house 68 and 70 dynamic leaves, respec- ter, while the maximum imaging field of view
tively, and the stacks are offset by half a leaf (FOV) is 54 cm. To facilitate patient positioning
width to increase the collimation resolution to using MRI images, the scanner and the linac
0.415 cm in the IEC-Y direction while also miti- share a common isocenter. The RF excitation B1
gating the tongue and groove effect. The mini- transmit pulse is generated using a 16-rung
mum and maximum field sizes are 0.2 × 0.415 cm2 quadrature birdcage coil with a 75 cm diameter.
and 27.4 × 24.1 cm2, respectively, projected at the The receiver signal can be collected using the
isocenter with a source-to-axial (SAD) distance body coil or site-specific coils if a higher SNR or
of 90 cm. The MR-Linac has a nominal dose rate parallel image acquisition is needed. The receiver
of 600 MU/min. The couch for this linac has a coils use radiolucent phased arrays placed inside
uniform thickness and is only capable of transla- low-density foam. The torso and the head and
tional motion limiting the delivery to coplanar neck (H&N) receiver coils contain 2 × 6 and
radiation beams. Schematic of linac and its sub- 2 × 5 channels, respectively, one for anterior and
components is shown in Fig. 9.3. posterior each. The gradient coils can generate
The MRIdian system allows 3D conformal strengths of 18 mT/m with a maximum slew rate
radiation therapy (3DCRT), also referred to as of 200T/m/s.
fixed conformal, and step-and-shoot intensity- The Bo field homogeneity is improved using
modulated radiation therapy (IMRT), referred to five higher-order superconducting shims as well
as optimized conformal. A combination of IMRT as passive and gradient shimming. The “tune-up”
and 3DCRT can also be used in some cases. Since mode for image acquisition uses a constant or
the beam attenuation through the couch can be zero gradient offset across all gantry angles
significant of up to 20%, gantry angles that shoot leading to varying Bo homogeneity depending on
the beam through the edges of the couch should the gantry angle. The “standard shim” mode
be avoided. reduces inhomogeneities by calculating gradient
The MRIdian MR-Linac uses split-bore super- offsets using a pre-scan acquisition. This method
conducting magnets that generate a uniform maximizes Bo homogeneity at the expense of
0.345T magnetic field along the superior-inferior increased scan time. The “MRI-GO” mode
154 P. Yadav et al.
Gantry and RT
Components - Redesigned
MRIS (Unchanged)
Patient Handling
System (unchanged)
Magnetron
Circulator
Rotating Shim
Tray Fixture
Accelerator
Pulse
Transformer
MLC
Gun Driver
Fig. 9.3 The schematic of the MRIdian MR-Linac (top). The linac subcomponents inside the shielding buckets (bot-
tom). Figure courtesy of ViewRay, Incorporated
tweaks the gradient current as a function of gan- rents depending on the magnitude. Peripheral
try angles based on homogeneity measurements nerve and muscle stimulation is also possible due
acquired with a spherical phantom. This mode to Eddy currents requiring a threshold ceiling on
reduces Bo inhomogeneity, albeit not as well as the rate of change of the gradient fields. The rap-
the standard shim mode, without significantly idly changing gradients also produce strong
increasing the scan time. acoustic noise, which can disrupt patient comfort
A localization scan can be performed within and may pose a concern to patients with audito-
15 s with an SNR of >30 to register the patient or rial complications. The RF energy transmitted to
a phantom. True Fast Imaging with Steady State the patient tissue to excite the magnetization can
Precession (TRUFI) pulse sequence is the default act as a localized heating source. The intensity of
acquisition sequence for both volumetric and the heating is proportional to the electrical con-
cine imaging utilizing a mixed T1 and T2 weighted ductivity of the tissue, patient size, and square of
contrast. Scans with T1 contrast can be acquired the electric field induced by the RF pulse.
using either a 3D gradient-echo pulse sequence Several measures can be taken to prevent or
or a turbo flash (TFL) sequence. The latter was reduce the hazardous effects produced by the
developed for thorax and abdomen scans and sig- MRI scanner. Since the magnetic field strength
nificantly reduces motion artifacts. Other pulse drops rapidly as a function of distance from the
sequences, such as turbo spin echo (TSE) and magnets, multiple zones can be created based on
diffusion-weighted imaging (DWI), are also the field strength to isolate regions of high mag-
available for image acquisition. The volumetric netic fields. The zones can then be categorized,
imaging can be acquired with a slice thickness of and zone-specific protocols can be used to restrict
1.5 mm or 3 mm with an in-plane resolution of or control different regions depending on the
1.5x1.5 mm2. Cine imaging has a lower spatial posed concerns. Figure 9.4 describes different
resolution of 3.5 × 3.5 mm2 or 2.4 × 2.4 mm2 with MRI zones recommended for the MRIdian sys-
slice thickness options of 5 mm, 7 mm, or 10 mm. tem. For diagnostic scanners, it is important to
Real-time 2D multi-planar imaging can be per- visually indicate the 5 Gauss line inside the treat-
formed with FOV options of 27 × 27 cm2, ment vault. The MRIdian system specifies the 50
35 × 35 cm2, and 40 × 40 cm2, with acquisition Gauss line to be near the foot of the couch. A
rates of 4 FPS or 8 FPS. visual indicator of the high magnetic field region
also aids in instruction to personnel regarding the
storage of equipment related to radiotherapy,
MR Safety for MRIdian such as immobilization or quality assurance (QA)
devices, that can generate RF noise or can be
It is pertinent that MR safety guidelines from rel- impacted by the magnetic field. Prior to the treat-
evant organizations, such as the American ment, the patient’s medical data must be reviewed
College of Radiology (ACR), are followed to to check for any implanted medical devices or
ensure safe operation of the MR-Linac [12]. The other ferromagnetic material inside the patient’s
patients can be exposed to several possible haz- body. Any questionable devices must be
ards from the operation of the MRI scanner in the researched thoroughly by the personnel, and
treatment room. The presence of a magnetic field strict screening protocols must be placed to only
can induce translational and rotational forces on allow equipment that is considered “MR safe.”
ferromagnetic objects leading to hazards related Besides reviewing medical records, it is good
to projectile damage. The rapid switching of the practice to employ a ferromagnetic detection
gradient fields generates Eddy currents in tissue system before each treatment fraction to ensure
depending on the rate of change of the field, the safe usage of the MRI.
electrical conductivity of the media, and the The heating effects, especially hotspots, due
cross-sectional area of the tissue. Heating or to Eddy currents and RF pulses can be reduced
burning effects might be induced by Eddy cur- by avoiding contact of conductors with the
156
Fig. 9.4 A safety protocol with a process map for ViewRay MRIdian. The colors of each box indicate the MRI zoning for that step
P. Yadav et al.
patient’s skin. Such a scenario can occur when ogy, or boost their academic status. An MR-Linac
using immobilization or other accessory devices vendor might reach out to a clinic with a proposal
for treatment or scanning. Compliance with the for implementing this technology at the clinic.
US Food and Drug Administration (FDA) or There are multiple MR-Linac vendors with
other relevant authorities must be maintained to distinct linacs and MRI scanning systems. Once
ensure that the rate of change of the gradient the decision of acquiring an MR-Linac is made,
fields is below the threshold of palpable periph- presentations and demos are provided by the
eral nerve stimulation and that the acoustic vendors to summarize the benefits and costs
noise levels are safe for each patient. The spe- associated with each system. Site visits are
cific absorption rate (SAR), defined as the scheduled by the vendors to evaluate vault
energy dissipated in tissue per unit mass, due to requirements for their MR-Linac, and any addi-
the RF pulse must be well below the established tional construction costs are added to the budget
thresholds and should be monitored for the uti- proposal. It is of note that the MRIdian MR-Linac
lized pulse sequence. The International can fit inside a conventional vault. The vendor
Electrotechnical Commission (IEC) limits the might facilitate site visits to existing users of
average SAR to 4 W/kg for the whole body and their MR-Linac for a Q&A session. The selec-
3.2 W/kg for the head. For the MRIdian system, tion criteria to choose a vendor can be dependent
the SAR value is typically <1.2 W/kg and there- on the (1) disease sites that will be treated with
fore is well below the specified tolerances. the machine, (2) clinical trial data showing out-
Besides health concerns, it is important to comes for patients treated with the MR-Linac,
ensure that the equipment inside the room has (3) initial cost related to the MR-Linac acquisi-
minimal impact on MRI image quality, noise, tion, (4) operational costs and staffing require-
and the presence of artifacts. ments to maintain the MR-Linac program, (5)
cost and maintenance of the auxiliary equipment
required, (6) potential research grants and proj-
Logistics and Economics ects that can be acquired, and (7) treatment time
of Acquiring an MR-Linac or physician time required at the treatment con-
sole. Following the purchasing order, it can take
Frequently, the purchase of an MR-Linac is con- 1–2 years until a patient is treated with the
sidered when an existing linac needs replacing. acquired MR-Linac.
Otherwise, the decision to instantiate an
MR-Linac program can be based on gaining stra-
tegic advantage by including advanced tech- Teams and Training for MRIdian
niques such as image-guided radiation therapy
(IGRT) and adaptive radiotherapy (ART). Some ViewRay Training
clinics decide to acquire MR-Linacs to expand on
the disease sites or the number of patients that ViewRay offers on-site and remote training pro-
can be treated at their clinic. Evaluation might be grams for MRIdian. The on-site training program
performed by the clinics to consider their market involves hands-on training sessions conducted by
share of the potential patients that require treat- trained staff members, covering MR simulation,
ment using MR-Linacs. Increase in patient load, data transferring, treatment planning, and treat-
financial reimbursement for these treatments, and ment procedures. Virtual training includes online
return on investment (ROI) data play an impor- courses and webinars on various aspects of
tant role when an acquisition of MR-Linacs is MRIdian, including system operation, clinical
considered. Often, physicians or physicists desire applications, maintenance, and QA. The training
to have an MR-Linac at their clinic to pursue programs are designed to ensure that the staff
ART. Academic centers might consider members possess the knowledge and skills nec-
MR-Linacs to increase their research output, gain essary to efficiently operate and maintain the
access to research grants related to this technol- MRIdian systems.
158 P. Yadav et al.
In-House Training for Physicians, tency evaluation, five cases should be selected,
Physicists, Therapists, including the case that will be used for treatment.
and Dosimetrists Members of each team, including radiation
oncologists, medical physicists, radiation thera-
The following topics can be covered in in-house pists, and dosimetrists, need to be involved in the
courses and hands-on training sessions for physi- training program to ensure that everyone under-
cians, physicists, therapists, and dosimetrists for stands their role in the workflow process.
ViewRay MRIdian: To start the program, various team members
should work together to generate checklists that
• MRI physics. cover each stage of the workflow process. The
• CT vs. MRI anatomy. following checklist may be helpful:
• MRI safety.
• Dose calculation under magnetic fields. 1. Simulation checklist: scanning with a low
• Appropriate beam arrangement for different resolution for the setup verification and field
treatment sites. of view, choosing a proper slice for tracking
• Beam angles to avoid. structure.
• Optimization techniques. 2. Planning checklist: correct site of treatment,
• Adaptive treatment workflow. appropriate contours, and optimization.
• ViewRay MRIdian hardware and software. 3. Adaptive checklist: checking the rules, elec-
• Quality assurance (patient-specific QA, peri- tron density, density override, plan review,
odic QA, etc.) and tracking algorithm.
• Troubleshooting and maintenance of the
device (support package, etc.) A flowchart describing (example Fig. 9.5
shows the workflow of ViewRay MRIdian in
terms of process map including an adaptive treat-
Practicing Contouring and Workflow ment workflow) the clinical workflow for the pro-
gram should be developed. The flowchart should
The contouring training session can be set up include information about each stage of the pro-
using a single box (simulation box provided by cess, including imaging, contouring, planning,
ViewRay) or MIM. The simulation box is an inte- and treatment delivery. It should also identify the
grated system that provides MR simulation and individuals responsible for each task and the
on-table adaptation. MIM is a comprehensive timeline for completing each task. The workflow
image analysis software that provides advanced practice should involve different group members
tools for contouring. Contouring practice should to ensure that everyone understands their role in
involve a hands-on approach, where each pro- the process. Note some tasks involve different
gram member should be given the opportunity to group members working together. During the
contour organs at risk for different clinical sites. workflow practice, each group member should be
The staff members should receive feedback given the opportunity to practice their respective
from their peers and the team leaders. For compe- tasks and receive feedback from their peers.
Fig. 9.5 Process map of ViewRay MRIdian workflow. The colors of each flowchart box indicate the responsible staff
member for that step. Multiple colors indicate that multiple staff may be responsible for completing that step
MRIdian Acceptance, tems remains intact. Shielding calculations are

Commissioning, and QAs performed prior to the installation to ensure
that the radiation exposures outside the vault
The MRIdian MR-Linac is a complex machine are below the National Council on Radiation
with multiple components that work in synergy Protection and Measurements (NCRP) speci-
to treat patients. It is crucial to ensure that all fied tolerances. RF penetration through the
individual subsystems are working as intended walls is also evaluated to ensure adequate RF
and the communication between the subsys- shielding inside the vault. Following the con-
160 P. Yadav et al.
struction of the treatment vault and installation The acceptance testing for the MRI scanner
of the machine, functional and acceptance test- can be performed using the equipment provided
ing must be performed by the team before any by the vendor and is divided into four distinct
patients are treated. Commissioning of the phases: (1) Bo field homogeneity, (2) SNR and
treatment planning system (TPS) and QA uniformity, (3) ACR phantom, and (4) spatial
devices is also performed at this stage. Baseline integrity.
data are also acquired to monitor the perfor-
mance of the MR-Linac over time. There is a Bo Homogeneity
paucity of protocols and guidelines related to The Bo field homogeneity can be evaluated using
the commissioning and QA of MR-Linacs. a 24 cm diameter of spherical volume (DSV)
Therefore, the following sections describe phantom, shown in Fig. 9.6a), aligned with the
either the vendor-recommended testing proto- isocenter. A localizer scan is used to align any
cols or guidelines provided by clinical physi- equipment placed on the couch. The spectral
cists based on their experience working with peak method is used to test field homogeneity,
the MRIdian MR-Linac. and a free induction decay (FID) signal is
obtained. The signal in the spectral domain is uti-
lized to measure the field homogeneity by calcu-
Acceptance Testing for MRI lating the full width at half maximum (FWHM)
of the signal profile. The Bo field homogeneity is
Once the installation is complete and the calculated by
MR-Linac is handed to the clinical team, the rel- Δf
evant team members should familiarize them- ppm =
fo
selves with the hardware components and any
connections between them as well as software where ∆f is the FWHM of the profile and fo is
interfaces. It’s also best to be aware of the acces- the central frequency. The tune-up mode is typi-
sory components and their place of storage. cally utilized for these measurements, and the
Before the acceptance testing protocol is homogeneity is evaluated over gantry angles
launched, it’s important to visually inspect the from 30° to 330° in 30° intervals. Since each of
machine and the accessory components for no the six buckets mounted on the gantry contains a
physical damage. different linac subcomponent, the Bo homogene-
a b c
Fig. 9.6 (a) The spherical phantom used for Bo field The spatial integrity phantom utilized in this work placed
homogeneity and SNR/uniformity testing. (b) The ACR in the axial orientation
phantom placed inside the 0.35sT MR-Linac is shown. (c)
ity is distinct at each gantry angle. The gantry where σ is the standard deviation of the ROI
angle with the greatest homogeneity is usually on the noise image and Smean, Smin, and Smax are the
selected for all pretreatment scans. Using the mean, minimum, and maximum on the signal
tune-up mode, the homogeneity is typically <5 ROI, respectively. The manufacturer-specified
parts per million (PPM) over all gantry angles as SNR thresholds are 12 for the body coil and 30
shown in Fig. 9.7. However, better homogeneity for the torso and H&N coils. These thresholds are
can be achieved using the standard shim mode or even greater for the SNRs measured with the
the MRI-GO mode. individual coil elements. SNRs for diagnostic
scanners with higher Bo field strengths are typi-
Receiver Coil SNR and Uniformity cally greater than these values. The vendor speci-
For all of the receiver coils, SNR and uniformity fies the uniformity thresholds to be 50% for the
are evaluated using the same spherical phantom image acquired with the body coil and 60% for
that was used for field homogeneity testing. The the images acquired with the torso and H&N
acquisition parameters are provided by the manu- coils. The SNR and uniformity for the MRIdian
facturer for this test and are listed in Table 9.1. machines are typically well above the specified
For the site-specific receiver coils such as the thresholds [13].
torso and H&N coils, the SNR is measured for
each coil element in addition to the cumulative
signal. The SNR and uniformity are evaluated in Table 9.1 Acquisition parameters for the SNR and uni-
all three anatomical planes. Region of interests formity measurements
(ROIs) are placed on the scanned images, and the Parameter Value
SNR/uniformity metrics are calculated by Repetition time 1500 ms
(TR)
0.66 x S mean Echo time (TE) 15 ms
SNR =
σ Resolution 1.2 × 1.2 mm2 in-plane and 10 mm
slice thickness
⎡ S - S min ┐ Flip angle 90° for signal image and 0° for noise
Uniformity = 100 x 1 - | max |
⎣ S max + S min ⎦ images
6
Bo homogeneity (ppm)
0 30 60 90 120 150 180 210 240 270 300 330

Gantry angle ( )
Fig. 9.7 Multi-institutional data on Bo field homogeneity as a function of gantry angle (adapted from Khan et al. [13])
162 P. Yadav et al.
ACR Phantom Tests lyze and calculate the ACR metrics following the
The ACR phantom comprehensively evaluates scan to improve the speed of the evaluation [14].
the image quality and positional accuracy of the Table 9.3 compiles ACR phantom results for a
MRI scanner using a single phantom. The phan- multi-institutional study [13].
tom itself has several fixtures and inserts that are
used to measure geometrical accuracy, slice posi- Spatial Integrity
tional and thickness accuracy, high contrast spa- Spatial integrity is crucial in MR-guided radio-
tial resolution, low-contrast object detectability, therapy (MRgRT) to accurately localize the tar-
image uniformity, and percent ghosting. For this get using the MR images. ViewRay offers a
test, the phantom is aligned using a localizer scan spatial integrity acrylic phantom, shown in
and imaged with the torso coil. Both T1 and T2 Fig. 9.6c, consisting of cylinders filled with
images are acquired with the parameters listed in signal-generating media, such as CuSO4 or min-
Table 9.2. The ACR specifies thresholds for each eral oil, forming a 20 × 20 grid with a spacing of
of the tests. The geometrical accuracy, deter- 1.45 cm. The location of the individual cylinders
mined by the difference in the known dimensions is known and verified using a CT scan. The
of the phantom and the measured dimensions on Magphan® RT phantom can be utilized as an
the images, must be greater than ±3 mm. The alternative to the ViewRay phantom to perform
slice positional and thickness accuracy is mea- this test. The spatial fidelity of each of the clinical
sured using two 45° wedges and must be within pulse sequences is verified by imaging the phan-
5.0 mm and 0.7 mm, respectively, of the known tom in all three anatomical planes and calculating
values. The high contrast spatial resolution the offsets between the actual and imaged loca-
threshold is 1 mm and is measured using hole tions of the cylinders. Spatial accuracy is also
arrays of different sizes ranging from 0.9 to evaluated in the sagittal planes with offsets of
1.1 mm in 0.1 mm increments. Several objects of ±7 cm and ±12.5 cm from the isocenter since this
varying contrast and sizes are embedded inside anatomical plane is used for real-time target
the ACR phantom that can be used to test the low- tracking. The vendor-provided software is used
contrast object detectability of the scanner. The to calculate the percentage of cylinders located
total number of visible spokes on the two slices accurately by the MR images in a 10 cm and a
must be greater than 9 to meet the ACR 17.5 cm radius ROI. Figure 9.8 displays spatial
specifications for this test. A large ROI on a slice
consisting of uniform media is used to measure
the uniformity of the images, which has a specifi- Table 9.3 The mean values of the multi-institutional
ACR measurements along with the ACR specifications
cation of 87.5%. Four background ROIs are
ACR
placed on the peripheral regions of the uniform
ACR test Mean specification
ROI to measure the percent ghosting effect. The Slice positional
intensity of the ghosting must be <2.50% to meet accuracy (mm)
the ACR criteria. ViewRay offers software to ana- Slice 1 0.97 (T1), <5.00
1.03 (T2)
Slice 11 1.08 (T1), <5.00
1.14 (T2)
Table 9.2 Acquisition parameters for the ACR phantom
Slice thickness 5.29 (T1), 5.0 ± 0.7
tests
accuracy (mm) 5.26 (T2)
Parameter T1 value T2 value High contrast spatial 1.00 (T1), ≤1.00
Repetition time 500 ms 2000 ms resolution (mm) 1.00 (T2)
(TR) Low-contrast object 31 (T1), 22 >9
Echo time (TE) 20 ms 20 ms detectability (T2)
Number of 25 9 Image uniformity (%) 97.55 (T1), >87.5
averages 97.79 (T2)
Resolution 0.98 × 0.98 mm2 0.98 × 0.98 mm2 Percentage ghosting 1.47 (T1), <2.50
Flip angle 90° 90° (%) 0.06 (T2)
Coronal Sagittal
3 3
2.5 2.5
Maximum offset (mm)

Maximum offset (mm)
2 2
1.5 1.5
1 1
0.5 0.5
0 0
10 30 50 70 90 110 130 150 170 190 10 30 50 70 90 110 130 150 170 190
Axial
3
2.5
Maximum offset (mm)
1.5
0.5
0
10 30 50 70 90 110 130 150 170 190
Off-axis distance (mm)
Fig. 9.8 The spatial integrity results, in the form of maximum offset between the actual and measured locations of the
cylinders, as a function of off-axis distance from the isocenter (adapted from Khan et al. [13])
fidelity results for a multi-institutional study. The model. Therefore, the acceptance testing protocol
spatial accuracy typically decreases as the dis- requires a comparison between the measured and
tance from the isocenter increases due to the non- the TPS-planned absorbed dose distributions to
linearity of the gradients. Therefore, the passing ensure a good match between the two systems.
criteria for accurately locating a cylinder are The beam characteristics of the MRIdian
1 mm and 2 mm for the 10 cm and the 17.5 cm MR-Linacs were previously found to be similar
radii ROIs, respectively. in a multi-institutional study [16].
Termination of Radiation
Acceptance Testing for Linac The radiation dose to deliver in linacs is quanti-
fied using charge accumulated inside the sensi-
Before initiating the acceptance testing for the tive volume of the monitor chamber defined as
linac, it is well worth it for the clinicians to famil- monitor units (MUs). Prior to reference and rela-
iarize themselves with the software and the pro- tive dosimetry, the reliability of the monitor
vided accessories. The vendor makes a chambers inside the MR-Linac head must be
recommendation for all of the equipment required evaluated. This test verifies the ability of the sec-
for this testing protocol. However, it is the ondary dose monitoring system to terminate the
responsibility of clinical physicists to understand beam based on the planned MUs. Various planned
the reasoning behind the recommendation and MUs are considered, and the delivered MUs are
consequences of using alternative equipment. It recorded to calculate the deviations between the
must also be ensured that all equipment is planned and delivered MUs. According to the
MR-safe and compatible. The vendor also pro- vendor-provided specifications, the radiation
vides specifications for each of the tests per- beam must terminate when the planned MUs
formed at this stage. Most of these criteria are exceed by
based on the AAPM TG-142 [15]. Unlike con-
ventional linacs where TPS beam modeling is ⎛| 1.10 MU planned + 1MU , MU planned < 10
performed by the clinical physicists using the {
acquired experimental data, the MRIdian vendor |⎝min {1.10 MU planned ,25MU } , MU planned 10
tunes the photon beam to match the TPS beam
164 P. Yadav et al.
Laser Alignment is aligned using the lasers. The vendor recom-

It is important to ensure that the lasers are cor- mends the use of the RIT® film analysis software
rectly aligned and isocenter coincidence is main- to calculate the size of the isocenter. The radius
tained. For this test, a laser alignment device, of the smallest tangent circle must be within
such as the ViewRay daily QA phantom or a 1.0 mm for the film embedded inside the phan-
detector array, must be used to align the lasers tom, and the deviation of the beam center along
accurately without any translational or rotational the IEC-Y direction for the film wrapped around
shifts along all three axes. It is of note that the the phantom must be within 1.0 mm of the center
laser isocenter is offset −155 cm along the IEC-Y value.
axis due to the presence of the bore. The lateral
lasers must be checked to ensure a tilt or shift of oincidence of Radiation, Imaging,
C
<1 mm across the couch width. Once the lasers and Laser Isocenters
are aligned, the laser coincidence can be visually Although the isocenters for the imaging system
evaluated using a thin semitransparent sheet of and the linac coincide nominally, an offset
paper and must be within 1 mm. between the two isocenters is possible but must
be within 1 mm according to the AAPM TG-142
Gantry Isocenter Tolerance and TG-179 reports [15, 17]. Since patients or
Due to the mechanical tolerances on the gantry phantoms can be aligned on the couch using the
rotation, the isocenter is typically an ellipsoid or lasers or the MRI images, it is crucial that the
a sphere rather than a point. AAPM TG-142 rec- laser, the MRI scanner, and the linac isocenters
ommends that the isocenter for a linac is within are coincident within 1 mm of each other. For the
1.0 mm considering radiation beams from all radiation isocenter measurement described above
gantry angles. The size of the isocenter can be using radiochromic films, the ViewRay QA phan-
evaluated using a star shot plan. The vendor rec- tom can be aligned on the couch using the lasers,
ommends using radiochromic films placed inside and the irradiated film can be marked to indicate
the ViewRay QA phantom provided by the manu- the phantom crosshairs. Therefore, the offset
facturer. The phantom is made of plastic and between the lasers and the radiation isocenters
houses a cylindrical volume filled with distilled can be measured using the same film analysis
water. The cylindrical volume contains a central software. The embedded film inside the water
insert for an ion chamber as well as several phantom can be used to determine the offsets in
peripheral inserts for additional chambers. The the IEC-X and IEC-Z directions, while the
isocenter size in the IEC-XZ plane can be deter- wrapped film can be used to calculate the offset
mined by irradiating a piece of circular film from in the IEC-Y direction. The offsets between the
various gantry angles embedded inside the cylin- MRI and the laser isocenters can be measured by
drical volume and calculating the radius of the aligning the ViewRay QA phantom using the
smallest tangent circle. The vendor suggests uti- lasers and acquiring MRI images of the phantom.
lizing a treatment plan consisting of five radiation The acquired images can be co-registered with
beams with a field size of 1.0 × 6.3 cm2 (IEC-X the planning images, and the offset between the
by IEC-Y) with gantry intervals of 72°. Since the lasers and the MRI isocenters can be calculated
photon beam generated by the MRIdian linac is in all three directions. The radiation, laser, and
FFF, the center of the beam in the IEC-Y direc- MRI isocenters are considered coincident if the
tion can be determined by the peak of the beam offset between the three is within 1 mm.
profile. Therefore, the variation in the beam’s
center as a function of gantry angle can be mea- ose and Timer Linearity
D
sured by wrapping a sheet of film around the The delivered absorbed dose must be propor-
cylindrical volume. Throughout this measure- tional to the planned MUs. Any deviations from
ment, the film pieces are marked based on the this linearity must be within 2% for prescriptions
crosshairs and scribes on the QA phantom, which over 5 MUs and within 5% for prescriptions
under 5 MUs. This relationship can be evaluated measurement to determine the PDD at 10 cm
by irradiating a dosimeter with various MUs and depth with an SSD of 100 cm due to a smaller
ensuring a constant response when normalized SAD of the MRIdian machine compared to con-
by the delivered MUs. It is crucial to employ a ventional linacs. Using the corrected beam qual-
dosimeter with a linear response over a large ity, the kQBo,Qo correction factor can be extracted. It
range of absorbed dose. The vendor recommends is currently assumed that the impact of a 0.35T
using an Exradin A28 MR safe ionization cham- magnetic field on the response of the A28 cham-
ber inside the water-filled cylindrical volume of ber is minimal and can be considered negligible
the ViewRay QA phantom. The constancy of the when the chamber is oriented along the direction
charge accumulated by the ion chamber of the magnetic field.
normalized by MUs must be verified over a range
of MUs.
Percent Depth Dose
Reference Dosimetry According to the vendor specifications, the PDD

at 10 cm and the depth of the maximum dose
No official guidelines currently exist for refer- must be within 1.0% and 1.5 mm of the planned
ence dosimetry under the influence of strong dose. These metrics are already measured when
magnetic fields. Reference class ion chambers performing reference dosimetry to determine the
remain absolute dosimeters for absorbed dose beam quality and can be compared with the plan-
measurements in radiotherapy. The absolute ning system. PDD for three field size is shown in
absorbed dose to water, Dw, for photon beams can Fig. 9.9. The vendor also recommends evaluating
be measured by the stability of the monitor chamber. This can be
achieved by comparing the output of the radia-
Dw = M Q N D , w kQBo,Qo
tion beam in a water tank over a long period. The
where MQ is the measured charge for a set num- output must be within 1% of the initial value.
ber of MUs, ND, w is the chamber-specific calibra-
tion coefficient typically measured in 60Co beams,
and kQBo,Qo corrects for the difference in the cham- Couch Transmission
ber response between the calibration beam qual-
ity and the MR-Linac beam quality in the For gantry angles between 140° and 220°, the
presence of a magnetic field [8]. Before the abso- photon beam traverses the couch before reaching
lute absorbed dose is measured, the clinical beam the patient. Therefore, the attenuation through
quality must be determined to calculate the kQBo,Qo the couch must be quantified and compared with
correction factor for the utilized ion chamber. the planned value. The relative attenuation can be
The reference dosimetry for the MR-Linac must measured by irradiating an ion chamber, placed
be performed in a water phantom. MR-safe 1D inside the ViewRay QA phantom, from various
water tanks (PTW, CNMC, etc.) and 3D tanks gantry angles using the same MUs. The charge
(PTW, LAP, etc.) are available for dosimetry. The readings as a function of gantry angle can then be
vendor recommends using an Exradin A28 ion normalized by the charge reading acquired with
chamber for reference dosimetry. After aligning the unattenuated beam. Couch attenuation for the
the chamber on the surface of the water tank visu- MRIdian MR-Linac can be up to 15% and must
ally by partially submerging it until a perfect cir- be considered during treatment planning. The
cle is formed, the beam quality is determined by vendor requires the measured and the planned
measuring the percent depth dose (PDD) at 10 cm couch attenuation values to be within 3% of each
depth with a source-to-surface distance (SSD) of other. Using the same irradiation setup, the out-
80 cm and a field size of 10.04 × 9.96 cm2. A put constancy as a function of gantry angle can be
Mayneord F factor is applied to the beam quality measured. For gantry angles not obstructed by
166 P. Yadav et al.
Fig. 9.9 The percent 100

depth dose (PDD) 4.2x4.2 cm2
measured for the 90
10x10 cm2
MRIdian MR-Linac for
three field sizes 24x24 cm2
80
70
60
PDD (%)
50
40
30
20
10
0
0 50 100 150 200
depth (mm)
the couch attenuation, the output must be within Electrotechnical Commission (IEC) requires the
1% of each other in accordance with AAPM average and maximum leakages to be <0.375%
TG-142 [15]. and 1.0%, respectively. The dual stack design of
the MRIdian system significantly reduces the
radiation leakage compared to other linear accel-
MLC Transmission and Leakage erators [16].
The radiation leakage through and between the

MLC leaves must be determined to ensure patient LC Linearity and Positional
M
safety and minimal out-of-field dose. This task is Accuracy
best achieved when using a 2D dosimeter such as
radiochromic films. The vendor recommends The positional accuracy of the MLC leaves must
irradiating an EBT3/EDR2 film sheet with the also be evaluated and compared to the TPS. The
upper and lower MLC stacks at −0.2 cm and maximum deviation between the TPS and the
0.2 cm, respectively. A large number of MUs, measured MLC positions must be within 1 mm,
such as 6000–10000 MUs, are delivered to the as specified in AAPM TG-142 [15]. The vendor
film with at least 2 cm buildup and 5 cm back- recommends using radiochromic films to mea-
scatter provided by solid water slabs. This irra- sure the positional accuracy since the MRIdian
diation is repeated at each cardinal gantry angle. MR-Linac lacks an electronic portal imaging
Using the same geometrical setup, a known device (EPID) for megavoltage (MV) imaging.
absorbed dose is delivered to another piece of The positional accuracy of the MLC banks can be
film using a 10 × 10 cm2 field size for reference. assessed by irradiating 1 × 24.1 cm2 (IEC-X by
Following the leakage irradiations, the average IEC-Y) rectangular fields centered at −10 cm,
leakage is calculated based on a large 30 × 30 cm2 −5 cm, 0 cm, 5 cm, and 10 cm IEC-X positions.
ROI on the film sheet, and maximum leakage is A treatment plan is created to irradiate a sheet of
calculated based on a small 1x1 cm2 ROI around film placed at 90 cm SAD with 5 cm depth and at
the hottest spot on the film sheet. The International least 5 cm backscatter using 40x40 cm2 solid
water slabs. This setup is repeated for 90° and identical. For the penumbra evaluation, the pro-
270° gantry angles. The film is exposed to the files for the larger field sizes are normalized to
planned dose distribution consisting of five the absorbed dose at the inflection point. The dis-
1 × 24.1 cm2 rectangular dose regions. The irra- tance between the 20% and the 80% isodose line
diation is repeated for the 0° gantry angle. A ref- is defined as the penumbra. The measured and
erence sheet of film is also typically irradiated planned penumbras must be within 1 mm of each
using an identical setup but with a single 10 × other for all of the beam profiles. Since the profile
10 cm2 field. Following the irradiations, the has a forward peaked shape, the conventional
absorbed dose on the film sheets is scanned and definition of flatness cannot be applied. Therefore,
analyzed using film analysis software such as the vendor specifies that the shape of the mea-
RIT. Several line profiles, parallel to the IEC-X sured and planned profiles must be compared.
axis, are placed on the scanned film with ade- This can be achieved by comparing measured
quate IEC-Y spacing. The deviation between the off-axis ratios at several off-axis distances with
measured beam centers and the planned beam the TPS-calculated off-axis ratios. Any devia-
centers is calculated for each line profile, and tions between the two values must be within 2%.
maximum deviations are noted. For all the mea- The beam symmetry can be determined by calcu-
surements, the deviations must be within 1 mm. lating the difference between the absorbed dose
values at off-axis points equidistant from the cen-
tral axis (CAX). The symmetry must be within
adiation Field Size, Shape,
R 2% according to the vendor recommendations.
Symmetry, and Penumbra Accuracy
The dosimetric metrics related to lateral beam IMRT Functionality

profiles must also be measured and compared to
the TPS. The measured beam shape, symmetry, The IMRT functionality must be evaluated to
width, and penumbra must match the TPS data ensure that the MR-Linac is able to deliver the
within the specified tolerances. A scanning ion planned absorbed dose within acceptable accu-
chamber and a 3D water tank are desired for this racy. This task requires measurement of the
task for fast acquisition of beam profiles. absorbed dose using a 2D or a 3D dosimeter. The
However, the vendor recommends the use of an vendor recommends the use of a diode array
ion chamber array, such as the IC profiler by embedded inside a cylindrical phantom such as
SunNuclear, if a 3D tank is unavailable. For field the ArcCHECK® or the Delta4® phantom.
sizes smaller than 4.16 × 4.16 cm2, small volume Therefore, this task can be performed following
detectors such as the EDGE diode, PTW micro- the installation and calibration of the diode-array
Diamond, or Exradin W2 can be used in a 1D phantom. The vendor recommends the use of two
water tank, and the profile can be measured by treatment plans for this task, but this number can
moving the couch laterally in small increments. be increased at the discretion of the physicists.
Ideally, the beam profiles would be measured at The planned dose must be recalculated using the
multiple depths; however, the vendor recom- CT images of the diode-array phantom. The
mends a single depth for each of the measure- absolute point absorbed dose can be measured
ments to reduce the total acquisition time. The and compared with the planned point dose by
field size is recommended to be measured at inserting an ion chamber inside the phantom. It is
90 cm SAD with at least 1.4 cm of buildup mate- of note that the volume averaging effect can lead
rial. Following the irradiations, the planned and to large deviations between the planned and mea-
measured field sizes are compared by calculating sured absorbed dose. The phantom can be aligned
the FWHM of the profiles which must be within on the couch using a conformal plan consisting of
2 mm of each other. The geometrical setup for the beams at 0° and 90°/270° gantry angles with a
measured and planned dose distributions must be 10 × 10 cm2 field size. The couch translational
168 P. Yadav et al.
shifts that maximize the gamma pass rate with a drical volume appears as a circle in the sagittal
1%/1 mm criterion can then be applied. Following plane and is a surrogate for tumors. The tumor
the setup, the QA plans must be delivered, and surrogate is driven by a long rod connected to a
the measured dose is obtained. During the treat- motion stage, which is kept far away from the
ment delivery, the MRIdian MR-Linac is capable isocenter to minimize the impact of the external
of recording the delivery parameters in a log file. magnetic field on the functionality of the stage.
These parameters can be imported into the TPS, Tumor motion can be replicated by using an
and the delivered absorbed dose can be predicted oscillating waveform for the motion stage. The
using the TPS beam model dose calculation. amplitude, phase shift, and starting position of
Using a gamma criterion of 3%/3 mm, the gamma the surrogate can be altered. The CIRS system
pass rate must be ≥95% when comparing the utilizes optical sensors for triggering when the
measured and planned absorbed dose and when surrogate volume is outside of the delineated
comparing the measured and log file predicted margins. This is achieved by using two optical
absorbed dose. The planned and measured sensors placed in the superior-inferior direction
absorbed dose using the ion chamber must be with a set spacing between them. At the home
within 3%. position, the distal edge of the rod is equidistant
from the superior and inferior sensors. When the
edge of the rod and the tumor surrogate volume
MR Tracking Latency moves outside of the bounds in either direction,
the sensors are triggered, and an electrical signal
Tracking moving tumors and accurate gating is transmitted to the oscilloscope. The CIRS
requires a sufficiently low latency so that the phantom is imaged using volumetric MRI, and a
radiation beam is turned off instantly following contour is drawn around the surrogate volume in
the target volume moving outside the delineated the sagittal plane. Before the latency measure-
boundaries. The MRIdian system tracks the ments are acquired, a preview cine scan is typi-
tumor motion using cine MRI in the sagittal plane cally acquired to evaluate the accuracy of the
and triggers the radiation beam off when the tracking algorithm. Following the preview scan,
tracked tumor margin is outside the reference real-time imaging is performed, and the tumor
margin. The vendor requires that the gating surrogate is driven using the desired waveform. A
latency of the MR-Linac is within 0.35 s over 20 multichannel oscilloscope is used to input the
gating cycles. The gating latency is affected by trigger signal from the gating phantom as well as
the MRI acquisition speed, image reconstruction the beam-off signal from the treatment control
and processing time, tracking algorithm process- cabinet or the dosimeter. The time difference
ing time, and radiation beam signal latency. In between the gating signal indicating the distal
case a dosimeter is employed to measure the edge of the rod moving out of margins and the
latency, the delay time between the radiation beam-off trigger is noted as the latency value.
beam turning off and the dosimeter readout sig-
nal must be considered as well. Alternatively, the
radiation beam trigger time can be extracted from Dosimetry Consistency with Gating
the beam control system directly. The vendor rec-
ommends the use of a CIRS motion phantom to In addition to measuring gating latency, it is
measure the system latency. Other phantoms, important that the dose delivered to a moving tar-
such as the MRI4D QUASAR motion phantom, get using gating remains similar to a stationary
are also commercially available to perform this target. This can be evaluated by irradiating an ion
test. The CIRS motion phantom consists of a chamber embedded inside the gating phantom
cylindrical phantom with ion chamber inserts and such as the CIRS phantom. A conformal plan is
small volumes embedded inside the large vol- created to deliver a set absorbed dose to the cav-
ume. The smaller volume inside the large cylin- ity of the ion chamber using a contour drawn
around the cavity. Two irradiations are per- outlined in the AAPM MPPG 5a and TG-53
formed. For the first irradiation, the ion chamber reports can be used to evaluate the TPS beam
is driven resembling tumor motion using an model [18, 19]. Dose calculation in heterogenous
oscillation waveform. The charge accumulated media should especially be evaluated due to the
over the irradiation is measured, and gating is presence of the ERE. A variety of complex MLC
enabled by turning off the beam when the ion shapes can be created and assessed using 2D pro-
chamber cavity moves outside the margins. The files and point dose measurements. In addition to
second irradiation is performed with the motion the two IMRT QA plans recommended by the
turned off. The difference between the two charge vendor for measurement-based QA evaluation,
readings must be within 2% of each other to sat- several additional IMRT QA plans can be created
isfy the TG-142 criteria. and measured using the diode-array phantom
depending on the commonly treated sites.
Further, independent end-to-end testing using
Commissioning of the MRIdian phantoms can be performed to ensure the accu-
MR-Linac racy of the TPS beam model.
The commissioning of the MRIdian MR-Linac is

performed following the acceptance testing and Patient-Specific QA
evaluates the linac using tests not performed dur-
ing acceptance testing. At this stage, the baseline An inverse treatment planning technique is used
data for ongoing QA are also acquired using the for IMRT deliveries that optimizes beam param-
equipment employed by the clinic. For the MRI eters based on user-defined dose objectives. The
component of the system, the photon beam atten- complexity of these treatment plans warrants pre-
uation through the receiver coils must be mea- treatment QA to assess the plan deliverability. A
sured. Output factors (OFs) must be measured comparison is performed between the patient-
and compared with the TPS during commission- specific planned and machine-delivered dose dis-
ing. The vendor recommends using either the tributions. The delivered dose distribution can
EDGE diode, the PTW microDiamond, or the either be measured or calculated using the
Exradin W2 detectors to measure output factors machine log files. Gamma analysis is used to
for small fields. The vendor also provides the compare the dose distributions considering both
field- and detector-specific correction factors cal- dose differences and distance to agreement.
culated using MC simulations. Although the gat- Similar to conventional IMRT QA, several
ing accuracy and latency are evaluated during the MR-safe diode-array phantoms are commercially
acceptance testing stage, a more extensive evalu- available, such as ArcCHECK and Delta4, to
ation of the gating tracking algorithm can be per- acquire the absorbed dose measurements. These
formed during commissioning using a range of phantoms must be installed and commissioned
motion traces. Any tilts in the MLC carriages according to vendor guidelines before use. For
must also be corrected using the beam profiles measurement-based QA, a QA plan is created for
acquired with the IC profiler. The profile shape each treatment plan before the delivery by recal-
and symmetry metrics can be used to assess the culating the planned dose distribution using the
alignment of the MLC banks. The mechanical CT images of the homogeneous phantom.
checks must also be performed during commis- Absorbed dose is sampled using an array of
sioning. The digital indicators must be verified diodes, and the planned and measured dose dis-
using independent measurements. Translational tribution is compared. A gamma pass rate is cal-
and rotational shifts in the couch can be evaluated culated to quantify the treatment plan quality, and
using handheld tools. a threshold value is used to approve the plan. The
Extensive validation of the TPS must be per- gamma criteria and the pass rate threshold vary
formed during commissioning. The guidelines between clinics and treatment techniques. For
170 P. Yadav et al.
example, most clinics consider a 95% pass rate Table 9.4 Suggested daily QA tasks for the MRIdian
with a 3%/3 mm criterion to be adequate. The MR-Linac
gamma criterion for SBRT is typically reduced to Procedure Tolerance
3%/2 mm. Dosimetry
MR-Linacs are often used for adaptive radio- X-ray output constancy 3%
Dose-rate constancy 2%
therapy (ART) during which the treatment plan
Mechanical and imaging
gets modified daily based on the changes in the Laser and imaging coincidence 1 mm
treatment parameters. During the adaptation Radiation beam and imaging 1 mm
process, the patient is typically on the table and coincidence
measurement-based QA is not feasible. In such Patient positional accuracy 1 mm
cases, calculation-based QA can be performed by Receiver coil functionality Functional
Safety Functional
extracting the delivery parameters from the log
Door interlocks Functional
file and using an independent dose calculation Radiation area monitor Functional
system. The MRIdian system offers an indepen- Beam-off functionality Functional
dent dose calculation software called Zeus that In-room camera Functional
can perform adaptive QA (AQA) [20]. Zeus uses In-room audio Functional
a phase space file from the TPS beam model Panic bulb Functional
above the MLC plane and uses an independent
fast Monte Carlo-based platform to calculate A1SL MR-safe ion chamber fits well inside the
absorbed dose in patient geometry. The effect of chamber insert of the phantom with its centroid
the external magnetic field is considered in the aligned with the radiation isocenter. A 10.04 ×
dose calculation. Since the MRIdian system lacks 9.96 cm2 field size is typically used to measure
daily CT data for dose calculation, the daily volu- the output, and the MU to time ratio is used to
metric MRI is co-registered with the CT simula- determine the dose rate. The DQA phantom is
tion data, and a deformation matrix is applied to aligned using the lasers and the scribes on the
the CT data. The calculated delivered dose distri- phantom. By acquiring an MRI image of the
bution is compared with the planned dose using a phantom and co-registering the acquired volume
3D gamma analysis. A QA report is created indi- with the reference volume, the positional shifts
cating the gamma pass rates and other calculation between the lasers and the MRI can be calcu-
metrics. lated. Following the localization of the ion cham-
ber using the MRI images, the radiation isocenter
and MRI isocenter coincidence can be evaluated.
Daily, Monthly, and Annual QA This is achieved by leveraging the cylindrical
shape of the ion chamber cavity with a diameter
Following the initial characterization of the of 4 mm and a length of 4.4 mm. By irradiating
MR-Linac, it is important to evaluate the machine the ion chamber with a small field size of
at regular intervals to ensure acceptable perfor- 0.40 × 0.83 cm2 at each cardinal gantry angle, a
mance. Table 9.4 shows the recommended daily partial volume irradiation effect is induced inside
QA tasks for the MRIdian MR-Linac based on the cavity, which is highly sensitive to any posi-
the AAPM TG-142 report [15]. All safety fea- tional offsets in the radiation beam isocenter [21].
tures must be evaluated and functional to ensure By introducing known shifts to the ion chamber,
the utmost patient safety. The output and dose the relationship between the charge readings and
rate of the photon beam can be evaluated using an the positional offset can be determined a priori.
ion chamber inserted into the ViewRay DQA Daily monitoring of the charge reading with the
phantom or by using other commercially avail- 0.40 × 0.83 cm2 field size can yield an estimate of
able solutions such as the Sun Nuclear Daily QA the positional offset between the radiation iso-
3 phantom. The ViewRay DQA phantom was center and the MRI isocenter. The patient’s posi-
described earlier in this chapter. The Exradin tional and repositional accuracy can be
determined by imaging the phantom after intro- employed to acquire lateral beam profiles. The
ducing a known shift and aligning the phantom CIRS or QUASAR gating phantoms can be uti-
based on the peripheral physical landmarks inside lized to measure the gating latency. TPS QA can
the phantom. The shifts between the acquired be performed by creating and comparing
images and the planned images determine the absorbed dose distributions for mock IMRT plans
patient positional accuracy as well as the spatial from AAPM TG-119 [22].
fidelity of the MR images. It is essential that the lasers are accurate and
Table 9.5 shows the recommended monthly coincident with the radiation beam and the imag-
QA tasks for the MRIdian MR-Linac. In lieu of ing system. The offsets between these systems
using the ViewRay DQA phantom for dosimetry can be measured using the procedure described
tasks, a 1D water tank or solid water slabs are in the previous sections. Phantoms with ruler
recommended for these measurements. Using the markings can be used to evaluate the couch posi-
baseline measurements acquired following the tion indicators. The MLC positional accuracy is
acceptance testing and commissioning of the typically assessed using a picket fence test.
MR-Linac, the constancy of the photon beam However, due to a lack of an imaging panel, this
must be verified. Output and dose rate can be test can be performed using radiochromic films,
verified by acquiring absorbed dose measure- which is an arduous process with a passive read-
ments at a single depth. The beam energy can be out. Therefore, the high sensitivity of the ion
quantified using either the PDD or the TPR met- chamber response to positional offsets when
ric by acquiring the ratio of dose at two different using a partial volume irradiation technique can
depths. If a scanning water tank is unavailable, be leveraged to estimate the positional accuracy
the IC profiler or other detector arrays can be of the MLC leaves. This test involves irradiating
half of each of the ion chambers in the array
using a half beam block (HBB) technique and
Table 9.5 Suggested monthly QA tasks for the MRIdian
MR-Linac
establishing a relationship between the detector
response and known induced MLC positional
Procedure Tolerance
shifts for all banks [23]. During monthly QA, the
Dosimetry
X-ray output 2% measured detector response can then be used to
constancy calculate the expected MLC positional error. This
Dose-rate constancy 2% test should be repeated at all four cardinal gantry
Beam profile 1% angles to evaluate the effect of differences in the
constancy gravitational force on the MLC banks. The MRI
Beam energy 1%
constancy
tasks mimic the acceptance testing protocol
Gating latency 100 ms of expected described in the earlier sections.
TPS QA Constancy In addition to the monthly and daily QA tasks,
Mechanical and safety the annual QA procedures include reference
Emergency stop Functional dosimetry in a water tank where an absorbed
Treatment couch 2 mm/1° dose to water calibration must be maintained
position indicators
within 1%. For the TG-51 output calibration, a
Localizing lasers 1 mm
MLC positional 1 mm water tank is recommended instead of solid water
accuracy slabs. The beam profiles and output factors can
MRI be measured using the procedures described ear-
Bo field Constancy lier. The beam profile size, flatness, and symme-
homogeneity
try constancy must be within 1% as well. Field
ACR image quality ACR specifications (see
Table 9.3) output factors should be measured and a con-
SNR and uniformity NEMA specifications stancy of within 1% should be ensured. The aver-
Spatial integrity 1 mm (10 cm radius), age and maximum MLC leakage must be
2 mm (17.5 cm radius) measured for the cardinal gantry angles to be
172 P. Yadav et al.
within 0.375% and 1%, respectively. The laser, addition to the fact that the PAR files are propri-
MRI, and radiation isocenter coincidence must etary and can only be accessed by authorized per-
be within 1 mm. sonnel using the ViewRay TPS, this format is
specific to the ViewRay MRIdian system. In
order to prevent data loss or corruption, it is
IT and Other Considerations imperative that PAR files are stored and managed
properly. The PAR files should be regularly
The hospital and/or department IT team is backed up and validated, as well as protected by
required to install relevant QA software for strict access controls to prevent unauthorized
ViewRay MRIdian as well as to set up the IP access. Performing of backups that usually take
address for the system. The QA software that about 1 hour and a full backup in Synology Drive
needs to be installed can be used to monitor the at the end of the day (fixed time) can help ensure
system performance and ensure main compo- that data is protected and can be quickly restored
nents are functioning properly. Setting the IP in the event of a data loss or corruption. Cine data
address refers to the process of assigning a unique as well as other types of patient data need to be
network address to the ViewRay MRIdian sub- backed up. Log files provide a detailed record of
systems so that it can communicate with other all treatment and imaging activities, including
devices on the network. For the QA software to timestamps and dose information. A software
be installed and the IP address to be set, the hos- team can identify areas for improvement by ana-
pital’s IT support team will need to be familiar lyzing log files and creating patches to resolve
with the ViewRay MRIdian system, as well as any issues that may arise. Performing this type of
with the installation and configuration of medical analysis will help ensure that the healthcare sys-
imaging equipment. In order to ensure that the tem is operating optimally and that patients are
software is installed and configured correctly, receiving the best possible care.
they may need to work with the vendor of the QA
software and the ViewRay MRIdian system.
A database management system is used to Summary
store patient data. The data is periodically backed
up to a Synology Drive with a capacity of 10 tera- Real-time MRI-guided radiation therapy is made
bytes, specifically Racket. Data is then trans- possible by the groundbreaking ViewRay
ferred to the hospital’s server for further storage MRIdian system, which combines magnetic res-
and management. A multiple backup system is a onance imaging (MRI) and linear accelerators
best practice for data protection and disaster (linac). The history and underlying physics of
recovery. Furthermore, it is imperative that the MR-Linac technology, as well as the practicali-
data is stored and transmitted securely in order to ties and financial aspects of purchasing and put-
prevent unauthorized access or theft. By utilizing ting into use a ViewRay MRIdian system, have
encryption, secure networks, and access controls, all been covered in this chapter. The impact of the
this can be accomplished. The hospital should magnetic field on the dose distributions and dosi-
also have policies and procedures in place to metric data was discussed in the context of the
ensure compliance with regulations such as electron return effect (ERE). The system design
HIPAA regarding the management and protec- of the 0.35 T MRIdian MR-Linac was described
tion of patient data. Data backup includes patient in detail including the ancillary components used
data (Images, Dose, Reoptimized, predicated, for treatment delivery and imaging. Concerns
and Cine, etc.), machine data: configuration file, related to MRI safety were accentuated including
software version, Core and Database computer. outlining a zoning system specific to the MRIdian
Using the ViewRay treatment delivery system system. Relevant training topics for all personnel
(TDS), patient data can be accessed and restored involved in the MR-Linac treatments were dis-
using PAR (plan analysis record) files only. In cussed including workflow maps and teams
involved at each step. Vendor-provided accep- 10. Mutic S, Dempsey JF. The ViewRay system: magnetic
tance testing protocols for both the linac and the resonance–guided and controlled radiotherapy. Semin
Radiat Oncol. 2014;24:196–9.
MRI portions of the machine were compiled. 11. Klüter S. Technical design and concept of a 0.35 T
Discussion of additional commissioning tasks MR-Linac. Clin Transl Radiat Oncol. 2019;18:98–101.
and ongoing QA recommendations based on 12. Hu Q, Yu VY, Yang Y, et al. Practical safety con-
existing protocols were included. The advantages siderations for integration of magnetic resonance
imaging in radiation therapy. Pract Radiat Oncol.
of the ViewRay MRIdian system are obvious, 2020;10:443–53.
despite the difficulties involved in putting such a 13. Khan AU, DeWerd LA, Das IJ, et al. A multi-
cutting-edge technology into practice: increased institutional comparison of acceptance testing
precision and accuracy in the delivery of radia- data for a 0.35 T MRI scanner. Phys Med Biol.
2022;67:235005.
tion therapy, improved imaging capabilities for 14. Gao Y, Lotey R, Low DA, et al. Technical note: vali-
tumor detection and localization, and reduced dation of an automatic ACR phantom quality assur-
radiation exposure to healthy tissue. ance tool for an MR-guided radiotherapy system.
Med Phys. 2021;48:1540–5.
15. Klein EE, Hanley J, Bayouth J, et al. Task Group 142
report: quality assurance of medical acceleratorsa.
References Med Phys. 2009;36:4197–212.
16. Khan AU, Lotey R, DeWerd LA, et al. A multi-
1. Lagendijk JJW, Raaymakers BW, van Vulpen M. The institutional comparison of dosimetric data for a 0.35
magnetic resonance imaging–linac system. Semin T MR-Linac. Phys Med Biol. 2022;67:05NT01.
Radiat Oncol. 2014;24:207–9. 17. Bissonnette J-P, Balter PA, Dong L, et al. Quality
2. Raaymakers BW, Jürgenliemk-Schulz IM, Bol GH, assurance for image-guided radiation therapy utiliz-
et al. First patients treated with a 1.5 T MRI-Linac: ing CT-based technologies: a report of the AAPM
clinical proof of concept of a high-precision, high- TG-179. Med Phys. 2012;39:1946–63.
field MRI guided radiotherapy treatment. Phys Med 18. Fraass B, Doppke K, Hunt M, et al. American
Biol. 2017;62:L41–50. Association of Physicists in Medicine Radiation
3. Uilkema S, van der Heide U, Sonke J-J, et al. A 1.5 Therapy Committee Task Group 53: quality assurance
T transverse magnetic field in radiotherapy of rectal for clinical radiotherapy treatment planning. Med
cancer: impact on the dose distribution. Med Phys. Phys. 1998;25:1773–829.
2015;42:7182–9. 19. Jacqmin DJ, Bredfeldt JS, Frigo SP, et al.
4. Menten MJ, Fast MF, Nill S, et al. Lung stereotactic Implementation of the validation testing in
body radiotherapy with an MR-Linac – quantifying MPPG 5.a “Commissioning and QA of treatment
the impact of the magnetic field and real-time tumor planning dose calculations—megavoltage pho-
tracking. Radiother Oncol. 2016;119:461–6. ton and electron beams”. J Appl Clin Med Phys.
5. Huang C, Yang B, Lam WW, et al. Magnetic field 2017;18:115–27.
induced dose effects in radiation therapy using 20. Khan AU, Simiele EA, Lotey R, et al. An indepen-
MR-linacs. Med Phys. 2023;50(6):3623. dent Monte Carlo–based IMRT QA tool for a 0.35 T
6. Woodings SJ, Bluemink JJ, de Vries JHW, et al. Beam MRI-guided linear accelerator. J Appl Clin Med Phys.
characterisation of the 1.5 T MRI-linac. Phys Med 2023;24:e13820.
Biol. 2018;63:085015. 21. Mittauer KE, Dunkerley DAPP, Yadav P, et al.
7. Khan AU, Simiele EA, Lotey R, et al. Development Characterization and longitudinal assessment of
and evaluation of a GEANT4-based Monte daily quality assurance for an MR-guided radio-
Carlo Model of a 0.35 T MR-guided radiation therapy (MRgRT) linac. J Appl Clin Med Phys.
therapy (MRgRT) linear accelerator. Med Phys. 2019;20:27–36.
2021;48:1967–82. 22. Ezzell GA, Burmeister JW, Dogan N, et al. IMRT
8. O’Brien DJ, Roberts DA, Ibbott GS, et al. Reference commissioning: multiple institution planning and
dosimetry in magnetic fields: formalism and ion- dosimetry comparisons, a report from AAPM Task
ization chamber correction factors. Med Phys. Group 119. Med Phys. 2009;36:5359–73.
2016;43:4915–27. 23. Mittauer KE, Yadav P, Paliwal B, et al. Characterization
9. O’Brien DJ, Dolan J, Pencea S, et al. Relative dosim- of positional accuracy of a double-focused and
etry with an MR-Linac: response of ion chambers, double-stack multileaf collimator on an MR-guided
diamond, and diode detectors for off-axis, depth radiotherapy (MRgRT) Linac using an IC-profiler
dose, and output factor measurements. Med Phys. array. Med Phys. 2020;47:317–30.
2018;45:884–97.
Elekta Unity System
10
Jochem W. H. Wolthaus, Eenas A. Omari,
Xinfeng Chen, and Bram van Asselen
Introduction 2017 [4], and it proved the geometrical and

dosimetric accuracy. The Elekta Unity
Accurate target delineation which provides both MR-Linac (Fig. 10.1) was released in 2018
actual anatomical information (superb soft-tissue worldwide for commercial use [5].
contrasts) and functional imaging information For the linear accelerator (linac) of Unity
(biological function and physiology) just prior to MR-Linac, a standing wave guide design is used
and during delivery is essential for radiation to produce a 7 MV FFF beam. The linac is
treatment. Including the MRI during radiation mounted on a ring rotating around a closed bore
delivery allows for reduction of the treatment MRI scanner with a maximum continuous rota-
margins and volumes and opens new possibilities tion speed of 6 rpm. The sensitive accelerator
for treatment delivery techniques such as hypo- parts, such as the waveguide and the magnetron,
fractionation. Furthermore, with fast image are placed in a shielded torus with negligible
acquisition, intrafraction motion can be moni- magnetic field strength. This area is created using
tored allowing for real-time adaptive radiother- active shielding. The magnetic field is perpen-
apy by correcting for short-term geometrical dicular to the photon beam and points out of the
changes, e.g., due to breathing or drift. bore along the -y direction (IEC61217). The
The MR-Linac, combining a high voltage source-to-isocenter distance in this configuration
linear accelerator with onboard 1.5T MR imag- is 143.5 cm. Gantry bore diameter is 70 cm. Since
ing system, was originally invented at the a closed bore design is used, the beam is attenu-
University Medical Center Utrecht [1, 2] and
was developed in collaboration with Elekta (lin-
ear accelerator) and Philips (MR unit) [3]. The
first patient treatment with MR-Linac was in
J. W. H. Wolthaus (*) · B. van Asselen

Department of Radiotherapy, University Medical
Center Utrecht, Utrecht, The Netherlands
[email protected]
E. A. Omari · X. Chen
Department of Radiation Oncology, Medical College
of Wisconsin, Milwaukee, WI, USA
e-mail: [email protected]; [email protected] Fig. 10.1 The Elekta Unity 1.5T MR-Linac system
176 J. W. H. Wolthaus et al.
ated by the cryostat. Dose rate at isocenter is electrons in this chapter. In a medium, the elec-
approximately 425 MU per minute, calibrated at trons lose energy while interacting with the
the dose maximum (1.3 cm depth). The MRI medium as can be described by the stopping
components have been optimized to create a win- power. The slowing down process of the elec-
dow with a uniform high transmission but reduce trons results in a lower Lorenz force interacting
the maximum field size to 22 cm in longitudinal on the electron. As a result, the electron moves
direction. The MLC is based on the design of the over a spiraling trajectory with a gyro-radius
Agility MLC but with a fixed collimator angle of which decreases along its path [6]. Therefore, the
90 degrees allowing for leaf travel to be only in change of local dose deposit results in a differ-
the cranio-caudal direction. Field sizes ranging ence in absorbed dose distribution relative to the
from 0.7 × 0.7 cm2 to 57 × 22 cm2 can be created absence of a magnetic field.
by the multi leaf collimator consisting of 80 leaf
pairs with a projection of 7.2 mm at the isocenter.
Currently, only step-and-shoot IMRT is available Beam Characteristics
on the MR-Linac system, but VMAT delivery is
being developed and will be available in future Several aspects of the beam change in the pres-
system control software releases. ence of a magnetic field [7]. The surface dose
changes due to the change in contamination elec-
trons and different trajectories of the electrons.
Basic Physics in a Transverse For the entrance of beam, the electrons created
Magnetic Field above the surface are captured by the magnetic
field moving out of the beam in a spiraling path
In MRI guided radiotherapy, patients are treated reducing the surface dose [8, 9]. However, the
in the presence of a magnetic field which affects secondary electrons at the surface are swept side-
the dose deposition in the patient. The most com- ways and backward due to the Lorenz force
mon photon-electron interactions in radiotherapy increasing the local dose [10]. The net effect
are photoelectric effect, Compton scatter, and gives a similar surface dose as without magnetic
pair production. Meanwhile, the primary high- field.
energy photons produced by the accelerator are The curved electron paths cause the dose to be
not influenced by the magnetic field. On the flip deposited closer to the photon-electron interac-
side, the high-energy electrons produced inside tion point and result in a shallower dose buildup
the patient will experience the Lorenz force [11, 12] and a shallow depth of dose maximum.
which changes their trajectories affecting
 the Furthermore, this implies that the dose is closer
dose deposition. The  Lorenz force ( FL ) for a pres- to the kinetic energy released in matter (KERMA)
ent magnetic field B is given by and therefore slightly lower than when the mag-
   netic field is absent [13]. The penumbra region of
FL = qvxB (10.1) beam profiles is changed since the electrons are

where q is the charge of the particle and v is its swept out of one side of the field, and they move
velocity. In a vacuum, with a magnetic field per- toward the field center [12, 14, 15]. This results in
pendicular to the velocity of the particle with asymmetrical lateral profiles as shown in
mass m, the electrons move in a circular path Fig. 10.2.
with gyro-radius (rg) given by On the exit side of the beam, electrons leaving
the patient are either returned to the patient or
mv
rg = (10.2) start spiraling around the magnetic field lines
qB resulting in a stream of electrons. The phenome-
Positrons are also produced in pair production non where the electrons return to the surface
interactions; however, we will only consider causing an increase in the exit dose is known as
10 Elekta Unity System 177
Fig. 10.2 Magnetic field effect on the inline (blue) and crossline/lateral (red) dose profiles of a 10 × 10 cm2 beam
the electron return effect (ERE) [10, 11, 16]. accuracy and deformable and/or moving targets
However, some electrons “escape” and stream [5]. Depending on the chosen workflow, the aver-
away from the beam and deposit dose elsewhere age time of treatment can be variable between 20
outside the patient [17, 18]; this effect is known and 60 min per patient [23, 24].
as the electron streaming effect (ESE). These dif- The rapid technological advances give access
ferences in dose deposition are taken into account to real-time MRgRT to all kinds of radiotherapy
in the treatment planning system with proper centers. However, note that introducing MRgRT
modeling by the Monte Carlo algorithm includ- into the department also involves adaptation to
ing the magnetic field. building constructions, new treatment planning
software, quality assurance (QA) systems, staff-
ing, and training and may include an introduction
Logistics and Economics of an MRI simulator. This may lead to additional
of Acquiring an MR-Linac cost depending on existing resources within the
department and the degree of desired efficiency
Economic Analysis and Cost and reliability of the clinical workflow. For
example, a dedicated MRI simulator for treat-
It is difficult to make a realistic estimate of the ment workflow preparation increases the patient
cost and effort for implementing a real-time throughput on the treatment machine.
MRgRT program in a department as it depends Furthermore, the MRI simulator can be used for
on various circumstances, for example, the avail- tumor staging and localization regardless whether
able facilities, staff, hospital size, and commit- the patients will be treated on the MR-Linac or a
ment to research [19–22]. Furthermore, revenue conventional system [25]. In addition, the MRI
depends on the patient treatment reimbursement scans can be converted to pseudo-CT images
that might change as cost per fraction increases allowing for MRI only treatment planning [26].
with the number of treatment fractions. A 5 frac- Reliability of the clinical workflow is
tion MR-Linac treatment is equivalent to a 20 to increased if there are backup MRI systems at the
30 fraction scheme [19]. Patient selection and radiology department that can be used in case of
fractionation schemes focus on patients that may system failures. However, these “fallback” MR
benefit from high doses with high geometrical systems have to be prepared to allow for radio-
therapy treatment simulation, e.g., flat tabletop, new introduced technique or innovation in X-ray
coils, and position indexing. Another fallback radiotherapy. However, expanding the number of
scenario in the case of MR simulator downtime is MR scanners or MRgRT systems makes use of
to use the MR-Linac as a backup simulator. knowledge from existing staff at the department.
When an MRgRT system enters a radiother- Synergy and efficiency can be achieved by imple-
apy department, staff should be trained or menting better patient workflows, treatment
recruited to support the usage of MR imaging. scheduling, and QA and maintenance scheduling.
Training involves safety aspects, physics, system It might also be advantageous to reconsider the
operation, workflows, treatment planning system vendor support agreements for maintenance.
(TPS), and connectivity. Well-trained staff is also
important for proper patient referral for groups
that may benefit the most from the advanced Architectural and Structural Planning
online MRgRT program. Training and staff
responsibilities will be discussed in section A generic structural outline for a Unity MRgRT
“Staffing and Training”. The reader can also refer program consists of the treatment room, machine
to Chaps. 2 and 4 of this book on this topic. room, maze, control room, technical room, and
Depending on the level of vendor support and patient dressing and waiting rooms, as shown in
service agreements, physics and engineering sup- Fig. 10.3. The treatment room contains the actual
port is crucial. MRgRT requires additional soft- MR-Linac, which is isolated from external RF
ware or modules for existing software to support noise by a Faraday cage. The Faraday cage is
MRgRT, for example, to account for the mag- custom-built close to the wall, floor, and ceiling
netic field, additional upgrades are required to the of the vault to optimize the functional inner
treatment planning software. In addition, licens- space. The machine room is located within the
ing will be required to support extra workstations treatment room within the space of the concrete
for the TPS or record-and-verify (R&V) systems vault and is accessed from via a separate RF door.
to improve workflow efficiency. Although operat-
ing the MR-Linac in an isolated environment is
possible, integration of the MRgRT system into
the department’s infrastructure is advised.
Architecture and construction should also be
considered when installing an MR-Linac system.
This is discussed in more detail in the next sub-
section. Finally, existing hardware and phantoms
for QA purposes are generally not compatible
with MR or MRgRT systems due to the magnetic
field and limited bore diameter, which requires
the purchase of new equipment. Examples of new
MRI compatible equipment are (1) water scan-
ning phantom, waterproof ion chambers, and
phantoms for reference dosimetry; (2) a 2D/3D
patient delivery QA phantom for acceptance test-
ing, commissioning, calibration, and verification
of the dose delivery; (3) secondary dose calcula-
tion software for daily treatment plan verifica-
tion; and (4) MRI imaging QA phantoms and
analysis software. Fig. 10.3 The layout of a Unity MR-Linac vault at the
UMC Utrecht, which includes the treatment room,
The initial cost of establishing an MRgRT machine room, maze, control room, technical room, and
program is significantly higher than any other patient dressing/waiting rooms
This room is utilized for machine maintenance. The use of medical gases (e.g., to support
The maze is the corridor between treatment room anesthesia) is optional for the Elekta Unity sys-
and control room. The maze may have reduced tem, and if there is any intent to use these gases in
number of corners by taking into consideration the future, supply pipes should be installed at the
the maximum energy, shielding, and leakage. All end of the machine room. The outlets for medical
user interface workstations for imaging, delivery gases are placed on the left side from the bore
and workflow controls, and patient CCTV moni- entrance of the treatment room. Locating the out-
tors are located in the control room. The control lets in this location allows for free access behind
room should be treated as MR-Zone 3. The tech- the bore in case of an emergency in order to avoid
nical room contains all computers, servers, tripping over cables and tubes.
switches, routers, and UPS systems used for the Even with proper (active) shielding, magnetic
MR-Linac system. Finally, waiting rooms and fringe fields could still be present outside the
patient dressing rooms are not explicitly part of treatment room, at a different level below or
the official layout, but it is recommended to con- above the vault, or outside the building [27]. All
sider them for workflow efficiency. adjacent areas have to be identified and magnetic
The floor plan of a Unity MR-Linac, as defined fringe fields measured. Additional attention must
by Elekta, is such that it can be retrofitted within be taken for the linear accelerators in adjacent
an existing conventional linac vault. Costs for the spaces to make sure that they lie outside of the
rebuilding and construction of an existing vault, 0.05 mT field. It is important to note that elec-
prepared to install a Unity MR-Linac, are slightly trons travel through tubes and can distort the tar-
higher. This is due to reconstruction since the get due to the magnetic fringe fields and can vary
foundation for a conventional linac has to be for different gantry angles; however, the remain-
removed and an additional pit (hole) has to be ing magnetic fringe fields can be considered
drilled to fit the large gantry ring and add the static and therefore be solved by gantry angular
Faraday cage. Furthermore, it might be required dependent calibration of the beam. In case of a
to reconstruct the maze to allow the delivery of quench, adjacent linacs should immediately be
the large components like MRI and gantry ring of considered out of service and may require recali-
the system. In preparation of the installation, bration of the beam. It is important to include this
environmental conditions of the facility regard- information in the emergency procedures.
ing interference from electromagnetic radiation,
vibrations, magnetic fringe fields, or radiation
leakage should be considered. A survey by the Construction and Installation
vendor must be performed prior to installation. Timeline
The dimensions of the existing vault might be
less than preferable. If the available space to store Regardless of using an existing or a new bunker,
equipment such as MRI crash cart, patient masks, the vendor will perform a site inspection follow-
phantoms, or MRI coils is limited, practical pro- ing the completion of construction. If the bunker
cedures and workflows should be adapted. In is considered to be ready, the first stage is the
case of a newly constructed facility, it is advised delivery and assembly of all MR-Linac basic
to consider extra space for these items. components: MRI, the gantry, and linac. After
A helium safety vent pipe, quench pipe, is approximately 10 weeks, the linac is mounted on
required to exhaust helium gas. The pipe and the gantry ring and aligned; this will allow for
venting system must be capable of exhausting the radiation beam production and to perform radia-
large amount of helium gas that may result from tion survey. Subsequently, in about 6 weeks, the
a quench of the MRI when the magnetic field is RF Faraday cage is built, the quench pipe and the
terminated quickly. Although these events are medical gases are connected, and some remain-
very rare, a quench can occur spontaneously or ing electrical works are completed. Finally, the
by manual activation in case of an emergency. couch, also known as ‘patient positioning sys-
tem’, is installed and the MR-Linac is ready to time. For example, the selection and purchase of
become “on field.” After the ramp-up of the mag- QA equipment (e.g., ion chambers, electrome-
netic field, all existing (conventional) linacs in ters, cables, water tank and/or 2D detectors, and
the adjacent areas must be recalibrated for gantry patient QA phantoms) should be considered at
angle dependent beam steering and dosimetry to an early stage as it might take a considerable
correct for the new emerged magnetic fringe amount of time to have all required equipment
fields. Following the installation of the accompa- delivered and validated for clinical use.
nied software (TPS, Mosaiq patient manage- Depending on the customer activities, after about
ments and R&V software, imaging), the device 15 weeks, the clinical readiness phase is com-
acceptance tests are performed as a joint activity pleted with an end-to-end workflow test validat-
between the vendor and customer. The ing the complete patient treatment workflow
MR-acceptance testing schedule includes B0 field from start to end.
strength, direction, stability, and homogeneity, as The installation of the Unity is mainly coordi-
well as an assessment of the B0 inhomogeneity nated and performed by Elekta engineers and
induced by the gantry. The gantry may introduce technical support. However, a local staff group of
field distortion, which is induced by the ferro- the various involved disciplines and expertise is
magnetic material on the gantry ring but can be advised to support the installation at any stage or
removed with a gantry shimming procedure. phase of the project, including the acceptance
Furthermore, cryogen helium level, RF cage seal- and commissioning according to local
ing, image orientation to ensure that the correct regulations.
patient positioning is propagated from the MRI
control system, and testing for ghosting artifacts
to validate the fidelity of the gradient system are Staffing and Training
performed. At this stage, testing the MR imaging
quality may be considered by imaging healthy eam Member Roles
T
volunteers or patients. This imaging data may be and Responsibilities
valuable to use in workflow development and
testing in the subsequent stages of the It is important to properly staff the MR-Linac
installation. program in order to guarantee seamless and effi-
The linac-acceptance testing schedule includes cient workflows. Depending on how many
safety, dosimetry (beam alignment), and mechan- patients are intended to be treated on the machine,
ical (MLCs, treatment table alignment, the minimum recommended staffing require-
MR-to-MV) checks. Acceptance of the separate ments for daily use are two radiation therapists,
parts of the MR-Linac, in this phase, will be two dosimetrists, two medical physicists, and two
according to guidelines and procedures of Elekta radiation oncologists.
(or provided by Philips regarding the MR The dosimetrist generates the offline reference
system). plans to be used for daily adaptation: adapt-to-
The second stage of the MR-Linac installa- position or adapt-to-shape. Once the reference
tion will be the ‘clinical readiness phase’ in plan during the initial planning process is
which training to operate the system (MR imag- reviewed and approved by the radiation oncolo-
ing, delivery, TPS, workflow, physics) will be gist, the dosimetrist’s role, depending on the
given, beam data is acquired to build a system clinic, can be extended into the online adaptive
specific beam model, and the physics of the sys- workflow process by being involved with online
tem is validated. In this phase, physics commis- contouring and optimization. The radiation
sioning is completed where the customer’s own oncologist’s role within the online adaptive plan-
QA procedures and tests can be performed. It is ning process is crucial; they focus on editing the
recommended that the customer’s QA activities planning target volumes and recontouring some
are planned at an early stage in order to save major organs at risk, for example, gastrointestinal
structures near the targets. The radiation oncolo- Staff Training and Education
gist also reviews the online adaptive plans and
makes the final approvals. Training and education for implementing any
The medical physicist plays an important role new technology comes at phases. First is manu-
in various aspects of the MR-Linac program. The facturer or vendor-based training and the second
medical physicist works closely with the vendor is institutional training. Vendor training is geared
team during acceptance testing and ensures the toward medical physicists, dosimetrists, and radi-
quality and the safety of the system through com- ation therapists. The medical physicist is expected
missioning and performing routine QA. They to play a major role for ongoing institutional
also develop various workflows and perform staff training, so it is highly recommended that they
training. Their role also extends into both the would be involved in all various manufacturer
offline and online adaptive process. The medical course training for all treatment team members.
physicist works closely with the dosimetrist to One major difference between a conventional
ensure the quality of the reference plan; they also linear accelerator and the MR-Linac is the exis-
perform patient-specific QA offline prior to tence of the magnetic field, adding another layer
patient treatment. During the online adaptive pro- for safety training which most therapy physicists
cess, the medical physicist assists with contour- are not familiar with [28, 29]. It is important to
ing by focusing on contours which could affect learn from the radiology department on the clini-
the dose calculation, such as high-density materi- cal safety practices pertaining to the existence of
als, air cavities, calcifications, and external con- a magnetic field within the environment. These
tours. They also perform the optimization and procedures can be transferred to the radiation
online QA. The radiation therapist’s roles include oncology and implemented within the MR-Linac
the performance of the daily quality assurance, program. If an existent MRI simulation program
patient safety screening, patient setup, image is present at the radiation oncology department,
acquisition and registration, and machine opera- similar practices are to be implemented. An insti-
tion. In more advanced programs, the radiation tutional annual MRI safety refresher training is
therapist can play a role in the online adaptive recommended for all MR-Linac team members
workflow by performing daily image segmenta- including the radiation oncologists, radiation
tion and contouring. therapists, and medical physicists. This annual
The availability of linac engineers is institu- training is generally conducted by a specialized
tional specific; however, direct access to techni- therapist or a medical physicist.
cal support from the vendor to address any It is also important for the radiation therapy
machine needs is of importance especially in the physicist to work closely with an MRI physicist
early phase of starting the MRgRT program. to implement a quality assurance program per-
With regard to IT support, depending on the com- taining to the MRI imaging system and to learn to
plexity of the program, IT support specialists identify artifacts which may occur, for example,
should work closely with the MR-Linac physics metal artifacts. It is also important to identify
team to reduce any communication issues that whether distortion corrections have been applied
may occur and to support with any software to the images, among other image quality
upgrades. IT should ensure that the MR-Linac is factors.
fully integrated within the hospital system and The medical physicist’s role is to train the
patient image data can be routinely archived to radiation therapists on daily QA procedures and
avoid storage issues which may cause software to set daily thresholds. The medical physicist is
speed slowness. The medical physicist works expected to also work closely with the dosime-
closely with the technical teams to ensure any trists and train them on various aspects of plan-
software and hardware upgrades are safely imple- ning which deviate from conventional planning
mented within the clinic and to provide any addi- due to construction differences and the presence
tional staff training. of a magnetic field.
Workflow Setup and Considerations After structure delineation (e.g., GTV/CTV,

organ-at-risks, body), the CT or MRI images,
Pretreatment Preparations including all structures, are imported into the
and Online Workflow Unity treatment planning system for offline plan-
ning. Next, couch and coil structures are added to
The pretreatment process (Fig. 10.4) is similar to the data set. Electron density information is
the conventional pretreatment process in radio- assigned to each structure to be used for online
therapy. It starts with the imaging of a patient, MRI planning using a bulk density override tech-
which can either be a CT or MRI as primary nique. Electron density information can be
image. Other image data sets such as CT, MRI, extracted from the simulation CT. An IMRT step-
and PET can be registered to help with tumor and and-shoot plan is then created and evaluated to be
organ delineation. This can all be performed in used as the daily reference plan (Fig. 10.4).
imaging or planning software. It is prudent to Prior to treatment, it is important to screen
invest in image registration software that can help patients daily for MRI compatibility and safety
streamline transfer, fusion, and segmentation to prior to entering the treatment vault. The patient
provide seamless operation. is then positioned on the couch using a couch
Fig. 10.4 MR-Linac clinical treatment workflows. the daily patient translations (adapt-to-position) or to
Above, the preparation phase to create the patient-specific incorporate the daily anatomy deformations
reference plan. Below, the online workflow to correct for (adapt-to-shape)
index system. Next, the coil is placed over the among team members to generate contours
target area for optimal MR imaging. After posi- simultaneously. Once the contours are gener-
tioning the patient, an MRI image is taken, with a ated, they are merged and transferred back to the
preselected patient-specific 3D sequence or an online treatment planning system [31]. The
average image reconstructed from a 4D MRI, and adapted plan is then created on the MRI starting
imported to the online treatment planning sys- either from fluence or from segments of the ref-
tem. The MRI is registered to the planning image erence plan. For the latter, the options are to use
(CT or MRI), and the plan is adapted to the cur- the original segments, adaptation of the seg-
rent anatomical situation. This can be done using ments using the changing projection of the struc-
two different workflows: adapt-to-position (ATP) tures, or optimizing the weight and/or shape of
or adapt-to-shape (ATS) [30]. the segments.
The ATP workflow is analogous to image After the plan (Fig. 10.5) is evaluated and
guided radiotherapy on a conventional linac as no approved by the radiation oncologist, the plan is
contouring and replanning is performed. ready to be delivered. At this point, there is an
However, since the Unity couch is static, the iso- option to evaluate the target position using a sec-
center on the pretreatment planning image is ondary verification scan taken during optimiza-
shifted using the online registration results. The tion and, if necessary, adapt the plan just before
plan is then adapted to the shifted isocenter by the actual radiation for any additional shifts.
using either the original segments, adaptation of Patient-specific QA is performed following plan
the segments using the changing projection of the acceptance [32]. During delivery, three-plane 2D
structures, or optimizing the weight and/or shape real-time imaging is used for tracking where it is
of the segments. The dose calculation is carried possible to use gating or to correct for target
out on reference plan image set. drifts; motion management has been described in
In the ATS workflow, the structures are propa- detail in Chap. 6. However, in an unconventional
gated to the online MR image using deformable workflow, tracking can be disabled if it’s not
registration, and the structures can be evaluated needed (depending on treatment site) to allow for
and edited if necessary. An unconventional alter- additional volumetric MRI sequence acquisition
native to the vendor’s standard workflow is to on the imaging system. Following treatment,
use external software for parallel contouring, posttreatment scans can be acquired for
where contour assignments can be divided evaluation.
a b
Fig. 10.5 Examples of MR-Linac patient treatment planning for (a) prostate and (b) rectum patient
Implementation of New Sites workflow should be trained with the personnel

and task should be clear to each group. Also sev-
Implementation of new workflows depends on eral tasks might be passed on to other professions
the availability of software, hardware, staff, and to smoothen the workflow, such as contour adap-
other resources. However, some general items tation or plan evaluation.
can be considered before treating patients on the
MR-Linac.
First, the patient positioning and immobiliza- Fallback Scenarios
tion devices being used should be considered.
Physical limitations such as the bore size should In some cases, the radiation delivery can be inter-
be taken into consideration when selecting rupted during treatment due to machine issues or
patients and creating/purchasing immobilization a patient emergency. Depending on the severity
devices. All devices must be MRI safe and com- of the issues, the radiation treatment may be
patible. MRI sequences can be chosen based on resumed if the issues can be resolved within a
the particular site where image quality, acquisi- reasonable time such as a quick system reset or
tion time, and geometrical fidelity are optimized. restarting the treatment control software while
Consensus for target and organs at risk delinea- maintaining the treatment plan and patient posi-
tion should be established in advance to save tion. However, if the issues can’t be resolved
time during the online workflow. Personnel within a reasonable time and the treatment must
training for contour delineation should take be aborted, a completion plan will need to be
place following establishing guidelines espe- generated by following vendor’s instruction. The
cially for those who lack experience. In the completion plan only contains the remaining
online process, contours can be propagated to gantry angles and MUs and uses the ATP work-
the online MRI in the ATS workflow, but if adap- flow assuming no anatomical changes. It is not
tation of the contours is necessary, the accept- possible to generate an adaptive plan using the
able time needed and who will adapt the contours completion plan. In such scenario, it is difficult to
should be decided up front. PTV margins are estimate the dosimetric quality of the completion
based on the total uncertainties in the online plan in the Monaco TPS, but it is possible to
planning process such as the remaining intrafrac- export the delivered plan dose and daily comple-
tion motion and might differ for different patient tion adaptive plan dose to a third-party image
groups. analysis software package to estimate the
During optimization, MLC field shapes are summed 3D dose distribution.
optimized for the predefined gantry angles, tak-
ing the magnetic field into account in the dose
calculation. For online MRI planning, an adap- Acceptance, Commissioning,
tive workflow has to be developed focusing on and Routine QA
plan quality and optimization time, thus consid-
ering ATP versus ATS but also the way to adapt Acceptance testing, commissioning, and routine
the plan [24, 33]. QA for MR-Linac include tests for the MRI
Various methods can be used to perform imaging system [34] and Linac radiation delivery
online patient QA. Most clinics use an indepen- system [35, 36]. Most of these tests and routine
dent calculation software or perform a plan QA for the radiation delivery system follow pub-
parameter check during the plan evaluation pro- lished clinical practice guideline such as that
cess. Also, plans could be measured after irradia- from the American Association of Physicists in
tion. Some backup procedures should be Medicine (AAPM) for medical accelerators used
implemented. The systems provide a procedure for radiation therapy. For example, task group
to finish a plan after the workflow is stopped dur- reports TG-45 [37] and TG-142 [38] can be used
ing radiation (described below). Finally, the for major guidance. However, several adaptions
to the existing protocols have to be applied to lished by Tijssen et al. [34] based on the MRI
account for the influence of the magnetic field performances of four MR-Linacs from members
from the onboard MRI, the nonstandard source of the Elekta MR-Linac Consortium. A set of
axis distance (SAD = 143.5 cm), a beamline tests for the onboard MRI system were suggested
passing through a cryostat wall, and an MR ante- for the commissioning, quality control (QC), and
rior coil before reaching patient. Furthermore, QA measurements as well as system configura-
the magnetic field will affect the properties of the tion and connectivity (SCC). The SCC includes
QA devices [6, 39]. Acceptance testing, commis- software configuration checks, receiving coils
sioning, and routine QA items and procedures for and other peripherals devices of the MR scanner,
radiation delivery system will need to be tailored DICOM image storing and transferring, and con-
or changed to account for these differences. QA nectivity to hospital picture archiving and com-
devices should be MR safe and MR compatible. munication systems (PACS). QC tests include
Setup of the dosimeters during testing and QA assessment of the main static B0 field inhomoge-
may be different due to the magnetic field effect neities, gradient field nonlinearity, and RF field
on the effective position of measurement. A cus- and hybrid tests to check the interference between
tomized water tank and nonstandard beam data the linac and MR scanner (e.g., direction of the
should be collected for the treatment planning B0 field, gantry rotation, RF interference). QA
system due to geometry limitations of the measurements periodically check the general
MR-Linac and magnetic field effect. Additional image quality, such as the signal to noise ratio,
test such as cryostat characteristics must be perlow contrast detectability, and gradient fidelity.
formed due to the beamline passing through the Most MR tests can be done following the guid-
cryostat wall. ance of TG-284. Current phantoms used on the
MR-Linac for MRI image quality QA include
Philips uniformity disc phantom, GNL phantom,
erifying the Onboard MRI Imaging
V periodical image quality test (PIQT) phantom,
System ACR phantom, and Modus 3D geometry distor-
tion phantom.
Contrary to diagnostic MRI, which is used for
detection, characterization, and staging of dis-
ease, the onboard MRI is used during daily Verifying the Linac Delivery System
patient treatment for contouring in adaptive
replanning, motion monitoring, and acquiring Machine QA of the MR-Linac includes geomet-
functional images for offline evaluation. It rical and dosimetric verification, but the effort is
requires high geometric fidelity and high image limited as there is just a single energy (7 MV), a
quality for patient registration and contour delin- beamline mounted on a stiff gantry ring, and a
eation. However, no formal guidelines on com- fixed collimator angle (90 degree). The delivery
missioning and routine QA for the onboard MRI system is aligned to the MRI coordinate system
system have been yet published. TG-284 [40] using a dedicated MR-to-MV alignment phantom
provides recommendations for an MR simulator (Fig. 10.6b).
for acceptance and commissioning testing, as PTW BEAMSCAN MR water phantom
well as periodic QA. Some tests can be tailored (Fig. 10.6a) is currently the only commercially
for QA of the onboard MR system of the available MR compatible water tank for radiation
MR-Linac. However, additional tests are needed beam data collection on the Unity MR-Linac.
to check interference and isocenter coincidence Due to the bore size limitation, the percentage
and to verify the communication between the depth dose (PDD) and beam profiles are collected
imaging system and the linac. not only at beam angle 0° (depth can only reach
A multi-institutional study on MRI commis- at about 10 cm with SSD =133.5 cm at this direc-
sioning of the 1.5T MR-Linac system was pub- tion) but also at 270°.
a b c
Fig. 10.6 (a) The PTW BEAMSCAN MR water phan- ter of 0.4 mm. The film was acquired on a Unity MR-Linac
tom. (b) MR-to-MV alignment phantom, provided by using a phantom with copper rings to avoid the influence
Elekta, used to align the delivery and the imaging system. of the Lorentz force on each spoke
(c) Spoke film showing a locus of radius of gantry isocen-
MR-to-MV alignment test is a unique mea- MLC leaf positions are verified using a stan-
surement for the Unity MR-Linac, which is used dard picket-fence test of adjacent small field slits
to determine the origin-shift of MR imaging to with a 1 mm gap in between (Fig. 10.7).
MV radiation. The origin-shift values on three- Measurements can be performed with film or the
dimensional orthogonal coordinate axis x, y, and onboard EPID, but due to the large maximum
z are measured during commissioning and field size (57 cm × 22 cm), only a portion of the
applied to the TPS for daily adaptive planning. full extent can be captured on the EPID (22 cm ×
This test needs to be repeated periodically to 8.5 cm). Analysis of the measurement quantifies
make sure the drifting of the values is within tol- deviation from this 1 mm for each leaf pair sepa-
erance and is stable over time. The phantom used rately over the leaf trajectory.
for MR-to-MV alignment consists of several Magnetic field must be taken into account dur-
ceramic balls surrounded with MR visible solu- ing output calibration of Unity MR-Linac with an
tion (Fig. 10.6b) and is provided with analysis ion chamber since the magnetic field will change
software by the vendor. the trajectory path of the secondary electrons in
Verification of the beam direction to the iso- the surrounding medium and inside the gas filled
center (isocenter accuracy) is performed by an volume of the detector, thus the response of the
adapted starshot film measurement [41]. The detector [13, 42]. Magnetic field correction fac-
adapted setup uses a film, sandwiched between tors are introduced to account for dose changes
two copper rings to prevent a dose print by the (fixed value for the 1.5T Unity setup) and detec-
secondary electrons affected by the magnetic tor response. Correction values for detector
field (Fig. 10.6c). The spokes, drawn between the response depend on the ion chamber type (avail-
two opposing dose prints at the copper ring, able for the commonly used detectors) and beam
should intersect the isocenter. The distance over quality kQ [43–45]. Both O’Brien [42] and
all spokes to the isocenter can be calculated and Malkov [46] pointed out that TPR20/10 (IAEA
is about half the distance compared to a conven- TRS-398 [47] suggestion for kQ determination) is
tional linac. more stable than PDD10 (TG-51 [48] suggestion
Fig. 10.7 Film and EPID image (covering the area within the blue rectangle) of the picket-fence test to verify the MLC
positions on the MR-Linac. The right plot shows the profile for a single leaf throughout its trajectory
for kQ); thus, it is recommended to use TPR20/10 ing). Additional daily QA can be performed
to obtain the kQ. No consensus of the location of including functionality checks for metal detec-
the reference point for the dose calibration has tors, helium level and MRI cold head, bore
been made; therefore, one can use a reference inspection for any metal parts, and linac running
point at the depth of the maximum dose, or 5 cm parameters (magnetron power, waveguide water
and 10 cm at beam angle 0° or 90°. temperatures, SF6 pressure). Due to the com-
A waterproof Farmer-type ion chamber should plexity of the treatment workflow, a daily work-
be used for reference dosimetry as measurements flow test is favorable in order to verify
have to be performed in water to prevent air gaps communication between every component of the
around the detector [46, 49–51]. prior to daily treatments.
A weekly image quality test with three scans
is recommended using the PIQT phantom to
Routine QA evaluate the key properties of the MR system,
i.e., the flood-field uniformity, spatial linearity,
The full list of checks and measurements, includ- slice profile, spatial resolution, and noise (accep-
ing their frequency and tolerances, was suggested tance levels are given [35]). The MR-to-MV
in the report of machine QA for the Elekta Unity alignment test can either be performed weekly or
system from Elekta MR-Linac Consortium [35] monthly, where the maximum deviation from
(see Table 10.1). For daily QA, photon beam out- baseline not to exceed ±0.5 mm and ±0.3°. Linac
put constancy (±3%) can be measured with an output and backup monitor consistency should be
ion chamber, daily QA device (e.g., Sun Nuclear verified weekly with an ion chamber or detector
DailyQA3 MR), onboard MV imaging detector array (acceptance level is ±2%). MLC and dia-
(center pixel value is representative and stable), phragm positions should be verified either by the
or the linac secondary (backup) monitor unit onboard EPID or film (tolerance ±1 mm at iso-
chamber. Safety checks include functionality of center plane).
the audio-video equipment, couch panel emer- Monthly linac QA includes photon beam pro-
gency button, interlocks, and beam on indicator. file and beam quality constancy verification.
MR checks include devices (coil and accessories, Mechanical QA includes checks on gantry angle,
call bulb) and image quality checks (SNR, scal- table position, MLC field size, leaf position and
radiation isocenter size, QA platform alignment, volume (DSV). It may also be feasible to sepa-
laser alignment (if any), and EPID image quality. rately measure the B0 field uniformity. The accep-
Safety checks include radiation door interlock tance levels for monthly QA checks are given by
checks, pause and terminate key function on the Roberts et al. [35].
keypad, and stop motors. Monthly MR checks Annual dosimetry QA is more comprehen-
include MR geometric accuracy tests in volumes sive and includes verification of the beam using
of 200, 300, and 400 mm diameter of spherical the 3D water tank and ionization chambers and/
Table 10.1 An overview of all checks and measurements, including their frequency and tolerances, as suggested in the
report of machine QA for the Elekta Unity system from Elekta MR-Linac Consortium [35]
Frequency Category Test Device Acceptance level Note
Daily Dosimetry Beam output EPID ±3% Suggested by
IC with water David Robert et al.
IC with solid water [35]
Backup dose monitor Linac console display ±3%
Safety Audio-visual monitor Functional
CCTV Functional
Intercome system Functional
Equipment check (coil, Functional
PPD)
Table emergency stop Functional
button
Radiation door interlock Functional
Radiation area monitor Functional
(if used)
Beam on indicator Functional
MR Signal to noise ratio PIQT phantom Vendor provided
Scaling test (TRA and Vendor provided
COR)
Call bulb Functional
Additional Metal detector Functional More tests
Cold head Functional performed in some
Helium level Check level clinics
End-to-end workflow test Functional
Weekly Dosimetry Beam output IC with 1D water ±2% Suggested by
tank David Robert et al.
IC with water [35]
equivalent plastic
Multi-axis linear or
2D detector array
Mechanicl MLC and diaphragm EPID ±1 mm
position Gafchromic film
Multi-axis linear or
2D detector array
MR PIQT PIQT phantom Vendor provided
MR-to-MV MR-to-MV aligment test MR-to-MV ±1 mm
alignment phantom
Table 10.1 (continued)

Frequency Category Test Device Acceptance level Note
Monthly Dosimetry Beam output IC with water ±2% Suggested by
equivalent plastic David Robert et al.
Beam profile EPID or multi-axis ±2% [35]
linear detetctor array
Beam quality IC with water ±1%
equivalent plastic
Mechanical Gantry angle Level 0.3 degree
Table position MV alignment ±1 mm
phantom
BLD field size check EPID, Gafchromic ±2 mm
film, 2D detector
array
Leaf position accuracy ±1 mm
Radiation isocenter size MV alignment ±0.5 mm
phantom
Safety Safety interlock Functional
MR GNL test Philips GNL 200 mm
distortion phantom DSV ≤ 1 mm
300 mm
DSV ≤ 2 mm
400 mm
DSV ≤ 4 mm
Additional B0 uniformity Philips 40 cm disc Ccompared to More tests
phantom baseline performed in some
clinics
Annual Dosimetry Beam output IC with 1D water ±1% Suggested by
tank David Robert et al.
Beam profile comparison IC with 3D water ±2% [35]
from baseline tank
Output factor IC with 3D water ±2% for field
tank size≤4 × 4 cm2
±1% for field
size>4 × 4 cm2
Beam quality IC with water ±1%
equivalent plastic
MU linearity IC with water 5% for 2–4 MU
equivalent plastic 2% for ≥5 MU
Output vs dose rate IC with water ±2% from baseline
equivalent plastic
Output vs gantry angle IC with water ±2% from baseline
equivalent plastic
Mechanical Similar to monthly QA ±0.5 mm
leaf position repeatability
MR B0 uniformity vs gantry Philips 40 cm disc Compared to More tests
angle phantom baseline performed in some
clinics
or microdiamond detector. Dosimetry related phantom. Results should be presented as absolute

checks are machine dose output, output factors deviations in position and dose with respect to
and profiles with a combination of field sizes the intended values.
and gantry angles, beam quality, MU linearity, Patient-specific QA is closely related to end-
X-ray output constancy vs. dose rate, and X-ray to-end testing but only verifies specific parts of
output constancy vs. gantry angle. Mechanical the workflow, namely, treatment planning and
checks are similar to the monthly QA. MR QA delivery. The generated plan (derived from pre-
is covered in the daily, weekly, and monthly QA treatment preparations or online workflow) is
items also in addition to the verification of the recalculated on the phantom geometry and com-
MR main frequency variation vs. gantry pared to the measurement of the delivered dose
rotation. on the phantom. Positional errors of the phantom
setup are generally disregarded. Phantoms for
patient-specific plan QA include PTW Octavius
End-to-End Verification and Patient- 1500MR, SNC ARcCHeckMR, ScandiDos
Specific Plan QA Delta4 Phantom, Gafchromic film, and radio-
chromic gel.
Due to the complexity of the daily MR-Linac Patient-specific plan QA is particularly chal-
system and daily treatment workflow, any lenging regarding online adaptive radiotherapy.
machine or communication issues could signifi- The patient cannot leave the treatment position
cantly increase patient-on-couch time or result in on the couch for any verification. Quality assur-
interruption/aborting of the patient treatment. ance is often performed on the pretreatment plan,
Issues might occur at and between the various being a representative plan for the online work-
parts of the linac and MR imaging system, the flow. Online QA might include visual inspection
treatment session manager, treatment planning of new contours and electron density map, inde-
system (Monaco), and record-and-verify system pendent dose calculation of the online treatment
(Mosaiq). Workflow tests are recommended to be plan using third-party software [53], and verifica-
performed daily to verify the functionality and tion of integrity of transferred data during the
communication between all parts of the workflow [32, 52]. Postdelivery verification can
MR-Linac used in the clinical workflow [52]. be performed using the treatment delivery log
After any major hardware or software updates, files [54–57] or physical measurements of the
modifications of the system, or implementation treatment plans [39, 58–61].
of a new workflow, an end-to-end test should be
performed using a phantom to mimic patient
treatment which includes testing the various Summary
components of the workflow. This includes
patient (or phantom) setup, MR pretreatment The Elekta MR-Linac, combining a linear accel-
imaging, adaptive RT plan generation, MR imag- erator with onboard 1.5T MR imaging system,
ing during treatment, delivery, and posttreatment allows for accurate target definition with accurate
MR imaging. Although not specified and ana- radiation delivery. Daily treatment adaptation,
lyzed separately, the “overall” end-to-end test based on the online acquired MRI, is performed
includes all uncertainties related to imaging, using an online workflow. The system is in its
beam model (e.g., beam profiles, MLC transmis- early phases of adoption of real-time gating and
sions, couch), dose delivery, and system commu- tracking. Although dose delivery is affected by
nication. Examples on phantoms which can be the magnetic field, it is considered in the dose
utilized for end-to-end testing are SNC calculation engine and plan optimization. The
StereoPHAN, Standard Imaging Lucy Phantom, Unity system has become mature over the last
IROC Houston phantom, MR-to-MV phantom, few years and is becoming more widely available
CIRS STEEV, and Anthropomorphic PRIME worldwide. Installation and quality assurance
procedures (described in literature and reports) 14. Raaymakers BW, et al. Integrating a MRI scanner
have been developed and implemented for both with a 6 MV radiotherapy accelerator: dose deposi-
tion in a transverse magnetic field. Phys Med Biol.
the imaging and the dose delivery components of 2004;49(17):4109–18.
the system. Appropriate training of the staff is 15. Ahmad SB, et al. Evaluation of a commercial MRI
necessary for MR-Linac operation, treatment Linac based Monte Carlo dose calculation algorithm
planning, safety due to the presence of the mag- with GEANT4. Med Phys. 2016;43(2):894–907.
16. Ahmad SB, et al. Backscatter dose effects for high
netic field, support and maintenance, online atomic number materials being irradiated in the pres-
replanning, and workflow efficiency. ence of a magnetic field: a Monte Carlo study for the
MRI linac. Med Phys. 2016;43(8):4665.
17. Malkov VN, et al. Monte Carlo simulations of out-
of-field surface doses due to the electron streaming
References effect in orthogonal magnetic fields. Phys Med Biol.
2019;64(11):115029.
1. Lagendijk JJW, Bakker CJG. MRI-guided radiother- 18. Park JM, et al. Air-electron stream interactions during
apy: a MRI based linear accelerator. Radiother Oncol. magnetic resonance IGRT: skin irradiation outside the
2000;56(S1):S60. treatment field during accelerated partial breast irra-
2. Lagendijk JJW, et al. MRI/linac integration. Radiother diation. Strahlenther Onkol. 2018;194(1):50–9.
Oncol. 2008;86(1):25–9. 19. Hehakaya C, et al. Early health economic analysis of
3. Raaymakers BW, et al. Integrating a 1.5 T MRI scan- 1.5 T MRI-guided radiotherapy for localized prostate
ner with a 6 MV accelerator: proof of concept. Phys cancer: decision analytic modelling. Radiother Oncol.
Med Biol. 2009;54(12):N229–37. 2021;161:74–82.
4. Raaymakers BW, et al. First patients treated with a 20. Palm RF, et al. Assessment of MRI-Linac Economics
1.5 T MRI-Linac: clinical proof of concept of a high- under the RO-APM. J Clin Med. 2021;10(20):4706.
precision, high-field MRI guided radiotherapy treat- 21. Yu JB, et al. National patterns of early adop-
ment. Phys Med Biol. 2017;62(23):L41–50. tion of magnetic resonance imaging-guided linear
5. Werensteijn-Honingh AM, et al. Feasibility of ste- accelerators in 2018 to 2019. Adv Radiat Oncol.
reotactic radiotherapy using a 1.5T MR-linac: 2023;8(3):101167.
multi-fraction treatment of pelvic lymph node oligo- 22. Hehakaya C, et al. Implementation of magnetic
metastases. Radiother Oncol. 2019;134:50–4. resonance imaging-guided radiation therapy in rou-
6. de Pooter J, et al. Reference dosimetry in MRI-linacs: tine care: opportunities and challenges in the United
evaluation of available protocols and data to establish States. Adv Radiat Oncol. 2022;7(5):100953.
a Code of Practice. Phys Med Biol. 2021;66(5):05tr02. 23. Jagt TZ, et al. Benchmarking daily adaptation using
7. Raaijmakers AJ, Raaymakers BW, Lagendijk fully automated radiotherapy treatment plan optimi-
JJ. Integrating a MRI scanner with a 6 MV radiother- zation for rectal cancer. Phys Imaging Radiat Oncol.
apy accelerator: dose increase at tissue-air interfaces 2022;24:7–13.
in a lateral magnetic field due to returning electrons. 24. Dunkerley DAP, et al. Clinical implementational and
Phys Med Biol. 2005;50(7):1363–76. site-specific workflows for a 1.5T MR-Linac. J Clin
8. Hackett SL, et al. Spiraling contaminant electrons Med. 2022;11(6):1662.
increase doses to surfaces outside the photon beam 25. Speight R, et al. IPEM topical report: a 2018 IPEM
of an MRI-linac with a perpendicular magnetic field. survey of MRI use for external beam radiotherapy
Phys Med Biol. 2018;63(9):095001. treatment planning in the UK. Phys Med Biol.
9. Malkov VN, et al. Monte Carlo simulations of out- 2019;64(17):175021.
of-field skin dose due to spiralling contaminant elec- 26. Kim J, et al. Implementation of a novel algorithm
trons in a perpendicular magnetic field. Med Phys. for generating synthetic CT images from mag-
2019;46(3):1467–77. netic resonance imaging data sets for prostate can-
10. Oborn BM, et al. High resolution entry and exit Monte cer radiation therapy. Int J Radiat Oncol Biol Phys.
Carlo dose calculations from a linear accelerator 6 2015;91(1):39–47.
MV beam under the influence of transverse magnetic 27. Kok JG, et al. Installation of the 1.5 T MRI accelera-
fields. Med Phys. 2009;36(8):3549–59. tor next to clinical accelerators: impact of the fringe
11. Raaijmakers AJ, Raaymakers BW, Lagendijk field. Phys Med Biol. 2009;54(18):N409–15.
JJ. Magnetic-field-induced dose effects in MR-guided 28. Cao M, Padgett KR, Rong Y. Are in-house diagnostic
radiotherapy systems: dependence on the magnetic MR physicists necessary for clinical implementation
field strength. Phys Med Biol. 2008;53(4):909–23. of MRI guided radiotherapy? J Appl Clin Med Phys.
12. Woodings SJ, et al. Beam characterisation of the 1.5 T 2017;18(5):6–9.
MRI-linac. Phys Med Biol. 2018;63(8):085015. 29. McGee KP, et al. Findings of the AAPM Ad Hoc com-
13. van Asselen B, et al. A formalism for reference dosim- mittee on magnetic resonance imaging in radiation
etry in photon beams in the presence of a magnetic therapy: Unmet needs, opportunities, and recommen-
field. Phys Med Biol. 2018;63(12):125008. dations. Med Phys. 2021;48(8):4523–31.
30. Winkel D, et al. Adaptive radiotherapy: The Elekta 46. Malkov VN, Rogers DWO. Sensitive volume effects
Unity MR-linac concept. Clin Transl Radiat Oncol. on Monte Carlo calculated ion chamber response in
2019;18:54–9. magnetic fields. Med Phys. 2017;44(9):4854–8.
31. Paulson ES, et al. 4D-MRI driven MR-guided online 47. Andreo P, et al. Absorbed dose determination in
adaptive radiotherapy for abdominal stereotactic body external beam radiotherapy - an international code of
radiation therapy on a high field MR-Linac: imple- practice for dosimetry based on standards of absorbed
mentation and initial clinical experience. Clin Transl dose to water, in technical report series. Vienna:
Radiat Oncol. 2020;23:72–9. International Atomic Energy Agency; 2000.
32. Chen X, et al. A daily end-to-end quality assurance 48. Almond PR, et al. AAPM’s TG-51 protocol for clini-
workflow for MR-guided online adaptive radiation cal reference dosimetry of high-energy photon and
therapy on MR-Linac. J Appl Clin Med Phys. electron beams. Med Phys. 1999;26(9):1847–70.
2020;21(1):205–12. 49. Hackett SL, et al. Consequences of air around an ion-
33. Gupta A, et al. Online adaptive radiotherapy for head ization chamber: are existing solid phantoms suitable
and neck cancers on the MR linear accelerator: intro- for reference dosimetry on an MR-linac? Med Phys.
ducing a novel modified Adapt-to-Shape approach. 2016;43(7):3961.
Clin Transl Radiat Oncol. 2022;32:48–51. 50. O’Brien DJ, et al. Small air-gaps affect the response
34. Tijssen RHN, et al. MRI commissioning of 1.5T of ionization chambers in the presence of a 1.5 T mag-
MR-linac systems - a multi-institutional study. netic field. Med Phys. 2015;42(6):3724.
Radiother Oncol. 2019;132:114–20. 51. O’Brien DJ, Sawakuchi GO. Monte Carlo study of the
35. Roberts DA, et al. Machine QA for the Elekta Unity chamber-phantom air gap effect in a magnetic field.
system: a report from the Elekta MR-linac consor- Med Phys. 2017;44(7):3830–8.
tium. Med Phys. 2021;48(5):e67–85. 52. Malkov VN, et al. MR-linac daily semi-automated
36. Woodings SJ, et al. Acceptance procedure for the lin- end-to-end quality control verification. J Appl Clin
ear accelerator component of the 1.5 T MRI-linac. J Med Phys. 2023;24(5):e13916.
Appl Clin Med Phys. 2021;22(8):45–59. 53. Wang Y, et al. Development of a fast Monte Carlo
37. Nath R, et al. AAPM code of practice for radio- dose calculation system for online adaptive radia-
therapy accelerators: report of AAPM Radiation tion therapy quality assurance. Phys Med Biol.
Therapy Task Group No. 45. Med Phys. 2017;62(12):4970–90.
1994;21(7):1093–121. 54. Agnew CE, Irvine DM, McGarry CK. Correlation of
38. Klein EE, et al. Task Group 142 report: qual- phantom-based and log file patient-specific QA with
ity assurance of medical accelerators. Med Phys. complexity scores for VMAT. J Appl Clin Med Phys.
2009;36(9):4197–212. 2014;15(6):4994.
39. Houweling AC, et al. Performance of a cylindrical 55. Menten MJ, et al. Automatic reconstruction of the
diode array for use in a 1.5 T MR-linac. Phys Med delivered dose of the day using MR-linac treatment
Biol. 2016;61(3):N80–9. log files and online MR imaging. Radiother Oncol.
40. Glide-Hurst CK, et al. Task group 284 report: mag- 2020;145:88–94.
netic resonance imaging simulation in radiother- 56. Lim SB, et al. An investigation of using log-file analy-
apy: considerations for clinical implementation, sis for automated patient-specific quality assurance in
optimization, and quality assurance. Med Phys. MRgRT. J Appl Clin Med Phys. 2021;22(9):183–8.
2021;48(7):e636–70. 57. Kontaxis C, et al. Delivered dose quantification in
41. van Zijp HM, et al. Minimizing the magnetic prostate radiotherapy using online 3D cine imaging
field effect in MR-linac specific QA-tests: the and treatment log files on a combined 1.5T magnetic
use of electron dense materials. Phys Med Biol. resonance imaging and linear accelerator system.
2016;61(3):N50–9. Phys Imaging Radiat Oncol. 2020;15:23–9.
42. O'Brien DJ, et al. Reference dosimetry in magnetic 58. Yang, B., et al. Initial clinical experience of patient-
fields: formalism and ionization chamber correction specific QA of treatment delivery in online adaptive
factors. Med Phys. 2016;43(8):4915. radiotherapy using a 1.5 T MR-Linac. Biomed Phys
43. de Prez L, et al. First water calorimetric Dw measure- Eng Express. 2021;7(3).
ment and direct measurement of magnetic field cor- 59. Xhaferllari I, et al. Clinical utility of Gafchromic film
rection factors, KQ, B, in a 1.5 T B-field of an MRI in an MRI-guided linear accelerator. Radiat Oncol.
Linac. Med Phys. 2016;43(6):3874. 2021;16(1):117.
44. de Prez LA, et al. Direct measurement of ion chamber 60. Mönnich D, et al. Quality assurance of IMRT treat-
correction factors, kQ and kB, in a 7 MV MRI-linac. ment plans for a 1.5 T MR-linac using a 2D ionization
Phys Med Biol. 2019;64(10):105025. chamber array and a static solid phantom. Phys Med
45. Malkov VN, Rogers DWO. Monte Carlo study of Biol. 2020;65(16):16nt01.
ionization chamber magnetic field correction factors 61. de Vries JHW, et al. Characterization of a prototype
as a function of angle and beam quality. Med Phys. MR-compatible Delta4 QA system in a 1.5 tesla
2018;45(2):908–25. MR-linac. Phys Med Biol. 2018;63(2):02NT02.
The Alberta Rotating Biplanar
Linac-MR, a.k.a., Aurora-RT™
11
B. Gino Fallone, Satyapal Rathee, Nicola de Zanche,
Eugene Yip, Keith Wachowicz, and Jihyun Yun
Introduction lel B0-field to radiation beam configuration. The

parallel configuration not only avoids electron-
MagnetTx Oncology Solutions (www.MagnetTx. return effect (ERE) [1] but with significant hori-
com, Edmonton, Alberta, Canada) which is a zontal and vertical shifts allows the positioning
start-up from the Alberta Health Services (AHS) of any tumor (including those offset) to be posi-
and the University of Alberta (UofA) has exclu- tioned at the 2D planned position or at isocenter.
sive license to commercialize the rotating open The parallel configuration also allows, with
biplanar linac-MR system developed and couch movement and gantry rotation to position
installed at the Cross Cancer Institute (CCI), as in 3D, any offset tumor to the planned location or
the Aurora-RT™. The open biplanar MR, with a to the isocenter.
patient bore size of 110 cm by 60 cm, improves
patient experience by reducing claustrophobia,
provides space for significant lateral and vertical Historical Milestone
shifts, and allows perpendicular as well as paral- and Development
Significant scientific and engineering milestones

were achieved in developing the Aurora-RT. The
B. G. Fallone (*) Alberta group reported the first proof of principle
Department of Medical Physics, Cross Cancer
Institute, Edmonton, AB, Canada of a successful linac-MR hybrid system in 2008
which consisted of an open biplanar 0.2T MRI
Department of Oncology, Medical Physics Division,
University of Alberta, Edmonton, AB, Canada with about a 30 cm gap between the poles con-
nected in perpendicular configuration to a Varian
MagnetTx Oncology Solutions,
Edmonton, AB, Canada 6 MV medical linac and a Varian 52 leaf MLC,
e-mail: [email protected] both decommissioned by the CCI [2]. The system
S. Rathee · N. de Zanche · E. Yip · K. Wachowicz · allowed extensive (1) experimental studies in
J. Yun basic linac RF noise [3, 4], RF shielding [5],
Department of Medical Physics, Cross Cancer magnetic shielding [2] and MLC operation within
Institute, Edmonton, AB, Canada a magnetic field, and (2) theoretical investiga-
Department of Oncology, Medical Physics Division, tions that side coupling of a 6 MV linac provided
University of Alberta, Edmonton, AB, Canada proof of the clinical applicability and the ability
e-mail: [email protected]; [email protected];
[email protected]; [email protected]; to study linac performance within B0 [6]. Linac
[email protected] performance was not affected by electron beam
194 B. G. Fallone et al.
loss within the accelerating waveguide due to shape the magnetic field optimally and to mini-
magnetic fields that persist in the perpendicular mize the focusing of the contaminate. A beam
configuration even after shielding the linac [7]. stop on the gantry also provides counterbalance
The active shielding design provides magnetic for the gantry. An unobstructed radiation beam
decoupling of the linac from the imager, while path goes through a 25-cm-diameter opening at
the use of magnetically insensitive MLC motors the center of MRI planar poles. The LMR 1000
would avoid the need of passive shielding [8, 9]. bore size is only 58 cm diameter.
The parallel configuration was proposed in A much larger bore size (110 × 60 cm) was
2010 [1] with the unobstructed linac beam going provided by the LMR 2000 assembled in around
through a circular opening in one of the poles 2020 as shown in Fig. 11.1. It received Class II
[10] to avoid ERE, all together. A maximum Certification by the Canadian Nuclear Safety
magnetic field that sufficiently avoided ERE was Commission in March 2021. Aurora-RT received
0.5 T, with >0.5 T B0 still resulting in larger US FDA clearance in May 2022. The LMR 2000
ERE. We then evaluated the performance of the uses a custom 120 double-focused MLC project-
linac due to the parallel fringe magnetic field of a ing 0.475 mm leaf width at 120 cm SAD, with
0.5 T open imager. The longitudinal magnetic two independent Y rotating jaws. The dose rate at
field on the electron beam only affected in the 120 cm SAD from the custom FFF 6 MV linac is
electron gun optics, which can be eliminated by a 600 MU/min.
redesign of the electron gun optics or through The linac-MR gantry has a rotational range of
magnetic shielding [9, 11]. ±185° from the 0° at 0.1° resolution and position
The preferred design was parallel configura- accuracy with maximum nominal speed of 6°/s.
tion with a B0 of 0.5 T which provided acceptable The two A-frames holding the gantry have a rear
avoidance of ERE and maximum contrast-to- 60-cm-diameter patient-size bore and a front
noise for fast MR imaging [12]. A first whole- 110-cm-diameter patient-size bore. This provides
body rotating biplanar linac-MR (LMR 1000) a useable rectangular-elliptical rotating bore of
was built at the CCI/UofA in 2013–2014 using a 110 cm by 60 cm. The Kevlar couch has three
high-temperature superconducting custom open translational degrees of freedom with maximum
biplanar 0.5 MRI system with a decommissioned vertical and horizontal speeds of 5 cm/s and
Varian 600C FFF and a Varian Mark 80 MLC for 15 cm/s, respectively. It provides a longitudinal
the LMR 1000. The linac-MR gantry is supported range of 260 cm, a vertical range of 65–140 cm
by two A-frames allowing +180 and –180 ° rota- above the floor, and, more importantly, lateral
tion. The MRI iron-yoke structure is designed to ranges of ±23 cm. The latter allows significant
a b
Fig. 11.1 (a) Photo of LMR 2000 (a.k.a. Aurora-RT) installed at the Cross Cancer Institute, Edmonton, Alberta,
Canada. (b) Schematic design, patient bore opening of 110 cm by 60 cm
11 The Alberta Rotating Biplanar Linac-MR, a.k.a., Aurora-RT™ 195
2D positioning, while when used in tandem with

gantry rotation, 3D positioning of offset targets to
any planned position (including isocenter). B B
Because of an open MR, while the gantry is at
90°, during pre-MR imaging used for positioning
patient prior to treatment. A patient lying on the I II
couch inside the bore has a direct view of the
vault ceiling reducing possible claustrophobia. In
addition, when the gantry is at 0°, the left and B B
right side of the patient lying on the couch inside
the bore is completely opened allowing staff the
direct access to the patient or to a phantom for
QA.
Real-time imaging during treatment is possi- III IV
ble at 4 fps. Being a high-temperature supercon-
ductive MRI, cryogens and exhaust vents are not
Fig. 11.2 Various geometrical orientations for radiation
required and the magnetic field B0 can be safely
detector’s long axis (central cylinder) magnetic field
turned on/off in minutes. (arrow) and radiation beam (transparent rectangular
prism) as used in Monte Carlo simulation and experimen-
tal measurements. Taken with permission from Ref. [13],
Med Phys. 2013;40(4):042102
hange in Radiation Response
C
of Detectors in Magnetic Field
ion chamber was also measured in both magnetic
Ion Chambers field orientation up to 0.2 T and compared to
PENELOPE simulations, to validate the electro-
Ion chambers are generally used for photon beam magnetic field implementation in PENELOPE
reference dosimetry to calibrate radiation beams. and our choice of simulation parameters within
Electrons traversing the sensitive volume ionize the code.
the air; these ions are collected as signal propor- Lorentz force has more of an effect on elec-
tional to the energy fluence. Magnetic fields exert trons traversing low density air than those in
Lorentz force on electrons and change their path chamber wall or buildup cap. Orientation I in
within the sensitive volume. Therefore, the rela- Fig. 11.2 suggests that the Lorentz force on elec-
tive response of ion chambers to photon beam trons will likely deflect them away from the
irradiation will change in the presence of mag- long, sensitive air column, thereby decreasing
netic field. Our group has investigated the relative the chamber’s relative response, while an oppo-
response of cylindrical ion chambers, NE2571 site effect can be expected in orientation II. In
(NE Technology Ltd) and PR06C (Capintec), to orientations III and IV, Lorentz force will focus
magnetic field in 6 MV photon beams [13]. Two the electron path along the magnetic field direc-
magnetic field orientations, either perpendicular tion and may bring few electrons deflecting away
(Fig. 11.2 top quadrants) or parallel (Fig. 11.2 from the sensitive volume back into the volume.
bottom quadrants) to the radiation beam direc- The results of this simulation are presented in
tion, with magnetic field strengths up to 2.0T Fig. 11.3. The response of PR06C is affected in
were used in Monte Carlo simulations using perpendicular magnetic field, and it can be as
PENELOPE. The long axis of chambers was high as 1.085 in orientation II and as low as
either parallel (Fig. 11.2 left quadrants) or per- 0.90 in orientation I. In parallel magnetic field,
pendicular (Fig. 11.2 right quadrants) to the radi- up to 0.5 T, the response of PR06C is nearly
ation beam travelling in both orientations of identical to no magnetic field case, and its maxi-
magnetic field. The relative response of PR06C mum value is 1.02 in orientation III at
Fig. 11.3 PR06C 1.10

relative dose response as
a function of magnetic
field in orientations I– 1.05
Relative Response (to no field)

IV. Measurements are
only for I–II. Taken with
permission from Ref. 1.00
[13], Med Phys.
2013;40(4):042102
0.95
0.90
Orientation I Simulations Orientation I Measurements
Orientation II Simulations Orientation II Measurements
Orientation III Simulations Orientation IV Simulations
0.85
0.00 0.25 0.50 0.75 1.00 1.25 1.50
Magnetic Field (Tesla)
1.5 T. These results are similar to those obtained beam scanning. Inspired by this work, there has
by other investigators for other cylindrical ion been significant amount of work to determine
chambers [14]. As such, no correction for mag- the magnetic field correction factors for perpen-
netic field induced response change is needed in dicular magnetic field linac- MR’s reference
parallel magnetic field below 0.5 T. Subsequent dosimetry calibration [17–19].
measurements in parallel oriented magnetic field
up to 0.22 T have further corroborated this con-
clusion [15]. Additionally, if these chambers are Solid State Detectors
used for reference dosimetry, a correction factor
is required in perpendicular magnetic field that Solid state detectors such as PTW 60003 dia-
depends on the magnetic field strength, chamber mond detector and IBA PFD diode detectors are
dimensions, and chamber orientation with generally used in relative beam dosimetry, i.e.,
respect to magnetic field and photon beam. beam scanning. The sensitive volume in these
However, our work further demonstrated that if detectors is of significantly higher density than
the chambers’ long axis is aligned with the mag- air in ion chambers; thus, the effect of Lorentz
netic field direction in the perpendicular mag- force may not be significant. With this hypothesis
netic field, the magnetic field response is close to in mind, we performed Monte Carlo simulation
1.0, and smaller angular deviation of chamber of magnetic field dose response of these detectors
from this orientation does not impact the mag- by constructing the fine dimensions and materials
netic field response of chamber. Please note that of the detectors in PENELOPE code system [20].
this orientation is not shown in Fig. 11.3. In In addition to the four orientations of long axis of
other conventional orientations, small 5° mis- detectors as shown in Fig. 11.2, we considered
alignments may change the chamber’s response two additional orientations. Orientation II-2 is
up to 2%. In parallel magnetic field, chamber’s the same as II, but detector is rotated by 180
angular alignment errors in standard orientations degree such that detectors’ stem becomes oppo-
III and IV result well below 1% change in mag- site to orientation II. Another orientation desig-
netic field response [16]. Although one orienta- nated as orientation V is obtained by rotating the
tion of chamber along the perpendicular detector in orientation I or II such that its long
magnetic field appears to be suitable in terms of axis becomes parallel to magnetic field. The sim-
chamber response and its insensitivity to align- ulated magnetic field response of diamond detec-
ment errors, this orientation may be limiting for tor is shown in Fig. 11.4 [20].
Fig. 11.4 Relative dose 1.25 1.25

response of PTW60003 Orientation I Simulations
diamond detector with Orientation II Simulations
magnetic field strength. 1.20 1.20
Orientation II-2 Simulations
Relative Response (to no field)

Taken with permission Orientation III Simulations
from Ref. [20], Med Orientation IV Simulations
1.15 1.15
Phys. 2014;41(9):09210
Orientation V Simulations
1.10 1.10
1.05 1.05
1.00 1.00
0.95 0.95
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Magnetic Field (Tesla)
Perpendicular field orientations appear to have F Coils for the Rotating Biplanar
R
the largest dose responses, and the parallel mag- Linac-MR
netic orientations show little response as a func-
tion of magnetic field. In the presence of a Standard MRI detectors (typically arrays of
longitudinal magnetic field, minimal correction loop-shaped surface coils) are often placed
is needed at 1.5 T, and no correction appears to be directly on the patient’s body to optimize SNR
required below a field strength of 1.0 T for both by placing coils as close to the imaging region
PTW60003 and IBA PFD diode (not shown) as possible. When collisions of the radiation
detectors. This contrasts with the cases with per- beam with RF coil materials cannot be avoided,
pendicular magnetic fields, where complete char- their effect on the dose delivered to the underly-
acterization of detectors on an individual basis is ing tissue must be evaluated. Our measure-
required as a function of relative orientations and ments [21] have shown that standard coil
field strengths. Our additional simulations mim- materials can increase the surface dose to >74%
icking profile measurements at the edges of the relative to Dmax, compared to 20% with an open
photon beam in perpendicular magnetic field fur- beam. This undesired “bolus” effect enhances
ther explain the errors, due to detectors’ own skin dose, thus increasing the risk of acute reac-
dose response in magnetic field, in reproducing tions, such as erythema, and long-term radia-
the relative dose to water under perpendicular tion damage [22].
magnetic field.
Coil Construction
Summary
For optimal detector performance, the conduc-
Minimal magnetic field responses, in parallel tors in an RF coil must have a minimum thick-
magnetic field orientations, ensure that conven- ness on the order of one electromagnetic “skin
tional commissioning and quality assurance depth” [23] which in copper at 0.48 T (20.6 MHz)
detectors and techniques can be used without is about 15 μm and for aluminum 18 μm. Typical
corrections due to magnetic field with a 0.5 T MRI coils, however, also include substantial
MRI. mechanical supports for the conductors: for
example, to maximize durability, allow weight- The relationship between coil thickness and
bearing, etc. A radiation beam may need to travel its resistance is a complicated function, espe-
through several millimeters of these support cially when thickness is near the skin depth. We
materials (low-loss dielectrics like polycarbonate studied these effects in more detail in Barta
or fiberglass composite), which thus contributes (2021) [26] by constructing coils of various
substantially to the bolus effect in an RF coil con- thicknesses of copper (17–127 μm) and alumi-
structed using standard methods. num (9–127 μm) and measuring their perfor-
Coils for linac-MR applications must there- mance (i.e., efficiency, η, obtained from loaded
fore be constructed with minimal support materi- and unloaded coil quality factors) on the bench
als. Potentially, they could be paper thin like and in the 0.48 T scanner (image SNR). For com-
those of Corea et al. [24] which are printed on parison, the thickness of copper in printed circuit
flexible dielectric substrates (50–150 μm thick) boards is 35 μm (standard 1 oz.) and multiples
using conductive inks. We measured the effects thereof (less common, and rarely >4). To repre-
of coils made of thin substrates and conductors sent manually assembled coil construction (and
with and without a 0.22 T magnetic field (created the maximum SNR achievable in practice), one
by a yokeless electromagnet) parallel to the beam was also made using 600 μm copper foil. Coil
[25]. Results show that for thin copper and alumi- efficiency was in the range of 0.56–0.74 (Cu) and
num foils, in the absence of magnetic field, the 0.34–0.59 (Al), and it correlated well with image
surface dose increases linearly with thickness,
SNR comparisons according to the relationship
with slopes of 0.339%Dmax/μm and 0.116%Dmax/
μm, respectively, starting from 22.2% of Dmax for
SNR A ηA
[27] = . Importantly, there were no
the unobstructed beam. With the magnetic field SNR B ηB
parallel to the beam, the slopes are lower differences in SNR achieved by coils made of
(0.16%Dmax/μm and 0.06%/μm, respectively) but 127 μm copper and aluminum conductors, and
start from a higher surface dose of 31.9% of Dmax with the much thicker 600 μm copper, the SNR
for the unobstructed beam. The 3.3 times lower improved only by 5%. Very thin conductors such
density of aluminum compared to copper as 17 μm copper and 13–20 μm aluminum
accounts for the lower slopes both with and with- achieved, respectively, 85% and 75–85% of the
out magnetic field, while the higher starting dose SNR with 127 μm conductors. If needed, these
in the presence of the magnetic field is due to limited SNR losses can be readily compensated
contaminant electrons trapped by the Lorentz for by corresponding adjustments to the imaging
force. sequences, such as using thicker slices, lower
For comparison, the dose beneath a hospital acquisition bandwidth, longer scan time, etc.
gown was measured to be 36.0% of Dmax with the
0.22 T field, which corresponds to approximately
the same increase as 25-μm-thick copper or Summary
70-μm-thick aluminum. These results suggest
that aluminum coil conductors can be chosen to Thin aluminum conductors can allow surface
have the same effect on surface dose while being dose to be controlled (e.g., an enhancement less
much thicker in terms of skin depth than copper. than that caused by a gown compared to open
Because image SNR depends directly on coil beam), with very limited loss in SNR (not more
resistance, the larger thickness of aluminum than 25% compared to coils made using the
could compensate for its slightly higher resistiv- thickest common PCB copper thickness).
ity and lead to equal or better SNR from coils Radiolucent coils made in this way are therefore
made of aluminum. a practical option for linac-MR applications.
ccelerated MR Imaging for Real-

A ner such that any Nyquist artifact appears inco-
Time Motion Management herent and noise like. CS reconstructed image is
the solution to an optimization function that max-
One of the benefits of having a MR/RT hybrid imizes fidelity of its k-space representation to the
system is the ability to use the online MRI capa- partially acquired k-space data, via L2 norm min-
bilities to provide real-time soft-tissue visualiza- imization, while simultaneously maximizing
tion, to monitor and manage the patient’s sparsity in a transform domain (e.g., wavelet, dis-
respiratory motion, particularly in the thorax or crete cosine, or total variations) via a L1 norm
abdominal areas during RT. Compared to current minimization, weighted by λ1. For a total varia-
state of the art Linacs (e.g., Varian Truebeam, tion transform, a split Bregman solver [29] can be
which is capable of real-time orthogonal kV pro- used efficiently to iteratively solve this equation,
jections), real-time MRI has a couple of key in approximately 100 ms in a typical CPU
advantages: flexibility of geometry of the imag- (3.4GHz, Intel i7–4770), using MATLAB.
ing slice volume, which is not dependent on the
Gantry’s position, and the ability to visualize soft
tissue, without inserted fiducials. Nevertheless, rior Data Assisted Compressed
P
MRI is inherently a slower imaging method com- Sensing (PDACS)
pared to kV imaging. For a typical breathing
respiratory motion monitoring, 4 fps is consid- For 2D Cartesian acquisitions, the ability to
ered to be adequate. For modern MR imagers, the undersample k-space is limited to the phase-
fastest dynamic sequences (i.e., balanced SSFP) encode direction only. As a result, the second
should already be capable of providing a single condition of incoherent sampling is not fully sat-
2D plane with 2–3 mm resolution at that frame isfied, usually resulting in poor quality images at
speed, without applying any specialized tech- high acceleration. An alternative approach is to
niques. Nevertheless, the development of special include dynamic 2D data in the L1 norm minimi-
techniques that further increase acquisition speed zation term (i.e., perform 3D sparsifying trans-
can improve the spatial or temporal resolution of form on the dynamic data in the optimization
2D images, and may even allow real time multi- function). Unfortunately, these 3D operations in
slice or 3D imaging that more accurately monitor an iterative solver are too slow for real-time
the 3D trajectory of a moving tumor. applications. The key feature of prior data
assisted compressed sensing (PDACS) [30] is to
take advantage of prior data in a different man-
ccelerated MR Acquisition
A ner, instead by adding an additional L2 norm
and Compressed Sensing MRI minimization term that enforces fidelity of the
unacquired k-space to the k-space of some aver-
There has been considerable work in the MR lit- aged prior data, in a 2D operation. PDACS gives
erature to improve acquisition speed. One very considerably better images compared to the orig-
well developed and well known is parallel acqui- inal CS method, while the split Bregman solver is
sition (i.e., SENSE or GRAPPA), which has spe- still able to provide the final image in fraction of
cific hardware requirements (i.e., multichannel a second. The weights given to λ1 and λ2 have a
array coils). Our group decided to pursue more significant impact on the quality of the final
general techniques that can be used in any MR image – a simple brute force optimization is per-
scanners. Compressed sensing (CS) in MRI [28] formed on some training data to give approxi-
allows an image to be recovered from significant mate values of λ1 and λ2 that are adequate.
undersampled k-space data, via optimization, if it As these images are ultimately used to moni-
(1) has a sparse representation in some known tor a patient’s tumor motion, we evaluated the
domain and (2) is acquired in an incoherent man- quality of PDACS reconstruction by evaluating
its tumor auto-contouring performance of an peripheral phase-encode locations are never

undersampled image and against a fully sampled refreshed. For the sliding window technique, the
image. Relevant metrics include Dice similarity sampling pattern for every dynamic is determined
index (DSI), as well as mean centroid displace- with a dynamic probability function that has
ment (MCD). A high DSI indicates a high quality three regions (central, intermediate, and periph-
undersampled image, whereas a low mean cen- eral). The central five phase-encode lines are
troid displacement indicates a higher quality acquired at 100%, while the intermediate region
undersampled image. is the region where the probability is falling off
from 1 to 0.25, with a falloff rate determined by
pdating Prior Data with Sliding
U the acceleration factor and the peripheral region
Window where p < 0.25. In the peripheral region, the ini-
By using fully acquired images prior to acceler- tial probability for each location is equalized.
ated image acquisitions, the PDACS technique Once a peripheral phase-encode location is sam-
was found to improve image quality of unders- pled, its probability is uniformly redistributed to
ampled images. However, we found the benefit other peripheral, non-sampled phase-encode
decreases gradually over time as the prior data positions for the next dynamic. This forced redis-
gets progressively more outdated compared to tribution of sampling probability ensures that no
the current data, due to shifts in the baseline MR peripheral dynamic will be acquired consecu-
signal unrelated to patient motion. Over the tively and that sampling pattern eventually cycles
course of a 3-min acquisition, in terms of tumor through all k-space location, ensuring all k-spaces
autocontouring accuracies for a 5× accelerated lines will be periodically refreshed. Prior data
acquisition are 0.93, 0.9, and 0.87 for the first, now derives from the average of this sliding win-
second, and third minutes’ data after the acquisi- dow, rather than from the very beginning of the
tion of prior data for the high SNR scenario and acquisition.
0.91, 0.88, and 0.86 in the low SNR scenario. The In our retrospective study using the same data,
mean centroid error increases from 1.1 mm, our result shows that by refreshing the prior data
1.5 mm, and 2.5 mm for the first, second, third via a sliding window approach, we are able to
minutes of data after the acquisition of prior data maintain a tumor-tracking accuracy of DSI = 0.93
in the high SNR scenario, whereas they are 1.2, and mean centroid error of 1.0 mm for the entire
1.9, and 3.2 mm in the low SNR scenario. One 3 min of data acquisition in the high SNR sce-
should note that when the prior data is 3 min out nario and a DSI of 0.91 and centroid displace-
of date, these results can be even poorer as the ment of 1.2 mm for a low SNR scenario.
2D-CS technique, without any prior data!
To address this, sliding window PDACS [31] rior Data Assisted Compressed
P
is a technique that requires a different k-space Sensing: Summary
sampling strategy. The CS requires a pseudoran- Accelerated MRI (reconstructing undersampled
dom sampling pattern, but the center of k-space MRI with special techniques) has many benefits
contributes more value to image quality and is for real-time MRI in RT. This added speed can be
sampled more often. In PDACS, the sampling used for increased temporal resolution, spatial
pattern is generated with a probability function resolution, or imaging multiple slices. Accelerated
such that the central phase encode is sampled at MRI in the context of real-time motion monitor-
100% probability, but the function falls off as it ing is quite challenging, in the sense that a long
moves to the peripheral phase-encode locations reconstruction time will reduce or negate any of
at a rate governed by acceleration factor (i.e., the the advantages you gain reducing image acquisi-
higher the acceleration, the faster the falloff). tion. On the other hand, this problem is made
Once a pattern is determined for a single dynamic, easier by having access to a pool of prior data
it is used for the entire series, and as such, some (due to this being a dynamic scan of a reasonably
repetitive motion pattern). The PDACS method is real-time reconstruction. Dietz et al. [32] from
one way in which prior data from a dynamic MR our laboratory employed a method to iteratively
acquisition can be used to improve the image determine a subset and corresponding amplitudes
estimate for a undersampled “current” image. of the PCs that best represented the undersam-
Ultimately, this technique can increase acquisi- pled data, analogous to minimizing the L1 norm
tion speed by 5×, with the cost of 100 ms recon- of a sparsifying transform in common com-
struction time, with the 100 ms being an pressed sensing implementations. In this work,
unavoidable aspect of the iterative nature of the although it could have technically been made part
compressed sensing technique. Other non- of the iterative procedure, the data consistency
iterative techniques that can take advantage of was forcibly maintained by reinserting the origi-
prior data for accelerated MRI reconstruction nally undersampled data at every iteration for the
with even shorter reconstruction time are highly purpose of computational efficiency.
desirable. This method was tested retrospectively on a
set of six patients with non-small cell lung tumors
with institutional ethics approval originally
Compressed Sensing and Principal acquired at four frames per second on a 3 T MRI
Component Analysis (CS-PCA) scanner using a balanced steady-state free pre-
cession (bSSFP) sequence. The first 30 images in
Principal component analysis (PCA) is a mathe- the dynamic set were used to create the database
matical tool frequently used for reducing the that generated representative PCs. A preliminary
dimensionality of large data sets. In this process, optimization was performed to select the optimal
data is mapped onto a new orthogonal basis set iteration parameters, including the maximum
defined to correspond with the vectors of maxi- number of iterations as well as the maximum
mum variation. In the context of imaging, each coefficient threshold below which the PC was
pixel or value of acquired k-space can be consid- dropped from the solution. As in the section
ered separate variables, and a database of images above, the visible tumors were auto-contoured on
can be processed through PCA to identify the both the original fully sampled images and the
most common modes of variation between them. retrospectively sampled reconstructions to test
Generally, the number of significant modes of for robustness in target delineation. Acceleration
variation is vastly smaller than the number of rates were tested across a range of 2–10× unders-
variables, and thus a limited subset of these ampling. The DSI and MCD of the contours were
modes or principal components (PCs) are respon- used to quantify differences between the unders-
sible for the bulk of differences between images. ampled reconstructions and those of the fully
On this assumption, if the database can be con- sampled data. Additional image quality metrics,
sidered representative of subsequently acquired including normalized mean squared error
images, those new images can be well-represented (NMSE) and structural similarity index (SSIM),
by a limited number of coefficients representing were calculated to aid in this comparison.
the magnitudes of significant PCs.
In the context of real-time MRI and recon- CS-PCA Performance
struction, if undersampled data can be accurately The mean DSI for this technique was reported
mapped to a set of PCs from a previously acquired above 0.9 for both high and low SNR states, even
fully sampled database, then the PCs could be at an acceleration rate of 10×. Specifically at an
used to reconstruct the full image content. acceleration of 5×, the DSI was greater than 0.92
Unfortunately, the fact that only a partial amount for both high and low SNR scenarios. The mean
of data is acquired in the undersampled image centroid displacement was also reported to be
means that a simple projection will not be suffi- robust up to the tested acceleration of 10×, with
cient to determine appropriate PC coefficients for the maximum MCD of 1.15 mm across the range.
At 5× acceleration, the MCD was reported at pled MRI data based on an initially acquired
0.9 mm. The image fidelity and robustness up to database. While its performance demonstrated
high rates of accelerations were quite favorable suitability for real-time target tracking applica-
for this technique. However, as we pointed out in tions, anatomic or physiologic changes after the
this work, there is a slow quality degradation over database acquisition could reduce the reconstruc-
time. Since the quality of any given reconstruction quality. To adapt the technique into one
tion depends on how well it can be represented by where the database is continually updated, Wright
the extracted features of an initial database, if the et al. presented a work that uses a PCA character-
database becomes less relevant due to physio- ization in the time domain based on undersam-
logic or anatomic motions, the resulting recon- pled data alone [34]. In this way, the database can
struction will be less successful. Nevertheless, adjust along with potential physiologic or muscle
for the patients examined over a 3-min period in motions to produce a stable output over extended
this study, the reported >0.9 DSI was similar or imaging times.
better to that reported for intra-observer contour
variability as drawn by expert radiation oncolo- Implementation Concept
gists when assessing the same data sets at differ- Incoherent undersampling schemes for acceler-
ent times (mean DSI = 088) [33]. Reconstruction ated MRI generally have high sampling rates in
times were reported to vary between 5 and 20 ms the central low-frequency region of k-space and
per dynamic image. scattered samples out to the periphery. This work
takes advantage of the consistent presence of the
CS-PCA Summary low-frequency data to generate a time-based
The CS-PCA strategy as reported had several characterization of k-space signal using
favorable attributes. Firstly, the algorithm is com- PCA. With the assumption that time-based evolu-
putationally efficient, with less than 20 ms recon- tion in central and peripheral k-space results from
struction times on a desktop CPU. This is critical the same physical motion sources, the principal
for real-time interventional applications which characterizations of motion can be used to fit to
depend on low lag times between acquisition and the less frequently sampled peripheral k-space to
image availability. Secondly, the method was fill in missing data with motion-based estimates.
reported to behave well (based on contouring To implement this idea, the authors generated
metrics) in lower SNR regimes as might be sampling masks that consistently acquired a core
expected at lower MR field strengths and of between 8 and 16 central phase encodes, with
sequences for rapid dynamic imaging and main- the periphery undersampled on a regular repeat-
tained its good performance for acceleration rates ing pattern. Masks for acceleration rates varying
of up to 10×. A reported limitation was a slow between 3× and 8× were generated and used for
degradation of metrics as the time between the testing. The time-domain core k-space database
database acquisition and accelerated dynamic was chosen to cover 60 dynamic frames, with the
frame increases. However, for the patient data last frame being the one to be reconstructed. As
sets analyzed in this work, the reported contour new data is acquired, this 60-frame time-domain
metrics were reliable over the observed 3-min window is shifted over, and a new database is
dynamic imaging interval. formed to reconstruct the new dynamic. Of the
PCs generated by the PCA process, only the first
few will contain the most valuable motion char-
ime-Domain PCA-Based Real-Time
T acterizations with the last ones representing pri-
Reconstruction for Dynamic marily noise.
Undersampled MRI The optimizations for this technique and sub-
sequent testing were performed on the same six
The previous section discussed a method for patient data sets as in the previous study with
using a PCA technique to reconstruct undersam- local institutional ethics approval. Both image-
based metrics, including NMSE and SSIM, were atient-Specific CNN Reconstruction
P
performed as well as metrics specifically relating for Accelerated MR Imaging in Real-
to contour accuracy, namely, DSI sand MCD. Time Radiotherapy Tracking
erformance of Time-Domain PCA

P eal-Time CNN Implementation
R
Reconstruction One of the first publications to apply 3D CNN to
The performance of this technique was assessed dynamic data with a practical implementation for
over multiple minutes to look for trends in terms real-time radiotherapy use was published in 2019
of degradation of image quality. While the initial from our laboratory by Dietz et al. [35]. This
dynamic frames generated using this method work identified a workflow that would be practi-
were slightly poorer in quality metrics than the cal for most radiotherapy patients and worked
previously described PCA technique, the perfor- within that framework to develop a CNN archi-
mance remained relatively stable over time, as tecture and training paradigm that would allow
expected due to the continually updated database. the radiotherapy process to play out without addi-
All the acceleration rates up to 6× were reported tional patient visits or requiring them to lay on
to have DSI over 0.9, with the acceleration rate of the couch for prolonged periods during training.
8× yielding a DSI of 0.88. These values are In this conceptualization, the patient would be
consistent with intra-observer variability when MR imaged with fully sampled dynamic frames
contoured by an expert radiation oncologist and on the day of their simulation. A CNN specific to
suggest the method is suitable for real-time targetthis patient would then be trained based on this
tracking. The reconstruction times were reported data and would be ready to implement on the day
at approximately 50 ms per dynamic frame on a of their first treatment to reconstruct accelerated
CPU implementation. This is about twice the data for target tracking. Provided the training
reconstruction time required for the previous could be easily performed within the span of a
method, likely due to the extra demands of recal- day, there would be no alteration of the treatment
culating the PCs from a new database for each schedule.
image. However, this reconstruction time is still The network itself consisted of a cascade of
well within an appropriate range for real-time alternating convolutional networks and data con-
radiotherapy applications and could likely be sistency layers. The number of filters and layers
improved with higher-spec hardware. in each convolution module and the cascade
depth were empirically optimized. Raw MRI data
Summary is complex, so the real and imaginary data were
The time-domain PCA approach for dynamic fed into the network in separate channels. The 3D
undersampled MR reconstruction in real time nature of the CNN takes advantage of redun-
was shown to perform within acceptable bounds dancy in much of the slowly varying k-space con-
for target tracking. This method does not rely on tent to improve the quality of the final 2D
any fully sampled data being acquired prior to the reconstruction, rather than having the network
dynamic acquisition. It functions solely based on process each dynamic frame independently. Data
the undersampled data itself and uses common for five dynamic frames were fed into the net-
modes of variation in the time domain to fully work at a time (the frame-of-interest preceded by
populate k-space for the current frame-of- data for the four previous consecutive frames) for
interest. Importantly, this method provides addi- processing. An additional data sharing layer was
tional confidence that the geometry has not been appended to the front of each convolutional mod-
misrepresented because of an out-of-date data- ule, combining adjacent frames to generate three
base in the event that there is an unexpected extra artificial dynamic time points for more
patient or organ motion. reconstruction stability.
Training and testing was performed on six of the CNN from the previous section by funda-
patient dynamic data sets with non-small cell car- mentally altering how the data is undersampled
cinoma, used with local institutional approval. [36]. Undersampling strategies such as those
The native frame rate of the source images was for compressed sensing generally require inco-
~250 ms/frame, with a matrix of 128 × 128. herently spaced samples to be distributed
There were 650 frames for each data set. To sim- throughout k-space into the high-frequency
ulate undersampling, an incoherent undersam- regions. This creates an incoherent artifact pat-
pling mask was applied along the phase-encode tern superimposed on top of an otherwise high
direction, nulling all the values except for an resolution image and allows for an iterative
undersampled fraction. Two different accelera- model-based algorithm to remove artifact from
tion rates were investigated in this study: 5× and the image while maintaining fidelity to the
10×. The first 300 of the available dynamic actual sampled data. However, since the CNN
frames were used for training each patient’s indi- above is trained on fully sampled data, the
vidualized CNN. The next 150 were split evenly authors proposed that the network can be
between validation and testing purposes. The trained to infer the higher-resolution regions
final 200 were used for metric evaluation to from a strictly low-resolution acquisition. The
determine image and tumor contour accuracy, by data in central k-space generally has much
comparing the results of the accelerated images higher SNR than does the high-frequency data,
to those of the fully sampled. The contour metrics potentially offering advantages in terms of
evaluated were DSI and MCD, which are particu- reconstruction stability. The same network
larly useful parameters for assessing applicability architecture as in the previous section was
for target tracking. The testing was performed employed, although the hyperparameters were
using a single NVIDIA GTX 1080Ti GPU. reoptimized. The acceleration rates of 5× and
10× were investigated. With this new sampling
CNN Performance strategy, the mean DSI was in general much
Training for this optimized network was reported improved across the patient data sets, in addi-
to take roughly 6 h, well within the time window tion to the standard deviation and stability
between initial simulation and first treatment day. across the dynamic frames.
The reconstruction time for each undersampled
frame was reported to be 65 ms. The mean DSI ummary of Real-Time CNN
S
across all CNN reconstructions was 0.94 for 5× Reconstruction
acceleration and 0.92 for 10× acceleration. This section demonstrates the applicability of
Similar results for 6× SNR-reduced data were CNN reconstruction strategies to accelerated
reported with 0.92 for 5× and 0.91 for 10× accel- MRI dynamic data. Reconstruction duration was
eration. One of the patient data sets was reported reported to be on the order of 65 ms, with DSI
to have a mean DSI below 0.85, potentially due metrics that suggest the images are well-suited
to an elongated tumor shape to which the auto- for target tracking purposes. While this strategy
contouring was particularly sensitive to changes. does require a 6-h training interval, the work
This was reported as a potential direction for points out that this may be feasible to fit into a
future work. clinic’s regular workflow. The algorithms and
hardware available for CNN training and execu-
Coherent Versus Incoherent tion are currently evolving, and as they do, the
Undersampling for CNN Reconstruction image quality and reconstruction times that such
Further work from the same group investigated a strategy can produce may continue to improve
an idea for improving the stability and accuracy as well.
Noninvasive Intra-Fractional Tumor- markers as tumor surrogates [43, 44]. These sys-
Tracked Radiotherapy (NifteRT) tems are used in conjunction with additional
using the Alberta Linac-MR motion management techniques including
abdominal compression, respiratory gating, and
Background breath hold to reduce the geometric margins
around the tumor [45].
If a tumor is static during irradiation (e.g., brain Despite the variety of techniques, all current
tumor), modern RT can deliver highly accurate tracking methods are based on the indirect track-
radiation to the tumor (<1 mm geometric error) ing using internal/external tumor surrogates.
without much harm to the adjacent normal tissue Reliance on surrogates, however, is known to be
[37]. Unfortunately, most tumors are mobile and problematic for accurate tumor tracking because
can deform significantly due to the respiratory (1) utilizing internal surrogates requires invasive
and cardiac motions (e.g., lung tumor) [38–40], implant procedures that may cause medical com-
and such accurate RT delivery is challenging due plications [46, 47]. Moreover, the implanted sur-
to the uncertainty in tumor localization. RT is rogates are known to migrate or completely
typically given over multiple fractions of daily dislodge from their initial positions during the
treatment, and the continuous tumor motion and course of RT (typically several weeks) [48, 49].
deformation occurring during the therapeutic (2) Tracking with external surrogates must rely
radiation delivery in a given fraction is known as on ambiguous correlations between the internal
the intra-fractional tumor motion. In the RT of tumor motion and the external surrogate dis-
mobile tumors, intra-fractional tumor motion placement [50]. More importantly, any deforma-
management and the surrounding critical struction of tumor shape during irradiation is
ture sparing during irradiation are of paramount completely unknown, e.g., >10 mm for lung and
importance. This is particularly true for the liver tumors [40, 51] and up to 6.6 mm for pan-
modern stereotactic radiosurgery/stereotactic creatic tumor [39, 52]. The additional motion

body radiotherapy (SRS/SBRT) techniques in management techniques also have disadvantages:
which much higher dose per fraction is delivered (1) Not all patients can tolerate abdominal com-
compared to the conventional 3DCRT or IMRT pression nor breath hold especially for those with
[41]. impaired pulmonary function. In some cases,
A traditional solution to manage the intra- abdominal compression is known to work nega-
fractional tumor motion is to add geometric mar- tively by increasing the tumor motion, and its
gins around the tumor ensuring adequate target success rate is largely dependent on the patient
coverage. This approach, however, can result in and the anatomical location of tumor [53, 54]. (2)
(1) unacceptable medical complications due to Respiratory gating cannot be used for patients
the excessive normal tissue irradiation adjacent with irregular breathing patterns, and it typically
to the tumor [42] or (2) poor disease control requires two to three times longer treatment time
caused by limiting the necessary therapeutic dose than continuous beam delivery, causing patient
to avoid those complications. Real-time tumor- discomfort and challenges in the clinical
tracking RT is an alternative approach that may workflow.
potentially improve tumor localization. Both Most tumors reside in the soft tissue. However,
“intra-fractional” and “real-time” tumor-tracking CT, fluoroscopy, and other X-ray based image
RT refer to the methods of continuously tracking guidance systems currently available on the RT
the tumor with radiation beam during beam devices (e.g., linacs) provide good bony but very
delivery. Several tracking systems are commer- poor soft-tissue imaging during the therapeutic
cially available, estimating the tumor location radiation delivery. This is the fundamental reason
during irradiation based on invasive internal why the indirect tumor-tracking methods have
(X-ray opaque implant surgery) and/or external long been used with the additional motion man-
agement techniques, despite their shortcomings range of intra-fractional motion and the patient
mentioned above. Newly emerging hybrid linac- population at the CCI (the three tumor sites take
MRI systems may overcome the limitation by 43% of all RT patients). Lung and liver tumors
providing simultaneous MR imaging and thera- are known to move up to 40 and 50 mm in
peutic radiation delivery [2, 55, 56]. Our labora- superior-inferior direction during normal breath-
tory at the Cross Cancer Institute (CCI) built the ing, respectively [58, 59]. Although intra-
world’s first hybrid system known as linac-MR in fractional prostate motion is known to be <7 mm
2008 [2] and a human whole-body system in in most cases [60], some patients may exhibit
2014 [10]. Recently, our latest linac-MR system substantially large 3D prostate displacements up
known as the Alberta linac-MR (LMR 2000) was to 18 mm [61]. In fact, we observed ~18 mm of
installed and commissioned for its clinical use at abrupt prostate displacement from our patient
the CCI. This linac-MR can provide continuous during 5 min of MR scan.
MR imaging of the mobile tumor and surround- In general, RT requires a simulation session
ing anatomy with excellent soft-tissue contrast, several days prior to the actual treatment day.
concurrent with the therapeutic radiation NifteRT workflow expects that the simulation
delivery. occurs in our linac-MR system, and a pretreat-
To overcome the abovementioned challenges ment MR scan is performed using the same MR
in RT for mobile tumors, we have been develop- sequence, patient setup, and imaging plane
ing a novel RT method known as noninvasive intended to be used during nifteRT. These MR
intra-fractional tumor-tracked radiotherapy (nif- images are used for our algorithm training and
teRT) using our linac-MR systems. In nifteRT, risk mitigation. An overall framework of nifteRT
while a patient is under free breathing, a thera- is shown in Fig. 11.5.
peutic radiation beam will track a moving and During nifteRT, our linac-MR acquires con-
deforming tumor by adapting the beam’s position tinuous MR images of a mobile tumor while the
and shape in real time. radiation is always on. In Step 1, an intra-
fractional MR image of a tumor is acquired. In
Step 2, the tumor shape is automatically detected
verview of Objectives of the NifteRT
O by our auto-contouring algorithm (red bound-
Development Program ary), and the centroid of tumor contour is consid-
ered as its position (a red dot). In Step 3, a future
The objective of nifteRT program is to enhance tumor position is predicted based on the current
the geometric accuracy of RT for mobile tumors and previous positions by our motion prediction
for the Alberta linac-MR. Specifically, we aim to algorithm. In Step 4, the multi-leaf collimator
achieve <3 mm geometric error in radiation (MLC) is driven to conform the treatment beam
delivery treating three mobile tumor sites: lung, to the auto-contoured tumor shape at its predicted
liver, and prostate. The 3 mm limit was calculated position. Steps 1–4 are continuously run each
using the van Herk margin formula [57] by inves- time a new MR image is acquired at 4 fps.
tigating the possible error sources in nif-
teRT. From the formula, by allowing 3 mm of a
planning target volume (PTV) margin for a ifteRT Step 1: Intra-Fractional MR
N
mobile target in nifteRT, we expect to cover a Imaging
clinical target volume (CTV) with >95% of pre-
scribed dose without the need of an internal tar- The Alberta linac-MR can provide 2D MR
get volume (ITV) margin. images at 4 fps, i.e., 250 ms acquisition time per
The three tumor sites were chosen with our frame, using a balanced steady-state free preces-
physician input, to maximize the clinical impact sion (bSSFP) sequence that is a widely used for
of nifteRT by reducing the PTV margin. The site fast sequence due to its high SNR potential [62].
selection was based on their potential for a large The 250 ms acquisition time addresses the pro-
Fig. 11.5 Overview of nifteRT framework in the linac-MR (radiation is always on during nifteRT)
cessing time in each step of nifteRT, following reconstruction [35, 36]. We trained FCNN to fill
the American Association of Physicists in the gaps in partial k-space; thus, the difference in
Medicine (AAPM) report recommending reconstructed images from partial k-space and
<500 ms time delay between the tumor detection fully sampled k-space is minimized. Since pre-
and RT system response to manage the serving auto-contouring accuracy at fast frame
respiration-related tumor motion [63]. In nifteRT, rate is critical, we used our in-house developed
the time delay is currently ~300 ms including tumor auto-contouring algorithm as a quantita-
image acquisition time (counted from the center tive tool measuring the fidelity of MR images
of k-space acquisition, i.e., 125 ms for 250 ms [65]. Ideally, the algorithm generated tumor con-
image acquisition), auto-contouring and motion tours (ROIauto) would be identical if the MR
prediction (~60 ms), and MLC motion (~ 100 ms) images reconstructed from partial k-space acqui-
[64]. sition are the same as the ones from fully sampled
Although our system can meet the AAPM rec- k-space. Our FCNN algorithm was validated
ommendation at 250 ms/frame image acquisition using six non-small cell lung cancer (NSCLC)
time, the AAPM report states that the future patients’ MR images (bSSFP, 250 ms/frame
tumor positions can be predicted (up to 80% acquisition). For each patient, ROIauto were gen-
accuracy) only if the time delay is less than erated in 200 images reconstructed from both
200 ms [63]. One obvious way to decrease the partial k-space and fully sampled k-space acqui-
main source of time delay is to decrease the sitions. We calculated the Dice coefficient (DC)
image acquisition time per frame. At the same between each pair of ROIauto as a measure of
time, because tumor auto-contouring is unavoid- agreement. Our new FCNN algorithm [35]
able to cope with the fast frame rate, our imaging showed superior performance over our previous
techniques must not degrade the auto-contouring CS-PCA algorithm [32]. In case of 5× accelera-
accuracy. To satisfy these requirements, we have tion in the image acquisition time, FCNN algo-
implemented various partial k-space acquisition rithm produced higher mean DC for all patients
schemes to reduce the scan time, combined with over CS-PCA algorithm as shown in Fig. 11.6.
novel image reconstruction algorithms to More importantly, FCNN algorithm showed
preserve the image quality, e.g., prior data
lower variance in DC, indicating more stable
assisted compressed sensing (PDACS) [30], slid- image reconstruction performance.
ing window PDACS [31], compressed sensing
and principal component analysis (CS-PCA)
[32], fully convolutional neural networks (FCNN) ifteRT Step 2: Tumor
N
[35, 36], and time-domain PCA [34]. Auto-Contouring
Our most accurate algorithm utilizes FCNN,
which is a type of deep-learning neural network The accuracy of tumor auto-contouring is crucial
(DLNN) applicable to MR image acquisition and in nifteRT, because it dictates the accuracy of
a b
Fig. 11.6 (a) DC comparison: FCNN vs. CS-PCA. (b) Image comparison: fully sampled vs. FCNN vs. CS-PCA
tumor localization. We developed and validated liver, and 24 prostate cancer patients, 130 images/
several auto-contouring algorithms against patient, 250 ms/frame image acquisition) to vali-
human experts’ ROImanual in both phantom and date the U-net algorithm. In current RT of lung or
in vivo MR images [65–70]. In our studies, we liver cancer, a localized tumor lesion is defined,
used various metrics to quantify the accuracy of and the therapeutic radiation is focused on the
ROIauto against ROImanual, e.g., Dice coefficient local lesion. For prostate cancer, however, the
(DC), Hausdorff distance (HD), and centroid dis- current standard of care in RT is to irradiate the
placement (Δcentroid) between ROIauto and ROImanual. entire prostate gland. Whether to irradiate the
Hereinafter, we use DC as a metric indicating the seminal vesicles or not, in addition to the gland,
accuracy of ROIauto, e.g., 90% auto-contouring is at each physician’s discretion reflected in their
accuracy refers to 0.9 DC between ROIauto and ROImanual. Hence, we tested our U-net algorithm
ROImanual. Also, to find the appropriate level of for both scenarios in prostate auto-contouring.
auto-contouring accuracy required for nifteRT, For each patient, the first 30 images were used
we investigated the inter- and intra-observer vari- for patient-specific U-net algorithm training. In
ations in ROImanual drawn by several experts [33]. practice, these images will be obtained during the
Our study showed 87% inter- and 88% intra- pretreatment RT simulation. The auto-contouring
observer variations in ROImanual drawn on the six accuracy was evaluated by calculating DC
NSCLC patients’ image data. Because human between ROIauto and ROImanual over the remaining
experts’ tumor contour is the gold standard in 100 images for each patient. From the 45 patients’
clinic, the accuracy of ROIauto needs to be compa- data, the U-net algorithm showed 91, 90, and
rable to ROImanual. Thus, >90% accuracy in ROIauto 93% mean accuracy in auto-contouring lung,
is considered reasonably acceptable for nifteRT liver, and prostate tumors, respectively.
and is set as our general goal in the development Figure 11.7 emphasizes the strength of U-net
of auto-contouring algorithms. algorithm, especially in challenging anatomies:
Previously we used pulse-coupled neural net- partial omission in tumor shape, diffuse tumor
work (PCNN), which is a unique 2D neural net- boundary, and low contrast situations. Moreover,
work applicable to image segmentation tasks [65, our algorithm successfully reproduced each
70]. Our PCNN algorithm performed well in lung expert’s preference whether to include or exclude
tumor cases, where the contrast between the seminal vesicles for prostate cases.
tumor and its background is relatively large.
However, the algorithm may produce inaccurate
contours in more challenging anatomies such as ifteRT Step 3: Tumor Motion
N
the liver or prostate, where the tumor contrast is Prediction
low with unclear boundary to its surroundings.
To overcome these challenges, our latest algo- Tumor motion prediction occurs immediately
rithm adopted U-net, which is a type of DLNN after the auto-contouring that localizes the cur-
applicable to image segmentation tasks [66, 69]. rent tumor position. The prediction is necessary
We used 12 patients’ MR images (11 lung, 10 to compensate for the tumor motion during the
Fig. 11.7 ROImanual (red contour) vs. ROIauto (green contour) comparisons in lung, liver, and prostate tumors
time delay as mentioned in Sect. “NifteRT step 1: 3D tumor positions in the study. The first 0.5 min
Intra-Fractional MR Imaging.” The range for our of data was used for LSTM training, and the
time delay during nifteRT was approximately remaining 7.5 min of data was used to validate
200–500 ms [64]. Without the motion prediction the motion prediction. The prediction error was
capability, therefore, the therapeutic radiation defined as the distance between true and pre-
beam will always be lagging behind the actual dicted tumor positions. We calculated the mean
tumor motion. Ideally, a future tumor shape error, as well as 99% confidence interval (CI),
should also be predicted if the tumor deforms over the 46 patients to address outliers. We
severely during the time delay. Although this may used = mean ± ( z _ score × standard deviation ) / sample size ,
be feasible by establishing a temporal tumor where z_score = 2.58, to calculate 99% CI. Using
deformation model during free breathings, it is the LSTM algorithm, the 99% CI for the mean
not addressed in the current nifteRT scheme. This error was from 0.9 ± 0.1 mm to 1.6 ± 0.2 mm for
is based on the observation that the deformation 120–520 ms time delays, respectively.
between each imaging (250 ms) is small, Figure 11.8 presents two patient examples of 3D
<0.6 mm in radius from our previous study [65], tumor motion prediction at 280 ms time delay
if each tumor shape is approximated as a circle. (pancreatic tumor: irregular motion; lung tumor:
Previously we used a feed forward artificial regular, cyclic motion).
neural network, specifically multilayer percep-
tron (MLP), to predict 1D tumor motion [71]. In
MLP, however, overfitting is difficult to avoid, ifteRT Step 4: Real-Time MLC
N
and no time factor (previous memory) is involved Control
in the prediction [72]. To address these short-
comings and add 3D motion prediction capabil- Immediately after the motion prediction, the
ity, our latest algorithm adopted long short-term MLC is driven to conform the radiation beam
memory (LSTM), which is a type of DLNN to the auto-contoured tumor shape at its pre-
applicable to time series predictions [73, 74]. dicted position. To control the MLC on-the-fly
Our algorithm was validated using various 3D during nifteRT, we developed the software/
tumor motion data from 46 patients (30 lung, 9 hardware components of MLC control system
pancreas, 2 liver, 2 retroperitoneum, 2 chest in-house and verified the system using a Varian
wall, 1 internal mammary nodes). In this initial 52 leaf MLC [64]. Also, we investigated the RF
study using LSTM, 46 patients were chosen noise shielding for proper MR imaging during
based on data availability. Each patient’s data the MLC operation and the magnetic field
contains 8 min of 3D tumor motion recorded at effect on the MLC motors [4, 8]. We confirmed
40 ms intervals, which was considered as true the proper MLC motor operation, as well as
Fig. 11.8 3D tumour motion prediction using LSTM algorithm, pancreatic and lung tumour patients. Note: Color
overlapping indicates high degree of agreement between predicted and true tumour positions
a b c e
Fig. 11.9 Experimental setup: (a) top-down view dia- get, (e) phantom and RF coil placed in the linac-
gram of entire setup, (b) motion phantom components, (c) MR. Taken from Ref. [64]. Med Phys. 2013;40(5):(12pp)
an assembled phantom, (d) MR image of the moving tar- 051718
MR imaging during nifteRT in our linac-MR While our phantom was driving a target repro-
system [64]. ducing 1D lung tumor motion within the linac-MR
system (motion range: 4 cm; period: 5 s), we were
able to (1) acquire intra-fractional MR images of the
Demonstration of NifteRT moving target at 250 ms/frame image acquisition
in a Prototype Linac-MR time, (2) automatically detect the shape and location
of the target from each MR image, (3) predict the
We reported the world’s first demonstration of future target position over the time delay, and (4)
nifteRT utilizing 2D intra-fractional MR imaging control the MLC on-the-fly to conform the radiation
and a motion phantom in our prototype linac-MR beam to the auto-contoured target shape at its pre-
system [64]. A brief diagram of the experimental dicted position. These four steps were continuously
setup, motion phantom, and MR image of a mov- run each time a new MR image was acquired, and
ing target acquired during irradiation is shown in the radiation beam was always on during nif-
Fig. 11.9. teRT. The result is summarized in Fig. 11.10.
Fig. 11.10 Film measurements and motion pattern comparisons between target and MLC. From Ref. [64] Med Phys.
2013;40(5):(12pp) 051718T
Our target, with Gafchromic film inserted, were modeled to score the phase space at 120 cm
was irradiated for 2 min receiving ~2.5 Gy. The isocenter and the PDDs results validated by mea-
radiation delivery to a static or a moving target surements. The contaminate electrons originate
would be identical if nifteRT tracks the target in air at 60 cm from target near the magnetic pole
perfectly. Thus, the film exposed in moving target plate and progressively increase to the isocenter
irradiation was compared to the static target one to produce the surface-dose increase [78].
via dose profiles. Also, two independently The LMR 2000 has an SAD of 120 cm, and
recorded motion patterns are compared from the the Bo can safely be on/off in minutes. A conven-
target and the MLC during irradiation. The feasi- tional TPS (Eclipse v13.6 TPS with doses calcu-
bility of nifteRT was verified by comparing the lated with the Pencil Beam Convolution algorithm
two dose profiles. We delivered highly accurate to accommodate the SAD) without B0 was com-
radiation to a moving target reproducing 1D lung missioned for the LMR 2000 with B0 on [80].
tumor motion with <1.7 mm error. Calculated tissue maximum ratios (TMR) and the
tissue phantom ratios (TPR10) agree with mea-
surements within 0.4% for source-to-point-
Calculated and Measured Radiation distance (SPD) of 120 cm and 10 × 10 cm2 fields
Doses with and without B0 at 1, 3, 5, and 10 cm depth. The calculated TMR
and TPR10 agree within 0.48% and 0.26%,
One of the characteristics of the Alberta linac- respectively, with measurements for the SPD of
MR (a.k.a. Aurora-RT) is that the MR’s Bo field 127 cm, at depths of 15 and 20 cm for
does not affect the response of the radiation 9.4 × 9.4 cm2 field sizes. PDD Gammas
detectors nor the dose distribution perturbations (3%|3 mm) are 94.9% and 100% if surface dose
[1, 75, 76]. This simplifies the commissioning is excluded. At all depths to 37 cm, inline and
process and provides the probability of using crossline profiles agree within 96.4 and 94.4%,
treatment planning systems that do not explicitly respectively. The Gammas of IMRT plans
require code to consider magnetic fields. (3%|3 mm) delivered at 0° are 95.3% (five-field
Surface D0.2 cc skin dose [77] is measured and lung), 95.9% (seven-field brain), and 92.5%
calculated to increase from 46% to 75% when the (five-field prostate). Other more advanced TPSs
field is turned on a 20 × 20 cm en face fields but that can accommodate the 120 cm SAD are being
increases only by 3–9% for tangents and five- investigated.
field cases [78]. The increase is caused by con- Reference dose calibration was performed
taminant electrons that may be captured and [81] with EGSnrc simulations to validate TG51
focused in the parallel configuration [79]. kQ,msr with external validation from IROC
BEAMnrc radiation and Opera3D magnetic field (Imaging and Radiation Oncology Core, Houston,
TX). TPR20,10 is converted to equivalent 6. 3D MR imaging is about 3 min, while real-

%dd10× to kQ,msr with EGSnrc (MC) validation time imaging is at 4 fps providing online
of the Exradin A12 used with and without the beam’s eye view due to the configuration of
magnetic field. Calibrated dose agrees for the the beam with respect to the MR system and
LMR 2000 within 1% for IROCs, OSLDs, and the rotation of the gantry.
TLDs. Measurements of all TG51 parameters
with field are essentially equivalent to that with-
out field. Thus, kQ is 1 which means no magnetic References
field effects on the dose. Please note that linac-
MR with perpendicular configuration can only 1. Kirkby C, Murray B, Rathee S, Fallone BG. Lung
dosimetry in a linac-MRI radiotherapy unit
use TLDs for routine output to IROC as OSLD with a longitudinal magnetic field. Med Phys.
reports have large variations [82]. 2010;37(9):4722–32.
The National Research Council (NRC), 2. Fallone BG, Murray B, Rathee S, Stanescu T, Steciw
Canada, performed Co-60 alanine calibration and S, Vidakovic S, et al. First MR images obtained
during megavoltage photon irradiation from a pro-
determined Dread/Dstated of 1.0092 and 0.9986 totype integrated linac-MR system. Med Phys.
for water and solid water. The level of agreement 2009;36(6):2084–8.
between measured and stated is within the k = 1 3. Burke B, Lamey M, Rathee S, Murray B, Fallone
uncertainty (within 1%), and the NRC ruled that BG. Radio frequency noise from clinical linear accel-
erators. Phys Med Biol. 2009;54(8):2483–92.
no transport correction to Dread for the dosimeters 4. Lamey M, Yun J, Burke B, Rathee S, Fallone
irradiated in the linac-MR beam is required. BG. Radio frequency noise from an MLC: a feasibil-
ity study of the use of an MLC for linac-MR systems.
Phys Med Biol. 2010;55(4):981–94.
5. Lamey M, Burke B, Blosser E, Rathee S, De Zanche
Conclusions N, Fallone BG. Radio frequency shielding for a linac-
MRI system. Phys Med Biol. 2010;55(4):995–1006.
The system offers several interesting 6. St. Aubin J, Steciw S, Fallone BG. The design of a
characteristics: simulated in-line side-coupled 6 MV linear accelera-
tor waveguide. Med Phys. 2010;37(2):466–76.
7. St Aubin J, Steciw S, Fallone BG. Waveguide detuning
1. NO ERE with B0-induced minimal dose per- caused by transverse magnetic fields on a simulated
turbation. Conventional commissioning and in-line 6 MV linac. Med Phys. 2010;37(9):4751–4.
QA can be performed without the need of B0 8. Yun J, St Aubin J, Rathee S, Fallone BG. Brushed per-
manent magnet DC MLC motor operation in an exter-
corrections, while conventional TPS can be
nal magnetic field. Med Phys. 2010;37(5):2131–4.
used. 9. St Aubin J, Steciw S, Fallone BG. Magnetic decou-
2. The large 110 cm by 60 cm rotating bore with pling of the linac in a low field biplanar linac-MR sys-
couch motion allows positioning of the target tem. Med Phys. 2010;37(9):4755–61.
10. Fallone BG. The rotating biplanar linac-magnetic
to any planned position including isocenter
resonance imaging system. Semin Radiat Oncol.
positioning for offset tumors. 2014;24(3):200–2.
3. Installation is simplified with installation 11. St Aubin J, Santos DM, Steciw S, Fallone BG. Effect
through conventional door or maze of the of longitudinal magnetic fields on a simulated in-line
6 MV linac. Med Phys. 2010;37(9):4916–23.
vault without the need of an exhaust vent. The
12. Wachowicz K, De Zanche N, Yip E, Volotovskyy
high-temperature superconducting MR allows V, Fallone BG. CNR considerations for rapid real-
safe on/off switching within minutes. time MRI tumor tracking in radiotherapy hybrid
4. Because of item 2 above, there is no need to devices: effects of B0 field strength. Med Phys.
2016;43(8):4903.
adapt for every patient because the tumor,
13. Reynolds M, Fallone BG, Rathee S. Dose response of
even those offset, can be placed at any planned selected ion chambers in applied homogeneous trans-
position. verse and longitudinal magnetic fields. Med Phys.
5. Because of item 1 above, any third party TPS 2013;40(4):042102.
14. de Pooter J, Billas I, de Prez L, Duane S, Kapsch
and adaptive therapy system can be used.
RP, Karger CP, et al. Reference dosimetry in MRI-
Gating is the current delivery system.
linacs: evaluation of available protocols and data ing: a novel MR imaging strategy for real time track-
to establish a code of practice. Phys Med Biol. ing of lung tumors. Med Phys. 2014;41(8):082301.
2021;66(5):05TR2. 31. Yip E, Yun J, Wachowicz K, Gabos Z, Rathee S,
15. Reynolds M, Fallone BG, Rathee S. Technical note: Fallone BG. Sliding window prior data assisted com-
response measurement for select radiation detectors pressed sensing for MRI tracking of lung tumors. Med
in magnetic fields. Med Phys. 2015;42(6):2837–40. Phys. 2017;44(1):84–98.
16. Reynolds M, Rathee S, Fallone BG. Technical note: 32. Dietz B, Yip E, Yun J, Fallone B, Wachowicz K. Real-
ion chamber angular dependence in a magnetic field. time dynamic MR image reconstruction using com-
Med Phys. 2017;44(8):4322–8. pressed sensing and principal component analysis
17. Malkov VN, Rogers DWO. Monte Carlo study of (CS-PCA): demonstration in lung tumor tracking.
ionization chamber magnetic field correction factors Med Phys. 2017;44(8):3978–89.
as a function of angle and beam quality. Med Phys. 33. Yip E, Yun J, Gabos Z, Baker S, Yee D, Wachowicz
2018;45(2):908–25. K, et al. Evaluating performance of a user-trained
18. Spindeldreier CK, Schrenk O, Bakenecker A, MR lung tumor autocontouring algorithm in the con-
Kawrakow I, Burigo L, Karger CP, et al. Radiation text of intra-and interobserver variations. Med Phys.
dosimetry in magnetic fields with farmer-type 2018;45(1):307–13.
ionization chambers: determination of magnetic 34. Wright M, Dietz B, Yip E, Yun J, Gabos Z, Fallone B,
field correction factors for different magnetic field et al. Time domain principal component analysis for
strengths and field orientations. Phys Med Biol. rapid, real-time 2D MRI reconstruction from unders-
2017;62(16):6708–28. ampled data. Med Phys. 2021;48(11):6724–39.
19. van Asselen B, Woodings SJ, Hackett SL, van Soest 35. Dietz B, Yun J, Yip E, Gabos Z, Fallone B, Wachowicz
TL, Kok JGM, Raaymakers BW, et al. A formal- K. Single patient convolutional neural networks for
ism for reference dosimetry in photon beams in real-time MR reconstruction: a proof of concept
the presence of a magnetic field. Phys Med Biol. application in lung tumor segmentation for adaptive
2018;63(12):125008. radiotherapy. Phys Med Biol. 2019;64(19):195002.
20. Reynolds M, Fallone BG, Rathee S. Dose response of 36. Dietz B, Yun J, Yip E, Gabos Z, Fallone B, Wachowicz
selected solid state detectors in applied homogeneous K. Single patient convolutional neural networks for
transverse and longitudinal magnetic fields. Med real-time MR reconstruction: coherent low-resolution
Phys. 2014;41(9):092103. versus incoherent undersampling. Phys Med Biol.
21. Ghila A, Fallone BG, Rathee S. Influence of standard 2020;65(8):08NT3.
RF coil materials on surface and buildup dose from 37. Park J-H, Han JH, Kim C-Y, Oh CW, Lee D-H,
a 6 MV photon beam in magnetic field. Med Phys. Suh T-S, et al. Application of the gamma evaluation
2016;43(11):5808. method in gamma knife film dosimetry. Med Phys.
22. Bray FN, Simmons BJ, Wolfson AH, Nouri 2011;38(10):5778–87.
K. Acute and chronic cutaneous reactions to ion- 38. Huang F, Ma C, Wang R, Gong G, Shang D, Yin
izing radiation therapy. Dermatol Ther (Heidelb). Y. Defining the individual internal gross tumor vol-
2016;6(2):185–206. ume of hepatocellular carcinoma using 4DCT and
23. Jackson JD. Classical electrodynamics, vol. xxi. 3rd T2-weighted MRI images by deformable registration.
ed. New York: Wiley; 1999. p. 808. Transl Cancer Res. 2018;7(1):151–7.
24. Corea JR, Lechene PB, Lustig M, Arias AC. Materials 39. Feng M, Balter JM, Normolle D, Adusumilli S, Cao
and methods for higher performance screen-printed Y, Chenevert TL, et al. Characterization of pancreatic
flexible MRI receive coils. Magn Reson Med. tumor motion using cine MRI: surrogates for tumor
2017;78(2):775–83. position should be used with caution. Int J Radiat
25. Barta R, Ghila A, Rathee S, Fallone BG, De Zanche Oncol Biol Phys. 2009;74(3):884–91.
N. Impact of a parallel magnetic field on radiation 40. Kyriakou E, McKenzie DR. Changes in lung
dose beneath thin copper and aluminum foils. Biomed tumor shape during respiration. Phys Med Biol.
Phys Eng Express. 2020;6(3):037002. 2012;57(4):919–35.
26. https://doi.org/10.1002/mrm.28540. 41. Zhang GG, Yu H-HM, Stevens CW, Dilling TJ, Hoffe
27. Edelstein WA, Glover GH, Hardy CJ, Redington SE, Moros EG, et al. Motion management in stereo-
RW. The intrinsic signal-to-noise ratio in NMR imag- tactic body radiotherapy. J Nucl Med Radiat Ther.
ing. Magn Reson Med. 1986;3(4):604–18. 2012;S6(012):6.
28. Lustig M, Donoho DL, Pauly JM. Sparse MRI: the 42. Hendry JH, Jeremic B, Zubizarreta EH. Normal tis-
application of compressed sensing for rapid MR sue complications after radiation therapy. Rev Panam
imaging. Magn Reson Med. 2007;58(6):1182–95. Salud Publica. 2006;20(2–3):151–60.
29. Goldstein T, Osher S. The split Bregman method 43. Vedam SS, Kini VR, Keall PJ, Ramakrishnan V,
for L1-regularized problems. Siam J Imaging Sci. Mostafavi H, Mohan R. Quantifying the predict-
2009;2(2):323–43. ability of diaphragm motion during respiration
30. Yip E, Yun J, Wachowicz K, Heikal AA, Gabos Z, with a noninvasive external marker. Med Phys.
Rathee S, et al. Prior data assisted compressed sens- 2003;30(4):505–13.
44. Nioutsikou E, Seppenwoolde Y, Symonds-Tayler 58. Plathow C, Fink C, Ley S, Puderbach M, Eichinger
JRN, Heijmen B, Evans P, Webb S. Dosimetric inves- M, Zuna I, et al. Measurement of tumor diameter-
tigation of lung tumor motion compensation with a dependent mobility of lung tumors by dynamic
robotic respiratory tracking system: an experimental MRI. Radiother Oncol. 2004;73(3):349–54.
study. Med Phys. 2008;35(4):1232–40. 59. Eccles CL, Patel R, Simeonov AK, Lockwood G,
45. Giraud P, Houle A. Respiratory gating for radio- Haider M, Dawson LA. Comparison of liver tumor
therapy: main technical aspects and clinical benefits. motion with and without abdominal compression
ISRN Pulmonol. 2013;2013:1–3. using cine-magnetic resonance imaging. Int J Radiat
46. Bhagat N, Fidelman N, Durack JC, Collins J, Gordon Oncol Biol Phys. 2011;79(2):602–8.
RL, LaBerge JM, et al. Complications associated 60. Tong X, Chen X, Li J, Xu Q, Lin M-H, Chen L,
with the percutaneous insertion of fiducial market al. Intrafractional prostate motion during exter-
ers in the thorax. Cardiovasc Intervent Radiol. nal beam radiotherapy monitored by a real-time
2010;33(6):1186–91. target localization system. J Appl Clin Med Phys.
47. Roberge D, Cabrera T. Liver biopsy in modern medi- 2015;16(2):51–61.
cine. UK: IntechOpen Ltd.; 2011. 61. Sihono DSK, Ehmann M, Heitmann S, Von
48. Kitamura K, Shirato H, Shimizu S, Shinohara N, Swietochowski S, Grimm M, Boda-Heggemann
Harabayashi T, Shimizu T, et al. Registration accu- J, et al. Determination of Intrafraction prostate
racy and possible migration of internal fiducial gold motion during external beam radiation therapy
marker implanted in prostate and liver treated with with a transperineal 4-dimensional ultrasound real-
real-time tumor-tracking radiation therapy (RTRT). time tracking system. Int J Radiat Oncol Biol Phys.
Radiother Oncol. 2002;62(3):275–81. 2018;101(1):136–43.
49. Imura M, Yamazaki K, Shirato H, Onimaru R, 62. Chavhan GB, Babyn PS, Jankharia BG, Cheng H-LM,
Fujino M, Shimizu S, et al. Insertion and fixation Shroff MM. Steady-state MR imaging sequences:
of fiducial markers for setup and tracking of lung physics, classification, and clinical applications.
tumors in radiotherapy. Int J Radiat Oncol Biol Phys. Radiographics. 2008;28(4):1147–60.
2005;63(5):1442–7. 63. Keall PJ, Mageras GS, Balter JM, Emery RS, Forster
50. Ozhasoglu C, Murphy MJ. Issues in respiratory motion KM, Jiang SB, et al. The management of respiratory
compensation during external-beam radiotherapy. Int motion in radiation oncology report of AAPM task
J Radiat Oncol Biol Phys. 2002;52(5):1389–99. group 76. Med Phys. 2006;33(10):3874–900.
51. Paulsson AK, Yom SS, Anwar M, Pinnaduwage D, 64. Yun J, Wachowicz K, Mackenzie M, Rathee S,
Sudhyadhom A, Gottschalk AR, et al. Respiration- Robinson D, Fallone BG. First demonstration of
induced intraorgan deformation of the liver: impli- intrafractional tumor-tracked irradiation using 2D
cations for treatment planning in patients treated phantom MR images on a prototype linac-MR. Med
with fiducial tracking. Technol Cancer Res Treat. Phys. 2013;40(5):051718.
2017;16(6):776–82. 65. Yun J, Yip E, Gabos Z, Wachowicz K, Rathee S,
52. Chen S, Lin C, Enke C, Zhou S. SU-E-J-122: quanti- Fallone BG. Improved lung tumor autocontour-
fication of respiratory-induced pancreatic head tumor ing algorithm for intrafractional tumor tracking
rotation and deformation using 4DCT and fiducial using 0.5 T linac-MR. Biomed Phys Eng Express.
markers. Med Phys. 2012;39(6Part8):3680. 2016;2(6):067004.
53. Gelder RV, Wong S, Le A, Podreka A, Briggs A, 66. Yun J, Yip E, Gabos Z, Baker S, Yee D, Wachowicz
Haddad C, et al. Experience with an abdominal com- K, et al. A deep learning-based tumor auto-contouring
pression band for radiotherapy of upper abdominal algorithm for real-time tumor tracking using linac-
tumours. J Med Radiat Sci. 2018;65(1):48–54. MR. Med Phys. 2018;45(6):E669.
54. Bouilhol G, Ayadi M, Rit S, Thengumpallil S, 67. Tahmasebi N, Boulanger P, Yun J, Fallone BG,
Schaerer J, Vandemeulebroucke J, et al. Is abdominal Punithakumar K. Tracking tumor boundary using
compression useful in lung stereotactic body radia- point correspondence for adaptive radio therapy.
tion therapy? A 4DCT and dosimetric lobe-dependent Comput Methods Prog Biomed. 2018;165:187–95.
study. Phys Med. 2012;29(4):333–40. 68. Yun J, Yip E, Wachowicz K, Rathee S, Mackenzie M,
55. Dempsey JF, Benoit D, Fitzsimmons JR, Haghighat Robinson D, et al. Evaluation of a lung tumor auto-
A, Li JG, Low DA, et al. A device for realtime 3D contouring algorithm for intrafractional tumor track-
image-guided IMRT. Int J Radiat Oncol Biol Phys. ing using low-field MRI: a phantom study. Med Phys.
2005;63(2):1095. 2012;39(3):1481–94.
56. Raaymakers BW, Wlagendijk JJ, Overweg J, Kok 69. Han G, Wachowicz K, Usmani N, Yee J, Wong J,
JGM, Raaijmakers AJE, Kerkhof EM, et al. Integrating Fallone BG, et al. Deep learning-based autocontour-
a 1.5 T MRI scanner with a 6 MV accelerator: proof ing algorithm for non-invasive intrafractional tumour-
of concept. Phys Med Biol. 2009;54(12):N229–N37. tracked radiotherapy (nifteRT) on Linac-MR. Med
57. van Herk M, Remeijer P, Rasch C, Lebesque Phys. 2022;49(8):5634.
JV. The probability of correct target dosage: dose- 70. Yun J, Yip E, Gabos Z, Wachowicz K, Rathee S,
population histograms for deriving treatment mar- Fallone BG. Neural-network based autocontouring
gins in radiotherapy. Int J Radiat Oncol Biol Phys. algorithm for intrafractional lung-tumor tracking
2000;47(4):1121–35. using Linac-MR. Med Phys. 2015;42(5):2296–310.
71. Yun J, Mackenzie M, Rathee S, Robinson D, Fallone based Boltzmann solver. Biomed Phys Eng Express.
BG. An artificial neural network (ANN)-based lung- 2020;6(2):025006.
tumor motion predictor for intrafractional MR tumor 77. Hilts M, Halperin H, Morton D, Batchelar D, Bachand
tracking. Med Phys. 2012;39(7):4423–33. F, Chowdhury R, et al. Skin dose in breast brachy-
72. Lawrence S, Giles CL, editors. Overfitting and neural therapy: defining a robust metric. Brachytherapy.
networks: conjugate gradient and backpropagation. 2015;14(6):970–8.
Proceedings of the IEEE-INNS-ENNS International 78. Oliver P, Yip E, Reynolds M, Burke B, Fallone
Joint Conference on Neural Networks IJCNN 2000 BG, Murray B, et al. Skin dose investigations on a
Neural Computing: New Challenges and Perspectives 0.5 T inline rotating biplanar linac-MR. Med Phys.
for the New Millennium; 2000 27-27 July 2000. 2022;49(8):5628.
73. Yun J, Rathee S, Fallone BG. A deep-learning based 79. Reynolds M, Oliver PAK, Wood T, Wachowicz
3D tumor motion prediction algorithm for non- K, Burke B, Fallone BG. Contaminant electron
invasive intra-fractional tumor-tracked radiotherapy origins on a 0.5T inline linac-MR. Med Phys.
(nifteRT) on Linac-MR. Int J Radiat Oncol Biol Phys. 2022;49(8):5681.
2019;105(1):S28. 80. Tari SY, Oliver P, Sinn D, Wood T, Steciw S, Murray B,
74. Johnson N, Wachowicz K, Rathee S, Fallone BG, Yun et al. Initial experience for treatment planning system
J. Accurate, on-demand neural networks for respira- commissioning of the 0.5 T inline rotating bi-planar
tory motion forecasting. Med Phys. 2022;49(8):5640. Linac-MR system. Med Phys. 2022;49(8):5693.
75. Yang R, Santos DM, Fallone BG, St-Aubin J. A novel 81. Yip E, Tari SY, Reynolds M, Sinn D, Murray B,
transport sweep architecture for efficient determinis- Fallone BG, et al. Clinical reference dosimetry on the
tic patient dose calculations in MRI-guided radiother- inline 0.5T rotating biplanar Linac-MR system. Med
apy. Phys Med Biol. 2019;64(18):185012. Phys. 2022;48(8):5630.
76. Yang R, Santos DM, Fallone BG, St-Aubin 82. Alvarez P, IROC, editors. Updates imaging and dosim-
J. Feasibility of energy adaptive angular meshing for etry relevant to MRgRT2021. Summer School of the
perpendicular and parallel magnetic fields in a grid American Association of Physicists in Medicine; 2021.
Clinical Applications of MR-Linac
in Oligometastatic Disease
12
Katie N. Lee, Martijn Intven, Filippo Alongi,
and Mai Anh Huynh
isease Overview: Oligometastatic

D tumor could translate into prolonged survival and
Disease even cure [1]. The oligometastatic state has been
increasingly recognized throughout the interven-
“Oligometastatic” disease (OMD) has been ing decades, and its definition continues to be
defined as an intermediate and distinct state of further refined to characterize different disease
metastatic spread, wherein metastatic lesions are states within the clinical spectrum of OMD [2].
confined to a single or limited number of organs, Currently, advances in methods of oncologic
and metastatic-directed therapy (MDT) can imaging, detection, and treatment have trans-
impact the natural history of cancer spread. formed cancer into an increasingly chronic dis-
Originally proposed in 1995 by Hellman and ease with most deaths (greater than 90%) from
Weichselbaum, they contended that this specific solid tumors caused by metastasis [3]. The under-
clinical scenario could deserve a radical approach standing of the metastatic disease continuum and
to localized therapy, as in this early state of meta- effective therapies for this disease state remains
static disease, the ability to ablate all detectable of utmost importance.
K. N. Lee · M. A. Huynh (*) Defining Oligometastatic Disease

Harvard Radiation Oncology Program, Brigham and
Women’s Hospital/Dana Farber Cancer Center/
Massachusetts General Hospital/Boston Children’s The definition of OMD continues to evolve since
Hospital, Boston, MA, USA its original conception as a limited number of
Department of Radiation Oncology, Dana-Farber/ metastases lying on the spectrum between local-
Brigham and Women’s Cancer Center and Harvard ized and more widespread tumor dissemination.
Medical School, Boston, MA, USA In clinical trials and meta-analyses, the number
e-mail: [email protected]; of oligometastatic sites has typically been limited
[email protected]
to a maximum of three to five lesions [4].
M. Intven However, some studies have included individual
Department of Radiotherapy, Division Imaging and
Oncology, University Medical Centre Utrecht, lymph nodes in this number, whereas others have
Utrecht, The Netherlands not, and some have limited the number of metas-
e-mail: [email protected] tases allowable in a single organ [5–8]. Following
F. Alongi a systematic review of the OMD literature, which
Advanced Radiation Oncology Department, Cancer has predominantly been comprised of retrospec-
Care Center, IRCCS Sacro Cuore Don Calabria tive experiences, the European Society for
Hospital, Negrar, VR, Italy
218 K. N. Lee et al.
Therapeutic Radiation Oncology and American However, there are likely additional factors
Society for Therapeutic Radiation Oncology characterizing OMD that go beyond this subclas-
(ESTRO-ASTRO) consensus guidelines defined sification of the OMD state and may affect patient
OMD as one to five lesions, where a controlled outcomes. These characteristics, such as lesion
primary is optional, but where all metastatic sites number, site, size, and histology, remain contro-
must be safely treatable. In the absence of clinical versial, and their exact contribution remains
data supporting a maximum number of metasta- unknown. Furthermore, rapid improvements in
ses and organs to define OMD, ESTRO-ASTRO diagnostic imaging, specifically in prostate can-
consensus guidelines further specify the ability to cer with prostate specific membrane antigen
deliver safe and clinically meaningful radiother- (PSMA)-PET, have led to earlier detection of
apy with curative intent to all metastatic sites as a lower oligometastatic disease burden [12]. With
minimum requirement [9]. most prior studies defining the OMD state by
Beyond a numeric view of OMD, it is impor- number of sites based on conventional imaging,
tant to recognize that OMD could be considered now with these improved detection methods, the
a dynamic state. The magnitude of benefit of OMD state and application of these studies will
MDT for patients may vary depending on the likely need to be further refined [5, 13, 14].
biology underlying the relative timing of meta-
static presentation, prior exposure to systemic
therapy, or how MDT compares as an alternative SBRT Treatment for Oligometastatic
to the toxicities associated with systemic therapy. Disease
Recognizing the heterogeneity represented
within OMD, the ESTRO and European The treatment paradigm for OMD, in general,
Organization for Research and Treatment of has been changing with studies demonstrating a
Cancer (EORTC) created a consensus recom- survival benefit for definitive treatment of
mendation of nine subgroups [10]. Categories metastases when limited to a maximum of three
associated with oligometastases include synchro- to five sites, both in the upfront setting and at
nous, in which the primary cancer and oligome- time of relapse and progression. Hellman and
tastases present at time of initial diagnosis (within Weichselbaum in 1995 insightfully hypothe-
6 months); metachronous, in which OMD devel- sized that the application of conformal radio-
ops following treatment of the primary cancer therapy, involving high doses of radiation
(>6 months); and induced, in which patients with delivered precisely to lesions with normal tissue
a history of polymetastatic disease are reduced to sparing, could be of benefit for those with OMD
an oligometastatic number of lesions following [1]. Advancements in radiotherapy with image
systemic treatment (Fig. 12.1). Further defining guidance, respiratory motion management, and
the OMD state are oligorecurrence, in which improved methods of treatment planning and
OMD develops while systemic therapy-free; oli- delivery enabled the development of stereotactic
goprogression, in which oligometastatic sites body radiotherapy (SBRT), where ablative
progress while under treatment with systemic doses of radiation can be given with rapid dose
therapy; and oligopersistence, in which there is falloff to surrounding normal tissues. SBRT has
stable disease or partial response while under increasingly been adopted to ablate common
treatment with systemic therapy. The prognostic areas of metastatic spread, including the lung,
significance of the ESTRO-EORTC classification liver, colorectal region, adrenal glands, and
system remains unknown, although some studies bone [13, 15–23]. Draining lymph node regions
have identified improved PFS and OS among are also common sites for cancer spread, and
patients with metachronous, synchronous, and SBRT for oligometastatic nodal disease has
repeat presentations vs. induced using this clas- been associated with greater than 80% local
sification system [11]. control [24]. Severe (greater than grade 3) toxic-
12 Clinical Applications of MR-Linac in Oligometastatic Disease 219
a b c
Synchronous Metachronous Induced
Oligopersistence
Oligorecurrence
Oligoprogression
Fig. 12.1 Oligometastatic disease classification. (a) ment of the primary (red), leads to OMD. (d)
Synchronous OMD where primary (red) and metastatic Oligorecurrence where OMD with new or regrowing
lesions (blue) present at time of initial diagnosis or within lesions develops following systemic therapy-free interval;
6 months. (b) Metachronous OMD where following treat- oligoprogression where OMD with new or regrowing
ment of primary (red) and systemic therapy-free interval lesions develops while undergoing systemic therapy; oli-
there is development of OMD (blue). (c) Induced OMD gopersistence where OMD lesions are persistent or have
where previously diagnosed polymetastatic disease fol- partial response while undergoing systemic therapy
lowed by systemic treatment, with or without local treat-
ity was mainly GI toxicity and occurred in the Supporting the role of metastasis-directed
setting of re-irradiation. For the purposes of this therapy (MDT) was the prospective phase II trial
book chapter, as other chapters address SABR-COMET, which randomized patients with
MR-guided treatments of the lung, liver, and a controlled primary solid tumor and up to five
bone specifically, we will focus on abdomino- distant metastases to either the palliative standard
pelvic oligometastatic lesions. of care (SOC) or standard therapy plus stereotac-
tic ablative body radiation (SABR) in a 1:2 ratio initiated for symptomatic progression, progres-
[25]. They demonstrated improved progression- sion of treated lesions, or progression to more
free survival (PFS) in patients treated with com- than three lesions. Most patients in the MDT arm
prehensive SABR for oligometastatic disease received SBRT as the mode of MDT (25 of 31
(5-year PFS 17.3% for control vs. not reached for patients). At a median follow-up of 3 years, ADT-
SABR, p = 0.001) and overall survival (OS), with free survival was improved in the MDT arm in
5-year OS being 17.7% for control vs. 42.3% for comparison with observation (21 months vs.
SABR (p = 0.006) [25]. Longer-term follow-up 13 months). Median time to PSA progression
at 8 years showed durable improvements in PFS, was also improved in the MDT arm (10 months
with estimated 21.3% in the SABR arm vs. 0% in vs. 6 months). Similarly, the ORIOLE trial ran-
the control arm (p < 0.001), as well as OS, with domized 54 men with HSPC to either observation
27.2% in the SABR arm compared to 13.6% in or SBRT to all sites of metastatic disease (1–3
the control arm (p = 0.008) [7]. sites based on conventional imaging) [28]. SBRT
In non-small cell lung cancer (NSCLC), improved 6-month PFS from 19% to 61% and
Gomez and colleagues looked at definitive treat- biochemical PFS from 11% to 50%. At baseline
ment of predominantly synchronous oligometa- and day 180, PSMA-PET was also performed,
static disease for patients with non-small cell and a post hoc analysis suggested an association
lung cancer (NSCLC) and showed benefit of both of significantly better PFS (hazard ratio [HR]
PFS (14.2 months vs. 4.4 months, p = 0.022) and 0.26; 95% CI, 0.09–0.76, p = 0.006) and distant
OS (41.2 months vs. 17 months, p = 0.017) in metastasis-free survival (HR 0.19; 95% CI, 0.07–
patients treated with local consolidative therapy 0.54, p < 0.001) among patients with no untreated
compared with observation or maintenance ther- lesions versus any untreated lesion by
apy [26]. In the SINDAS phase III trial, patients PSMA-PET.
with epidermal growth factor receptor (EGFR) In renal cell carcinoma (RCC), a large number
mutated NSCLC and one to five synchronous of patients present with OMD as shown in a
metastatic lesions without brain metastases were large, population-based analysis of 11,157
randomized to either SABR followed by first- patients where 61% had a single metastasis com-
generation EGFR tyrosine kinase inhibitor (TKI) pared to 39% with 2 or more metastatic sites [13,
versus TKI alone [27]. Due to the efficacy results 29]. SBRT has been an attractive alternative for
at the prespecified interim analysis, the trial was treating metastatic RCC given its relative radio-
stopped at enrollment of 133 patients as the resistance, which may be overcome by delivering
median PFS of 20.2 months in the RT + TKI arm a higher dose per fraction. In a prospective phase
vs. 12.5 months in the TKI only arm (p < 0.001). II single-arm trial of 23 patients with OMD
Improvement in median OS was also demon- defined as 1–3 extracranial metastatic sites,
strated of 25.5 months vs. 17.4 months 1-year PFS was 82.6% (95% CI, 60.1–93.1) and
(p < 0.001). freedom from systemic therapy was 91.3% (95%
In the phase II randomized STOMP and CI, 69.5–97.8) with 100% LC [30]. Concordantly,
ORIOLE trials, there was an association of MDT, in a phase II feasibility trial at MD Anderson
either surgery or RT, with improved PFS and Cancer Center with 30 RCC patients with 5 or
androgen deprivation therapy (ADT)-free sur- fewer oligometastatic lesions where all sites were
vival in patients with hormone-sensitive oligo- treated with SBRT, including subsequent sites of
metastatic prostate cancer (HSPC) [5, 6, 28]. In progression, median PFS was 22.7 months and
the STOMP trial, 62 men with asymptomatic oli- 1-year systemic therapy-free probability was
gorecurrent HSPC (1–3 metastases on choline 82% [31]. A prospective phase II multicenter trial
PET) were randomized to either MDT (SBRT or investigated the role of SBRT after TKI ther-
surgical resection) or surveillance [6]. ADT was apy for patients with oligopressive RCC with up
to five progressive metastases [32]. The median quently asymptomatic, and shared decision-
duration of TKI therapy prior to study entry was making between clinicians and patients regarding
18.6 months, and tumors treated with SBRT had an acceptable risk of toxicity in the pursuit of
1-year LC of 93%. Following SBRT, the median cure is critical. There is limited prospective evi-
PFS was 9.3 months (95% CI, 7.5–15.7 months), dence to justify comprehensive SBRT as superior
and median time to change in systemic therapy to the palliative SOC, and the ability to treat all
was 12.6 months. One-year OS was 92%. Hannan sites safely remains of utmost importance in a
and colleagues have also reported a phase II pro- population where the intent of treatment fre-
spective trial of 20 patients with oligoprogressive quently remains palliative despite the elusive
RCC with up to 3 progressive lesions on first- to goal of cure.
fourth-line systemic therapy [33]. At a median Important clinical considerations in determin-
follow-up of 10.4 months, SBRT extended the ing those who would most benefit from SBRT in
duration of ongoing systemic therapy by more OMD are prognosis, performance status (PS),
than 6 months in 14 patients (70%), and the comorbidities, and age. Patients who would ben-
median time from SBRT to new systemic therapy efit from more intensive techniques, such as
or death was 11.1 months. One patient developed SBRT, should have good life expectancy, whereas
possibly treatment-related grade 3 gastrointesti- those with poorer prognosis may derive palliative
nal toxicity. benefit from conventional radiotherapy. Overall
PS of the patient is an important measure when
deciding on whether to treat oligometastatic dis-
Patient Selection Characteristics ease as most studies have demonstrated a strong
correlation between PS and OS [34, 35].
As with any treatment choice, appropriate patient Another important consideration in deciding
selection is critical. In the OMD setting, this is whether to treat OMD with SBRT is the histology
even more imperative given the need to identify of the primary tumor. While promising results
those patients that could benefit from a more have been seen in the treatment of OMD in a
aggressive approach with MDT and balancing number of primary cancers including prostate,
the potential benefit with the risk of toxicity. In NSCLC, and renal cancer, recent prospective
the treatment of OMD, there are a number of data suggests that MDT does not improve PFS or
options including surgery, ablative radiotherapy, OS for patients with oligometastatic breast can-
interventional radiology (IR)-guided approaches, cer [36, 37]. Recently reported is the NRG-BR002
or systemic therapy. Treatment decisions should randomized phase II/III trial (NCT02364557)
be made with a multidisciplinary team to identify wherein in the phase II portion, 129 patients with
the optimal local approach in the context of a oligometastatic breast cancer with 1–4 extracra-
controlled primary tumor and a plan to address nial metastases were randomized to SOC sys-
all sites of disease. SBRT as a treatment option temic therapy (chemotherapy, endocrine therapy,
for local ablation of OMD offers a number of or anti-HER2) with or without SBRT to all meta-
attractive advantages in comparison with static lesions [36]. At median follow-up of
metastasis-directed surgery, including the ability 30 months, no difference was seen in PFS
to target multiple metastatic lesions, limited side (23 months in patients with SOC alone vs.
effect profile, and the less invasive nature of 20 months in patients with SBRT and SOC).
SBRT, which may expand treatment options for Given this result, the trial was terminated and
patients with comorbidities and also may reduce will not proceed to the phase III component. Tsai
time off of systemic therapy. However, it is and colleagues presented the results of their
important to note that SBRT to metastatic lesions phase II Consolidative Use of Radiotherapy to
is not without risk. Patients with OMD are fre- Block (CURB) study where patients with NSCLC
or breast cancer and oligoprogressive disease motion, and registration with high-resolution
(one to five oligometastatic sites) were random- imaging for stereotactic treatment planning are
ized to SBRT to all sites of progressive disease required. However, standard SBRT relies on
vs. palliative SOC and found an association of cone-beam CT (CBCT) image guidance, and
SBRT with improved PFS (median PFS 22 weeks while good results can be obtained, there are
for SBRT vs. 10 weeks for palliative SOC, limits in defining the soft- tissue boundary
p = 0.005) [37]. This benefit was driven by between target and normal tissues. In cases
patients with NSCLC [37]. No difference in where imaging quality is insufficient to delin-
median PFS was seen in the breast patient cohort eate the target or adjacent organs at risk, reli-
(18 weeks for SBRT vs. 17 weeks for palliative ance on rigid registration, such as in the form of
SOC, p = 0.5). This underscores the possible bio- bony landmarks or fiducial markers, may be
logical and histological differences in metastatic required. Fiducial marker placement can pro-
lesions from different primary cancers. The value long the patient’s timeline to treatment, and
of MDT may diminish in the presence of there is risk for minor and major complications,
improved systemic therapy or if there is a high such as fiducial marker migration, bleeding,
competing risk of radiographically occult micro- infection, and pneumothorax [40]. Further
metastatic disease. complicating the accurate delivery of SBRT are
In the absence of clinical biomarkers for variations in the anatomy on a day-to-day or
selecting patients with OMD, a long metastasis- even minute-to-minute basis (e.g., bowel peri-
free interval (MFI) has consistently been stalsis, bladder filling, rectal gas, etc.). Given
associated with improved outcomes among
the limits of soft-tissue definition and uncer-
patients treated with SBRT. In a large case series tainties with positioning, the use of CT can
of 1033 patients with oligometastatic disease result in the necessity of increasing planning
(1–5 lesions) who were treated with SBRT, mul- target volume (PTV) margins, thereby increas-
tivariate analysis identified more than 24 months ing the dose to adjacent OARs. Conversely, the
of MFI since initial diagnosis as a survival factor use of planning organ at risk volume (PRV)
(HR 0.63; 95% CI, 0.49–0.80, p < 0.001) [38]. In margins to buffer for the uncertainty of OAR
consideration of patients that would benefit from position and intra- and inter-fraction motion
SBRT or MDT in the oligometastatic setting, can limit the dose prescribed to target. The
these factors should be considered as part of the close approximation of oligometastatic sites to
multidisciplinary discussion for treatment nearby critical or radiation sensitive organs,
options. such as loops of bowel, esophagus, visceral
organs, or nerve structures like the lumbar
sacral plexus, frequently limits the potential
Overcoming Challenges therapeutic benefit of SBRT by preventing dose
with Traditional Approaches escalation to ablative doses due to the risk of
dose-limiting toxicity, including risk of perfo-
Advances in technology have increasingly ration or fistula, which can be fatal, or perma-
allowed for lesions to be targeted with high nent nerve injury. Underscoring the importance
doses of radiation with a steep dose falloff in all of safe delivery and potential toxicities to nor-
directions to minimize toxicity to surrounding mal tissues, in the SABR-COMET trial, there
tissues [39]. SBRT requires precise delivery were three treatment- related deaths in the
with submillimeter accuracy given these abla- SABR treatment arm, which included a subdu-
tive doses. Secure immobilization, accurate ral hemorrhage following surgery for a perfo-
repositioning, rigorous accounting for tissue rated gastric ulcer [25].
The advent of MRI-guided radiation, which fraction motion management based on real-time
has only recently been developed with Food and soft-tissue imaging can be further refined using
Drug Administration (FDA) approval in 2017, gating and tracking techniques for improved tar-
provides a new and improved ability to visualize geting of the lesion and eliminating the need for
the lesion and normal tissues with the capability an internal target volume (ITV) and larger mar-
of adapting treatment in real time. MRI provides gins around the target. As a result, surrounding
superior soft-tissue discrimination compared to normal tissues are able to receive a reduced radia-
CT [41]. In oligometastatic disease where meta- tion dose.
static lymph nodes and lesions may be close to Utilizing MR to visualize soft-tissue targets
organs at risk or vessels, improved imaging with and organs at risk in real time at time of treatment
high tissue contrast may be beneficial, especially delivery mitigates the issues of needing larger
in the abdomen and pelvis, and allow for highly margins around both target and organs at risk to
accurate radiation treatments that can spare sur- account for this uncertainty. This benefit is espe-
rounding normal tissues and reduce toxicities. cially relevant in the OMD setting where patients
Further, diagnostic metabolic imaging, such as often have had prior radiation treatment and
PSMA-PET in prostate cancer, is identifying ear- increased precision in treatment planning and
lier, smaller metastatic lesions that may be better delivery can further reduce the potential toxici-
visualized by MRI in comparison with standard ties associated with re-irradiation.
CT [12].
Additionally, online-adaptive MRI-guided
radiotherapy (MRgRT) enables daily adaptation Contouring, PTV, and OAR
so that volumes and plan can be changed accord-
ing to the appearance of the target and OARs on Defining the gross tumor volume (GTV), plan-
that day compensating for inter-fraction anatomic ning target volume (PTV), and organs at risk
changes. For example, within the abdomen and (OARs) is an important and critical step for safe
pelvis, bowel loops can fall within the treatment and effective delivery of stereotactic MR-guided
field. MRgRT can adapt for these changes and radiation therapy. During simulation for a 0.35
thereby theoretically allow for safer delivery of Tesla (T) MR-linac, patients undergo simulation
high doses of radiation therapy. A retrospective supine on both MR simulator and CT simulator
dosimetric analysis of 150 sessions for 23 patients with motion management via breath hold to
who received MRgSBRT on a 1.5 Tesla (T) reduce motion when necessary, which is typical
MR-linac for lymph node oligometastases found for abdominal targets. For pelvic targets or on
bowel Dmax to be significantly reduced with a 1.5 T MR-linac, no breath hold is needed. CT
daily adaptive strategy and single intestinal loop simulation may be omitted at centers with an MR
sparing [42]. pseudo-CT solution [43]. Contouring of the GTV
Further, there is potential for oligometastatic is done with reference to any diagnostic imaging
lesions, such as abdominopelvic lymph nodes or that is available, including PET-CT and diagnos-
soft-tissue deposits, to move with respiratory or tic quality MRI with or without contrast, prefer-
patient motion. In traditional CT-based radiother- ably with fused images, in order to identify
apy, four-dimensional CT is often utilized to targets.
depict the motion of the lesion through the Using an adaptive MR-guided workflow, typi-
breathing cycle and to generate an internal target cally, contouring is done on the MRI sequences
volume (ITV). Other strategies to reduce tumor which are used during treatment. A breath hold
motion include breath hold and abdominal com- technique with light inspiration or end inspiration
pression. With MR-guided radiotherapy, intra- is frequently used for abdominal lymph nodes on
a 0.35 T MR-linac to reduce respiratory motion DVH Constraints

and to optimize the capabilities of the MR-linac
for real-time monitoring. In most cases, a 3-mm With the treatment of oligometastatic disease
isotropic expansion from the GTV, with or with- with the goal of ablative therapy, the target pre-
out a clinical target volume (CTV) at the discre- scription can be between 20 and 50 Gy typically
tion of the treating physician, to the planning between one to five fractions dependent on the
target volume (PTV) is used. A CTV may be tolerance of the surrounding OARs (see Dose
indicated depending on the relative location of Escalation section for further discussion of dose).
multiple oligometastases, particularly lymph OAR constraints are a critical part of treatment
nodes or within the same organ, to encompass planning, especially in the delivery of SBRT;
areas at increased risk of regional relapse [44]. In however, there is wide variability in these con-
cases of spine oligometastases, a CTV should be straints in both prior studies and current clinical
contoured as per international consensus con- trials [47]. A key principle in the determination of
touring guidelines [45]. Organs at risk (OARs) OAR constraints includes evaluation of the organ
that are contoured vary by the location of the tar- type. Serial (e.g., spinal cord) and parallel (e.g.,
geted lesion but can include the spinal cord, duo- lungs, liver) organs differ in their radiation risks.
denum, small bowel, large bowel, stomach, For serial OARs, constraints are based around a
esophagus, kidneys, heart, liver, femoral heads, maximal threshold dose that is allowable to a
bladder, and rectum. OARs are contoured on the small volume (i.e., point dose). In contrast, OAR
breath hold sequence and used for treatment constraints for parallel organs typically relate to
planning. For organs that extend outside of the the maximum organ volume that can receive a
treatment field, the relevant organ is contoured specified threshold dose and the minimum criti-
within 2 cm of PTV in all directions or larger if cal volume of the OAR that must receive a speci-
indicated based on clinical scenario (to encom- fied threshold dose or lower [48]. The United
pass relevant lower dose isodose lines in setting Kingdom consensus for SABR constraints was
of stricter OAR constraints). The spinal cord is updated from their 2017 publication, in which
expanded with a PRV to 3 mm. Typically, gastro- they took into account the work of the
intestinal OARs are grouped together with the Hypofractionated Treatment Effects in the Clinic
small bowel, large bowel, stomach, duodenum, (HyTEC) and the systematic review by Gerhard
and esophagus and evaluated for overlap with the and colleagues (Table 12.1) [47–49]. In the sys-
GTV. If necessary due to overlap, the GTV will tematic review, 53 trials assessing the role of
then have the GI OARs subtracted from the final SBRT in the treatment of oligometastatic disease
GTV volume. In a study of adaptive MR-linac were examined for OAR constraints and the
treatment of pancreatic cancer, grouping of GI modal, median, and range reported. Although
OARs (stomach, duodenum, small bowel, and there was considerable variability, the modal and
large bowel) resulted in outperformance of con- median constraints reported were often closely
ventional plans and simplified treatment planning aligned to that of other commonly used dose con-
workflow [46]. Similarly, a PTV optimization straint guidelines in SBRT, the American
(PTV_opt) structure can be applied for ultra- Association of Physicist in Medicine Task Group
hypofractionation cases (one and three fraction (AAPM-GT) 101, and the Timmerman con-
treatments) where the spinal cord PRV (PRV_ straints [50–52]. As these oligometastatic SBRT
spinal cord) and GI OARs are subtracted if there trials mature, more information will likely
is overlap with the PTV. be gleaned as toxicities are reported.
12
Table 12.1 DVH constraints (adapted from UK 2022 consensus) [48]

Number of fractions Endpoint
1 3 5
Organ at risk Constraint Optimal Mandatory Optimal Mandatory Optimal Mandatory
Bowel_Small D0.1 cc 15.4 Gy 25.2 Gy 35 Gy Grade 3+ enteritis/obstruction
D5 cc 11.9 Gy 17.7 Gy
D10 cc 25 Gy
Duodenum D0.1 cc 12.4 Gy 22.2 Gy 33 Gy 35 Gy Grade 3+ ulceration
D10 cc 9 Gy 11.4 Gy 25 Gy
Kidney_Cortex (individual/combined) Dmean 8.5 Gy 10 Gy Grade 3+ renal dysfunction
Kidney_Cortex (combined) D≥200 cc 8.4 Gy 16 Gy 17.5 Gy
If solitary kidney or one Kidney_Cortex V10 Gy 33% 33% 10% 45%
Dmean constraint exceeded
Liver (non-liver) lesions and liver – GTV D≥700 cc 9.1 Gy 15 Gy 17 Gy 15 Gy Grade 3+ liver dysfunction, radiation induced
(liver lesions) liver disease (classic or non-classic)
V10 Gy 70%
Dmean 13 Gy 15 Gy 13 Gy 15.2 Gy
Stomach D0.1 cc 12.4 Gy 22.2 Gy 33 Gy 35 Gy Grade 3+ ulceration/fistula
D10 cc 11.2 Gy 16.5 Gy 25 Gy
Clinical Applications of MR-Linac in Oligometastatic Disease
D50 cc 12 Gy
BileDuct D0.1 cc 30 Gy 50 Gy 50 Gy
Esophagus D0.1 cc 15.4 Gy 25.2 Gy 35 Gy Grade 3+ stenosis/fistula
Bladder D0.1 cc 18.4 Gy 28.2 Gy 38 Gy Grade 3 cystitis/fistula
Bowel_Large D0.1 cc 18.4 Gy 28.2 Gy 38 Gy Grade 3+ colitis/fistula
Femur_HeadNeck D10 cc 14 Gy 21.9 Gy 30 Gy Grade 3+ necrosis
LumbSacPlex D0.1 cc 16 Gy 24 Gy 32 Gy Grade 3+ neuritis
D5 cc 14.4 Gy 22.5 Gy 30 Gy
Rectum D 18.4 Gy 28.2 Gy 38 Gy Grade 3+ proctitis/fistula
Ureter D0.1 cc 35 Gy 40 Gy
SpinalCord_PRV D0.035 cc 12.4 Gy 14 Gy 20.3 Gy 25.3 Gy Radiation myelopathy (1–5% risk)
CaudaEquina and SpinalCanal (below D0.035 cc 16 Gy 24 Gy 32 Gy Grade 3+ neuritis
level of cord)
D5 cc 14 Gy 21.9 Gy 30 Gy
225
Dose Escalation tion. In a proof of concept, Henke and colleagues

characterized five patients with non-liver abdo-
Ablative radiotherapy doses to oligometastatic men oligometastatic disease and five patients
lesions is one of the tenets of localized treatment with central thorax disease treated by MR-guided
of OMD, and ablation using SBRT to prolong SBRT and compared day-to-day anatomy varia-
PFS and overall survival is being increasingly tions to simulate adaptive planning [56]. Adaptive
recognized [6, 7, 26, 28]. However, the exact dos- planning allowed for increased PTV coverage in
ing for lesions remains an area of continued 70% (21 of 30 treatments), including 14 treat-
investigation as most randomized trials investi- ments in which hard OAR constraints were vio-
gating optimal radiotherapy dosing were con- lated by the non-adaptive plan. Dose escalation in
ducted in polymetastatic disease with nonmobile treatment of primary pancreatic cancer with use
targets (e.g., bone) with the goal of palliation of of MRgRT and real-time target localization has
symptoms, namely, pain. In determining the dose been demonstrated, allowing for dose escalation
for OMD lesions, the forefront priority remains by 36 Gy on average to at least 50 Gy [57]. Initial
minimizing risk of toxicity to surrounding OARs results from a multicenter, prospective study of
while delivering a therapeutic dose to the target. ablative adaptive MRgSBRT with a prescribed
An additional important factor in consideration BED10 of 100 Gy (50 Gy in five fractions) for
of dose and fractionation is histology, as exempli- pancreatic cancer demonstrate the safety and fea-
fied by RCC. Prior adjuvant RT studies with con- sibility of this approach, with a 2.2% rate of
ventional fractionation in RCC patients grade 3 toxicity [58]. Extrapolating this to oligo-
demonstrated no benefit and rather increased tox- metastatic lesions, MRgRT may allow for further
icity; however, with additional evidence of the increase in dose while sparing normal tissues. In
radioresistant nature of RCC and with the advent a study of 20 patients with oligometastatic dis-
of SBRT, meta-analyses of SBRT treatment for ease or unresectable disease of the abdomen,
oligometastatic RCC have now shown excellent PTV dose escalation was able to be achieved in
local control rates and safety for select patients 20 of 97 treatments with online daily adaptive
[53]. Further, a biologically effective dose (BED) plans [59].
greater than 115 Gy has also been associated with Further, there is motivation for utilizing fewer
improved outcomes [54]. Another important fac- fractions for the delivery of ablative radiation
tor in determination of dose and fractionation is therapy using MRgRT. Some patients travel from
size of the lesion and proximity to OARs. If not far distances specifically for the ability to receive
able to safely deliver a SBRT fractionation, mod- MRgRT and its potential advantages, and fewer
erately hypofractionated regimens with or with- fractions may reduce the logistical burdens on
out a simultaneous integrated boost (SIB) can patients (NCT04115254, DF/HCC Protocol
also be considered. In unresectable cholangiocar- #:19-353 SMART-STOP; NCT04939246 SMART
cinoma, a retrospective series of 79 patients who ONE). It also reduces the number of sessions that
were mainly treated with moderate hypofraction- a patient must be on the treatment table and rela-
ation and a central tumor boost of 75 Gy in 15 tively immobilized, which for adaptive planning
fractions or 100 Gy in 25 fractions to GTV, sub- can be upward of 90 min per session. The
tracted from a 5 mm PRV expansion on adjacent MR-linac is also a limited resource for which
OAR, found that a BED greater than 80.5 Gy there is an increasing demand and utility for mul-
correlated with improved LC and OS, with LC tiple cancer types. Utilizing fewer fractions may
rates (78%) comparable to resection [55]. potentially allow for resource optimization and
Online-adaptive MRgSBRT may allow for an allow a greater number of patients to access this
increase in PTV coverage while simultaneously novel technology. Additionally, there is emerging
improving OAR sparing compared to non- evidence that ultra-high dose fractionation,
adaptive SBRT, thereby allowing for dose escala- beginning at a threshold of 12 Gy, may provide
distinct advantages compared to lower doses per greater dose delivery to the OMD site with real-
fraction by inducing microvascular vasoactive time monitoring of OARs. Given the abilities of
dysfunction and inhibition of homologous adaptive planning, OARs can be differentially
recombination [60]. contoured into regions and set for different con-
A recent randomized phase III trial explored straints to account for prior radiation (Fig. 12.2).
whether a single fraction of 24 Gy would be In determining the area needed for adaptive
superior in local control compared to a 3 × 9 Gy recontouring, a ring structure expansion com-
fractionation for oligometastatic disease of non- monly is used, and if OARs fall within this
mobile lesions (bone and stationary lymph nodes) region, these differential constraints can then be
[61]. They found that the single fraction increased applied. By utilizing adaptive recontouring,
dose was superior in local control compared to stricter constraints of OARs can typically be
the conventional SBRT regimen (3-year local achieved that further mitigate the risk of toxicity.
recurrence 5.8% vs. 22%, p = 0.0048). Depending on anticipated patterns of failure
Provocatively, there was also an improvement in or recurrence based on tumor histology, a CTV
distant progression, with the 3-year distant pro- may be considered to address the risk of micro-
gression being 5.3% vs. 22.5% (p = 0.010). metastatic disease in intervening nodal regions.
However, it is important to recognize the limita- This may be particularly relevant for patients
tions given that most lesions were bone lesions who experience a regional nodal relapse out of
and could not be in close proximity to OARs field of prior nodal irradiation (Fig. 12.3). The
given the CT-based modality of treatment. It is technical capabilities of MRgSBRT to allow for
intriguing that potentially increasing dose could longer areas to be treated while also visualizing
allow for improved oncologic outcomes and nearby OARs have expanded options for patients
advances in MRI-guided technologies could help in whom a protracted course of pelvic nodal irra-
facilitate this dose escalation safely. diation is not clinically or logistically feasible.
Given the limited prospective data to compare the
benefit of comprehensive nodal radiation to
On Table Adaptation SBRT for oligometastatic nodal recurrence, this
should be offered on clinical protocol whenever
Adaptive radiotherapy (ART) allows for adjust- possible.
ment of treatment plans based on anatomical and/ OMD treatment is also unique in that multiple
or biological changes which may occur during lesions tend to be treated contemporaneously for
treatment, such as respiratory movement, changes maximal therapeutic benefit. In determining
in location of bowel loops, and lesion size whether more than one lesion can be treated in
changes. Online plan adaption allows for changes the same course, the ability to track the lesions,
to be made based on the anatomy of the day while movement of the lesions, proximity to OARs, and
the patient is in the treatment position. As a result, placement of the isocenter must be considered. If
the process for online plan adaptation needs to be two or more lesions are adjacent and either are
performed efficiently and fast with priorities of stationary or move together, then creating a sin-
safety (see Chap. 8 for further technical details). gle PTV structure incorporating the targets could
In OMD, there are a number of special cases that be employed. However, many lesions move dif-
highlight the unique challenges and advantages ferentially with relation to each other. Until
of on-table adaptation. recently, technology for adaptive planning was
In OMD treatment, patients may have a his- limited to single planar view for gaiting and
tory of prior radiation therapy to their primary tracking. Dependent on the sites of the oligome-
cancer and/or other metastatic sites necessitating tastases and whether they move differentially
stricter dose cutoffs for OARs. In this setting, with relation to each other, a single plane may not
adaptive planning can be useful to allow for be optimal for tracking of more than one lesion
Ureter_R Bowel
LSPlexusSup
LSPlexusInf
Ring
PTV
b Structure Min (Gy) Mean (Gy) Max (Gy) Dose to Volume

Rx ≥ 95% 18 Gy
PTV_opt
Plan 16.60 21.55 24.52 98.77% 18 Gy
Rx ≥ 99% 16 Gy
PTV
Plan 16.60 21.55 24.52 100% 16 Gy
Rx ≤ 0.03 cc 12.4 Gy
Bowel
Plan 0.01 0.59 11.87 0 cc 12.4 Gy
Rx ≤ 5 cc 11.2 Gy
Bowel
Plan 0.01 0.59 11.87 0.03 cc 11.2 Gy
Rx ≤ 0.03 cc 14 Gy
LSPlexusSup
Plan 0.04 1.23 13.24 0 cc 14 Gy
Rx ≤ 0.35 cc 10 Gy
LSPlexusSup
Plan 0.04 1.23 13.24 0.25 cc 10 Gy
Rx ≤ 0.03 cc 12 Gy
LSPlexusInf
Plan 0.03 0.26 1.76 0 cc 12 Gy
Rx ≤ 5 cc 8 Gy
LSPlexusInf
Plan 0.03 0.26 1.76 0 cc 8 Gy
Rx ≤ 0.03 cc 26 Gy
Skin-Organ
Plan 0 0.06 8.98 0 cc 26 Gy
Rx ≤ 1.2 cc 7 Gy
Cauda Equina
Plan 0.06 0.20 0.90 0 cc 7 Gy
Rx ≤ 10 cc 23 Gy
Skin-Organ
Plan 0 0.06 8.98 0 cc 23 Gy
Rx ≤ 0.03 cc 16 Gy
Bladder
Plan 0.03 0.15 0.64 0 cc 16 Gy
Rx ≤ 15 cc 11.4 Gy
Bladder
Plan 0.03 0.15 0.64 0 cc 11.4 Gy
Rx ≤ 0.03 cc 14 Gy
Rectum
Plan 0.04 0.12 0.30 0 cc 14 Gy
Rx ≤ 20 cc 12 Gy
Rectum
Plan 0.04 0.12 0.30 0 cc 12 Gy
Rx ≤ 10 cc 12 Gy
FemurHead_L
Plan 0.01 0.08 0.25 0 cc 12 Gy
Rx ≤ 10 cc 12 Gy
FemurHead_R
Plan 0.06 0.18 0.40 0 cc 12 Gy
Rx ≤ 0.03 cc 16 Gy
UreterR
Plan 0.50 5.23 15.35 0 cc 16 Gy
Rx ≤ 1.2 cc 7 Gy
LSPlexusSup
Plan 0.04 1.23 13.24 2.82 cc 7 Gy
Rx ≤ 10 cc 9 Gy
Bowel
Plan 0.01 0.59 11.87 0.28 cc 9 Gy
Fig. 12.2 Ultra-hypofractionation for oligometastatic accounting for prior radiation. A 2 cm ring structure was
right pelvic node. (a) Axial and sagittal view of a patient applied to define organs in need of adaptive recontouring
with oligorecurrent prostate cancer with a PSMA-avid on day of treatment. (b) Stricter constraints were applied
right pelvic node treated with a dose of 18 Gy × 1 on to LSPlexInf due to overlap with prior prostate bed/pelvic
SMART-STOP Protocol 19–353. A 3 mm PTV margin EBRT. The adaptive treatment plan met all clinically
was applied to GTV. Organs at risk including the bowel, defined hard constraints. The soft constraint for
right ureter, and a lumbar sacral plexus superior LSPlexusSup was slightly exceeded but clinically
(LSPlexSup) and lumbar sacral plexus inferior (LSPlexInf) acceptable
were contoured to allow for differential constraints,
b
Structure Min (Gy) Mean (Gy) Max (Gy) Dose to Volume
a PTV_35
Rx ≥ 95% 35 Gy
LumboSacral Plexus PTV_35

Plan
Rx
29.90 38.25 43.67 86.74%
≥ 99%
35 Gy
30 Gy
Plan 29.90 38.25 43.67 99.95% 30 Gy
PTV25 Spinal Canal Rx ≤ 0.03 cc 35 Gy
Liver Stomach
Plan 0.29 3.48 21.05 0 cc 35 Gy
Ring PTV35 Stomach
Rx
Plan 0.29 3.48 21.05
≤ 0.5 cc
0 cc
33 Gy
33 Gy
Rx ≤ 0.03 cc 35 Gy
Bowel
Stomach Plan
Rx
0.31 6.70 35.07 0.01 cc
≤ 0.5 cc
35 Gy
33 Gy
Bowel
Plan 0.31 6.70 35.07 0.50 cc 33 Gy
Rx ≤ 0.03 cc 25 Gy
SpinalCanal
Plan 0.52 6.06 15.38 0 cc 25 Gy
Rx ≤ 0.5 cc 23 Gy
SpinalCanal
Plan 0.52 6.06 15.38 0 cc 23 Gy
Bowel KidneyR
Rx
Plan 0.72
≤ 12
4.40 8.77
≤ 33%
0.38%
8 Gy
8 Gy
Rx ≤ 12 ≤ 33% 8 Gy
KidneyL
Plan 0.50 3.95 10.17 4.64% 8 Gy
Rx ≤ 20 ≤ 235.4 cc 15 Gy
zPriorSum_45Gy Liver
Plan
Rx
0.14 0.77 10.48 0 cc
≤ 0.03 cc
15Gy
32 Gy
Skin-Organ
Plan 0.03 0.60 15.14 0 cc 32 Gy
Rx ≤ 0.03 cc 25 Gy
zPriorSum_45Gy LumboSacralPlex
Plan
Rx
9.14 12.11 15.83 0 cc
≤ 0.05 cc
25 Gy
23 Gy
LumboSacralPlex
Plan 9.14 12.11 15.83 0 cc 23 Gy
Rx ≥ 99% 25 Gy
PTV_25
Plan 24.32 28.79 43.67 99.64% 25 Gy
c
Ring
Duodenum d
Bowel
PTV40
PTV25
Fig. 12.3 MRgSBRT to abdominal nodes including a straint of V30 ≥ 99% was achieved. (c) Axial, sagittal, and
CTV. (a) Axial and sagittal view of a patient with vulvar coronal views of a patient with oligorecurrent prostate
cancer with oligoprogression in the para-aortic lymph cancer involving two PSMA-avid lymph nodes above
nodes abutting prior field of pelvic nodal intensity modu- prior field of pelvic nodal IMRT. An SIB was applied to
lated radiation therapy (IMRT). A simultaneous integrated deliver a dose of 40 Gy to the PSMA-avid lymph nodes
boost (SIB) was applied to deliver a dose of 35 Gy to the with a 3 mm margin (PTV40) and a dose of 25 Gy to a
FDG-avid lymph nodes with a 3 mm margin (PTV35) and CTV inclusive of the para-aortic region and involved
a dose of 25 Gy to a CTV inclusive of the para-aortic nodes, cropped from duodenum and bowel anteriorly,
region and involved lymph nodes, cropped from bowel with a 3 mm margin (PTV25). A 3 cm expansion from
and duodenum anteriorly, with a 3 mm margin (PTV25). GTV and 1 cm expansion from CTV were used to gener-
A ring using a 2 cm expansion from PTV35 and 1 cm ate a ring structure for adaptive daily recontouring. OARs
expansion from PTV25 was applied to define the region in included bowel and duodenum. (d) A sum plan account-
which organs at risk (OARs) required adaptive recontouring for all prior radiation treatment plans should be gener-
ing. zPriorSum_45 Gy and zPriorSum_50 Gy were ated when possible to account for cumulative dose to
applied at time of initial contouring to guide target volume OARs. The sum plan graphically displays the dose
delineation and constraints applied to OAR. OARs defined received to patient treated in (c), including prior salvage
included the lumbosacral plexus, spinal canal, liver, stom- prostate bed EBRT (68.4 Gy/28 F, 2014), pelvic nodal
ach, and bowel. (b) The treatment plan was able to meet IMRT (50.4–61.6 Gy/28 F, 2020), and abdominal node
all predefined hard constraints. PTV35 coverage did not MRgSBRT (25–40 Gy/5 F, 2023)
meet the soft constraint of V35 ≥ 95%, but the hard con-
(Fig. 12.4). If determined that sequential treat- gle course (Fig. 12.5). Further, overlapping hot
ment of multiple oligometastases would be pref- spots should be avoided, and plan optimization
erential, then there must be accounting for with beam arrangements should be carefully
cumulative OAR dose as though treated in a sin- evaluated.
a
PTV1 PTV2 PTV1 PTV2
Isodose_10Gy Isodose_10Gy
Ring 4cm Ring 4cm
bLarge Bowel Small Bowel
LumboSacral Plexus
Fig. 12.4 Single isocenter, multi-metastasis treatment on was applied each day with alignment checked relative to
MR-linac. (a) Axial, sagittal, and coronal views of a the vertebral body anatomy to allow for differential dos-
patient with oligoprogressive renal cell cancer involving a ing to the spinal canal and lumbosacral plexus, based on
left para-aortic node (PTV1) and left psoas muscle degree of overlap. (b) The adaptive treatment plan to
(PTV2) following prior SBRT to a left psoas mass. A 4 cm deliver 30 Gy in five fractions for this patient was able to
ring was used to define the organs at risk in need of daily spare dose to the adjacent bowel and lumbosacral plexus
adaptive recontouring due to stricter dose constraints and left kidney
applied in setting of re-irradiation. An “Isodose_10 Gy”
Fig. 12.5 Sequential MRgSBRT for oligometastatic bowel and rectum PRV allowed for further sparing of
recurrence. (a) Axial, sagittal, and coronal view of a treat- OARs. BowelRectum_Superior, which defined the OAR
ment plan for a patient with oligorecurrent ovarian cancer region that overlapped with Isodose_10 Gy, received min-
s/p prior radical resection and chemotherapy with a vagi- imal dose from the vaginal cuff course. (d) Axial view of
nal cuff mass and presacral node involvement. The vagi- the sequential treatment plan used to target two oligopro-
nal cuff mass was defined with a 3 mm margin and treated gressive presacral nodes. The OARs included the lumbo-
to a dose of 35 Gy in five fractions (PTV). Organs at risk sacral plexus, cauda equina, and bowel. The presacral
(OARs) including the bowel, bladder, and rectum were nodes were treated to a dose of 40 Gy in five fractions. (e)
defined. An “Isodose_10 Gy” structure was created to The beam arrangement for the presacral nodes was
allow for differential dose constraints to be generated designed to bisect patient to minimize divergence with
daily for the bowel and rectum, anticipating overlap from prior vaginal cuff treatment plan. (f) The adaptive treat-
sequential treatment to the presacral nodes. (b) Beam ment plan was able to optimize coverage to the two oli-
arrangement to treat the vaginal cuff mass was intention- goprogressive lymph nodes (PTVsup and PTVinf) while
ally designed to bisect patient to minimize divergence meeting the hard constraints for the bowel, lumbosacral
with presacral treatment plan. (c) The treatment plan met plexus, and cauda equina
all hard constraints. PTV coverage was a soft constraint. A
Fig. 12.5 (continued)

Evaluation expansion into investigations of oligometastatic

lesions specifically. In a retrospective review of
Generally, in the treatment of oligometastatic dis- MRgRT for abdominopelvic oligometastases of
ease, OAR constraints are prioritized over PTV 101 patients, online adaptation on a 0.35 T
coverage for safety and to minimize harm to the MR-linac resulted in a clinically significant
patient. In practice, this can be described with the improvement in coverage of the lesion or organ
usage of hard DVH constraints that cannot be sparing in 86.6% of delivered fractions and
breached (where the plan is unacceptable if not showed excellent safety, feasibility, and local
met) as well as priority ranking of DVH con- control [44]. Henke and colleagues published a
straints, with PTV coverage at a lower ranking phase I trial investigating the use of MRgRT on a
compared to essential OAR constraints, such as 0.35 T MR-linac in oligometastatic or unresect-
GI and spinal cord OARs. Soft constraints may able primary disease of the abdomen and found
also be used, where the OAR constraint is a goal that daily online-adaptive plans were feasible and
but may not be met. In cases where daily treat- created for 81 of 97 patients due to initial plan
ments are handled by a treating physician and violation of OAR constraints [59]. All patients in
team for the day, a safe handoff of the targeted the study had at least one treatment that required
lesions and OAR priorities must be carefully adaptive planning. In non-liver lesions, 88% of
done to ensure treatment of the proper oligometa- required plan adaption was due to one or more
static site and OAR constraints are appropriately OAR violations, most commonly of small bowel.
met. Within 6 months of treatment, there were no
In OMD, many patients require multiple grade 3 or greater toxicities. Similarly, a phase I
courses of MDT throughout the lifetime of their study of MRgRT in oligometastatic lesions of ten
disease. Accounting for prior radiation and calcu- ovarian cancer patients was conducted where
lation of a sum plan for future radiation treat- oligometastatic disease was classified as three or
ments is necessary for accounting of dose to less lesions and all radiographic sites were
OARs and to reduce potential toxicities treated, with three lesions in the thorax, six
(Figs. 12.2, 12.3, 12.4, and 12.5). However, cal- lesions in the abdomen, and eight lesions in the
culating dose accumulation for MRgRT is chal- pelvis [62]. This study showed excellent feasibil-
lenging given that there are daily adaptive plans. ity and significantly improved OAR sparing.
To generate an accurate dose, the daily plans Online adaptation was able in 32 of 65 treatments
would need to be exported and calculated for to mitigate unintended OAR constraint viola-
total dose accumulation and should be consid- tions, most commonly in the pelvis followed by
ered in plan evaluation. the abdomen, which was generally well tolerated.
Dose escalation was allowed, and while most
lesions received 35 Gy in five fractions, some
Outcome received doses upward of 50 Gy in five fractions.
Tumor control per lesion at 3 months was 94%
The majority of prospective studies that have and was maintained for 1 year.
been published thus far on the role of MRgRT for Recently the first multicenter experience in
oligometastatic disease have been phase I trials the treatment of oligometastatic disease, with
addressing feasibility; however, long-term impact lymph node and liver lesions, on the Elekta 1.5
on oncologic outcomes and toxicity remain under Tesla MR-linac, was published [63]. In total, 168
investigation. Since the first clinical applications patients with lymph node metastases were treated
of MR-radiation therapy systems, there has been mainly in the pelvis (72%), with the primary can-
cers originating from prostate (60%) or colorec- Future Directions

tal (12%). Between institutions, there was
variation in workflow patterns and differences in The treatment of OMD is continuing to be refined
dose prescription and coverage criteria, but all and phase III clinical trials are ongoing, which
within the range of prescriptions set forth in the will guide recommendations for therapy. These
SABR-COMET trial [7]. Nicosia et al. recently trials will also hopefully evaluate the role of
published the largest prospective study of lymph MDT for different histologies, amount of dis-
node SBRT on a 1.5 Tesla MR-linac in patients ease, and OMD state (Table 12.2). The role of
with prostate cancer [64]. A total of 118 lymph MRgRT for MDT in OMD is promising with the
node oligometastases were treated in 63 oligo- initial results of the phase I and phase II trials
metastatic patients. Of the patients, 63.5% were showing feasibility and safety. The advantages of
oligorecurrent and 36.5% were oligoprogressive. MRgRT with improved soft-tissue visualization
Lymph node SBRT by 1.5 T MR-linac also pro- and the ability to perform real-time tracking and
vides high local control rates with an excellent gating to achieve greater ablative doses while
toxicity profile. Furthermore, the oligorecurrent sparing critical OARs are clear benefits for
state during ADT was correlated with a lower patients receiving MDT in the proper clinical
nodal progression-free survival than was the oli- settings.
goprogressive state. The same group have previ- Improvements in the technology continue to
ously reported preliminary data on feasibility and be developed, with most recently the release of
patient-reported outcomes following PSMA- MRIdian A3i, which allows for multi-planar
PET/CT guided SBRT for oligometastatic tracking. This could potentially allow for ease of
castration-sensitive prostate cancer by means of treatment of multiple oligometastatic sites in the
1.5 T MR-Linac [65]. In this series, 25 lesions in same radiation course. Further technological
20 castration-sensitive oligorecurrent patients advances will also likely improve clinical work-
were treated: The most commonly treated ana- flows and time on table with improved efficiency.
tomic sites were nodal (n = 16) and pelvic bone The role of artificial intelligence to automate cur-
(n = 9), and encouraging findings in terms of rent manual operations efficiently is being
effectiveness and tolerability were confirmed explored [66].
[65]. Utilizing MRgRT for the treatment of OMD is
These studies demonstrate the feasibility and becoming increasingly recognized with the
safety of the MR-guided MDT approaches for promising results in these initial phase I and
oligometastases. Nevertheless, more data is phase II trials. Currently, trials for the role of
needed for understanding longer-term oncologic MRgRT in OMD are ongoing, and their results
outcomes, optimization of processes, optimal will provide necessary information regarding
dose and prescription, long-term toxicities, and whether there is a benefit to MRgRT in the OMD
patient-reported measures. setting (Table 12.3).
12
Table 12.2 Selected phase II and III ongoing clinical trials investigating metastasis-directed therapy in oligometastatic cancer
Estimated
MDT study
Trial name/ Target ablative Primary completion
identifier Histology Description Key inclusion criteria enrollment treatment endpoint date
SABR- Multiple Phase III multicenter trial (randomized Controlled primary site, one to 330 SBRT OS 2028
COMET-3 1:2) comparing SOC vs. SOC + SBRT three metastases with all
NCT03862911 metastases amenable to SBRT
[67]
SABR- Multiple Phase III multicenter trial (1:2 ratio) Controlled primary site, four to ten 204 SBRT OS 2029
COMET-10 comparing SOC vs. SOC + SBRT metastases with all metastases
NCT03721341 amenable to SBRT
[68]
EORTC 1945 Multiple (exclude Phase III multicenter trial comparing Controlled primary site, one to five 200 SBRT OS 2030
OligoRARE lung, breast, SOC + palliative RT vs. SOC + SBRT metastases
NCT04498767 colorectal, prostate)
CORE Breast, prostate, Phase II/III multicenter trial comparing Controlled primary site, one to 245 SBRT PFS 2024
NCT02759783 NSCLC SOC vs. SOC + SBRT three extracranial metachronous
metastases with all amenable to
SBRT and limited to two different
organ systems; no prior systemic
therapy for metastatic disease
PLATON Prostate Phase III multicenter trial comparing Castration-sensitive with primary 410 SBRT FFS 2025
NCT03784755 SOC ST vs. SOC ST + SBRT site controlled or amenable to
definitive or RT or surgery, one to
five synchronous or metachronous
metastases (detected on
conventional imaging or PET)
with all amenable to SBRT
START-MET Prostate Phase III trial comparing SOC Castration-sensitive, one to three 266 SBRT rPFS 2027
NCT05209243 (ADT + RT to primary tumor if not metastases (detected by
previously treated) + second generation conventional imaging) or one to
hormonal agent vs. SOC + SBRT five metastases (detected by PET)
(continued)
235
Table 12.2 (continued)
236
Estimated
MDT study
PRESTO Prostate Phase III multicenter trial comparing Castration-sensitive, primary site 350 SBRT CRPC-free 2028
NCT003678025 SOC vs. SOC + SBRT controlled or amenable to survival
definitive RT, one to five
synchronous or metachronous
metastases (detected on PET) with
all amenable to SBRT
METRO Prostate Phase III multicenter trial (1:2 ratio) Castration-sensitive, primary site 114 SBRT FFS (time 2029
NCT04983095 comparing SOC vs. SOC + SBRT controlled or amenable to to CRPC)
definitive RT, one to three
synchronous or metachronous
metastases (detected on PSMA-
PET) with all amenable to SBRT
ARTO Prostate Randomized phase II trial comparing CRPC; one to three metastases 174 SBRT PSA 2022
NCT03449719 abiraterone vs. abiraterone + SBRT response
PILLAR Prostate Randomized phase II trial comparing CRPC with no prior ST for CRPC; 60 SBRT PSA 2024
NCT03503344 apalutamide + SBRT vs. SBRT one to five metastases (detected by response
PSMA-PET)
PCS IX Prostate Randomized phase II/III trial comparing CRPC; one to five metastases 130 SBRT rPFS 2025
NCT02685397 LNRH agonist + enzalutamide vs. LNRH (detected by CT, bone scan, and/or
agonist + enzalutamide + SBRT MRI)
OMEGA NSCLC Phase III multicenter trial comparing Primary site controlled or suitable 195 Surgery or OS 2022
NCT03827577 SOC vs. SOC + LAT for definitive treatment, one to SBRT or
three metastases with all amenable RFA
to SBRT
NRG LU-002 NSCLC Phase II/III randomized multicenter trial One to three extracranial 400 SBRT Phase II: 2031a
NCT03137771 comparing chemotherapy vs. synchronous or metachronous PFS
chemotherapy + SBRT metastases with no progression Phase III:
after first line systemic therapy OS
with all amenable to SBRT; if
synchronous, primary site
amenable to SBRT or
hypofractionated RT
K. N. Lee et al.
12
Estimated
MDT study
TARGET-02 NSCLC with Phase III trial comparing SMT vs. One to five metastases with no 190 SBRT OS 2028
NCT05278052 negative oncogene SMT + LCRT (SBRT + hypofractionated progression following four to six
[69] mutation status RT to primary/nodal) cycles of ST
ECOG-ACRIN Esophageal and Phase III multicenter trial comparing One to three synchronous or 314 RT OS 2023
EA2183 gastric FOLFOX vs. FOLFOX+RT vs. CAPOX metachronous metastases all
NCT04248452 adenocarcinoma vs. CAPOX+RT amenable to RT; no evidence of
[70] (HER2 negative) disease progression following
induction CT
NCT04944914 Nasopharyngeal Phase III trial comparing camrelizumab Controlled primary site, one to five 188 SBRT PFS 2026
carcinoma vs. camrelizumab + SBRT metastases after first line
chemotherapy all amenable to
SBRT; exclude patients with
immunotherapy failure
NCT02956798 RCC Phase II single-arm trial investigating Systemic treatment naïve with one 23 SBRT FFST (time 2023
SBRT to all metastases to three extracranial metastases to initiation
of systemic
therapy)
GETUG-StORM Clear cell renal Randomized phase II trial (2:1 ratio) One to three oligoprogressive 114 SBRT PFS 2023
01 cancer comparing ST vs. ST + SBRT metastases all amenable to SBRT
NCT04299646
Abbreviations: CAPOX capecitabine and oxaliplatin, CRPC castration-resistant prostate cancer, FFS failure-free survival, FOLFOX fluorouracil and leucovorin calcium and
oxaliplatin, LAT local ablative therapy, LCRT local consolidative radiotherapy, OS overall survival, PFS progression-free survival, PSA prostate specific antigen, RCC renal cell
carcinoma, rPFS radiological progression-free survival, RT radiation therapy, SBRT stereotactic body radiotherapy, SOC standard of care, ST systemic therapy, SMT standard
maintenance therapy
a
Recruitment completed phase II and now paused for scheduled interim analysis to determine if PFS benefit before reopening for phase III recruitment
237
Table 12.3 Selected ongoing clinical trials investigating treatment of MR-guided radiation therapy in oligometastatic
cancer
Estimated
MDT study
Trial name/ Included Target ablative completion
identifier condition(s) Description enrollment treatment Primary endpoint date
SMART Oligoprogressive Phase I/II trial 1000 MRIdian Phase I: Delivery 2024
NCT04115254 nodal metastases investigating (master success rate,
Metachronous nodal feasibility and protocol) tumor
metastases efficacy of visualization, plan
Synchronous nodal SMART in creation
metastases multiple cancer Phase II: Rate of
types improvement in
tumor control
SMART ONE Inoperable primary Phase I multisite 30 MRIdian Number of 2024
NCT04939246 or metastatic trial investigating successfully
carcinoma feasibility and delivered single
safety of one fraction
fraction SMART treatments;
tolerability (no
greater than 4 of
30 patients with
grade3+ toxicity
within 90 days, no
grade 5 toxicity)
SOFT Metastatic cancer Phase II 121 MRIdian Cumulative 2023
NCT04407897 with 1-5 multicenter trial toxicity grade 4 or
synchronous or investigating greater
metachronous safety of
metastates or 1-3 MR-guided
oligoprogressive or infra-
induced metastases diaphragmatic
oligometastatic
disease
MOMENTUM Oligometastatic Multi-institutional 6000 Unity PFS, OS, DFS, 2024
NCT04075305 cancer observational MR-linac patient QoL,
[71] cohort study of toxicity, clinical
multiple cancer tumor response,
types pathologic tumor
response
NCT04172753 Oligometastatic Feasibility study 472 Unity Feasibility of 2027
[72] cancer investigating MR-linac online MR-guided
imaging and RT
treatment on 1.5 T
MR-linac of
multiple cancer
types
Abbreviations: MRgRT magnetic resonance (MR) guided radiation therapy, QoL quality of life, SMART stereotactic
magnetic resonance guided adaptive radiotherapy
References PSMA PET diagnostic imaging for prostate cancer. J

Nucl Med. 2021;62(5):596–604. https://pubmed.ncbi.
nlm.nih.gov/33712536/
1. Hellman S, Weichselbaum RR. Oligometastases. J
13. Huynh MA, Tang C, Siva S, Berlin A, Hannan R,
Clin Oncol. 1995;13(1):8–10. https://pubmed.ncbi.
Warner A, et al. Review of prospective trials assess-
ing the role of stereotactic body radiation therapy for
2. Weichselbaum RR, Hellman S. Oligometastases
metastasis-directed treatment in oligometastatic geni-
revisited. Nat Rev Clin Oncol. 2011;8(6):378–82.
tourinary cancers. Eur. Urol Oncol. 2023;6(1):28–38.
https://pubmed.ncbi.nlm.nih.gov/21423255/
3. Ganesh K, Massagué J. Targeting metastatic cancer.
14. Pacifico P, Colciago RR, de Felice F, Boldrini L,
Nat Med. 2021;27(1):34–44. https://pubmed.ncbi.
Salvestrini V, Nardone V, et al. A critical review on
oligometastatic disease: a radiation oncologist’s
4. Reyes DK, Pienta KJ. The biology and treatment of
perspective. Med Oncol. 2022;39(12):181. https://
oligometastatic cancer. Oncotarget. 2015;6(11):8491–
pubmed.ncbi.nlm.nih.gov/36071292/
524. https://pubmed.ncbi.nlm.nih.gov/25940699/
15. Chetty IJ, Doemer AJ, Dolan JL, Kim JP, Cunningham
5. Liu W, Bahig H, Palma DA. Oligometastases: emerg-
JM, Dragovic J, et al. MRI-guided radiotherapy
ing evidence. J Clin Oncol. 2022;40(36):4250–60.
(MRgRT) for treatment of oligometastases: review
of clinical applications and challenges. Int J Radiat
6. Ost P, Reynders D, Decaestecker K, Fonteyne V,
Oncol Biol Phys. 2022;114(5):950–67. https://
Lumen N, DeBruycker A, et al. Surveillance or
pubmed.ncbi.nlm.nih.gov/35901978/
metastasis-directed therapy for oligometastatic
16. Fleckenstein J, Petroff A, Schäfers HJ, Wehler T,
prostate cancer recurrence: a prospective, ran-
Schöpe J, Rübe C. Long-term outcomes in radi-
domized, multicenter phase II trial. J Clin Oncol.
cally treated synchronous vs. metachronous oligo-
2018;36(5):446–53. https://pubmed.ncbi.nlm.nih.
metastatic non-small-cell lung cancer. BMC Cancer.
gov/29240541/
2016;16(1):348. https://pubmed.ncbi.nlm.nih.
7. Palma DA, Olson R, Harrow S, Gaede S, Louie AV,
gov/27255302/
Haasbeek C, et al. Stereotactic ablative radiotherapy for
17. Milano MT, Katz AW, Zhang H, Okunieff
the comprehensive treatment of oligometastatic can-
P. Oligometastases treated with stereotactic body
cers: long-term results of the SABR-COMET phase II
radiotherapy: long-term follow-up of prospective
randomized trial. J Clin Oncol. 2020;38(25):2830–8.
study. Int J Radiat Oncol Biol Phys. 2012;83(3):878–
8. Wang XS, Bai YF, Verma V, Yu RL, Tian W, Ao R,
86. https://pubmed.ncbi.nlm.nih.gov/22172903/
et al. Randomized trial of first-line tyrosine kinase
18. Bezjak A, Paulus R, Gaspar LE, Timmerman RD,
inhibitor with or without radiotherapy for synchro-
Straube WL, Ryan WF, et al. Safety and efficacy of
nous oligometastatic EGFR-mutated NSCLC. J Natl
a five-fraction stereotactic body radiotherapy sched-
Cancer Inst. 2023;115:742–8. https://pubmed.ncbi.
ule for centrally located non-small-cell lung can-
cer: NRG oncology/RTOG 0813 trial. J Clin Oncol.
9. Lievens Y, Guckenberger M, Gomez D, Hoyer M,
Iyengar P, Kindts I, et al. Defining oligometastatic
gov/30943123/
disease from a radiation oncology perspective: an
19. Romesser PB, Tyagi N, Crane CH. Magnetic resonance
ESTRO-ASTRO consensus document. Radiother
imaging-guided adaptive radiotherapy for colorectal
Oncol. 2020;148:157–66. https://pubmed.ncbi.nlm.
liver metastases. Cancers (Basel). 2021;13(7):1636.
nih.gov/32388150/
10. Guckenberger M, Lievens Y, Bouma AB, Collette
20. Rosenberg SA, Henke LE, Shaverdian N, Mittauer
L, Dekker A, deSouza NM, et al. Characterisation
K, Wojcieszynski AP, Hullett CR, et al. A multi-
and classification of oligometastatic disease: a
institutional experience of MR-guided liver stereo-
European Society for Radiotherapy and Oncology and
tactic body radiation therapy. Adv Radiat Oncol.
European Organisation for Research and Treatment
of Cancer consensus recommendation. Lancet Oncol.
gov/30706022/
2020;21(1):e18–28. https://pubmed.ncbi.nlm.nih.
21. Borghesi S, Casamassima F, Aristei C, Grandinetti
gov/31908301/
A, di Franco R. Stereotactic radiotherapy for adre-
11. Thomas MC, Chen YH, Fite E, Pangilinan A, Bubelo
nal oligometastases. Rep Pract Oncol Radiother.
K, Spektor A, et al. Patient and treatment factors
associated with improved local control and survival
gov/35402020/
in oligometastatic bone disease: results from a large
22. de Ruysscher D, Wanders R, Hendriks LE,
single-institution experience using stereotactic body
van Baardwijk A, Reymen B, Houben R, et al.
radiation therapy. Int J Radiat Oncol Biol Phys.
Progression-free survival and overall survival beyond
5 years of NSCLC patients with synchronous oligo-
gov/35840113/
metastases treated in a prospective phase II trial (NCT
12. Farolfi A, Calderoni L, Mattana F, Mei R, Telo S, Fanti
01282450). J Thorac Oncol. 2018;13(12):1958–61.
S, et al. Current and emerging clinical applications of
23. Chmura S, Winter KA, Robinson C, Pisansky TM, 2021;80(6):693–700. https://pubmed.ncbi.nlm.nih.

Borges V, Al-Hallaq H, et al. Evaluation of safety gov/34399998/
of stereotactic body radiotherapy for the treat- 33. Hannan R, Christensen M, Hammers H, Christie A,
ment of patients with multiple metastases: findings Paulman B, Lin D, et al. Phase II trial of stereotac-
from the NRG-BR001 phase 1 trial. JAMA Oncol. tic ablative radiation for oligoprogressive metastatic
2021;7(6):845–52. https://pubmed.ncbi.nlm.nih. kidney cancer. Eur Urol Oncol. 2022;5(2):216–24.
gov/33885704/ https://pubmed.ncbi.nlm.nih.gov/34986993/
24. Jereczek-Fossa BA, Ronchi S, Orecchia R. Is stereo- 34. Yamamoto T, Niibe Y, Aoki M, Shintani T, Yamada
tactic body radiotherapy (SBRT) in lymph node oligo- K, Kobayashi M, et al. Analyses of the local control
metastatic patients feasible and effective? Rep Pract of pulmonary oligometastases after stereotactic body
Oncol Radiother. 2015;20(6):472–83. https://pubmed. radiotherapy and the impact of local control on sur-
ncbi.nlm.nih.gov/26696788/ vival. BMC Cancer. 2020;20(1):997. https://pubmed.
25. Palma DA, Olson R, Harrow S, Gaede S, Louie AV, ncbi.nlm.nih.gov/33054721/
Haasbeek C, et al. Stereotactic ablative radiotherapy 35. Fode MM, Høyer M. Survival and prognostic fac-
versus standard of care palliative treatment in patients tors in 321 patients treated with stereotactic body
with oligometastatic cancers (SABR-COMET): radiotherapy for oligo-metastases. Radiother Oncol.
a randomised, phase 2, open-label trial. Lancet. 2015;114(2):155–60. https://pubmed.ncbi.nlm.nih.
2019;393(10185):2051–8. https://pubmed.ncbi.nlm. gov/25583567/
nih.gov/30982687/ 36. Chmura SJ, Winter KA, Woodward WA, Borges VF,
26. Gomez DR, Tang C, Zhang J, Blumenschein GR, Salama JK, Al-Hallaq HA, et al. NRG-BR002: A
Hernandez M, Jack Lee J, et al. Local consolidative phase IIR/III trial of standard of care systemic ther-
therapy vs. maintenance therapy or observation for apy with or without stereotactic body radiotherapy
patients with oligometastatic non-small-cell lung can- (SBRT) and/or surgical resection (SR) for newly
cer: long-term results of a multi-institutional, phase II, oligometastatic breast cancer (NCT02364557). J
randomized study. J Clin Oncol. 2019;37(18):1558– Clin Oncol. 2022;40(16_suppl):1007. https://doi.
65. https://pubmed.ncbi.nlm.nih.gov/31067138/ org/10.1200/JCO20224016_suppl1007.
27. Wang XS, Bai YF, Verma V, Yu RL, Tian W, Ao R, 37. Tsai CJ, Yang JT, Guttmann DM, Shaverdian N,
et al. Randomized trial of first-line tyrosine kinase Shepherd AF, Eng J, et al. Consolidative use of radio-
inhibitor with or without radiotherapy for synchro- therapy to block (CURB) oligoprogression ― interim
nous oligometastatic EGFR-mutated non-small cell analysis of the first randomized study of stereotactic
lung cancer. J Natl Cancer Inst. 2023;115(6):742–8. body radiotherapy in patients with oligoprogressive
https://doi.org/10.1093/jnci/djac015/6517507. metastatic cancers of the lung and breast. Int J Radiat
28. Phillips R, Shi WY, Deek M, Radwan N, Lim SJ, Oncol Biol Phys. 2021;111(5):1325–6.
Antonarakis ES, et al. Outcomes of observation vs 38. Poon I, Erler D, Dagan R, Redmond KJ, Foote M,
stereotactic ablative radiation for oligometastatic Badellino S, et al. Evaluation of definitive stereotac-
prostate cancer: the ORIOLE phase 2 randomized tic body radiotherapy and outcomes in adults with
clinical trial. JAMA Oncologia. 2020;6(5):650. /pmc/ extracranial oligometastasis. JAMA Netw Open.
articles/PMC7225913/ 2020;3(11):e2026312. https://jamanetwork.com/
29. Bianchi M, Sun M, Jeldres C, Shariat SF, Trinh QD, journals/jamanetworkopen/fullarticle/2772997
Briganti A, et al. Distribution of metastatic sites in 39. Timmerman RD, Kavanagh BD, Cho LC, Papiez L,
renal cell carcinoma: a population-based analysis. Xing L. Stereotactic body radiation therapy in mul-
Ann Oncol. 2012;23(4):973–80. https://pubmed.ncbi. tiple organ sites. J Clin Oncol. 2007;25(8):947–52.
nlm.nih.gov/21890909/ https://pubmed.ncbi.nlm.nih.gov/17350943/
30. Hannan R, Christensen M, Christie A, Garant A, 40. Kord M, Kluge A, Kufeld M, Kalinauskaite G, Loebel
Pedrosa I, Robles L, et al. Stereotactic ablative radia- F, Stromberger C, et al. Risks and benefits of fiducial
tion for systemic therapy-naïve oligometastatic kid- marker placement in tumor lesions for robotic radio-
ney cancer. Eur Urol Oncol. 2022;5(6):695–703. surgery: technical outcomes of 357 implantations.
https://pubmed.ncbi.nlm.nih.gov/35985982/ Cancers (Basel). 2021;13(19):4838. /pmc/articles/
31. Tang C, Msaouel P, Hara K, Choi H, Le V, Shah AY, PMC8508340/
et al. Definitive radiotherapy in lieu of systemic ther- 41. Devic S. MRI simulation for radiotherapy treatment
apy for oligometastatic renal cell carcinoma: a single- planning. Med Phys. 2012;39(11):6701–11. https://
arm, single-centre, feasibility, phase 2 trial. Lancet pubmed.ncbi.nlm.nih.gov/23127064/
Oncol. 2021;22(12):1732–9. https://pubmed.ncbi. 42. Cuccia F, Rigo M, Gurrera D, Nicosia L, Mazzola
nlm.nih.gov/34717797/ R, Figlia V, et al. Mitigation on bowel loops daily
32. Cheung P, Patel S, North SA, Sahgal A, Chu W, variations by 1.5-T MR-guided daily-adaptive SBRT
Soliman H, et al. Stereotactic radiotherapy for for abdomino-pelvic lymph-nodal oligometastases.
oligoprogression in metastatic renal cell cancer J Cancer Res Clin Oncol. 2021;147(11):3269–77.
patients receiving tyrosine kinase inhibitor therapy: https://pubmed.ncbi.nlm.nih.gov/34268583/
a phase 2 prospective multicenter study. Eur Urol. 43. Ma X, Chen X, Wang Y, Qin S, Yan X, Cao Y, et al.
Personalized modeling to improve pseudo-computed
tomography images for magnetic resonance imaging- 55. Tao R, Krishnan S, Bhosale PR, Javle MM, Aloia
guided adaptive radiation therapy. Int J Radiat Oncol TA, Shroff RT, et al. Ablative radiotherapy doses lead
Biol Phys. 2022;113(4):885–92. https://pubmed.ncbi. to a substantial prolongation of survival in patients
nlm.nih.gov/35462026/ with inoperable intrahepatic cholangiocarcinoma: a
44. Yang DD, Brennan VS, Huynh E, Williams CL, Han retrospective dose response analysis. J Clin Oncol.
Z, Ampofo N, et al. Stereotactic magnetic resonance- 2016;34(3):219–26. https://pubmed.ncbi.nlm.nih.
guided adaptive radiation therapy (SMART) for gov/26503201/
abdominopelvic oligometastases. Int J Radiat Oncol 56. Henke L, Kashani R, Yang D, Zhao T, Green O,
Biol Phys. 2022;114(5):941–9. https://pubmed.ncbi. Olsen L, et al. Simulated online adaptive magnetic
nlm.nih.gov/35598799/ resonance–guided stereotactic body radiation ther-
45. Cox BW, Spratt DE, Lovelock M, Bilsky MH, Lis E, apy for the treatment of oligometastatic disease of
Ryu S, et al. International spine radiosurgery consor- the abdomen and central thorax: characterization of
tium consensus guidelines for target volume definition potential advantages. Int J Radiat Oncol Biol Phys.
in spinal stereotactic radiosurgery. Int J Radiat Oncol 2016;96(5):1078–86. http://www.redjournal.org/
Biol Phys. 2012;83(5):e597–605. https://pubmed. article/S0360301616331510/fulltext
ncbi.nlm.nih.gov/22608954/ 57. Hawranko R, Sohn JJ, Neiderer K, Bump E, Harris T,
46. Olberg S, Green O, Cai B, Yang D, Rodriguez V, Fields EC, et al. Investigation of isotoxic dose escala-
Zhang H, et al. Optimization of treatment planning tion and plan quality with TDABC analysis on a 0.35
workflow and tumor coverage during daily adaptive T MR-Linac (MRL) system in ablative 5-fraction ste-
magnetic resonance image guided radiation therapy reotactic magnetic resonance-guided radiation ther-
(MR-IGRT) of pancreatic cancer. Radiat Oncol. apy (MRgRT) for primary pancreatic cancer. J Clin
2018;13(1):1–8. /pmc/articles/PMC5866525/ Med. 2022;11(9):2584. https://pubmed.ncbi.nlm.nih.
47. Gerhard SG, Palma DA, Arifin AJ, Louie AV, Li GJ, gov/35566712/
Al-Shafa F, et al. Organ at risk dose constraints in 58. Parikh PJ, Lee P, Low D, Kim J, Mittauer KE, Bassetti
SABR: a systematic review of active clinical trials. MF, et al. Stereotactic MR-guided on-table adaptive
Pract Radiat Oncol. 2021;11(4):e355–65. radiation therapy (SMART) for patients with border-
48. Diez P, Hanna GG, Aitken KL, van As N, Carver A, line or locally advanced pancreatic cancer: primary
Colaco RJ, et al. UK 2022 consensus on normal tis- endpoint outcomes of a prospective phase II multi-
sue dose-volume constraints for oligometastatic, pri- center international trial. Int J Radiat Oncol Biol
mary lung and hepatocellular carcinoma stereotactic Phys. 2022;114(5):1062–3. http://www.redjournal.
ablative radiotherapy. Clin Oncol. 2022;34(5):288– org/article/S0360301622032680/fulltext
300. http://www.clinicaloncologyonline.net/article/ 59. Henke L, Kashani R, Robinson C, Curcuru A,
S0936655522000942/fulltext DeWees T, Bradley J, et al. Phase I trial of stereo-
49. Hanna GG, Murray L, Patel R, Jain S, Aitken KL, tactic MR-guided online adaptive radiation therapy
Franks KN, et al. UK consensus on normal tissue dose (SMART) for the treatment of oligometastatic or
constraints for stereotactic radiotherapy. Clin Oncol unresectable primary malignancies of the abdo-
(R Coll Radiol). 2018;30(1):5–14. https://pubmed. men. Radiother Oncol. 2018;126(3):519–26. https://
ncbi.nlm.nih.gov/29033164/ pubmed.ncbi.nlm.nih.gov/29277446/
50. Benedict SH, Yenice KM, Followill D, Galvin JM, 60. Bodo S, Campagne C, Thin TH, Higginson DS, Vargas
Hinson W, Kavanagh B, et al. Stereotactic body HA, Hua G, et al. Single-dose radiotherapy disables
radiation therapy: the report of AAPM task group tumor cell homologous recombination via ischemia/
101. Med Phys. 2010;37(8):4078–101. https://doi. reperfusion injury. J Clin Invest. 2019;129(2):786–
org/10.1118/1.3438081. 801. https://pubmed.ncbi.nlm.nih.gov/30480549/
51. Timmerman RD. An overview of Hypofractionation 61. Zelefsky MJ, Yamada Y, Greco C, Lis E, Schöder H,
and introduction to this issue of seminars in radiation Lobaugh S, et al. Phase 3 multi-center, prospective,
oncology. Semin Radiat Oncol. 2008;18(4):215–22. randomized trial comparing single-dose 24 Gy radia-
52. Kim DWN, Medin PM, Timmerman RD. Emphasis tion therapy to a 3-fraction SBRT regimen in the treat-
on repair, not just avoidance of injury, facilitates pru- ment of Oligometastatic cancer. Int J Radiat Oncol
dent stereotactic ablative radiotherapy. Semin Radiat Biol Phys. 2021;110(3):672–9. https://pubmed.ncbi.
Oncol. 2017;27(4):378–92. nlm.nih.gov/33422612/
53. Christensen M, Hannan R. The emerging role of 62. Henke LE, Stanley JA, Robinson C, Srivastava A,
radiation therapy in renal cell carcinoma. Cancers Contreras JA, Curcuru A, et al. Phase I trial of ste-
(Basel). 2022;14(19):4693. https://pubmed.ncbi.nlm. reotactic MRI-guided online adaptive radiation
nih.gov/36230615/ therapy (SMART) for the treatment of oligometa-
54. Wang CJ, Christie A, Lin MH, Jung M, Weix D, static ovarian cancer. Int J Radiat Oncol Biol Phys.
Huelsmann L, et al. Safety and efficacy of stereotac- 2022;112(2):379–89. https://pubmed.ncbi.nlm.nih.
tic ablative radiation therapy for renal cell carcinoma gov/34474109/
extracranial metastases. Int J Radiat Oncol Biol Phys. 63. Janssen TM, Aitken K, Alongi F, Barry A, Bernchou
2017;98(1):91. /pmc/articles/PMC5555369/ U, Boeke S, et al. First multicentre experience of
SABR for lymph node and liver oligometastatic
disease on the unity MR-Linac. Tech Innov Patient study protocol for a randomized phase III trial. BMC
Support Radiat Oncol. 2022;22:50–4. https://pubmed. Cancer. 2019;19(1):1–15. https://doi.org/10.1186/
ncbi.nlm.nih.gov/35586786/ s12885-019-5977-6.
64. Nicosia L, Trapani G, Rigo M, Giaj-Levra N, Mazzola 69. Tibdewal A, Agarwal JP, Srinivasan S, Mummudi N,
R, Pastorello E, et al. 1.5 T MR-guided daily adapted Noronha V, Prabhash K, et al. Standard maintenance
SBRT on lymph node oligometastases from pros- therapy versus local consolidative radiation therapy
tate cancer. J Clin Med. 2022;11(22):6658. https:// and standard maintenance therapy in 1-5 sites of
pubmed.ncbi.nlm.nih.gov/36431135/ oligometastatic non-small cell lung cancer: a study
65. Mazzola R, Cuccia F, Figlia V, Rigo M, Nicosia L, protocol of phase III randomised controlled trial. BMJ
Giaj-Levra N, et al. Stereotactic body radiotherapy Open. 2021;11(3):e043628. https://pubmed.ncbi.nlm.
for oligometastatic castration sensitive prostate cancer nih.gov/33727268/
using 1.5 T MRI-Linac: preliminary data on feasibil- 70. Uboha NV, Graham N, Rajdev L, Gibson MK,
ity and acute patient-reported outcomes. Radiol Med. Fisher GA, Lin SH, et al. EA2183: A phase III
2021;126(7):989–97. https://pubmed.ncbi.nlm.nih. study of consolidative radiotherapy in patients with
gov/33835309/ oligometastatic HER2-negative esophageal and
66. Cusumano D, Boldrini L, Dhont J, Fiorino C, Green gastric adenocarcinoma. J Clin Oncol. 2022;40(16_
O, Güngör G, et al. Artificial intelligence in magnetic suppl):TPS4162. https://doi.org/10.1200/
resonance guided radiotherapy: medical and physical JCO20224016_supplTPS4162.
considerations on state of art and future perspectives. 71. de Mol van Otterloo SR, Christodouleas JP, Blezer
Phys Med. 2021;85:175–91. https://pubmed.ncbi. ELA, Akhiat H, Brown K, Choudhury A, et al. The
nlm.nih.gov/34022660/ MOMENTUM study: an international registry for the
67. Olson R, Mathews L, Liu M, Schellenberg D, Mou B, evidence-based introduction of MR-guided adaptive
Berrang T, et al. Stereotactic ablative radiotherapy for therapy. Front Oncologia. 2020;10:1328.
the comprehensive treatment of 1-3 oligometastatic 72. Wegener D, Thome A, Paulsen F, Gani C, Boldt
tumors (SABR-COMET-3): study protocol for a ran- J, Butzer S, et al. First experience and prospective
domized phase III trial. BMC Cancer. 2020;20(1):1– evaluation on feasibility and acute toxicity of online
12. https://doi.org/10.1186/s12885-020-06876-4. adaptive radiotherapy of the prostate bed as salvage
68. Palma DA, Olson R, Harrow S, Correa RJM, treatment in patients with biochemically recurrent
Schneiders F, Haasbeek CJA, et al. Stereotactic abla- prostate cancer on a 1.5T MR-Linac. J Clin Med.
tive radiotherapy for the comprehensive treatment of 2022;11(16):4651. https://pubmed.ncbi.nlm.nih.
4-10 oligometastatic tumors (SABR-COMET-10): gov/36012885/
Management of Pancreatic Cancer
with MRI-Guided Adaptive
13
Radiotherapy
Ritchell van Dams, Sylvia S. W. Ng, John Hayes,

and Tarita O. Thomas
Introduction and Disease Overview denum to the tail that typically extends toward
the splenic hilum (Fig. 13.1).
The pancreas is a retroperitoneal organ located in More than 90% of pancreas cancer derives
the upper, central, posterior abdomen between from ductal epithelium resulting in ductal adeno-
vertebral levels L1–L2 whose function is com- carcinoma. Globally, pancreatic ductal adenocar-
plex with both endocrine and exocrine function. cinoma is the 14th most common cancer type but
It resides in close proximity to multiple adjacent given its high mortality rate represents the sev-
abdominal organs such as the duodenum as well enth most common cause of cancer-related death
as blood vessels and the celiac nerve plexus. [1]. Pancreatic cancer has one of the lowest
Critical vasculature that extends through this area 5-year survival rates of all cancers at 11% and
includes the celiac artery and its main branches even among patients with localized disease
(left gastric, splenic, and hepatic arteries), the 5-year survival is only 42% [2].
superior mesenteric artery, and the superior mes- Early diagnosis is uncommon due to the lack
enteric and portal veins. Anatomically designated of symptoms and signs of earlier stage disease as
as having four sections—head, neck, body, and symptoms are more common when there is local
tail—the gland is functionally continuous from progression of tumor causing obstructive symp-
its head that abuts the second portion of the duo- toms. Pancreatic ductal carcinomas are also rec-
ognized as having a high capacity for early
systemic dissemination, so prognosis is guarded
even for those patients diagnosed with seemingly
R. van Dams (*)
Department of Radiation Oncology, Dana-Farber localized disease [3]. General consensus is there-
Brigham Cancer Center, Harvard Medical School, fore that adenocarcinoma of the pancreas should
Boston, MA, USA be treated as a systemic disease even when appar-
ently localized [4]. Yet appreciating the natural
S. S. W. Ng history of the disease necessitates the recognition
Department of Radiation Oncology, Sunnybrook
that about one in three patients dies due to pro-
Odette Cancer Centre, University of Toronto,
Toronto, ON, Canada gression at the primary site of disease [5].
e-mail: [email protected] Optimal treatment strategies for patients are thus
J. Hayes · T. O. Thomas dependent on harnessing the benefits of surgery,
Department of Radiation Oncology, Northwestern systemic therapy, and radiation therapy at the
Memorial Hospital, Northwestern University appropriate time with a coordinated approach.
e-mail: [email protected]; [email protected]
244 R. van Dams et al.
Fig. 13.1 Anatomy of the pancreas and surrounding vasculature
Although the American Joint Committee on tumors typically demonstrate either encasement
Cancer (AJCC) has developed a traditional (contact ≥180°) of the CA or SMA or have un-
tumor, nodal, and metastasis (TNM) staging sys- reconstructable involvement of the SMV or
tem for pancreatic cancer, patients are generally PV. Involvement of the first jejunal branches of
stratified and treated according to a surgical the SMA or SMV, nonregional lymph node
classification system that establishes whether a involvement, and distant metastases also place a
patient has resectable, borderline resectable, or tumor in the unresectable category [6].
unresectable disease [6]. These differences are
critical for appropriate clinical decisions.
Resectable tumors are generally lesions with- Treatment Background
out metastases that do not contact major arteries
including the celiac artery (CA), superior mesen- Patients with resectable pancreatic cancer are
teric artery (SMA), or common hepatic artery treated with upfront resection as surgery offers
(CHA). They also have no or minimal contact the only potentially curative treatment strategy
with the superior mesenteric vein (SMV) or por- for this cancer [7]. Early trials of adjuvant radia-
tal vein (PV) and do not deform the normal tion therapy (RT) for resected pancreatic cancer
course of the vein. Pancreatic tail lesions may showed mixed findings. These trials, including
involve the splenic vasculature in resectable cases GITSG 9173, EORTC 40891, and ESPAC-1,
as the tumor can be removed with a distal pancre- were inconclusive regarding whether adjuvant
atectomy and splenectomy. RT provided an overall survival (OS) benefit, no
Various definitions of borderline resectable benefit, or decreased OS [8–10]. These historical
tumors are utilized. Generally, borderline resect- trials are difficult to interpret as split-course RT
able tumors consist of lesions with only abutment was utilized—a technique no longer used in the
(contact ≤180°) of the CA or SMA and/or modern era—did not mandate quality assurance
involvement of the CHA, SMV, or PV that is methods to ensure RT was delivered as per proto-
amenable to vascular reconstruction. Unresectable col, and had crossover of treatment arms. RTOG
13 Management of Pancreatic Cancer with MRI-Guided Adaptive Radiotherapy 245
9704 was one of the first trials to mandate quality atin 85 mg/m2, irinotecan 180 mg/m2, leucovorin
assurance to radiation field design and dosimetry 400 mg/m2, and infusional 5-fluorouracil
[11]. Although no overall survival difference was 2400 mg/m2 over 46 h) followed by chemoradio-
detected among the entire cohort for the addition therapy to 50.4 Gy in 28 fractions with concur-
of adjuvant chemoradiotherapy compared to rent capecitabine [15]. A large majority (68%) of
chemotherapy alone, an unplanned secondary
patients were able to undergo surgical resection,
analysis demonstrated statistically improved sur- and 93% had a margin-negative resection. Among
vival if treatment adhered to radiation therapy patients who underwent surgical resection, the
protocol highlighting the importance of careful 18-month overall survival rate was 67% com-
target delineation and treatment planning [12]. pared to 43% for those unable to proceed to
To date, the single most effective adjuvant resection [15]. Finally, the ESPAC5 trial [16] ran-
therapy for resected pancreatic cancer has been a domized 90 patients with borderline resectable
regimen of modified fluorouracil, leucovorin, iri- disease into 4 cohorts: (1) upfront surgery, (2)
notecan, and oxaliplatin (mFOLFIRINOX), neoadjuvant gemcitabine/capecitabine, (3) neo-
which increased median survival from 36 to adjuvant FOLFIRINOX, or (4) neoadjuvant
54 months and 5-year OS from 31 to 43% com- chemoradiotherapy to 50.4 Gy with concurrent
pared to single-agent gemcitabine [13]. capecitabine. There was no difference in the rate
Borderline resectable pancreatic cancer com- of resection between those randomized to upfront
prises a wide spectrum of disease, from relatively surgery vs. neoadjuvant therapy. Overall survival
minimal contact with veins that can undergo at 1 year was significantly improved among
reconstruction during resection to nearly 180° patients receiving neoadjuvant therapy (39% for
involvement of major arteries that are at very upfront surgery vs. 60–84% for neoadjuvant ther-
high risk of a positive margin. Several studies apy). Although the trial was not powered to com-
support the role of neoadjuvant therapy for bor- pare the neoadjuvant arms to each other, there
derline resectable pancreatic cancer; however, was a numerical improvement in overall survival
the specific benefit of preoperative radiation for FOLFIRINOX (84%) compared to gem-
remains unclear. citabine/capecitabine (78%) and chemoradio-
Three studies are worth noting in evaluating therapy (60%) [16]. These three studies, among
the overall role of neoadjuvant therapy for others, established the safety, feasibility, and effi-
patients with borderline resectable disease. First, cacy of neoadjuvant therapy for borderline resect-
Katz et al. [14] were the first to provide a formal able pancreatic cancer.
definition of borderline resectable disease and Randomized trials investigating the specific
reported their institutional experience treating role of radiation in borderline resectable pancre-
these patients with neoadjuvant therapy. Among atic cancer have shown a more mixed picture.
160 patients identified with borderline resectable PREOPANC-1 was a randomized study that
pancreatic cancer, 78% were able to complete enrolled both resectable and borderline resect-
neoadjuvant therapy with chemotherapy fol- able patients to either upfront surgery or neoadju-
lowed by chemoradiation, and 41% of the total vant gemcitabine for 3 cycles with radiation
cohort was able to undergo whipple resection (36 Gy in 15 fractions) given concurrently with
with a 94% rate of negative-margin resection. the second cycle of chemotherapy [17]. Patients
Furthermore, patients who were able to complete in the upfront surgery arm were planned to
all neoadjuvant therapy and undergo surgical receive six cycles of adjuvant gemcitabine, while
resection had a median overall survival of patients in the preoperative therapy arm were
40 months compared to patients who were unable planned for four cycles of adjuvant chemother-
to complete all planned therapy [14]. The Alliance apy. Among the cohort with borderline resectable
A021101 was a single-arm prospective trial that disease, use of neoadjuvant chemoradiotherapy
treated 22 patients with borderline resectable dis- increased the margin-negative resection rate to
ease with neoadjuvant mFOLFIRINOX (oxalipl- 79% from 13% and the median overall survival to
18 months from 13 months [17]. One limitation evaluate the role of conventional chemoradio-
of this study is that single-agent gemcitabine is therapy for patients with LAPC, but results have
no longer the standard of care in treatment of been mixed. The ECOG 4201 trial randomized
pancreatic cancer, raising the question of how patients to either gemcitabine chemotherapy
chemoradiotherapy compares to alone or gemcitabine with conventionally frac-
mFOLFIRINOX. As a partial answer to this
tionated radiation [19]. This trial identified a
question, the Alliance A021501 trial found no mere 2-month survival advantage for chemora-
benefit to the addition of neoadjuvant stereotactic diotherapy at the cost of a fourfold increase in
body radiotherapy (SBRT) for patients with bor- grade 4 toxicity [19]. In contrast, the much larger
derline resectable pancreatic cancer compared to LAP07 trial randomized 449 patients to gem-
neoadjuvant mFOLFIRINOX [18]. This multi- citabine vs. chemoradiotherapy with concurrent
center trial randomized 126 patients to either capecitabine and identified no significant differ-
8 cycles of neoadjuvant mFOLFIRINOX fol- ence in overall survival between the 2 cohorts
lowed by surgery or 7 cycles of neoadjuvant with a local control benefit seen with chemora-
mFOLFIRINOX followed by radiation therapy diotherapy compared to chemotherapy alone
(SBRT to 33–40 Gy in 5 fractions or hypofrac- which was statistically significant [20].
tionated image-guided RT to 25 Gy in 5 frac- Challenges to identifying a benefit to convention-
tions) followed by surgery. All patients were ally fractionated radiation in LAPC have led to
planned to receive four cycles of adjuvant development of methods for overcoming those
mFOLFOX (oxaliplatin 85 mg/m2, leucovorin traditional challenges as well as technological
400 mg/m2, and infusional 5- fluorouracil improvements allowing for dose escalation to
2400 mg/m2 over 46 h). The SBRT arm was improve clinical outcomes.
closed to accrual early due to a futility endpoint
defined by R0 (margin-negative) resection rate.
Patients in the SBRT arm experienced a higher Overcoming Challenges
rate of development of distant metastatic disease, of Radiotherapy in the Treatment
which was classified as margin positive on the of Pancreatic Cancer
study. Of note, patients in the SBRT arm had
higher carbohydrate antigen (CA) 19–9 scores, a Approximately 30% of pancreatic ductal adeno-
surrogate marker in PDAC used to assess treat- carcinoma (PDAC) patients die with locally
ment response, than those in the neoadjuvant destructive disease [5], as progressive locore-
chemotherapy arm (260 vs. 167, respectively). gional disease can cause pain which may be asso-
With these caveats in mind, the Alliance A021501 ciated with celiac plexus infiltration, gastric
trial identified a worse overall survival rate at outlet obstruction, and/or duodenal bleeding.
18 months for SBRT compared to chemotherapy These cancer-related complications negatively
alone (47% vs. 66%, respectively) [18]. Given impact the patient quality of life. As distant meta-
the disparate results for neoadjuvant radiation for static disease is better controlled with current
borderline resectable pancreatic cancer, practice standard-of-care multi-agent systemic therapy,
patterns vary widely by institution. Of note, the providing durable locoregional control with
radiation therapy doses employed in this study radiotherapy is integral to the long-term manage-
were non-ablative with a biological effective ment of PDAC patients. As technical advances in
dose (BED) of 37.5–72 Gy. radiation therapy delivery have been made, there
Locally advanced pancreatic cancer (LAPC) are now further studies demonstrating dose-
is defined as nonmetastatic but unresectable dis- escalated radiation therapy to a BED of 100 Gy
ease. It typically presents with either extensive has benefit for patients with PADC. Locoregional
peripancreatic lymph node involvement or sub- control can be improved by escalating radiation
stantial vascular involvement of the SMA, celiac dose using modern treatment techniques includ-
axis, IVC, or aorta. Attempts have been made to ing stereotactic body radiotherapy (SBRT) and
magnetic resonance imaging (MRI)-guided adap- CBCT, thereby enabling better delineation of soft
tive radiotherapy [21]. Early studies evaluating tissue target(s) and normal luminal gastrointesti-
the role of single fraction SBRT in locally nal organs [31] as shown in Chap. 2 of this book.
advanced PADC demonstrated local control with Using MRI for image-guided radiotherapy, the
first progression at distant sites however with MR-Linac (MRL) is well suited for dose-
significant toxicities [22]. Further studies with escalated SBRT in the treatment of
fractionated SBRT to lower BED doses reported PDAC. Importantly, the MRL allows the imple-
excellent local control with acceptable toxicity in mentation of online adaptive treatment planning,
these [23–28]. Furthermore, cancer-related pain whereby the radiation treatment plan can be
and quality of life were shown to improve in adapted in real time on the day of treatment based
patients with borderline resectable and locally on the patient’s luminal gastrointestinal anatomy
advanced PDAC at 4 months post SBRT [25, 26]. of the day just prior to radiation delivery. Dose
The 2019 American Society for Radiation delivery to the intended target is maximized for
Oncology (ASTRO) clinical practice guidelines cancer control while minimizing toxicities to the
for radiation therapy in PDAC conditionally rec- surrounding luminal gastrointestinal organs. In a
ommended a definitive treatment regimen con- single-institution retrospective analysis of inop-
sisting of systemic chemotherapy followed by erable PDAC patients who received induction
concurrent chemoradiation using conventional chemotherapy followed by SBRT to a total dose
fractionation or dose-escalated radiotherapy or of 50 Gy in five fractions (biological equivalent
multi-fraction SBRT alone in locally advanced dose or BED = 100 Gy) delivered using MRI-
disease that is not appropriate for downstaging to guided adaptive radiation therapy. Chuong et al.
eventual surgery [29]. In the 2023 NCCN guide- reported 1-year local control, progression-free
lines, both concurrent chemoradiation and SBRT survival, and overall survival of 87.8%, 52.4%,
were included as radiotherapy options to manage and 58.9%, respectively [32]. Updated results by
locally advanced/unresectable PDAC [30]. the same authors described median progression-
It is well recognized that the close proximity free survival of 20 months, median overall sur-
of the pancreatic tumor to many of the radiosen- vival of 23 months, and 2-year local control,
sitive luminal gastrointestinal structures (duode- progression-free survival, and overall survival of
num, small bowel, stomach, and large bowel) 68.8%, 40%, and 45.5%, respectively, in this
combined with the significant mobility and cohort of PDAC patients [33]. Acute and late
highly variable positioning of these structures grade 3 or higher toxicity rates were both 4.8%
makes radiation dose escalation extremely chal- [33]. Parikh et al. are the first to report a prospec-
lenging. On most conventional linacs, image- tive phase II trial of MRgRT utilizing SBRT for
guided radiotherapy is delivered using cone beam borderline resectable and locally advanced PDAC
computed tomography (CBCT) for daily image which demonstrated 1 year LC of 82.9% with
guidance and radiation treatment verification. 32.4% of patient undergoing surgery following
The CBCT has suboptimal soft tissue contrast radiotherapy and acute toxicity ≥3 GI of 8.8%
and is not ideal for image-guided SBRT in the for the cohort of 136 patients [34]. These data are
treatment of PDAC because the pancreatic tumor promising and have led to a phase III randomized
cannot be clearly differentiated from the sur- controlled trial titled “Locally Advanced
rounding normal luminal gastrointestinal struc- Pancreatic Cancer Treated with ABLAtivE
tures. As such, endoscopic ultrasound guided Stereotactic MRI-guided Adaptive Radiation
placement of fiducials within the pancreatic Therapy (LAP-ABLATE)” which was launched
tumor is required for target localization as well as in 2022 to evaluate overall survival among other
breath hold techniques for improved visualiza- endpoints in PDAC patients who receive chemo-
tion during safe delivery of SBRT on a conven- therapy followed by dose-escalated, MRL-based
tional linac. Magnetic resonance imaging (MRI) SBRT compared to those who receive chemo-
offers superior soft tissue contrast compared to therapy alone [35].
Contouring, PTV, and OARs and restaging during initial chemotherapy.

Special note is taken of areas of significant vas-
Contouring, PTV, and OARs cular contact, as well as areas initially involved
for ViewRay MRIdian by tumor at the time of diagnosis that may have
regressed on chemotherapy. Suspicious nodes at
Patients treated on the ViewRay MRIdian MRL risk are also contoured as the GTV. Critical
platform described in Chap. 9 of this book are regional vessels are then contoured separately,
simulated in the supine position with arms down most often the celiac artery, common hepatic
for both their CT and MRI simulation on the artery, superior mesenteric artery, superior mes-
same day. Contours are drawn on the true fast enteric vein, and portal vein. The splenic artery
imaging with steady-state free precession and vein are occasionally contoured when there
(TRUFISP) MR imaging sequence, with refer- is significant radiographic involvement with the
ence to the CT simulation sequences with 20-s primary tumor.
and 70-s post-contrast administration scans. Prior To create a structure identified as the Drawn_
work has identified sites of local failure along the CTV, the GTV and vessels have a margin applied
retroperitoneal margin and nearby vasculature in to account for microscopic tumor spread. Most
patients treated to small fields of pancreas SBRT commonly, a 3 mm margin is used for the GTV,
[36]. As a result, the Dana-Farber Brigham common hepatic artery, and portal vein. The
Cancer Center typically employs a clinical target celiac artery and superior mesenteric artery and
volume (CTV) approach informed by the work of vein are typically expanded with a 7 mm margin.
Hill et al. who defined a “triangle volume” of at- In addition to the expansions above, manual cor-
risk disease near pancreatic head, neck, and unci- rections are made to encompass the retroperito-
nate tumors [37]. An example of this triangle neal margin, especially in the pericaval and
volume as contoured on the ViewRay MRIdian portacaval spaces as well as subtraction of the
can be seen in Fig. 13.2. Drawn_CTV where it intersects with the aorta
The gross tumor volume (GTV) is drawn and renal vasculature. OARs are identified
using simulation TRUFISP and CT sequences in including the esophagus, stomach, duodenum,
addition to diagnostic triple-phase pancreatic small bowel, and large bowel within 2 cm supe-
protocol CTs obtained at the time of diagnosis rior and inferior to all slices of the Drawn_
Fig. 13.2 Various

color-coded anatomical
structures in the vicinity
of the pancreas.
“Triangle volume”
contour as described by
Hill et al. [37] on
ViewRay MRIdian.
GTV (pink), vasculature
(magenta), CTV after
expansion (blue), final
drawn CTV with manual
retroperitoneal margin
inclusion (orange),
stomach (green),
duodenum (yellow),
small bowel (purple),
and large bowel (brown)
Fig. 13.3 Locally

advanced pancreatic
adenocarcinoma plan on
ViewRay MRIdian.
Drawn_CTV (orange),
CTV (blue), PTV (red),
stomach (green),
duodenum (yellow);
small bowel (purple),
and large bowel (brown).
The 40 Gy isodose line
is shown in yellow
colorwash
CTV. In addition, the spinal canal, liver, and ontouring, PTV, and OARs for Elekta
C
kidneys are included as OARs for each treatment. Unity
No planning risk volume (PRV) expansion is
used for OARs. At the Sunnybrook Odette Cancer Centre, PDAC
The Drawn_CTV is permitted to overlap patients who are deemed appropriate for
with the adjacent organs-at-risk, including GI MR-Linac-based SBRT are treated on the Elekta
OARs. The Drawn_CTV has all GI OARs sub- Unity (Elekta AB, Stockholm, Sweden) that has
tracted to create the final CTV for treatment been described in Chap. 10 of this book. In
planning objectives, and a 3 mm margin is added general, most patients have locally advanced/

for a final PTV. The use of a Drawn_CTV allows unresectable disease who are not candidates for
for online adaptive recontouring of the target at chemotherapy due to advanced age and/or medi-
risk as well as GI OARs without having to man- cal comorbidities, who decline chemotherapy,
ually subtract structures, thereby saving time. A who tolerate chemotherapy poorly, or who have
rule can be implemented in the treatment deliv- completed up to 12 cycles of standard-of-care
ery software to automatically subtract all GI chemotherapy with stable disease or locally pro-
OARs from the Drawn_CTV, thereby recreating gressive disease. Patients with contraindications
the final CTV and PTV through automation to MRI, who are unable to lie supine for up to
rather than through manual implementation. 120 min, or who are >140 kg and/or have a cir-
The PTV is treated to a dose of 33 Gy in five cumference >60 cm are excluded. The radiation
fractions for borderline resectable cases planned treatment workflow and planning and radiation
for surgical exploration and resection or a dose dose-volume constraints described herein repre-
of 40–50 Gy in five fractions for patients with sent a high-level overview of the current local
locally advanced disease who are not undergo- institutional practice and should be considered as
ing a planned surgical exploration or resection. a guide only and are subject to change as new
An example of dose coverage achieved on the information and advances may become available
ViewRay MRIdian can be seen in Fig. 13.3. in the future.
Other examples of volume delineation are All patients are positioned supine with both
described in the literature as well as ongoing arms down and immobilized in a Vac-Lok cush-
clinical trials as described above. ion. Four-dimensional computed tomography
a b
Fig. 13.4 A representative case of locally advanced/ PTV = ITV + 5 mm (lavender); ITV25 (aqua);
unresectable pancreatic adenocarcinoma being treated on PTV25 = ITV25 + 5 mm (turquoise); duodenum (orange);
the Elekta Unity. (a, b) 4DCT simulation with intravenous duodenum PRV = duodenum +3 mm (light green color-
contrast (left), MR simulation (b3DVANE XD SPAIR wash); small bowel (white); small bowel PRV = small
sequence) on the Elekta Unity (right). ITV (red), ITV25 bowel +3 mm (dark green colorwash); large bowel (pink);
(cyan), duodenum (orange), small bowel (white). (c) large bowel PRV = large bowel +3 mm (pink colorwash)
Expansion of target and OAR volumes. ITV (red);
(4DCT) with intravenous contrast and MR sim- nodal regions at risk of harboring microscopic
ulation is performed on the CT simulator disease [38]. PTV25 is a 5 mm expansion of
(Philips Healthcare, USA) and the Elekta Unity ITV25 and is prescribed a dose of 25 Gy in five
(Elekta, Stockholm, Sweden), respectively, on fractions. Organs-at-risk (OARs) including the
the same day. The 4DCT image datasets and the stomach, duodenum, small bowel, large bowel,
3DVANE image datasets (b3DVANE XD SPAIR liver, kidneys, and spinal canal are contoured on
and T1 3DVANE XD) are used for contouring of the CT datasets. The planning risk volume
targets and organs-at-risk. Internal target vol- (PRV) margin for the stomach, duodenum, small
ume (ITV) is the gross tumor volume (GTV) bowel, and large bowel is 3 mm. A representa-
delineated using the 4DCT and 3DVane image tive case of locally advanced/unresectable
datasets. PTV is a 5 mm expansion of the PDAC with target and OAR volumes is illus-
ITV. PTV is prescribed a dose of 40–50 Gy in trated in Fig. 13.4. Reference plans are gener-
five fractions at the discretion of the treating ated in Monaco 5.4 (Elekta, Stockholm,
physician. The structure ITV25 is a 1 cm expan- Sweden) and optimized to maximize PTV cov-
sion of the ITV, celiac artery, superior mesen- erage by the prescription dose (i.e., 40–50 Gy),
teric artery, and portal vein and represents the with at least 99% of the PTV being covered by
Fig. 13.5 An Elekta Unity reference plan is generated PTV25 (turquoise colorwash). Table illustrating dose-
and optimized to maximize target coverage while respect- volume constraints of the reference plan (bottom panel);
ing all OAR constraints (top panel). Duodenum (orange), these constraints are used in local institutional practice
small bowel (white), PTV (lavender colorwash), and and should be considered as a guide only
35 Gy while respecting all OAR constraints. An erence plan with a prescription of 40 Gy in five
arrangement of 12 beams is used for reference fractions for the representative case is shown in
plans with 3 mm grid size and 1% statistical Fig. 13.5, where 72.7% of the PTV is covered
uncertainty per plan [39]. Target undercoverage by 40 Gy and 100% of the PTV is covered by
is accepted to respect OAR constraints. The ref- 35 Gy while meeting all OAR constraints.
n-Table Adaptation for ViewRay

O n-Table Adaptation for Elekta
O
MRIdian Unity
All patients are treated with an online adaptive All PDAC patients are treated with online adapt-
workflow. A daily TRUFISP sequence is to-shape (ATS) workflow. Clinical ATS workflow
obtained after aligning the patient on the is initiated by acquiring the 3DVANE imaging
machine. Shifts are identified by aligning the sequences, b3DVANE XD SPAIR, and T1
patient to the GTV, taking note of areas of rota- 3DVANE XD. Body contour, vertebrae, spinal
tion or other deformation of either the GTV or canal, liver, and kidneys are deformably or rig-
surrounding vasculature. The GTV is rigidly idly propagated, visually inspected, and edited.
propagated, and most OARs including the GI Subsequently, ITV, ITV25, celiac artery, superior
structures, spinal cord, liver, and kidneys are mesenteric artery, portal vein, stomach, duode-
deformably registered to the daily setup image. num, small bowel, and large bowel are contoured
On the MRIdian A3i system, the physician, online by the treating physician. PTV, PTV25,
physicist, and therapist role groups can simulta- and PRV margins are recreated from the modified
neously interact with the treatment plan, saving ITV, ITV25, and luminal gastrointestinal struc-
time through parallel processing. The physician tures based on the anatomy of the day. A new
recontours the Drawn_CTV structure (Fig. 13.2) ATS plan is generated for each fraction using flu-
and all OARs within a 2–3 cm boundary of the ence optimization and bulk electron density
PTV. This boundary is chosen due to the high assignment derived from the reference plan.
probability that any dose- limiting constraint Following evaluation and approval of the ATS
will be found within 2 cm of the PTV. The phys- plan by the treating physician, physics check is
icist ensures fidelity of the dose calculation by performed. Pretreatment verification images are
ensuring the skin contour is properly drawn and acquired for visual inspection of the plan and tar-
that any new areas of air or soft tissue within the get location. Adapt-to-position workflow is to
body contour are accounted for. The therapist follow only if the target falls outside of PTV on
role group is simultaneously able to modify any pretreatment verification images. Otherwise,
tracking structures needed for treatment track- treatment is delivered with CINE motion moni-
ing and gating, as well as being available to the toring. The ATS plan for fraction 5 of the repre-
patient for any concerns that may arise while in sentative case is shown in Fig. 13.6. On the day
the MRL bore. of fraction 5, improved PTV coverage is achieved
Once all contours are finalized and rules are with 86.5% of the PTV being covered by 40 Gy
applied, including automated generation of the (vs. 72.7% on the reference plan in Fig. 13.5) as
CTV and PTV from the Drawn_CTV, the origi- the small bowel is located farther away from the
nal treatment plan is reoptimized. An assess- PTV.
ment is made regarding the dosimetric In summary, the ATS workflow on the MRL
parameters of the so-called predicted plan, for pancreas SBRT harnesses the superior soft
which calculates the dosimetry to adapted target tissue contrast of MRI and supports adaptive
and OAR structures using the original plan planning in real time based on the patient’s anat-
MLC positions and beam fluence. In most cases, omy of the day, thereby maximizing target cover-
the predicted plan results either in undercoverage while minimizing toxicities to OARs. The
age of the PTV or dose to an OAR in excess of a results from the LAP-ABLATE and similar trials
strict constraint. In these cases, the plan is reop- are highly anticipated. In addition, it may be
timized by a physicist, often in an iterative fash- interesting to explore whether MRL-based,
ion in consultation with the physician, until an dose-escalated SBRT rivals surgery in the man-
acceptable plan is obtained. agement of resectable pancreatic cancer.
Fig. 13.6 Adapt-to-shape plan on Elekta Unity for the colorwash). The dose-volume constraints are also shown
fifth and last fraction. Duodenum (orange), small bowel in the figure
(white), PTV (lavender colorwash), and PTV25 (turquoise
Plan Evaluation mary disease. Incorporating molecular markers

such as SMAD4, among others, to stratify
Plans are evaluated with a goal of maximizing patients who would benefit from radiation ther-
coverage while respecting OAR constraints. apy would be one approach in future clinical
When there is a conflict, priority is given to trials. Ongoing prospective randomized studies
the OAR constraint while sacrificing target such as LAP-ABLATE will further inform us as
coverage. This isotoxic approach ensures that to whether dose-escalated MRgRT will improve
each adaptive fraction of radiation does not survival as well as improve quality of life.
exceed tolerances to critical adjacent OARs MRgRT has overcome the technical challenges
while attempting to maximize coverage of the of tumor location with dose-escalated radio-
target to the maximum safe dose possible each therapy in PDAC. We eagerly await further data
day. This approach is especially prudent in the to support the clinical benefit of this tool in
management of locally advanced, unresectable PDAC.
PDAC as the treatment intent is generally Special thanks go to Hyebeen Kim for design
understood to be palliative in nature. of Fig. 13.1.
Secondary objectives include managing the
hotspot of the plan to be reasonably close to
the original plan hotspot as well as ensuring References
the plan can be delivered within a reasonable
amount of time and/or monitor units. 1. Sung H, Ferlay J, Siegel RL, Laversanne M,
Soerjomataram I, Jemal A, et al. Global Cancer
Statistics 2020: GLOBOCAN estimates of incidence
and mortality worldwide for 36 cancers in 185 coun-
Summary tries. CA Cancer J Clin. 2021;71:209–49. https://doi.
org/10.3322/caac.21660.
2. Siegel RL, Miller KD, Fuchs HE, Jemal A. Cancer
The limitations for radiation therapy for pancreas
statistics, 2022. CA Cancer J Clin. 2022;72:7–33.
cancer have long been recognized as the risk of https://doi.org/10.3322/caac.21708.
damage to multiple critical organs intimate with 3. Jiang S, Fagman JB, Ma Y, Liu J, Vihav C, Engstrom
the pancreas. MR-Linac technology allows radia- C, et al. A comprehensive review of pancreatic cancer
and its therapeutic challenges. Aging. 2022;14:7635–
tion oncologists to maintain their respect for
49. https://doi.org/10.18632/aging.204310.
those adjacent organs-at-risk while significantly 4. Wolff RA, Crane CH, Li D, Evans DB, Maitra A, Tsai
increasing the amount of energy administered to S. Neoplasms of the exocrine pancreas. In: Bast RC,
the target disease. In the care of patients with Croce CM, Hait WN, Hong WK, Kufe DW, Piccart-
Gebhart M, et al., editors. Holland-Frei cancer
adenocarcinoma of the pancreas, there is mount-
medicine. 9th ed. Hoboken: Wiley Blackwell; 2017.
ing evidence showing improved local control p. 1129–50.
with the use of dose-escalated radiation therapy 5. Iacobuzio-Donahue CA, Fu B, Yachida S, Luo M,
given in this way. This has translated in better Abe H, Henderson CM, et al. DPC4 gene status of
the primary carcinoma correlates with patterns of
symptom control and therefore quality of life for
failure in patients with pancreatic cancer. J Clin
patients with locally advanced disease and resect- Oncol. 2009;27:1806–13. https://doi.org/10.1200/
ability and negative-margin resections in border- JCO.2008.17.7188.
line resectable cases. 6. Soloff EV, Zaheer A, Meier J, Zins M, Tamm
EP. Staging of pancreatic cancer: resectable, bor-
Dose-escalated radiotherapy with improved
derline resectable, and unresectable disease. Abdom
imaging and adaptation capabilities has Radiol. 2018;43:301–13. https://doi.org/10.1007/
improved local control and symptoms related to s00261-017-1410-2.
progressive local disease in PDAC but, as yet, 7. Shaib Y, Davila J, Naumann C, El-Serag H. The
impact of curative intent surgery on the survival
has not shown a clear survival benefit. Treatment
of pancreatic cancer patients: a U.S. population-
selection beyond optimized systemic therapy is based study. Off J Am Coll Gastroenterol ACG.
burdened by a lack of patient and tumor specific 2007;102:1377.
data indicating which groups are more likely to 8. Kaiser MH, Ellenberg SS. Pancreatic cancer: adjuvant
combined radiation and chemotherapy following cura-
have their survival compromised by their pri-
tive resection. Arch Surg. 1985;120:899–903. https:// gery for resectable and borderline resectable pancre-
doi.org/10.1001/archsurg.1985.01390320023003. atic cancer: long-term results of the Dutch randomized
9. Smeenk HG, van Eijck CHJ, Hop WC, Erdmann J, PREOPANC trial. J Clin Oncol. 2022;40:1220–30.
Tran KCK, Debois M, et al. Long-term survival and https://doi.org/10.1200/JCO.21.02233.
metastatic pattern of pancreatic and periampullary 18. Katz MHG, Shi Q, Meyers J, Herman JM,
cancer after adjuvant chemoradiation or observa- Chuong M, Wolpin BM, et al. Efficacy of pre-
tion: long-term results of EORTC trial 40891. Ann operative mFOLFIRINOX vs mFOLFIRINOX
Surg. 2007;246:734–40. https://doi.org/10.1097/ plus hypofractionated radiotherapy for borderline
SLA.0b013e318156eef3. resectable adenocarcinoma of the pancreas: the
10. Neoptolemos JP, Stocken DD, Friess H, Bassi C, A021501 phase 2 randomized clinical trial. JAMA
Dunn JA, Hickey H, et al. A randomized trial of Oncol. 2022;8:1263–70. https://doi.org/10.1001/
chemoradiotherapy and chemotherapy after resection jamaoncol.2022.2319.
of pancreatic cancer. N Engl J Med. 2004;350:1200– 19. Loehrer PJ, Feng Y, Cardenes H, Wagner L, Brell JM,
10. https://doi.org/10.1056/NEJMoa032295. Cella D, et al. Gemcitabine alone versus gemcitabine
11. Regine WF, Winter KA, Abrams R, Safran H, Hoffman plus radiotherapy in patients with locally advanced
JP, Konski A, et al. Fluorouracil based chemoradiation pancreatic cancer: an eastern cooperative oncology
with either gemcitabine or fluorouracil chemotherapy group trial. J Clin Oncol. 2011;29:4105–12. https://
following resection of pancreatic adenocarcinoma: doi.org/10.1200/JCO.2011.34.8904.
5-year analysis of the US intergroup/RTOG 9704 20. Hammel P, Huguet F, van Laethem J-L, Goldstein D,
phase III trial. Ann Surg Oncol. 2011;18:1319. https:// Glimelius B, Artru P, et al. Effect of chemoradiotherapy
doi.org/10.1245/s10434-011-1630-6. vs chemotherapy on survival in patients with locally
12. Abrams RA, Winter KA, Regine WF, Safran H, advanced pancreatic cancer controlled after 4 months
Hoffman JP, Lustig R, et al. Failure to adhere to of gemcitabine with or without Erlotinib: the LAP07
protocol specified radiation therapy guidelines randomized clinical trial. JAMA. 2016;315:1844–53.
was associated with decreased survival in RTOG https://doi.org/10.1001/jama.2016.4324.
9704—a phase III trial of adjuvant chemotherapy 21. Koay EJ, Hanania AN, Hall WA, Taniguchi CM,
and chemoradiotherapy for patients with resected Rebueno N, Myrehaug S, et al. Dose-escalated
adenocarcinoma of the pancreas. Int J Radiat Oncol radiation therapy for pancreatic cancer: a simul-
Biol Phys. 2012;82:809–16. https://doi.org/10.1016/j. taneous integrated boost approach. Pract Radiat
ijrobp.2010.11.039. Oncol. 2020;10:e495–507. https://doi.org/10.1016/j.
13. Conroy T, Castan F, Lopez A, Turpin A, Ben prro.2020.01.012.
Abdelghani M, Wei AC, et al. Five-year outcomes of 22. Koong AC, Le QT, Ho A, Fong B, Fisher G, Cho C,
FOLFIRINOX vs gemcitabine as adjuvant therapy for et al. Phase I study of stereotactic radiosurgery in
pancreatic cancer: a randomized clinical trial. JAMA patients with locally advanced pancreatic cancer. Int
Oncol. 2022;8:1571–8. https://doi.org/10.1001/ J Radiat Oncol Biol Phys. 2004;58:1017–21. https://
jamaoncol.2022.3829. doi.org/10.1016/j.ijrobp.2003.11.004.
14. Katz MHG, Pisters PWT, Evans DB, Sun CC, Lee 23. Mahadevan A, Miksad R, Goldstein M, Sullivan
JE, Fleming JB, et al. Borderline resectable pancre- R, Bullock A, Buchbinder E, et al. Induction gem-
atic cancer: the importance of this emerging stage citabine and stereotactic body radiotherapy for locally
of disease. J Am Coll Surg. 2008;206:833–46.; advanced nonmetastatic pancreas cancer. Int J Radiat
(discussion 846–848). https://doi.org/10.1016/j. Oncol Biol Phys. 2011;81:e615–22. https://doi.
jamcollsurg.2007.12.020. org/10.1016/j.ijrobp.2011.04.045.
15. Katz MHG, Shi Q, Ahmad SA, Herman JM, de 24. Pollom EL, Alagappan M, von Eyben R, Kunz PL,
Marsh RW, Collisson E, et al. Preoperative modified Fisher GA, Ford JA, et al. Single- versus multifrac-
FOLFIRINOX treatment followed by capecitabine- tion stereotactic body radiation therapy for pancre-
based chemoradiation for borderline resectable atic adenocarcinoma: outcomes and toxicity. Int J
pancreatic cancer. JAMA Surg. 2016;151:e161137. Radiat Oncol Biol Phys. 2014;90:918–25. https://doi.
https://doi.org/10.1001/jamasurg.2016.1137. org/10.1016/j.ijrobp.2014.06.066.
16. Ghaneh P, Palmer D, Cicconi S, Jackson R, Halloran 25. Rosati LM, Kumar R, Herman JM. Integration of
CM, Rawcliffe C, et al. Immediate surgery com- stereotactic body radiation therapy into the mul-
pared with short-course neoadjuvant gemcitabine tidisciplinary management of pancreatic cancer.
plus capecitabine, FOLFIRINOX, or chemoradio- Semin Radiat Oncol. 2017;27:256–67. https://doi.
therapy in patients with borderline resectable pan- org/10.1016/j.semradonc.2017.02.005.
creatic cancer (ESPAC5): a four-arm, multicentre, 26. Herman JM, Chang DT, Goodman KA, Dholakia
randomised, phase 2 trial. Lancet Gastroenterol AS, Raman SP, Hacker-Prietz A, et al. Phase 2 multi-
Hepatol. 2023;8:157–68. https://doi.org/10.1016/ institutional trial evaluating gemcitabine and stereo-
S2468-1253(22)00348-X. tactic body radiotherapy for patients with locally
17. Versteijne E, van Dam JL, Suker M, Janssen QP, advanced unresectable pancreatic adenocarcinoma.
Groothuis K, Akkermans-Vogelaar JM, et al. Cancer. 2015;121:1128–37. https://doi.org/10.1002/
Neoadjuvant chemoradiotherapy versus upfront sur- cncr.29161.
27. Moningi S, Dholakia AS, Raman SP, Blackford 34. Parikh PJ, Lee P, Low DA, Kim J, Mittauer KE,
A, Cameron JL, Le DT, et al. The role of stereo- Bassetti MF, et al. A multi-institutional phase 2
tactic body radiation therapy for pancreatic can- trial of ablative 5-fraction stereotactic magnetic
cer: a single- institution experience. Ann Surg resonance-guided on-table adaptive radiation ther-
Oncol. 2015;22:2352–8. https://doi.org/10.1245/ apy for borderline resectable and locally advanced
s10434-014-4274-5. pancreatic cancer. Int J Radiat Oncol Biol Phys.
28. Petrelli F, Comito T, Ghidini A, Torri V, Scorsetti M, 2023;117(4):799–808. https://doi.org/10.1016/j.
Barni S. Stereotactic body radiation therapy for locally ijrobp.2023.05.023.
advanced pancreatic cancer: a systematic review and 35. Locally advanced pancreatic cancer treated with
pooled analysis of 19 trials. Int J Radiat Oncol Biol ABLAtivE stereotactic MRI-guided adaptive radia-
Phys. 2017;97:313–22. https://doi.org/10.1016/j. tion therapy-Full Text. View-ClinicalTrials.gov.
ijrobp.2016.10.030. https://clinicaltrials.gov/ct2/show/NCT05585554.
29. Palta M, Godfrey D, Goodman KA, Hoffe S, Dawson Accessed 29 June 2023.
LA, Dessert D, et al. Radiation therapy for pancre- 36. Zhu X, Ju X, Cao Y, Shen Y, Cao F, Qing S, et al.
atic cancer: executive summary of an ASTRO clinical Patterns of local failure after stereotactic body radia-
practice guideline. Pract Radiat Oncol. 2019;9:322– tion therapy and sequential chemotherapy as ini-
32. https://doi.org/10.1016/j.prro.2019.06.016. tial treatment for pancreatic cancer: implications
30. National Comprehensive Cancer Center Network. of target volume design. Int J Radiat Oncol Biol
Pancreatic adenocarcinoma (version 2.2023). 2023. Phys. 2019;104:101–10. https://doi.org/10.1016/j.
https://www.nccn.org/professionals/physician_gls/ ijrobp.2019.01.075.
pdf/pancreatic.pdf. Accessed 29 June 2023. 37. Hill CS, Fu W, Hu C, Sehgal S, Reddy AV, He J, et al.
31. Caravatta L, Cellini F, Simoni N, Rosa C, Niespolo Location, location, location: what should be targeted
RM, Lupattelli M, et al. Magnetic resonance imaging beyond gross disease for localized pancreatic ductal
(MRI) compared with computed tomography (CT) adenocarcinoma? Proposal of a standardized clini-
for interobserver agreement of gross tumor volume cal tumor volume for pancreatic ductal adenocarci-
delineation in pancreatic cancer: a multi-institutional noma of the head: the “triangle volume”. Pract Radiat
contouring study on behalf of the AIRO group for Oncol. 2022;12:215–25. https://doi.org/10.1016/j.
gastrointestinal cancers. Acta Oncol Stockh Swed. prro.2022.01.005.
2019;58:439–47. https://doi.org/10.1080/02841 38. Kharofa J, Mierzwa M, Olowokure O, Sussman J,
86X.2018.1546899. Latif T, Gupta A, et al. Pattern of marginal local fail-
32. Chuong MD, Bryant J, Mittauer KE, Hall M, Kotecha ure in a phase II trial of neoadjuvant chemotherapy
R, Alvarez D, et al. Ablative 5-fraction stereotactic and stereotactic body radiation therapy for resect-
magnetic resonance-guided radiation therapy with able and borderline resectable pancreas cancer. Am J
on-table adaptive replanning and elective nodal irra- Clin Oncol. 2019;42:247–52. https://doi.org/10.1097/
diation for inoperable pancreas cancer. Pract Radiat COC.0000000000000518.
Oncol. 2021;11:134–47. https://doi.org/10.1016/j. 39. Paulson ES, Ahunbay E, Chen X, Mickevicius
prro.2020.09.005. NJ, Chen G-P, Schultz C, et al. 4D-MRI driven
33. Chuong MD, Herrera R, Kaiser A, Rubens M, MR-guided online adaptive radiotherapy for abdomi-
Romaguera T, Alvarez D, et al. Induction chemo- nal stereotactic body radiation therapy on a high field
therapy and ablative stereotactic magnetic resonance MR-Linac: implementation and initial clinical experi-
image-guided adaptive radiation therapy for inoper- ence. Clin Transl Radiat Oncol. 2020;23:72–9. https://
able pancreas cancer. Front Oncol. 2022;12:888462. doi.org/10.1016/j.ctro.2020.05.002.
https://doi.org/10.3389/fonc.2022.888462.
The Role of MR-Guided Radiation
Therapy in the Management
14
of Liver Malignancies
Nicki Karimi-Mostowfi, Luca Boldrini, Cihan Gani,

and Ann Raldow
Introduction and Disease Overview not in the top 10 of most commonly occurring
cancers [2].
Primary liver cancer—most commonly hepato- HCC usually arises in the setting of liver cir-
cellular carcinoma (HCC) and intrahepatic chol- rhosis and is often associated with alcohol abuse
angiocarcinoma (ICC)—was the sixth most and hepatitis B and C infections [3]. In the West,
diagnosed cancer and third leading cause of can- nonalcoholic steatohepatitis (NASH) associated
cer mortality worldwide in 2020 [1]. For decades, with metabolic syndrome and diabetes mellitus is
liver cancer had the fastest increasing mortality also a risk factor [3]. ICC is a malignancy that
in the United States, but the recent 5 years has arises within the cells of the peripheral bile duct
shown a plateau in these rates [2]. In the United and has similar risk factors to HCC [4]. Additional
States in 2022, it was the fifth leading cause of risk factors include primary sclerosing cholangi-
mortality in men and seventh in women, but was tis (PSC)—characterized by biliary inflammation
and fibrosis—congenital malformation of the bile
duct, and hepatic parasitic infections, particularly
N. Karimi-Mostowfi Opisthorchis viverrini and Clonorchis sinensis in
Lewis Katz School of Medicine at Temple University, East Asia [4].
Philadelphia, PA, USA
e-mail: [email protected] Both HCC and ICC are more classically asso-
ciated with older males [3, 4]. In 2016 in the
L. Boldrini
Department of Bioimaging, Radiation Oncology and United States, American Indians/Alaskan Natives
Hematology, Fondazione Policlinico Universitario had the highest incidence of HCC, followed by
“Agostino Gemelli” IRCCS, Rome, Italy Hispanics, Asians/Pacific Islanders, non-
e-mail: [email protected] Hispanic Blacks, and then non-Hispanic Whites
C. Gani [5]. Worldwide, in 2018, the age-standardized
Department of Radiation Oncology, University incidence and mortality of HCC was highest in
Hospital Tübingen, Tübungen, Germany
Eastern Asia, followed by Southeast Asia and
German Cancer Research Center (DKFZ) Heidelberg Africa [1]. The lowest age-standardized inci-
and German Consortium for Translational Cancer
Research (DKTK), Partner Site Tübingen, dence and mortality rates are in South Central
Tübingen, Germany Asia, followed by Europe and Western Asia [1].
e-mail: [email protected] Asians had a higher risk of an ICC diagnosis
A. Raldow (*) than Whites, who had a higher risk than Blacks
Department of Radiation Oncology, University of [6]. However, despite this trend, Black patients
California, Los Angeles, Los Angeles, CA, USA showed an increased risk of death with less sur-
258 N. Karimi-Mostowfi et al.
vival time [6]. Worldwide, the incidence of ICC and symptoms can vary; the primary means of
is also increasing—in 2017, higher rates were diagnosing liver metastases is via imaging includ-
seen in Eastern Asia when compared to Western ing US, CT, MRI, and PET [12]. For patients
countries [4]. with limited disease and resectable liver metasta-
The clinical presentation of HCC and ICC can ses, the preferred treatment is surgical resection
vary. Patients may be asymptomatic, present with [12]. However, radiation therapy with ablative
nonspecific symptoms—such as abdominal pain, intent is increasingly used with minimal toxicity
lethargy, nausea/vomiting, and weight loss—or and high local control [12].
with specific symptoms of jaundice, pruritis, clay-
colored stool, and dark urine. Carcinoembryonic
antigen (CEA) and carbohydrate antigen (CA) Overcoming Challenges
19–9 and 125 are commonly checked biomarkers with Traditional Approaches
in cholangiocarcinoma, and elevations in both cor-
relate with disease [7]. Alpha-fetoprotein (AFP) is Liver cancer is best managed with a multidisci-
useful for HCC diagnosis and screening, although plinary approach. Surgery, liver transplant, and
it lacks accuracy, with 32–59% of HCC patients thermal ablation are considered curative treat-
having normal levels [8]. As a result, the American ments for patients with HCC without macrovas-
Association for the Study of Liver Diseases cular invasion [13]. The most effective therapeutic
(AASLD) has a series of recommendations for option for HCC is liver transplantation—how-
surveilling HCC, depending on the risk factors of ever, this is an option available for less than 20%
the individual. Patients with cirrhosis are recom- of HCC patients and 15% of patients with ICC
mended to undergo ultrasound (US) surveillance, [11, 14]. Surgery may not be an option due to
with or without AFP, every 6 months as it improve advanced local extent of disease, insufficient
survival [9]. The diagnosis of HCC can be made liver function, or medical inoperability. Liver
via CT/MRI using Liver Reporting and Data metastases are only amenable to surgical resec-
System (LIRADS) criteria [10]. Major criteria tion in 10–20% of cases [15]. As a result, external
include (non-rim) arterial phase hyperenhance- beam radiation therapy (EBRT) has emerged for
ment, non-peripheral “washout,” enhancing cap- the treatment of liver metastases.
sule/pseudocapsule, and threshold growth. A Historically, liver malignancies were some-
LIRADS 5 lesion is diagnostic of HCC and does times treated palliatively with low-dose whole
not require a biopsy. ICC is typically diagnosed in liver radiation [16]. HCC, being one of the most
later stages when the tumor is unresectable; CT, radioresistant cancers, is a poor choice for low-
MRI, magnetic resonance cholangiopancreatog- dose radiation [13]. An additional challenge with
raphy (MRCP), and positron emission tomogra- liver-directed EBRT is the risk of radiation-
phy (PET) can help delineate the extent of disease induced liver disease (RILD), a phenomenon
[11]. associated with a high mortality rate in patients
Secondary malignancies are also common in with liver cancer [14]. There are two types of
the liver. Cancers that frequently metastasize to RILD—classic and non-classic. Classic RILD
the liver include colorectal, pancreatic, lung, typically occurs between 2 weeks and 3 months
breast, and melanoma. In fact, up to 50% of of therapy and is characterized by hepatomegaly,
patients with cancer can present with liver metas- anicteric ascites, and more than a twofold
tases at some point during their disease course increase in alkaline phosphatase (ALP) [14, 17].
[12]. More common than primary liver tumors, Classic RILD usually occurs in patients with
liver metastases are seen in approximately 5% of fairly normal pretreatment liver function. Non-
patients with cancer at the time of primary diag- classic RILD typically occurs between 1 week
nosis, although the percentage varies with cancer and 3 months after therapy, develops more often
type and disease progression [12]. Like cancers in patients with underlying chronic hepatic dis-
that originate in the liver, disease manifestation ease, and is characterized by elevated liver trans-
14 The Role of MR-Guided Radiation Therapy in the Management of Liver Malignancies 259
aminases or a decline in liver function in the too small to be properly targeted, and/or patients
absence of classic RILD [14, 17]. RILD is patho- with tumors that are in close proximity to muco-
logically similar to sinusoidal obstruction sal structures such as bowel [23]. The treatment
syndrome (SOS), which leads to hepatocellular requires accurate image guidance to properly
necrosis, terminal hepatic venule fibrotic scar- evaluate the position of the tumor, which can
ring, and architectural distortion [18]. The sever- change with breathing motion and the movement
ity of RILD also depends on factors such as total of adjacent organs, in order to ensure accurate
RT dose and fractionation, the volume of liver radiation delivery.
radiated, and the baseline liver function [19]. In Most facilities deliver SBRT using 4D imag-
both classic and non-classic RILD, treatment is ing and treatment under cone-beam CT image
limited, and both can progress to liver failure and guidance, which has limited soft tissue contrast
death [17]. To limit RILD, there have been dose and is susceptible to motion degradation and
constraints controlling the mean radiation dose to relies on surrogate tracking with fiducial mark-
the liver [17]. The Child–Pugh (CP) scoring sys- ers, as detailed below [23]. As a result of respira-
tem, a system used to assess hepatic function, has tory motion, gastric filling, peristalsis, bowel
been a helpful parameter in establishing radiation movements, and intrinsic deformation, abdomi-
dose constraints. In a 2014 study, an increase in nal organs and tumors are constantly moving and
CP score by at least 2 points within 3 months of changing position, which can lead to missing of
RT completion was determined to be RILD and the target and/or higher doses to organs at risk
shown to affect overall survival, with less sur- (OARs) [23]. Poor inter- and intra-fraction image
vival time and a higher relative risk of mortality guidance can compromise target and OAR visu-
[20]. It was determined that in order to minimize alization, potentially rendering the treatment less
such incidence, patients need to be carefully effective and more prone to normal tissue dam-
selected and given suitable treatment plans. In age. Traditionally, an approach used to partially
liver stereotactic body radiation therapy (SBRT), ameliorate this problem was the placement of
a classification score of 5 or 6 (class A) is desir- intra-lesion or peri-lesion fiducial markers to
able, as it indicates well-compensated disease allow for accurate image guidance before treat-
[21]. ment and for tracking the target during treatment
EBRT has been deemed to be relatively safe [22, 24]. Fiducials are small metal markers made
and conditionally recommended with additional of gold or platinum that are inserted in the tumor
treatments in both HCC and ICC for a variety of [25]. However, the problem with fiducials is that
differing severities, particularly those individuals they increase the risk of complications such as
who are not candidates for curative therapy [13]. bleeding, increase the cost of the treatment, and
SBRT delivers focal and intense doses of RT may migrate which renders treatment ineffective
(fraction size >500 cGy) in 1–5 fractions [13]; [22].
however, definitions may vary internationally.
Compared to low-dose whole liver radiation,
there is less destruction of surrounding normal MRgRT Advantages
tissue resulting in better outcomes and less toxic-
ity for the patient [13]. It can be used for both There are particular advantages to the magnetic
primary and metastatic malignancies and, at high resonance image-guided radiotherapy (MRgRT)
biological doses used with ablative intent (bio- system over traditional approaches. MRgRT has
logically effective doses BED ≥100 Gy10), is improved upon traditional CT-guided SBRT ther-
associated with improved local control in the apy and has been used in the abdomen with suc-
treatment of metastasis [22]. cess [16]. Broadly speaking, MRI does not
There are several challenges associated with subject patients to an increased radiation expo-
SBRT, and it can be especially difficult in patients sure that would be seen with CT. Given the supe-
with large, invasive tumors, patients with tumors rior soft tissue contrast of MR images, MRgRT
enables providers to directly visualize and local- with a gantry rotation maximum speed of 6.0 rpm
ize the tumor and surrounding organs and track [28]. A sample treatment plan using these sys-
the motion of these structures with the patient’s tems is shown below (Fig. 14.1).
breathing, without the need for fiducial markers Contouring is the process of outlining the tar-
[16]. In the liver specifically, organ-specific con- get tumor and OARs slice by slice on a simula-
trast agents such as gadoxetic acid allow for tion scan and thus is a critical step in the RT
improved differentiation of the liver, liver lesions, treatment plan [29]. A systematic review of con-
and the biliary tree [26]. Therefore, liver tumors touring guidelines published after 1995 demon-
are more easily differentiated on MRI, as are strated that characteristics of consensus
some liver OARs, such as the common bile duct contouring guidelines have not been described
and distal biliary tree. It also better allows for and that, of the 142 guidelines included in the
adaptive treatment planning and real-time imag- study, there could be major deviations in tumor
ing [26]. Additionally, MRgRT alleviates the target delineation, which can result in increased
need for multiphase contrast CT simulation scans toxicity and decreased patient survival [29]. An
in institutions that would otherwise opt to use added component is that, as of now, the majority
them [27]. of current guidelines give recommendations for
From published data, MRgRT appears to be a CT only. To ascertain optimal contouring guide-
safe alternative to traditional therapies. As delin- lines for MRgRT machines, further studies must
eated further in the Evaluation and Outcomes be done and a consensus among already available
sections below, MRgRT has shown promising guidelines must be reached. In an effort to stan-
clinical outcomes—patients are able to receive dardize OAR contouring, an international collab-
higher doses of radiation with a lower incidence orative group published a contouring atlas for
of severe toxicity to the liver and surrounding MR-based segmentation of OAR [26].
organs of interest. This has been associated with OARs with liver-directed therapy include the
high rates of local control and increased rates of normal liver, duodenum, small bowel, large
overall survival. However, further prospective bowel, stomach, esophagus, kidneys, spinal cord,
clinical trials are necessary to validate all biliary tree, large vessels, and heart [30]. Using
findings. MRI compared to CT adds superior soft tissue
contrast and organ/tumor motion information
that can help with positional information as well
RgRT Contouring, Planning Target
M as the target and delivery of the radiation [30]. As
Volumes, and Organs at Risk a result, planning target volumes (PTVs), mar-
gins encircling the clinical target, can be reliably
The two main MRgRT platforms commercially decreased due to increased confidence in posi-
available are the ViewRaw MRIdian system, tioning [31]. At UCLA, 3-mm gross target vol-
which uses a 0.35 Tesla MRI scanner, and the ume (GTV) to PTV margins are routinely
Elekta Unity system, which uses a 1.5 Tesla MRI employed. Additionally, gadoxetate can be used
scanner [28]. The main differences fall in con- during MRI-guided therapy as a contrast agent to
struction, imaging, gating, and treatment [28]. improve localization, tracking, and treatment of
The MRIdian is constructed using a split magnet the tumor [26]. MRI delineated target volumes
design, uses Trufi Sequence for imaging, has have been shown to be different—often smaller—
real-time tracking and automatic gating, and from those contoured with CT [31].
treats with a gantry rotation maximum speed of When available, respiratory gating by MRI is
0.5 rpm [28]. The Elekta Unity system conversely a technique used during radiation that can allow
is constructed using a single magnet design, uses for the adjustment of the radiation beam depend-
range of imaging sequence for imaging and has ing on breathing and the movement of body
real-time motion monitoring and introduced structures during a respiratory cycle [32]. It could
gated treatments in 2023. The Elekta Unity treats be advantageous for high-dose SBRT because it
Clinical Evaluaon and Eligibility

Paent is clinically evaluated and assessed for eligibility, including their ablity to hold their breath
for a minimum of 17 seconds
Simulaon Scan
Paent is immobilized for a reproducible setup and undergoes simulaon scan, both with free-breathing
and with a deep inspiratory breath hold.
Offline Countouring
GTV and OARs are manually contoured.
PTV Creaon
PTV is created by expanding GTV isotropically
Planning
Fig. 14.1 Sample workflow for the initial setup and creation of a baseline treatment plan
can allow reduction of margins (no internal target

volume necessary) and OAR exposure, thereby DVH Constraints
allowing for a more accurate dose delivery [33].
During treatment with a MR-Linac system, Dose-volume histograms (DVHs) summarize the
the GTV and PTV are defined, and a gating radiation distribution and are important in checking
boundary is defined. When the target structure adequate dose levels within a volume of interest in a
moves outside the boundary, the beam is auto- patient [37]. DVH constraints using MRgRT are
matically stopped. More work is needed to evalu- more accurate due to the nature of the system allow-
ate PTV margins in specific cases, particularly ing for adaptive planning based on the daily anat-
when asymmetric boundaries should be imple- omy of the patient [38]. Study protocols define
mented [34]. Visual examples of MRgRT are parameters that must be adhered to prevent any
located in the “Sample Patient Plans” below. unwanted effects of radiation [39]. OAR constraints
For the 1.5 T MR-Linac systems, ITV or mid- may vary from center to center or depending on the
position based systems to ensure sufficient target recommendations used. Please see Table 14.1 for an
volume coverage have been used in the past example for 5-fraction liver plan constraints [22].
before the implementation of automated gating At UCLA, the ideal normal liver constraint is
[35, 36]. >1000 cm3 of the normal liver receiving <15 Gy;
Table 14.1 Structure and OAR dose-volume constraints more compromised liver. It is important to note
[22]
that an increase of CP score within 6 months of
Structure Goal for three-fraction SBRT treatment completion was shown to be correlated
Liver with all-cause and liver-related mortality [40].
Single-lesion >1000 cm3 of normal liver receive
plan <15 Gy overall
Son et al. found that when the total liver volume
Multiple-lesion >800 cm3of normal liver receiving receiving <18 Gy was less than 800 cm3, the
plan <15 Gy overall probability of CP class progression was abruptly
OARs increased. As a result, they suggested that the
Spinal cord Dmax <3 Gy per fraction total liver volume receiving a dose less than
Esophagus D.5 cc <32 Gy
18 Gy should be >800 cm3 to reduce the risk to
Stomach V20 Gy <20 cm3
hepatic function [41].
Bowel bag V20 Gy <20 cm3
Bowel loop V35 Gy <0.35 cm3 Other limits should be established. Dose esca-
Kidney Dmax <3 Gy per fraction lation in the abdomen can only be done if the pro-
Skin D.05 cc <35 Gy vider can constantly localize and track the
anatomy of the patient—this is due to the con-
stant motion in the abdomen, from respiration
Table 14.2 UCLA SMART constraints
and peristalsis—as well as due to the large treat-
Structure Goal ment volumes present [42]. MRgRT provides
Liver Mean dose <20 Gy
certain advantages over traditional therapy
Keep 700 cc under 15 Gy
Duodenum V33 ≤0.5 cm3
regarding dose escalation. Due to its superior
Stomach V33 ≤0.5 cm3 imaging and smaller treatment volumes, it allows
Small bowel V33 ≤0.5 cm 3 for dose escalation to tumors while keeping the
Large bowel V33 ≤0.5 cm3 dose to normal tissue and surrounding OARs at a
Spinal canal V33 ≤0.5 cm3 minimum [43]. Doses are more accurately tai-
Kidney (each) Mean <12 Gy lored and restricted to appropriate locations [16].
2/3 of each kidney <8 Gy It also demonstrates an increased accuracy in
ascertaining the anatomical shifts of the abdomen
however, plans where 700 cm3 of the normal liver [16]. This can especially be a benefit if the target
receive <15 Gy are also accepted. For gastroin- is close to a sensitive OAR [44].
testinal mucosal structures (i.e., small bowel, In patients with pancreatic cancer, MRgRT
large bowel, esophagus, and stomach), the high- was shown to be safe for dose escalation, unlike
dose constraint is V35 Gy <0.5 cm3. Table 14.2 SBRT, which led to dose escalation but with an
has the current UCLA liver constraints. increase in toxicity [43]. Patients with higher
doses of radiation also had a significant survival
advantage over lower-dose groups, showing
Dose Escalation promise for the results of other cancer treatments
[43]. Patients with liver cancer could also poten-
While SBRT is generally well tolerated, there is a tially have an improved local control rate [44].
risk of toxicity to OARs, which has a historically Hoegen et al. [44] are running a prospective, ran-
limited dose escalation, especially in the case of domized, three-arm phase II clinical trial—
liver tumors in the proximity of luminal organs. dubbed the MAESTRO trial—which aims to
There have been many reports of damage to the provide evidence for the potential clinical benefit
GI organs in radiation treatments, and while of MRgRT for dose escalation, compared to
uncommon, potential toxicity is very important traditional SBRT, given the costs and lack of
to consider in treatment [16]. long-term data on MRgRT. In the first step of the
Higher volumes of radiation have linked with trial, they are planning on demonstrating the
increased toxicity, especially in patients with a safety and non-inferiority of MRgRT compared
Table 14.3 Dose constraint values for various thorax and tasis underwent daily MR-guided target localiza-
abdominal structures
tion and on-table treatment plan reoptimization,
Dose constraints (Gy) with GTVs and OARs adapted every day [38].
3 6 10 PTV coverage was modified to fit the new con-
Structure fractions fractions fractions
Uninvolved liver <19.2 <26 <32
straints and plans were then compared. The reop-
(=liver CTV) ≥700 cc timized plan showed improved PTV V100% and
Stomach D.5cc <22.2 <37 <43.5 V95% coverage in 47 and 45 of the applied frac-
Duodenum D.5cc <22.2 <37 <43.5 tions, respectively. PTV coverage increased by
Bowel D.5cc <25.2 <37 <43.5 4% in patients who had liver metastases close to
Sigmoid/rectum D.5cc <28.2 <37 <44 an OAR (<0.2 cm away), indicating a benefit to
Esophagus D.5cc <25.2 <36 <43.5
such patients. For metastasis distant from an
Heart D.5cc <25.0 <31.5 <66.0
Kidney mean dose <8.5 <10.8 <12.0 OAR, PTV V100% coverage was not significantly
Spinal cord D.1cc <21.6 <29.0 <35 improved, indicating that for such patients, daily
adaptive planning is not necessary. This is also
supported by a report by Uder et al. who showed
to standard SBRT, and in the second step, they a local control of irradiated liver metastasis after
are exploring the possibility of dose escalation in 12 months of 89.6% using the “adapt to position”
critically located hepatic metastasis, where stan- workflow on the 1.5 T MR-Linac [46]. Using the
dard SBRT would not be as achievable [44]. “adapt to position” workflow on the 1.5 T
Dose constraints for the OARs depend on the MR-Linac, the target volume and OARs are not
number of fractions being delivered. For an unin- resegmented and a virtual shift is performed. In
volved liver, for example, the dose constraint is contrast to this, in the “adapt to shape” workflow,
<19.2 Gy given in 3 fractions or <32 Gy given in a full reoptimization with new segmented struc-
10 fractions. The values are shown in Table 14.3, tures does take place [47].
adapted from the table published in Hoegen [44]. The process of on-table adaptation is as fol-
CTV will be GTV + 5 mm and PTV will be lows [22, 38, 47, 48]:
CTV + 3 mm. The dose escalation portion of the
study will be in study arm C and has not been 1. Creation of a robust “baseline” plan using the
delineated yet. Furthermore, at the time of this dedicated planning system.
chapter, this study has yet to be completed and, as 2. Patient arriving for treatment is positioned in
a result, has not drawn any conclusions [44]. the treatment position and MR images are
acquired during a breath-hold scan.
3. Newly acquired MR images are “matched” to
On-Table Adaptation the original baseline images and any neces-
sary shifts are performed. A decision regard-
In on-table adaptive radiotherapy (On-ART), an ing the necessity for generating a new online
initial radiation plan for a patient is created offline adapted plan when inter-fractional anatomical
and is then adapted for each fraction depending changes are clinically significant is made in
on the relationship between the target and OARs real time.
[45]. A sample flow chart for the process is shown 4. If the decision is made to adapt, the target(s)
below (Fig. 14.2). In the liver, On-ART with and OARs are recontoured or contours are
MRgRT improves PTV coverage in 63% of the adjusted based on the pretreatment plan to
applied fractions, a value consistent with previous account for the patient’s anatomy of the day.
studies [38]. In Mayinger’s study, 15 patients 5. The plan is delivered using respiratory gating
assigned to SBRT for oligometastatic liver metas- with breath hold if necessary.
Fig. 14.2 An example

flowchart for on-table
adaptation
Evaluation ticularly in the liver. The biggest one is that it

allows for superior imaging of liver structures
MRgRT has shown itself to be a promising addi- and is well suited to soft tissues and organs,
tion to RT that can be used for a variety of malig- unlike CT [33, 50]. Providers can directly visual-
nancies, including cancers of the breast, lung, ize the tumor and anatomy of the patient, allow-
prostate, and liver [49]. There are many advan- ing for better tracking of the tumor during
tages of MRgRT over traditional therapies, par- radiation [33]. Abdominal organs are constantly
fluctuating due to breaths, gastric filling, and in patients spending approximately 10 min lon-
bowel movements—MRgRT can track these ger in the treatment vault per fraction, which can
movements and allow better plan adaptation dur- result in other concerns for patient scheduling
ing treatment, although it is still unclear whether and waiting list management [38, 51].
that improved dosimetry translates into superior MRgRT is also more resource-intensive,
outcomes [38, 49]. There is no doubting, how- requiring increased clinician education to inter-
ever, that plan adaptation does present the oppor- pret images and more teams trained to operate the
tunity for the patient to have a more personalized systems and be mindful of the safety concerns—
treatment [51]. MRgRT systems are currently operated by physi-
The increased accuracy, along with gating that cians, radiation therapists, dosimetrists, and
automatically stops the radiation beam when a physicists in multidisciplinary teams operating
target structure moves outside the boundary, also simultaneously [56].
allows for smaller PTVs and less OAR exposure,
thus sparing normal healthy tissues [31, 33, 52].
In a study analyzing 20 consecutive liver SABR Outcome
patients, it was found that implementing tech-
niques to compensate the movement of tumors Table 14.4 details current evidence evaluating
and OARs resulted in a reduction in PTV size and clinical outcomes for MRgRT in primary hepato-
ability to escalate dose. The observed increase in biliary tumors and/or liver metastases. One of the
BED was up to 21.4 Gy10 for a 5-fraction sched- first studies delineating the MRgRT system was
ule and 37.5 Gy10 for a 3-fraction prescription in 2008 at the University Medical Center Utrecht,
schedule, corresponding to TCP gains of 5.7% which aimed to use this prototype system to
and 7.8%, respectively [53]. deliver radiation doses with millimeter precision,
The disadvantages of MRI treatment start with based on diagnostic quality 1.5 T MRI images
contraindications to getting an MRI. Non-MR- [55]. MRgRT has been clinically implemented
compatible cardiac implantable electronic from 2014, and since then, it has been used for
devices, metallic intraocular foreign bodies, numerous cancers, from head and neck cancers to
implantable neurostimulation systems, cochlear abdominal cancers to ovarian cancers [57].
implants/ear implants, drug infusion pumps, Recently, there has been more clinical studies
catheters with metallic components, metallic done for abdominal tumors, and it has been
fragments, cerebral artery aneurysm clips, mag- shown that such tumors are responsive to this
netic dental implants, tissue expanders, artificial treatment.
limbs, and hearing aids are absolute contraindica- A large retrospective study was done by
tions [54]. Other limitations include geometric Henke and colleagues at Washington University
distortion, electron density disruption, and sus- School of Medicine in 2018. There were 642
ceptibility artifacts [49]. Geometric distortions patients recruited between January 2014 and
can cause inaccurate target volumes and mis- March 2018, 275 of which were in the abdomen
placed radiation doses, which would limit the with 87 being in the hepatobiliary system [57].
effectiveness of the treatment and even poten- One of the studies that was analyzed was a pro-
tially harm the patient [51]. spective phase I trial, also done by Henke and
The added cost of MRI treatment has to be colleagues in 2017. Twenty patients with unre-
considered as well. One study found that the sectable primary liver and non-liver abdominal
direct costs associated with prostate SBRT were malignancies were recruited and received a
almost $1500 higher with MRgRT than with median radiation dose of 50 Gy in five treat-
CTgRT [32]. MRI is also an intrinsically slow ments. Within 1 year, overall survival (OS) was
imaging modality, taking longer than radiogra- 75%, and 15-month local control (LC) of the can-
phy to acquire images [55]. That, in addition to cer was 90%. Through this study, it was con-
respiratory gating and on-table adaptation, results cluded that SMART therapy was clinically
Table 14.4 The prospective and retrospective evidence evaluating clinical outcomes for MRgRT in primary hepatobiliary tumors and/or liver metastases
266
Number Late toxicity

Study of Median prescribed Median fraction Overall (non-
references Histology Year Design patients dose number survival (OS) Local control (LC) hematologic)
[22] Primary (HCC or ICC) and 2022 Prospective 20 54 (range: 11.5–60) 3 (range: 1–5) 2 years; 1 year; 94.7% 0 grade ≥3, 1
secondary liver (1–3 phase I 50.7% 2 years; 79.6% late grade 3/4
metastatic lesions) tumors toxicity
[23] Abdominal and pelvic 2021 Retrospective 106 40 Gy 5 2 years; 57% 2 years; 74% 1 grade ≥3
tumors (primary, locally 96% for lesions that 0 grade ≥4
recurrent, oligometastatic) were treated with 7 late
BED10 ≥100 versus toxicities
69% for BED10 <100
[36] Hepatic metastasis 2021 Prospective 10 38.5 Gy 3 or 5 N/A N/A 0 grade ≥2
[42] Unresectable malignancies 2021 Prospective 25 35, 40 5 N/A N/A 0 grade ≥3
in the upper abdomen
(primary and secondary)
[43] Hepatic metastasis 2022 Phase II 82 19–32 Gy <10 1 year; 84% 1 year; 88.1% N/A
[57] Unresectable primary liver 2017 Phase I 20 50 Gy 5 1 year; 75% 15 months; 90% 0 grade ≥3
and non-liver abdominal
malignancies
[58] Unresectable locally 2020 Prospective 17 40 Gy 5 1 year; 76% 1 year; 85.6% 1 grade ≥3
advanced 2 years; 2 years; 73.3% 1 late toxicity
cholangiocarcinoma 46.1%
[59] Hepatic metastasis or HCC 2021 Prospective 20 50 Gy (range: 8 (range: 3–12) 1 year; 84% 1 year; 88.1% 0 grade ≥3
45–60 Gy)
[60] Hepatic primary and 2019 Prospective 29 45–50 Gy in 5 fractions 5 or 3 N/A N/A N/A
metastatic tumors for 27 patients;
27–39 Gy in 2 fractions
for 3 patients
[61] Primary (HCC or ICC) and 2019 Retrospective 26 50 Gy (range: 30–60) 5 1 year; 69% 21.2 months; 80.4% 2 grade ≥3
secondary liver tumors 2 years; 60% 0 grade ≥4
[62] Liver metastatic lesions, 2020 Prospective 10 30–60 Gy 3–5 N/A N/A 0 grade ≥3, 1
HCC late grade 3
toxicity
[63] Hepatic and pancreatic 2022 Prospective 16 (11 27.50 Gy and 45 Gy in 5 (except 2 N/A N/A N/A
cancers liver) 5 fractions or 16 Gy liver patients
and 24 Gy in single who received
fractions 1)
N. Karimi-Mostowfi et al.
[64] HCC 2021 Retrospective 10 50 Gy 5 80% (at time 90% (at time of 0 grade ≥3
of analysis) analysis)
feasible, and ablative radiation doses were able to A trial at Heidelberg University Hospital in
be delivered with low rates of toxicity [65]. Germany by Hoegen et al. [44] treated 20
In 2021, a study by van Dams et al. [22] at the patients, 18 with liver metastasis and 2 HCC with
University of California, Los Angeles (UCLA) ablative MR-guided SBRT between January
involved 23 patients with primary or metastatic 2019 and February 2020. At 1 year, LC was
hepatic cancers who underwent liver-directed 88.1% and OS was 84.0%. Grade 1 toxicity
SBRT from 2015 to 2017. Of these 23, 20 under- occurred in 30.0% and grade 2 in 5.0% of the
went MR-guided SBRT and received a median patients. No patient had grade >3 toxicity. The
radiation dose of 54 Gy in three treatments. These treatment was overall rated positively, and it was
patients demonstrated an estimated 2-year LC of concluded that MRgRT is a well-tolerated and a
80%, 2-year OS of 50.7%, and no acute grade ≥3 well-accepted treatment modality [59]. A phase
toxicities [22]. A difference in LC between single II study at the same hospital recruited 82 patients
and multiple lesions or BED ≥100 Gy10 and with hepatic metastasis, delivering a range of
BED <100 Gy10, respectively, was observed. doses (between 19 and 32) in less than 10 frac-
Luterstein et al. [58] at UCLA reached a similar tions [44]. This study was simply to demonstrate
result, but with 17 patients with unresectable MRgRT safety, particularly regarding dose esca-
locally advanced cholangiocarcinoma treated lation. Further studies will analyze the LC, OS,
between May 2015 and August 2017. These and toxicity. Similar work has been published
patients received a median radiation dose of without delineating LC, OS, and toxicity.
40 Gy in five treatments, with a 2-year LC of Feldman et al. for example, evaluated a treatment
73.3% and a 2-year OS of 46.1%. One patient plan of 45–50 Gy in five fractions and were able
experienced an acute grade >3 toxicity and one to assess the use of the gated breathing technique
patient experience a late toxicity. on MRI [60]. One-year posttreatment confirmed
A retrospective study at UCLA by Yoon et al. either stable or decreased size of treated lesions
[23] involved 106 patients with abdominal and in all but one patient and toxicities were mild
pelvic tumors treated from November 2014 to [60].
August 2017. The median radiation dose was
40 Gy in five treatments, and the patients demon-
strated an overall 2-year LC of 74%, with a Sample Patient Plans
breakdown of 96% for lesions that were treated
with BED10 ≥100 versus 69% for BED10 <100, Figures 14.3 and 14.4 demonstrate sample patient
with a 2-year OS of 57% [23]. There was 1 plans in a patient with metastatic anal cancer and
patient who experienced an acute grade ≥3 toxic- HCC, respectively. As was discussed previously,
ity and 7 patients who experienced late grade 3 the MR system allows for smaller targets to be
and 4 toxicities. All 3 UCLA studies concluded identified, for OARs to be delineated, and for
that MRgRT is viable and safe and demonstrates plans to be adapted on-table in order to deliver
diminished toxicity with high rates of LC and OS the maximal dose to the target while keeping the
[22, 23, 58]. surrounding dose to a minimum.
Fig. 14.3 Patient with metastatic anal cancer to the liver. The target is adjacent to bowel (yellow) and liver (orange)—
the OARs in this patient
Fig. 14.4 Patient with HCC status post ablation with recurrence within the ablation zone. Target is adjacent to the
esophagus (light blue) and heart (orange)
Summary 2021;7(1):1–28. https://doi.org/10.1038/

s41572-020-00240-3.
4. Gupta A, Dixon E. Epidemiology and risk factors:
RT is a less invasive alternative to surgery and is intrahepatic cholangiocarcinoma. Hepatobiliary Surg
possible for patients who are poor candidates for Nutr. 2017;6(2):101–4. https://doi.org/10.21037/
surgery. MRgRT is a promising area of treatment hbsn.2017.01.02.
5. McGlynn KA, Petrick JL, El-Serag HB. Epidemiology
for liver malignancies, which primarily include of hepatocellular carcinoma. Hepatology.
HCC and ICC. CT-guided RT has traditionally 2021;73(S1):4–13. https://doi.org/10.1002/
been the mainstay, but it is an especially difficult hep.31288.
treatment modality in patients with large tumors, 6. Ali H, Tedder B, Waqar SH, Mohamed R, Cate EL,
Ali E. Changing incidence and survival of intrahe-
patients with tumors close to other organs, and patic cholangiocarcinoma based on surveillance,
patients with tumors too small to be properly tar- epidemiology, and end results database (2000–2017).
geted. There is also the problem with shifting Ann Hepato Biliary Pancreat Surg. 2022;26:235.
abdominal organs, which can lead to missing of https://doi.org/10.14701/ahbps.21-173.
7. Rodrigues PM, Vogel A, Arrese M, Balderramo
the target and/or inadvertently giving higher DC, Valle JW, Banales JM. Next-generation bio-
doses to nearby organs. MRgRT has improved markers for cholangiocarcinoma. Cancers (Basel).
upon these flaws, allowing physicians to directly 2021;13(13):3222. https://doi.org/10.3390/
visualize the tumor and surrounding organs with cancers13133222.
8. Muscari F, Maulat C. Preoperative alpha-fetoprotein
higher-quality imaging that can track the move- (AFP) in hepatocellular carcinoma (HCC): is
ment of the internal structures. On-table adapta- this 50-year biomarker still up-to-date? Transl
tion is particularly valuable in the delivery of Gastroenterol Hepatol. 2020;5:46. https://doi.
radiation. With MRgRT, targets can be optimized org/10.21037/tgh.2019.12.09.
9. Heimbach JK, Kulik LM, Finn RS, et al. AASLD
to the “anatomy of the day,” making it possible to guidelines for the treatment of hepatocellular carci-
deliver higher, tumoricidal doses of radiation to noma. Hepatology. 2018;67(1):358–80. https://doi.
patients, with minimal toxicity to OARs, which org/10.1002/hep.29086.
has been correlated with improvement in tumor 10. American College of Radiology Committee on
LI-RADS® (Liver). CT/MRI LI-RADS® v2018
control and an increase in survival for unresect- CORE. Accessed 28 Mar 2023.
able liver tumors [33, 49]. While all studies men- 11. Buettner S, van Vugt JL, IJzermans JN, Groot
tioned showed varying levels of success, they all KB. Intrahepatic cholangiocarcinoma: current per-
agree that future prospective clinical trials to vali- spectives. Onco Targets Ther. 2017;10:1131–42.
https://doi.org/10.2147/OTT.S93629.
date their findings are warranted [23]. Future 12. Tsilimigras DI, Brodt P, Clavien PA, et al. Liver
studies additionally have to evaluate MRgRT metastases. Nat Rev Dis Primers. 2021;7(1):1–23.
considering cost, effectiveness, posttreatment https://doi.org/10.1038/s41572-021-00261-6.
quality of life, and long-term outcomes. As of 13. Apisarnthanarax S, Barry A, Cao M, et al. External
beam radiation therapy for primary liver cancers:
now on the NIH Clinical Trials website, there are an ASTRO clinical practice guideline. Pract Radiat
5 actively recruiting studies and 3 completed Oncol. 2022;12(1):28–51. https://doi.org/10.1016/j.
study to assess MRgRT on hepatic cancers. prro.2021.09.004.
14. Kim J, Jung Y. Radiation-induced liver disease: cur-
rent understanding and future perspectives. Exp
Mol Med. 2017;49(7):e359. https://doi.org/10.1038/
References emm.2017.85.
15. Goodman KA, Kavanagh BD. Stereotactic body
1. Singal AG, Lampertico P, Nahon P. Epidemiology radiotherapy for liver metastases. Semin Radiat
and surveillance for hepatocellular carcinoma: new Oncol. 2017;27(3):240–6. https://doi.org/10.1016/j.
trends. J Hepatol. 2020;72(2):250–61. https://doi. semradonc.2017.02.004.
org/10.1016/j.jhep.2019.08.025. 16. Witt JS, Rosenberg SA, Bassetti MF. MRI-guided
2. Siegel RL, Miller KD, Fuchs HE, Jemal A. Cancer adaptive radiotherapy for liver tumours: visualising
statistics, 2022. CA Cancer J Clin. 2022;72(1):7–33. the future. Lancet Oncol. 2020;21(2):e74–82. https://
https://doi.org/10.3322/caac.21708. doi.org/10.1016/S1470-2045(20)30034-6.
3. Llovet JM, Kelley RK, Villanueva A, et al. 17. Pan CC, Kavanagh BD, Dawson LA, et al. Radiation-
Hepatocellular carcinoma. Nat Rev Dis Primers. associated liver injury. Int J Radiat Oncol Biol Phys.
2010;76(3):S94–100. https://doi.org/10.1016/j. Phys. 2020;107(4):827–35. https://doi.org/10.1016/j.

ijrobp.2009.06.092. ijrobp.2020.04.011.
18. Lo EC, Rucker AN, Federle MP. Hepatocellular 30. Chandarana H, Wang H, Tijssen RHN, Das
carcinoma and intrahepatic cholangiocarcinoma: IJ. Emerging role of MRI in radiation therapy. J Magn
imaging for diagnosis, tumor response to treat- Reson Imaging. 2018;48(6):1468–78. https://doi.
ment and liver response to radiation. Semin Radiat org/10.1002/jmri.26271.
Oncol. 2018;28(4):267–76. https://doi.org/10.1016/j. 31. Hunt A, Hansen VN, Oelfke U, Nill S, Hafeez
semradonc.2018.06.010. S. Adaptive radiotherapy enabled by MRI guid-
19. Toesca DAS, Ibragimov B, Koong AJ, Xing L, Koong ance. Clin Oncol. 2018;30(11):711–9. https://doi.
AC, Chang DT. Strategies for prediction and mitiga- org/10.1016/j.clon.2018.08.001.
tion of radiation-induced liver toxicity. J Radiat Res. 32. Respiratory Gating | OncoLink. https://www.
2018;59(Suppl 1):i40–9. https://doi.org/10.1093/jrr/ oncolink.org/cancer-t reatment/radiation/support/
rrx104. respiratory-gating. Accessed 10 July 2022.
20. Son SH, Jang HS, Jo IY, et al. Significance of an 33. Corradini S, Alongi F, Andratschke N, et al.
increase in the Child–Pugh score after radiotherapy MR-guidance in clinical reality: current treat-
in patients with unresectable hepatocellular car- ment challenges and future perspectives. Radiat
cinoma. Radiat Oncol. 2014;9:101. https://doi. Oncol. 2019;14(1):92. https://doi.org/10.1186/
org/10.1186/1748-717X-9-101. s13014-019-1308-y.
21. Tanaka O, Kojima T, Ohbora A, et al. Scores of 34. Green OL, Rankine LJ, Cai B, et al. First clinical imple-
Child–Pugh classification impact overall survival mentation of real-time, real anatomy tracking and
after stereotactic body radiation therapy for primary radiation beam control. Med Phys. 2018;45(8):3728–
and metastatic liver tumors. J Clin Exp Hepatol. 40. https://doi.org/10.1002/mp.13002.
2020;10(2):101–5. https://doi.org/10.1016/j. 35. van de Lindt TN, Nowee ME, Janssen T, et al.
jceh.2019.05.002. Technical feasibility and clinical evaluation of
22. van Dams R, Wu TC, Kishan AU, et al. Ablative 4D-MRI guided liver SBRT on the MR-linac.
radiotherapy for liver tumors using stereotactic Radiother Oncol. 2022;167:285–91. https://doi.
MRI-guidance: a prospective phase I trial. Radiother org/10.1016/j.radonc.2022.01.009.
Oncol. 2022;170:14–20. https://doi.org/10.1016/j. 36. Gani C, Boeke S, McNair H, et al. Marker-less online
radonc.2021.06.005. MR-guided stereotactic body radiotherapy of liver
23. Yoon SM, Luterstein E, Chu FI, et al. Clinical out- metastases at a 1.5 T MR-Linac—feasibility, work-
comes of stereotactic magnetic resonance image- flow data and patient acceptance. ctRO. 2021;26:55–
guided adaptive radiotherapy for primary and 61. https://doi.org/10.1016/j.ctro.2020.11.014.
metastatic tumors in the abdomen and pelvis. Cancer 37. Drzymala RE, Mohan R, Brewster L, et al.
Med. 2021;10(17):5897–906. https://doi.org/10.1002/ Dose-volume histograms. Int J Radiat Oncol
cam4.4139. Biol Phys. 1991;21(1):71–8. https://doi.
24. Parikh NR, Lee PP, Raman SS, et al. Time-driven org/10.1016/0360-3016(91)90168-4.
activity-based costing comparison of CT-guided 38. Mayinger M, Ludwig R, Christ SM, et al. Benefit
versus MR-guided SBRT. JCO Oncol Pract. of replanning in MR-guided online adaptive radia-
2020;16(11):e1378–85. https://doi.org/10.1200/ tion therapy in the treatment of liver metastasis.
JOP.19.00605. Radiat Oncol. 2021;16(1):84. https://doi.org/10.1186/
25. Knoll MA, Jagsi R, Rosenzweig K. Focal radiation s13014-021-01813-6.
therapy for cancer. JAMA Oncol. 2019;5(3):442. 39. Panchal A, Zhang Y, Ruffer J. A dose volume histo-
https://doi.org/10.1001/jamaoncol.2018.5868. gram (DVH) constraint analysis system for clinical
26. Lukovic J, Henke L, Gani C, et al. MRI-based upper protocol treatment plan evaluation using DICOM
abdominal organs-at-risk atlas for radiation oncology. RT. Int J Radiat Oncol Biol Phys. 2010;78(3):S754.
Int J Radiat Oncol Biol Phys. 2020;106(4):743–53. https://doi.org/10.1016/j.ijrobp.2010.07.1746.
https://doi.org/10.1016/j.ijrobp.2019.12.003. 40. Lee P, Ma Y, Zacharias I, et al. Stereotactic body radi-
27. Romesser PB, Tyagi N, Crane CH. Magnetic resonance ation therapy for hepatocellular carcinoma in patients
imaging-guided adaptive radiotherapy for colorectal with Child–Pugh B or C cirrhosis. Adv Radiat
liver metastases. Cancers (Basel). 2021;13(7):1636. Oncol. 2020;5(5):889–96. https://doi.org/10.1016/j.
https://doi.org/10.3390/cancers13071636. adro.2020.01.009.
28. Ng J, Gregucci F, Pennell RT, et al. MRI-LINAC: 41. Son SH, Choi BO, Ryu MR, et al. Stereotactic body
a transformative technology in radiation oncol- radiotherapy for patients with unresectable primary
ogy. Front Oncol. 2023;13:1117874. https://doi. hepatocellular carcinoma: dose-volumetric param-
org/10.3389/fonc.2023.1117874. eters predicting the hepatic complication. Int J Radiat
29. Lin D, Lapen K, Sherer MV, et al. A systematic review Oncol Biol Phys. 2010;78(4):1073–80. https://doi.
of contouring guidelines in radiation oncology: analy- org/10.1016/j.ijrobp.2009.09.009.
sis of frequency, methodology, and delivery of con- 42. Daamen LA, de Mol van Otterloo SR, van Goor
sensus recommendations. Int J Radiat Oncol Biol IWJM, et al. Online adaptive MR-guided stereotac-
tic radiotherapy for unresectable malignancies in the 55. Lagendijk JJW, Raaymakers BW, Raaijmakers
upper abdomen using a 1.5T MR-linac. Acta Oncol. AJE, et al. MRI/linac integration. Radiother
2022;61(1):111–5. https://doi.org/10.1080/02841 Oncol. 2008;86(1):25–9. https://doi.org/10.1016/j.
86X.2021.2012593. radonc.2007.10.034.
43. Rudra S, Jiang N, Rosenberg SA, et al. Using adaptive 56. Chin S, Eccles CL, McWilliam A, et al. Magnetic
magnetic resonance image-guided radiation therapy resonance-guided radiation therapy: a review. J Med
for treatment of inoperable pancreatic cancer. Cancer Imaging Radiat Oncol. 2020;64(1):163–77. https://
Med. 2019;8(5):2123–32. https://doi.org/10.1002/ doi.org/10.1111/1754-9485.12968.
cam4.2100. 57. Henke LE, Contreras JA, Green OL, et al. Magnetic
44. Hoegen P, Zhang KS, Tonndorf-Martini E, et al. resonance image-guided radiotherapy (MRIgRT):
MR-guided adaptive versus ITV-based stereo- a 4.5-year clinical experience. Clin Oncol.
tactic body radiotherapy for hepatic metastases 2018;30(11):720–7. https://doi.org/10.1016/j.
(MAESTRO): a randomized controlled phase II trial. clon.2018.08.010.
Radiat Oncol. 2022;17(1):59. https://doi.org/10.1186/ 58. Luterstein E, Cao M, Lamb JM, et al. Clinical
s13014-022-02033-2. outcomes using magnetic resonance-guided ste-
45. Finazzi T, Palacios MA, Spoelstra FOB, et al. Role reotactic body radiation therapy in patients with
of on-table plan adaptation in MR-guided ablative locally advanced cholangiocarcinoma. Adv Radiat
radiation therapy for central lung tumors. Int J Radiat Oncol. 2020;5(2):189–95. https://doi.org/10.1016/j.
Oncol Biol Phys. 2019;104(4):933–41. https://doi. adro.2019.09.008.
org/10.1016/j.ijrobp.2019.03.035. 59. Weykamp F, Hoegen P, Klüter S, et al. Magnetic
46. Uder L, Nachbar M, Butzer S, et al. Local control and resonance-guided stereotactic body radiotherapy of
patient reported outcomes after online MR guided ste- liver tumors: initial clinical experience and patient-
reotactic body radiotherapy of liver metastases. Front reported outcomes. Front Oncol. 2021;11:610637.
Oncol. 2023;12:1095633. https://doi.org/10.3389/ https://doi.org/10.3389/fonc.2021.610637.
fonc.2022.1095633. 60. Feldman AM, Modh A, Glide-Hurst C, Chetty IJ,
47. Winkel D, Bol GH, Kroon PS, et al. Adaptive radio- Movsas B. Real-time magnetic resonance-guided
therapy: the Elekta Unity MR-linac concept. Clin liver stereotactic body radiation therapy: an institu-
Transl Radiat Oncol. 2019;18:54–9. https://doi. tional report using a magnetic resonance-linac system.
org/10.1016/j.ctro.2019.04.001. Cureus. 2019;11(9):e5774. https://doi.org/10.7759/
48. Bohoudi O, Bruynzeel AME, Senan S, et al. cureus.5774.
Fast and robust online adaptive planning in ste- 61. Rosenberg SA, Henke LE, Shaverdian N, et al. A
reotactic MR-guided adaptive radiation therapy multi-institutional experience of MR-guided liver
(SMART) for pancreatic cancer. Radiother Oncol. stereotactic body radiation therapy. Adv Radiat
2017;125(3):439–44. https://doi.org/10.1016/j. Oncol. 2019;4(1):142–9. https://doi.org/10.1016/j.
radonc.2017.07.028. adro.2018.08.005.
49. Tchelebi LT, Zaorsky NG, Rosenberg J, Latifi K, 62. Hal WA, Straza MW, Chen X, et al. Initial clinical
Hoffe S. Integrating MR-guided radiation therapy experience of stereotactic body radiation therapy
into clinical practice: clinical advantages and practi- (SBRT) for liver metastases, primary liver malignancy,
cal limitations. Lung Cancer. 2021;9:1289. and pancreatic cancer with 4D-MRI based online
50. Vu LN, Morelli JN, Szklaruk J. Basic MRI for adaptation and real-time MRI monitoring using a 1.5
the liver oncologists and surgeons. J Hepatocell Tesla MR-Linac. PLoS One. 2020;15(8):e0236570.
Carcinoma. 2018;5:37–50. https://doi.org/10.2147/ https://doi.org/10.1371/journal.pone.0236570.
JHC.S154321. 63. Stanescu T, Shessel A, Carpino-Rocca C, et al. MRI-
51. Keall PJ, Brighi C, Glide-Hurst C, et al. Integrated guided online adaptive stereotactic body radiation
MRI-guided radiotherapy—opportunities and chal- therapy of liver and pancreas tumors on an MR-Linac
lenges. Nat Rev Clin Oncol. 2022;19(7):458–70. system. Cancers. 2022;14(3):716. https://doi.
https://doi.org/10.1038/s41571-022-00631-3. org/10.3390/cancers14030716.
52. Boldrini L, Corradini S, Gani C, et al. MR-guided 64. Boldrini L, Romano A, Mariani S, et al. MRI-
radiotherapy for liver malignancies. Front guided stereotactic radiation therapy for
Oncol. 2021;11:616027. https://doi.org/10.3389/ hepatocellular carcinoma: a feasible and safe inno-
fonc.2021.616027. vative treatment approach. J Cancer Res Clin Oncol.
53. Gargett M, Haddad C, Kneebone A, Booth 2021;147(7):2057–68. https://doi.org/10.1007/
JT, Hardcastle N. Clinical impact of removing s00432-020-03480-8.
respiratory motion during liver SABR. Radiat 65. Henke L, Kashani R, Robinson C, et al. Phase I trial
Oncol. 2019;14(1):93. https://doi.org/10.1186/ of stereotactic MR-guided online adaptive radiation
s13014-019-1300-6. therapy (SMART) for the treatment of oligometastatic
54. Ghadimi M, Sapra A. Magnetic resonance imaging or unresectable primary malignancies of the abdo-
contraindications. StatPearls Publishing; 2022. http:// men. Radiother Oncol. 2018;126(3):519–26. https://
www.ncbi.nlm.nih.gov/books/NBK551669/ doi.org/10.1016/j.radonc.2017.11.032.
Magnetic Resonance
Imaging-Guided Radiation
15
Therapy: Advances
and Opportunities for Thoracic
Malignancies
Julius Weng, Jinzhong Yang, Raymond H. Mak,

Mohamed Abazeed, and Percy Lee
Disease Overview static disease, demonstrating an improvement in

overall survival in phase II trials [3–5]. For
Non-small cell lung cancer (NSCLC) is the lead- tumors located away from the proximal bronchi
ing cause of cancer-related death in both men and and mediastinum (i.e., in the lung periphery),
women in the United States [1]. Stereotactic SBRT has demonstrated excellent local tumor
body radiation therapy (SBRT) is a treatment control with limited toxicity [6–8]. However, for
modality that delivers high doses of conformal centrally located tumors that approximate the
radiation and is considered standard of care for proximal bronchial tree, SBRT has been limited
early-stage NSCLC patients who are medically by significant toxicity due to complications from
inoperable or refuse surgery [2]. SBRT is also adjacent organs at risk (OARs), including the
increasingly utilized for treatment of oligometa- proximal bronchial tree, esophagus, heart/peri-
cardium, and great vessels [9–11]. SBRT for cen-
tral lung tumors can result in 2-year rates of
J. Weng · J. Yang
Division of Radiation Oncology, The University of
significant toxicity as high as 46%, including
Texas MD Anderson Cancer Center, treatment-related deaths [9–11]. As a result, cen-
Houston, TX, USA tral lung tumors are typically treated with more
e-mail: [email protected]; JYang4@ fractionated regimens (8–15 fx) [12, 13].
mdanderson.org
RTOG 0813 was a dose escalation study to
R. H. Mak determine the maximum tolerated dose (MTD) of
Department of Radiation Oncology, Brigham and
Women’s Hospital, Dana-Farber Cancer Institute,
SBRT for central tumors and included 120
Harvard Medical School, Boston, MA, USA patients [14]. Although the highest dose level of
e-mail: [email protected] 12 Gy in 5 fractions SBRT schedule was reached,
M. Abazeed this was associated with a high rate of severe
Department of Radiation Oncology, Northwestern treatment-related toxicity (21% grade ≥3, 7%
University Feinberg School of Medicine, Chicago, grade 5). In comparison, RTOG 0236 evaluated
IL, USA
SBRT for 59 patients with early-stage peripheral
lung tumors and had no grade 5 toxicities [15].
P. Lee (*)
Department of Radiation Oncology, City of Hope
The risk of SBRT may be higher among patients
National Medical Center, Los Angeles, CA, USA with ultracentral lesions, which have either gross
e-mail: [email protected] tumor volume (GTV) or planning treatment vol-
274 J. Weng et al.
ume (PTV) abutting critical mediastinal OARs CE mark and is used clinically in Europe, but
[16]. The phase 2 HILUS trial studied SBRT only recently received FDA approval.
(56 Gy/8 fx) in patients with ultracentral lesions There are several theoretical advantages of
defined as ≤1 cm from the proximal bronchial MRgRT that can increase the therapeutic ratio of
tree [17]. HILUS similarly demonstrated a high radiation, especially for central tumors adjacent
probability of significant treatment-related toxic- to multiple OARs, high-risk patients undergoing
ity (34% grade ≥3, 15% grade 5). Although re-irradiation, or patients with pulmonary comor-
HILUS utilized a more fractionated SBRT bidities [21, 22]. These advantages address sev-
regimen than RTOG 0813, HILUS exceeded
eral of the limitations of CT-based SBRT
standard OAR constraint doses for small/large including online adaptation for inter-fraction
bronchi and permitted hot spots of 150%. There anatomical changes, intra-fraction motion man-
are also several limitations of CT-based SBRT agement, soft tissue visualization, and margin
that may have contributed to the high rates of tox- reduction. In this chapter, we discuss the poten-
icity observed on these studies. To account for tial benefit, clinical workflow, current limitations,
respiratory and cardiac motion, large internal tar- and future directions of MRgRT for thoracic
get volumes (ITVs) and planning treatment vol- tumors.
umes (PTVs) are often utilized. For central
tumors, these large treatment volumes may over-
lap with OARs and, thus, necessitate the use of Overcoming Challenges
very high dose gradients near critical structures. of Traditional Approaches
However, inter- and intra-fraction tumor motion
can be highly variable, and the baseline planned A significant advantage of MRgRT is the capabil-
radiation treatment may not reflect the true dose ity for online adaptive radiotherapy, which is the
accumulated by critical OARs over the course of process of optimizing the initial radiation treat-
treatment. ment plan based upon “anatomy of the day”
Magnetic resonance imaging-guided radia- obtained from daily onboard MRI. Given the
tion therapy (MRgRT) is a recent technological close proximity of central tumors to critical struc-
innovation integrating a linear accelerator tures, radiation plans must incorporate a sharp
within an MRI machine. The two commercially dose falloff to ensure that OARs meet dose con-
available MR-linear accelerators are the straints. Even small changes in the position of
MRIdian (ViewRay Inc., USA) and Unity tumor or OARs between fractions may result in
(Elekta, Sweden) systems [18]. The first itera- underdosing of tumor or excessive dose to OARs.
tion of the MRIdian utilized a low-field 0.35 T Online adaptive MRgRT accounts for inter-
MRI with cobalt sources. It was FDA approved fraction changes in patient and tumor/OAR posi-
in 2012 and was in clinical use by 2014. The tion by applying the original radiation plan to the
second MRIdian system replaced the cobalt daily scan and recalculating dose metrics. The
sources with a 6MV linear accelerator and was plan can then be reoptimized to increase target
FDA approved in 2017. The Unity system uti- coverage or further spare normal tissue. An initial
lizes a high-field 1.5 T MRI with a 7MV linear in silico study compared adaptive and nonadap-
accelerator. The first patient was treated in tive SBRT plans delivered by MRgRT for 5
2018 in Europe with CE mark. It received FDA patients with centrally located oligometastatic or
approval in 2018 and the first patient treated in unresectable early-stage lung cancer [23].
North America was in 2019. Both the MRIdian Despite the initial plan meeting all constraints on
and Unity are capable of online adaptive therapy baseline CT simulation, all 5 patients had at least
and have similar workflows [19, 20]. The 1 fraction with ≥1 OAR constraint violation
MRIdian has received FDA approval for treat- based on daily MR imaging. These constraint
ment gating functionality and is being used in violations included the trachea (n = 6), esophagus
the clinic. At the time of writing, the Unity (n = 5), and heart/pericardium (n = 1). After
motion management has received the European adaptation, 100% of constraint violations were
15 Magnetic Resonance Imaging-Guided Radiation Therapy: Advances and Opportunities for Thoracic… 275
reversed and, furthermore, PTV dose coverage optical flow-based deformable registration on a
was improved in 66% of fractions. Based on graphics processing unit cluster. The MR ITV
these promising findings, Henke et al. conducted demonstrated differences in intra- and inter-
a subsequent phase I trial of MRgRT for ultracen- fractional motion that may not otherwise be
tral thoracic tumors and included 5 patients accounted for by standard 4D CT imaging,
treated with 50 Gy in 5 fractions. In this trial, 4 which suggests that real-time MR imaging for
out of 5 patients required at least 1 adaptive thoracic cancers may be necessary to ensure
fraction and 10 out of 25 fractions (40%) were accurate delivery of SBRT. In addition, there is
adaptive. Of the adaptive plans, 70% were for ongoing research into the development of 4D
constraint violations and 30% were to increase MRI, which may further improve resolution of
PTV coverage. The OARs that exceeded dose tumor and OAR motion [25].
constraints included the heart, trachea, esopha- An alternative to an ITV motion management
gus, stomach, and spinal cord. All the constraint approach is DIBH with gating [26]. Instead of an
violations were resolved with the adaptive plans ITV to cover the entire path of the tumor during
and overall PTV coverage was improved. respiration, DIBH with gating allows for smaller
Importantly, there were no acute grade ≥3 toxici- margins by limiting the treatment volume to a
ties in the first 6 months and only one radiation- designated phase in the respiratory cycle
related grade 3 toxicity at 15 months. These (Fig. 15.1). The reduction in treated normal tis-
studies highlighted the safety and dosimetric sue is especially relevant for patients with preex-
benefits of online adaptive MRgRT for high-risk isting medical comorbidities as decreasing the
thoracic tumors. treatment volume may preserve cardiopulmonary
MRgRT allows for continuous, real-time function. Patients can be coached for optimal
imaging during radiation delivery, which per- breath holds through audio-video methods or
mits tracking of intra-fraction tumor motion due even directly view their tumor motion using a
to respiration. Whereas CT-based motion man- monitor displaying real-time MR-cine imaging
agement strategies often require fiducial mark- [20]. A treatment boundary margin can be cre-
ers or external surrogates, MR cine imaging ated, so that treatments are automatically gated if
allows for direct tumor tracking without the a predefined percentage of the GTV is outside of
need for an invasive procedure. Management of the boundary. Although breath hold may not be
respiratory motion is a significant challenge in feasible for some patients due to poor lung func-
treating thoracic malignancies. Accounting for tion, if tolerated, it can also reduce the frequency
changes in tumor position is particularly impor- of transitions between beam delivery being
tant for centrally located tumors in which small turned on and off.
differences in adjacent OAR position may An additional potential advantage of MRgRT
greatly impact toxicity. Current strategies, such is the use of single-modality, simulation-free,
as contouring an ITV on 4DCT, generating an MR-only treatment planning. In general, a CT
ITV from the maximum intensity projection simulation is necessary for radiation therapy
(MIP), or utilizing deep inspiratory breath-hold treatment planning as an MRI does not include
(DIBH) technique, assume that tumor position(s) electron density data needed for dose calcula-
are relatively stable from baseline simulation. tions. However, there is increasing interest in the
However, there is likely deviation of tumor posi- generation of synthetic CT scans from MRI using
tion both across and within treatment fractions atlas-based, voxel-based, or deep-learning meth-
[24]. In a pilot study, 5 lung tumor patients ods [27]. Eliminating the need for a CT simula-
being treated on MRgRT were simulated with tion has several benefits including reduced
4DCT and MRI (real-time cine imaging for resource utilization, financial costs, patient radia-
172 s with −0.35 T MRI). During each fraction, tion exposure, and errors in image registration.
a 20-min cine-mode MR free-breathing scan The MR-linear accelerator is well-suited for a
was obtained. These images were then used to MR-only approach given the onboard MR imag-
contour and track GTV motion using offline ing and online adaptive therapy capabilities.
276 J. Weng et al.
Fig. 15.1 Comparison of an ITV vs DIBH motion man- gus. By utilizing an MR-Linac with DIBH (lower panels),
agement approach. For this patient with a 5-cm, mobile, target volumes could be significantly reduced. The
central tumor, an ITV approach using 4DCT (upper pan- decreased proximity of the target to the esophagus allowed
els) resulted in a large target volume abutting the esopha- for treatment with 5-fraction SBRT
Contouring, PTV, and OAR suggests MRI may improve the consistency of
target delineation for difficult cases. Similarly,
Accurate tumor and OAR delineation is an Zhang et al. compared the contours of 5 radia-
essential aspect of thoracic radiation treatment tion oncologists for 15 patients with lung
planning but can be challenging with conven- tumors adjacent to atelectasis [32]. They found
tional CT-based imaging alone. The advent of that MRI improved the delineation of GTV,
positron emission therapy (PET)/CT scans has resulting in reduced inter-observer variability
improved the ability to identify malignant met- and decreased GTVs. Thoracic MRI may be
abolically active tumor; however, PET/CT particularly helpful for challenging clinical
requires injection of a radiotracer, has limited scenarios such as tumors located in the superior
spatial resolution, and is susceptible to false- sulcus, adjacent to atelectatic lung, abutting
positive findings [28]. In comparison, MRI sub- central mediastinal structures, invading the
stantially improves soft tissue visualization, chest wall, or if the patient has a contraindica-
can highlight functional tumor characteristics, tion to contrast. The improved soft tissue reso-
and is obtained without the use of ionizing radi- lution of MRgRT is further highlighted when
ation (Fig. 15.2) [29, 30]. Preliminary studies compared to onboard cone-beam CT (CBCT).
have demonstrated improvements in the consis- An initial study compared OAR and target visu-
tency of GTV contours using thoracic alization using low-field onboard MRI and
MRI. Basson et al. conducted a study including CBCT for 221 OARs and 10 targets [29]. Three
5 physicians who independently contoured radiation oncologists reviewed the images and
GTV for 14 challenging lung cases. GTV was ultimately rated the onboard MRI images as
contoured in three phases: using CT only, add- superior for 63% of the OARs of interest, infe-
ing PET/CT, and MRI with PET/CT [31]. The rior for the ribs, and equivalent for the remain-
study demonstrated decreased inter-observer ing structures. The improved image quality of
variability in GTV contour with the addition of MRgRT may provide greater assurance of accu-
MRI compared to CT or PET/CT alone, which rate targeting and delivery of SBRT.
a b c
Fig. 15.2 Comparison of diagnostic CT with IV contrast (a), CT simulation (avg) (b), and Unity onboard MR imaging
(c) for a patient with ultracentral lung cancer
Given the ability to obtain continuous, real- cantly decreased acute genitourinary toxicity
time imaging, MRgRT enables tumor tracking [36]. A similar approach of decreasing PTV
with treatment gating and may allow for reduc- margins for central lung tumors may reduce
tion of PTV margins due to decreased treatment severe toxicity for these high-risk patients.
uncertainties. Finazzi et al. conducted a study The high precision of MRgRT may also be
comparing PTVs generated using 4DCT and beneficial for single-fraction treatment by
MRgRT for 14 patients with peripheral lung adapting for geographic misses or excessive
tumors [33]. The 4DCT PTVs were created OAR dose [37]. Finazzi et al. reported on their
using a motion-encompassing internal target initial clinical experience evaluating 17
volume with a 5-mm expansion, while the patients for single- fraction SBRT utilizing
MRgRT PTVs were an isotropic 5-mm expan- MRgRT [38]. Seven of the patients were deter-
sion from GTV that functioned as a gating mined to be unsuitable due to difficulty track-
boundary. They found that MRgRT PTVs were ing their small tumors. Out of the remaining
approximately 50% reduction compared to the 10 patients, nine were able to successfully
volume of 4DCT PTVs. A similar study com- complete their treatment in a single session
pared 41 patients treated with online adaptive and one patient required two sessions due to
MRgRT with treatment gating and 40 patients back pain. Although breath-hold patterns were
treated with conventional SBRT [34]. They variable and mean duty cycle efficiency was
found that the PTVs in the online adaptive 51%, GTV coverage was 99.6%. A study con-
MRgRT cohort were 40% smaller and did not ducted by Palacios et al. investigated
compromise local control at 1-year follow-up. MR-guided single-fraction SBRT, but with
The reduction in treated normal tissue may be consultation, simulation, and treatment con-
most beneficial to patients who have multiple ducted in a single day [39]. All 10 patients in
lung lesions or preexisting pulmonary comor- their cohort successfully completed their treat-
bidities, such as interstitial lung disease [35]. ment with excellent coverage of the GTV
There is evidence in other disease sites that (94.4%) using treatment gating. Of note,
reducing PTV margins may also decrease treat- patients were surveyed and reported overall
ment-related toxicity. A prospective randomized high levels of satisfaction with their experi-
trial in prostate cancer patients of MRgRT vs ence. By leveraging the advantages of MRgRT,
conventional CT-based SBRT demonstrated that it may be possible to more safely and precisely
using MRgRT to reduce PTV margins signifi- deliver single-fraction lung SBRT.
278 J. Weng et al.
Dose Escalation online adaptive therapy significantly reduced

OAR constraint violations [46]. A larger prospec-
There have been several significant imaging tive study of online adaptive MRgRT in
advancements that have increased the ability for 50 Gy/5 fx or 60 Gy/8 fx for central and ultracen-
radiation oncologists to safely dose-escalate radi- tral lesions is currently ongoing (NCT04917224).
ation treatments for lung tumors. With the advent There is also an ongoing phase I dose escalation
of PET/CT, 4DCT, and volumetric onboard trial using online adaptive MRgRT for ultracen-
image guidance, it became feasible to deliver tral tumors, which will deliver 55 Gy/10 fx and
highly conformal SBRT regimens that achieve up to 65 Gy/10 fx (NCT04925583). An additional
high biologically effective dose (BED). These potential advantage of MRgRT for dose escala-
innovations in imaging technology and SBRT tion is the incorporation of functional imaging in
delivery have translated to improved clinical out- treatment adaptation. MR sequences can provide
comes for patients. The initial studies utilizing not only anatomical, but also functional informa-
CT-based SBRT demonstrated very promising tion that may be utilized as guidance for treat-
local control rates of over 90% [15, 40]. In addi- ment adaptation and to evaluate treatment
tion, the CHISEL study randomized 101 patients response. For example, diffusion weighted imag-
with early-stage peripheral NSCLC to standard ing (DWI) and apparent diffusion coefficients
RT (66 Gy/33 fx or 50 Gy/20 fx) or SBRT (ADCs) have been studied in improving tumor
(54 Gy/3 fx or 48 Gy/4 fx). At 2 years, SBRT was delineation, assessing treatment response, and
found to improve local control (89% vs 65%) and identifying residual tumor [47]. As an increase
OS (88% vs 59%) [41]. Although SBRT for ADC values during therapy can reflect treatment-
peripheral tumors has excellent local control with related cell death, it may be a useful biomarker to
minimal toxicity, central and ultracentral tumors guide dose escalation for poorly responding
have been associated with severe treatment- tumors [48]. Dynamic contrast-enhanced (DCE)
related toxicity [9–11]. Given these concerning MRI is used to evaluate tissue perfusion, blood
rates of toxicity, patients with central tumors flow, and tumor vasculature. Early studies in
often are treated with more prolonged radiation other disease sites have suggested that increased
regimens (10–15 fractions), which may not enhancement during treatment may be associated
achieve equivalent BED. This is concerning as with improved outcomes [49]. The use of MRgRT
studies have demonstrated that local control of is novel and dose escalation using MR anatomi-
NSCLC has a dose-dependent relationship with cal and functional imaging in the context of his-
BED and that a BED of at least 100 Gy is neces- torically high-risk central lung tumors warrants
sary to achieve 90% local control [42–45]. Thus, prospective study.
the reduction of OAR doses while maintaining
high BED would represent a meaningful advance
in the treatment of central lung tumors. On-Table Adaptation (ViewRay
The advanced imaging capabilities of MRgRT and Unity)
such as improved soft tissue visualization, online
adaptive therapy, and real-time tumor tracking The following outlines the general workflow for
may allow for more precise and safer delivery of on-table adaptation, but this may vary depending
SBRT to central lesions. Potentially, OAR viola- on the system and individual institutional prac-
tions that occur during traditional treatment tice. On the MRIdian system, adaptation begins
delivery could be prevented using online adaptive by rigid registration of the online MRI to the
therapy. A study by Finazzi et al. included 25 simulation scans. If needed, the treatment table is
patients with central lung tumors who underwent then shifted to account for translational positional
online adaptive MRgRT in 60 Gy/8 fx or differences. The reference plan contours can be
55 Gy/5 fx. They found that physicians selected a rigidly or deformably propagated to the online
reoptimized plan in 92% of fractions and that MRI for modification. On the Unity system,
adaptation also starts with a rigid registration of patients, it is critical to thoroughly evaluate that
online MRI to the simulation scan. Based on the OAR constraints are being met with each fraction
rigid registration, clinicians can determine to use and target doses are being optimized. There is
either an adapt to position (ATP) or adapt to increasing interest in developing methods to
shape (ATS) approach [19]. The ATP method is accurately assess dose accumulation over the
suitable for cases in which contours do not need course of treatment [51]. This would allow for
to be edited and the relative positions of target/ monitoring of dose per fraction and cumulative
OARs are consistent with the reference treatment dose to OARs and tumor. Daily dose accumula-
plan. The rigid shift from the initial rigid tion can be used to inform optimization parame-
registration is used to move the plan isocenter ters of subsequent adaptive plans and adjust
accordingly. The reference plan is then reopti- prioritization of OAR sparing vs target coverage
mized with the new plan isocenter based on the [52].
original simulation CT and unedited contours. In
contrast, the ATS method is utilized for cases in
which the “anatomy of the day” differs signifi- Outcomes
cantly from the initial simulation. The initial con-
tours are transferred to the online MRI using Although there are several completed and ongo-
deformable registration and edited as necessary ing clinical trials of CT-based SBRT of central
within a 2–3 cm region adjacent to the tumor lung tumors, few studies have evaluated the role
based on a clinical decision. An adaptive plan of of MRgRT with daily adaptive plans for these
the day is then reoptimized based upon the new patients [14, 53, 54]. Prior simulation trials and
contours using the online MRI scan. Dose calcu- retrospective studies of MRgRT have demon-
lation is based on a synthetic CT created using strated dosimetric improvements including
the bulk density override method. For both the increased PTV coverage and OAR sparing [23,
MRIdian and Unity systems, after the adaptive 46]. A phase I trial including 5 ultracentral
plan is generated, dose metrics are compared patients treated with online adaptive MRgRT in
with the initial baseline plan and a treatment plan 50 Gy/5 fx met 100% of OAR constraints using
is selected. Prior to dose delivery, on-table qual- online adaptive therapy and demonstrated no
ity assurance is performed and tumor respiratory acute grade ≥3 toxicities in the first 6 months of
motion is reverified using cine imaging. follow-up [55]. A retrospective study of SBRT
using 0.35 T MRgRT reviewed outcomes for 50
consecutive patients with high-risk lung tumors
DVH Metrics/Evaluation defined as central location, prior thoracic radia-
tion, or interstitial lung disease [22]. A BED
It is well-established that a high BED of ≥100 Gy ≥100 Gy to 95% PTV was achieved in 93% of
is required to ensure optimal local control for patients and local control was 95.6% at 1 year.
lung tumors [42–44]. Although it is preferred that Importantly, at a median follow-up of
PTV is fully covered by tumoricidal prescription 21.7 months, there were only 8% grade 3 and no
dose, it may be necessary to compromise cover- grade ≥4 toxicities. Lastly, a prospective study of
age of PTV to meet adjacent critical OAR con- online adaptive MRgRT evaluated 21 patients
straints. Reassuringly, retrospective data suggests with peripheral (n = 10), central (n = 2), and
local control rates are maintained despite reduced ultracentral (n = 11) tumors [56]. Prescriptions
PTV coverage so long as ITV dose is adequate were risk-adapted and were 50 Gy/5 fx for
[50]. Prioritizing critical OAR doses and second- peripheral, 60 Gy/8 fx for central, and 50–60 Gy
arily maximizing tumor dose is particularly rele- in 10 fx for ultracentral tumors. Overall, 93% of
vant for patients with central lung tumors fractions were adapted (154/165). The most com-
undergoing online adaptive therapy given the mon reason for plan adaptation was to increase
potential for multiple plans. For these high-risk PTV coverage (86%), whereas 17% of adapta-
280 J. Weng et al.
tions for OAR constraint violation. Of note, all k-space, also known as 3DVaneXD or multi-
OAR constraint violations were reversed except VaneXD, has been found to greatly reduce the
for one. These promising results have prompted impact from respiratory motion in free-breathing
the authors to initiate a prospective, phase I dose acquisition [61, 62]. Breath-hold scans using a
escalation trial of MRgRT for central lung tumors balanced FFE sequence can acquire a 3D MRI
(NCT04925583). There are also other prospec- scan within 20 s with reasonable image quality,
tive treatment registries and ongoing clinical tri- thus reducing the impact from respiratory motion.
als utilizing MRgRT approaches for thoracic Also, the compressed SENSE technique has been
tumors (NCT04917224, NCT05609331, used in clinic for fast MRI acquisition to further
NCT04939246, NCT04115254, NCT04789486). reduce the impact from respiratory motion [63].
While early outcomes are promising, there Dosimetric drawbacks of MRgRT include the
remains a need for additional prospective studies lack of volumetric modulated arc therapy
to better characterize treatment outcomes, ensure (VMAT) capabilities, which limits dose confor-
that dosimetric optimization results in meaningful mality and extends treatment duration. In addi-
improvement in clinical outcomes, and guide tion, MR-gated treatments introduce beam-off
patient selection for thoracic MRgRT. latency dosimetric errors [64]. These errors occur
due to the time required for the MR cine image to
be acquired and the beam to be turned off appro-
Challenges priately when the tumor moves beyond the gating
boundary. Beam-off latency may reduce dose to
While there are significant advantages of MRgRT the PTV and inadvertently increase dose to
for lung tumors, there remain several limitations nearby OARs; however, this can be minimized
unique to thoracic MRI [57]. Unlike many other through development of faster MR imaging and
disease sites, the lungs inherently have a low MR limiting the transition between gated and ungated
signal to noise ratio (SNR) due to low tissue den- states. An additional dosimetric drawback of
sity and high air content [58]. The visualization MRgRT is that the Lorentz force produced by
of small parenchymal lesions and ground-glass magnetic fields leads to an electron returning
opacities can be suboptimal, particularly for effect. The electron returning effect results in
smaller tumors <1 cm. Low field strength (0.35 T) higher doses at air-tissue interfaces, such as the
on a MRIdian system also limits the SNR of lung, and must be corrected for by the treatment
MRI. In addition, MRI body coil design is lim- planning algorithm. Lastly, MRgRT with online
ited by the immobilization devices used during plan adaptation workflow usually requires an
treatment, which results in a body coil that is sub- extended treatment time with the patient on the
optimal for MRI image quality. Furthermore, treatment table [65]. This extended treatment
respiratory, cardiac, and great vessel pulsatile time is not only the result of complicated work-
motion degrades MR image quality due to motion flow with online plan adaptation and patient-
artifacts and image blurring. These field distor- specific QA, but also due to a low dose rate and
tion artifacts can be particularly problematic for low duty cycle with treatment gating during beam
tumors located at soft tissue and air interfaces delivery. Patients are more likely to move when
(e.g., the pericardium or diaphragm) and impact laying on the couch for a long time, resulting in
tumor/OAR delineation for initial contouring as unexpected movement and making motion man-
well as treatment gating. There are ongoing agement more difficult. Real-time motion moni-
efforts to develop and optimize MR sequences toring and treatment gating can alleviate the
for low-field and standard-field MRI to address impact from this unexpected motion. Additional
these challenges for MRgRT [59, 60]. Specifically, fast plan adaptation may be necessary to account
the turbo field-echo (TFE) or fast field-echo for large changes in position during treatment
(FFE) sequences with a radial sampling in delivery.
Future Directions References
MRgRT is a promising technologic innovation 1. Siegel RL, Miller KD, Wagle NS, Jemal A. Cancer
statistics, 2023. CA Cancer J Clin. 2023;73(1):17–48.
with several key advantages for thoracic radia- 2. NCCN Clinical Practice Guidelines. Non-small cell
tion therapy. This includes clinical scenarios lung cancer. (Version 1.2023). https://www.nccn.org/
that require precise OAR dose monitoring such professionals/physician_gls/pdf/nscl.pdf.
as patients with high-risk tumor locations (cen- 3. Bergsma DP, Salama JK, Singh DP, Chmura SJ,
Milano MT. Radiotherapy for oligometastatic lung
tral/ultracentral) or recurrent tumors requiring cancer. Front Oncol. 2017;7:210.
reirradiation. MRgRT may also facilitate the 4. Gomez DR, Tang C, Zhang J, Blumenschein GR,
simultaneous treatment of primary NSLCL or Hernandez M, Lee JJ, et al. Local consolidative ther-
oligometastatic disease with multiple thoracic apy vs. maintenance therapy or observation for patients
with oligometastatic non-small-cell lung cancer: long-
lesions with a single isocenter [66]. In addition term results of a multi-institutional, phase II, random-
to SBRT, patients with locally advanced ized study. J Clin Oncol. 2019;37(18):1558–65.
NSCLC undergoing longer courses of RT may 5. Palma DA, Olson R, Harrow S, Gaede S, Louie AV,
benefit from weekly treatment adaptation as Haasbeek C, et al. Stereotactic ablative radiotherapy
for the comprehensive treatment of oligometastatic
tumor regresses on therapy or due to anatomical cancers: long-term results of the SABR-COMET
distortions. The precision of MRgRT may also phase II randomized trial. JCO. 2020;38(25):2830–8.
facilitate delivery of hypofractionated regimens 6. Timmerman R, Paulus R, Galvin J, Michalski J,
for patients with locally advanced NSCLC who Straube W, Bradley J, et al. Stereotactic body radia-
tion therapy for inoperable early stage lung cancer.
have poor prognostic factors [67]. As many JAMA. 2010;303(11):1070–6.
candidates for radiation therapy are not surgical 7. Baumann P, Nyman J, Høyer M, Wennberg B,
candidates due to poor lung function or pulmo- Gagliardi G, Lax I, et al. Outcome in a prospective
nary comorbidities such as interstitial lung dis- phase II trial of medically inoperable stage I non-
small-cell lung cancer patients treated with stereotac-
ease, MRgRT with real-time tracking and gated tic body radiotherapy. J Clin Oncol. 2009;27:3290–6.
treatment delivery may be beneficial to reduce 8. Ricardi U, Filippi AR, Guarneri A, Giglioli FR,
the overall integral dose to the lungs. Future Ciammella P, Franco P, et al. Stereotactic body
research directions also include MRI sequence radiation therapy for early stage non-small cell lung
cancer: results of a prospective trial. Lung Cancer.
optimization and development for lung imag- 2010;68(1):72–7.
ing. Recent advances in MRI have greatly 9. Timmerman R, Mcgarry R, Yiannoutsos C, Papiez
improved the image SNR for lung MRI; how- L, Tudor K, DeLuca J, et al. Excessive toxicity
ever, there is still room for further improvement when treating central tumors in a phase II study of
stereotactic body radiation therapy for medically
(e.g., improving the MR body coil design for inoperable early-stage lung cancer. J Clin Oncol.
better image SNR and radiation treatment guid- 2006;24:4833–9.
ance). Another potential improvement is the 10. Bral S, Gevaert T, Linthout N, Versmessen H, Collen
latency of cine MR imaging. Although there are C, Engels B, et al. Prospective, risk-adapted strategy
of stereotactic body radiotherapy for early-stage non-
currently strategies to reduce the impact of small-cell lung cancer: results of a phase II trial. Int J
latency, the duty cycle for treatment gating or Radiat Oncol Biol Phys. 2011;80(5):1343–9.
breath-hold treatment is low and further extends 11. Song SY, Choi W, Shin SS, Lee SW, Ahn SD, Kim
treatment time. In the future, development of JH, et al. Fractionated stereotactic body radiation
therapy for medically inoperable stage I lung can-
faster cine MR imaging is desirable. Lastly, in cer adjacent to central large bronchus. Lung Cancer.
the future, integration of functional MRI 2009;66(1):89–93.
sequences may aid in tumor delineation, moni- 12. Chang JY, Li QQ, Xu QY, Allen PK, Rebueno N,
toring treatment response, and guiding dose Gomez DR, et al. Stereotactic ablative radiation ther-
apy for centrally located early stage or isolated paren-
escalation or de-escalation studies [68]. chymal recurrences of non-small cell lung cancer:
Overall, MRgRT for thoracic tumors has dem- how to fly in a “no fly zone”. Int J Radiat Oncol Biol
onstrated promising technical results with sev- Phys. 2014;88(5):1120–8.
eral active investigations actively seeking to 13. Haasbeek CJA, Lagerwaard FJ, Slotman BJ, Senan
S. Outcomes of stereotactic ablative radiotherapy for
improve the therapeutic ratio.
282 J. Weng et al.
centrally located early-stage lung cancer. J Thorac 26. Dhont J, Harden SV, Chee LYS, Aitken K, Hanna
Oncol. 2011;6(12):2036–43. GG, Bertholet J. Image-guided radiotherapy to man-
14. Bezjak A, Paulus R, Gaspar LE, Timmerman RD, age respiratory motion: lung and liver. Clin Oncol.
Straube WL, Ryan WF, et al. Safety and efficacy of 2020;32(12):792–804.
a five-fraction stereotactic body radiotherapy sched- 27. Olberg S, Zhang H, Kennedy WR, Chun J, Rodriguez
ule for centrally located non-small-cell lung cancer: V, Zoberi I, et al. Synthetic CT reconstruction
NRG Oncology/RTOG 0813 trial. J Clin Oncol. using a deep spatial pyramid convolutional frame-
2019;37(15):1316–25. work for MR-only breast radiotherapy. Med Phys.
15. Timmerman RD, Hu C, Michalski JM, Bradley JC, 2019;46(9):4135–47.
Galvin J, Johnstone DW, et al. Long-term results 28. Cobben DCP, de Boer HCJ, Tijssen RH, Rutten
of stereotactic body radiation therapy in medically EGGM, van Vulpen M, Peerlings J, et al. Emerging role
inoperable stage I non-small cell lung cancer. JAMA of MRI for radiation treatment planning in lung can-
Oncol. 2018;4(9):1287–8. cer. Technol Cancer Res Treat. 2016;15(6):NP47–60.
16. Chen H, Laba JM, Zayed S, Boldt RG, Palma 29. Noel CE, Parikh PJ, Spencer CR, Green OL, Hu Y,
DA, Louie AV. Safety and effectiveness of ste- Mutic S, et al. Comparison of onboard low-field mag-
reotactic ablative radiotherapy for ultra-central netic resonance imaging versus onboard computed
lung lesions: a systematic review. J Thorac Oncol. tomography for anatomy visualization in radiother-
2019;14(8):1332–42. apy. Acta Oncol. 2015;54(9):1474–82.
17. Lindberg K, Grozman V, Karlsson K, Lindberg S, Lax 30. Raaymakers BW, Lagendijk JJW, Overweg J, Kok
I, Wersäll P, et al. The HILUS-trial—a prospective JGM, Raaijmakers AJE, Kerkhof EM, et al. Integrating
Nordic multicenter phase 2 study of ultracentral lung a 1.5 T MRI scanner with a 6 MV accelerator: proof
tumors treated with stereotactic body radiotherapy. J of concept. Phys Med Biol. 2009;54(12):N229.
Thorac Oncol. 2021;16(7):1200–10. 31. Basson L, Jarraya H, Escande A, Cordoba A,
18. Liney GP, Whelan B, Oborn B, Barton M, Keall Daghistani R, Pasquier D, et al. Chest magnetic reso-
P. MRI-linear accelerator radiotherapy systems. Clin nance imaging decreases inter-observer variability
Oncol. 2018;30(11):686–91. of gross target volume for lung tumors. Front Oncol.
19. Winkel D, Bol GH, Kroon PS, van Asselen B, Hackett 2019;9:690. https://doi.org/10.3389/fonc.2019.00690.
SS, Werensteijn-Honingh AM, et al. Adaptive radio- 32. Zhang H, Fu C, Fan M, Lu L, Chen Y, Liu C, et al.
therapy: the Elekta Unity MR-linac concept. Clin Reduction of inter-observer variability using MRI and
Transl Radiat Oncol. 2019;18:54–9. CT fusion in delineating of primary tumor for radio-
20. Bryant JM, Sim AJ, Feygelman V, Latifi K, Rosenberg therapy in lung cancer with atelectasis. Front Oncol.
SA. Adaptive hypofractionted and stereotactic body 2022;12:841771.
radiotherapy for lung tumors with real-time MRI 33. Finazzi T, Palacios MA, Haasbeek CJA, Admiraal MA,
guidance. Front Oncol. 2023;13:1061854. https://doi. Spoelstra FOB, Bruynzeel AME, et al. Stereotactic
org/10.3389/fonc.2023.1061854. MR-guided adaptive radiation therapy for peripheral
21. Padgett KR, Simpson GN, Llorente R, Samuels MA, lung tumors. Radiother Oncol. 2020;144:46–52.
Dogan N. Feasibility of adaptive MR-guided ste- 34. Kang HJ, Kwak YK, Kim M, Lee SJ. Application of
reotactic body radiotherapy (SBRT) of lung tumors. real-time MRI-guided linear accelerator in stereotac-
Cureus. 2018;10(4):e2423. tic ablative body radiotherapy for non-small cell lung
22. Finazzi T, Haasbeek CJA, Spoelstra FOB, Palacios cancer: one step forward to precise targeting. J Cancer
MA, Admiraal MA, Bruynzeel AME, et al. Clinical Res Clin Oncol. 2022;148(11):3215–23.
outcomes of stereotactic MR-guided adaptive radia- 35. Chen H, Senan S, Nossent EJ, Boldt RG, Warner A,
tion therapy for high-risk lung tumors. Int J Radiat Palma DA, et al. Treatment-related toxicity in patients
Oncol Biol Phys. 2020;107(2):270–8. with early-stage non-small cell lung cancer and coex-
23. Henke L, Kashani R, Yang D, Zhao T, Green O, isting interstitial lung disease: a systematic review. Int
Olsen L, et al. Simulated online adaptive magnetic J Radiat Oncol Biol Phys. 2017;98(3):622–31.
resonance-guided stereotactic body radiation ther- 36. Kishan AU, Ma TM, Lamb JM, Casado M, Wilhalme
apy for the treatment of oligometastatic disease of H, Low DA, et al. Magnetic resonance imaging-
the abdomen and central thorax: characterization of guided vs computed tomography-guided stereotactic
potential advantages. Int J Radiat Oncol Biol Phys. body radiotherapy for prostate cancer: the MIRAGE
2016;96(5):1078–86. randomized clinical trial. JAMA Oncol. 2023;9:365.
24. Thomas DH, Santhanam A, Kishan AU, Cao M, Lamb https://doi.org/10.1001/jamaoncol.2022.6558.
J, Min Y, et al. Initial clinical observations of intra- 37. Ng SSW, Ning MS, Lee P, McMahon RA, Siva S,
and interfractional motion variation in MR-guided Chuong MD. Single-fraction stereotactic body radia-
lung SBRT. Br J Radiol. 2018;91(1083):20170522. tion therapy: a paradigm during the coronavirus dis-
25. Paulson ES, Ahunbay E, Chen X, Mickevicius NJ, ease 2019 (COVID-19) pandemic and beyond? Adv
Chen GP, Schultz C, et al. 4D-MRI driven MR-guided Radiat Oncol. 2020;5(4):761–73.
online adaptive radiotherapy for abdominal stereotac- 38. Finazzi T, van de Koste JRS, Palacios MA, Spoelstra
tic body radiation therapy on a high field MR-Linac: FOB, Slotman BJ, Haasbeek CJA, et al. Delivery of
implementation and initial clinical experience. Clin magnetic resonance-guided single-fraction stereotac-
Transl Radiat Oncol. 2020;23:72–9.
tic lung radiotherapy. Phys Imaging Radiat Oncol. tive stereotactic body radiation therapy. Br J Radiol.
2020;14:17–23. 2016;89:20150963.
39. Palacios MA, Verheijen S, Schneiders FL, Bohoudi O, 51. McDonald BA, Zachiu C, Christodouleas J, Naser
Slotman BJ, Lagerwaard FJ, et al. Same-day consulta- MA, Ruschin M, Sonke JJ, et al. Dose accumulation
tion, simulation and lung stereotactic ablative radio- for MR-guided adaptive radiotherapy: from practical
therapy delivery on a magnetic resonance-linac. Phys considerations to state-of-the-art clinical implementa-
Imaging Radiat Oncol. 2022;24:76–81. tion. Front Oncol. 2022;12:1086258.
40. Timmerman RD, Paulus R, Pass HI, Gore EM, 52. Lim SY, Tran A, Tran ANK, Sobremonte A, Fuller
Edelman MJ, Galvin J, et al. Stereotactic body radia- CD, Simmons L, et al. Dose accumulation of daily
tion therapy for operable early-stage lung cancer. adaptive plans to decide optimal plan adaptation
JAMA Oncol. 2018;4(9):1263–6. strategy for head-and-neck patients treated with
41. Ball D, Mai GT, Vinod S, Babington S, Ruben J, MR-Linac. Med Dosim. 2022;47(1):103–9.
Kron T, et al. Stereotactic ablative radiotherapy ver- 53. Adebahr S, Collette S, Shash E, Lambrecht M,
sus standard radiotherapy in stage 1 non-small-cell Le Pechoux C, Faivre-Finn C, et al. LungTech, an
lung cancer (TROG 09.02 CHISEL): a phase 3, open- EORTC phase II trial of stereotactic body radiother-
label, randomised controlled trial. Lancet Oncol. apy for centrally located lung tumours: a clinical per-
2019;20(4):494–503. spective. Br J Radiol. 2015;88(1051):20150036.
42. Onishi H, Araki T, Shirato H, Nagata Y, Hiraoka M, 54. Giuliani M, Mathew AS, Bahig H, Bratman SV, Filion
Gomi K, et al. Stereotactic hypofractionated high-dose E, Glick D, et al. SUNSET: stereotactic radiation for
irradiation for stage I nonsmall cell lung carcinoma: ultracentral non-small-cell lung cancer—a safety and
clinical outcomes in 245 subjects in a Japanese multi- efficacy trial. Clin Lung Cancer. 2018;19(4):e529–32.
institutional study. Cancer. 2004;101(7):1623–31. 55. Henke LE, Olsen JR, Contreras JA, Curcuru A,
43. Guckenberger M, Wulf J, Mueller G, Krieger T, Baier DeWees TA, Green OL, et al. Stereotactic MR-guided
K, Gabor M, et al. Dose–response relationship for online adaptive radiation therapy (SMART) for ultra-
image-guided stereotactic body radiotherapy of pul- central thorax malignancies: results of a phase 1 trial.
monary tumors: relevance of 4D dose calculation. Int Adv Radiat Oncol. 2019;4(1):201–9.
J Radiat Oncol Biol Phys. 2009;74(1):47–54. 56. Regnery S, Buchele C, Weykamp F, Pohl M, Hoegen
44. Onimaru R, Fujino M, Yamazaki K, Onodera Y, P, Eichkorn T, et al. Adaptive MR-guided stereotac-
Taguchi H, Katoh N, et al. Steep dose–response rela- tic radiotherapy is beneficial for ablative treatment
tionship for stage I non-small-cell lung cancer using of lung tumors in high-risk locations. Front Oncol.
hypofractionated high-dose irradiation by real-time 2022;11:757031.
tumor-tracking radiotherapy. Int J Radiat Oncol Biol 57. Sim AJ, Kaza E, Singer L, Rosenberg SA. A review
Phys. 2008;70(2):374–81. of the role of MRI in diagnosis and treatment of
45. Zhang J, Yang F, Li B, Li H, Liu J, Huang W, et al. early stage lung cancer. Clin Transl Radiat Oncol.
Which is the optimal biologically effective dose of 2020;24:16–22.
stereotactic body radiotherapy for stage I non-small- 58. Wild JM, Marshall H, Bock M, Schad LR, Jakob PM,
cell lung cancer? A meta-analysis. Int J Radiat Oncol Puderbach M, et al. MRI of the lung (1/3): methods.
Biol Phys. 2011;81(4):e305–16. Insights Imaging. 2012;3(4):345–53.
46. Finazzi T, Palacios MA, Spoelstra FOB, Haasbeek 59. Groot Koerkamp ML, Vasmel JE, Russell NS,
CJA, Bruynzeel AME, Slotman BJ, et al. Role of on- Shaitelman SF, Anandadas CN, Currey A, et al.
table plan adaptation in MR-guided ablative radiation Optimizing MR-guided radiotherapy for breast cancer
therapy for central lung tumors. Int J Radiat Oncol patients. Front Oncol. 2020;10:1107.
Biol Phys. 2019;104(4):933–41. 60. Kurz C, Buizza G, Landry G, Kamp F, Rabe M,
47. Datta A, Aznar MC, Dubec M, Parker GJM, O’Connor Paganelli C, et al. Medical physics challenges in
JPB. Delivering functional imaging on the MRI- clinical MR-guided radiotherapy. Radiat Oncol.
Linac: current challenges and potential solutions. Clin 2020;15(1):93.
Oncol. 2018;30(11):702–10. 61. Eccles CL, Adair Smith G, Bower L, Hafeez S,
48. Chin S, Eccles CL, McWilliam A, Chuter R, Walker Herbert T, Hunt A, et al. Magnetic resonance imag-
E, Whitehurst P, et al. Magnetic resonance-guided ing sequence evaluation of an MR Linac system;
radiation therapy: a review. J Med Imaging Radiat early clinical experience. Tech Innov Patient Support
Oncol. 2020;64(1):163–77. Radiat Oncol. 2019;12:56–63.
49. van Houdt PJ, Yang Y, van der Heide UA. Quantitative 62. Serai SD, Hu HH, Ahmad R, White S, Pednekar
magnetic resonance imaging for biological image- A, Anupindi SA, et al. Newly developed methods
guided adaptive radiotherapy. Front Oncol. for reducing motion artifacts in pediatric abdomi-
2021;10:615643. https://doi.org/10.3389/ nal MRI: tips and pearls. AJR Am J Roentgenol.
fonc.2020.615643. 2020;214(5):1042–53.
50. Shaverdian N, Tenn S, Veruttipong D, Wang J, Hegde 63. Kocaoglu M, Pednekar AS, Wang H, Alsaied T, Taylor
J, Lee C, et al. The significance of PTV dose coverage MD, Rattan MS. Breath-hold and free-breathing
on cancer control outcomes in early stage non-small quantitative assessment of biventricular volume and
cell lung cancer patients treated with highly abla- function using compressed SENSE: a clinical valida-
284 J. Weng et al.
tion in children and young adults. J Cardiovasc Magn with stereotactic MR-guided adaptive radiation ther-
Reson. 2020;22(1):54. apy: a case report. Medicine. 2023;102(2):e32626.
64. Hu P, Li X, Liu W, Yan B, Xue X, Yang F, et al. 67. Eze C, Lombardo E, Nierer L, Xiong Y, Niyazi M,
Dosimetry impact of gating latency in cine magnetic Belka C, et al. MR-guided radiotherapy in node-
resonance image guided breath-hold pancreatic can- positive non-small cell lung cancer and severely
cer radiotherapy. Phys Med Biol. 2022;67(5):055008. limited pulmonary reserve: a report proposing a new
65. Botman R, Tetar SU, Palacios MA, Slotman BJ, clinical pathway for the management of high-risk
Lagerwaard FJ, Bruynzeel AME. The clinical intro- patients. Radiat Oncol. 2022;17(1):43.
duction of MR-guided radiation therapy from a RTT 68. Keall PJ, Brighi C, Glide-Hurst C, Liney G, Liu
perspective. Clin Transl Radiat Oncol. 2019;18:140–5. PZY, Lydiard S, et al. Integrated MRI-guided radio-
66. Chen F, Cheng YK, Chiang CH, Lu TY, Huang therapy—opportunities and challenges. Nat Rev Clin
CJ. Simultaneous treatment of two lung cancer lesions Oncol. 2022;19(7):458–70.
MRI-Guided Radiation Therapy
for Prostate Cancer: Less Is More
16
When Technology and Evidence
Intersect
Luca Valle, Rosalyne Westley, Filippo Alongi,

Sean Sachdev, and Amar Kishan
Disease Overview (RP) or radiation therapy (RT), both of which have

been found to yield equivalent long-term survival
Prostate cancer is the most common non-cutaneous outcomes [2]. The natural history of prostate can-
malignancy affecting men, with an estimated cer is protracted, with very high overall survival
268,490 cases diagnosed in the United States in rates and a number of salvage treatment options
2022 [1]. The definitive treatment options for local- available for men who undergo management with
ized prostate cancer include radical prostatectomy either definitive RP [3] or definitive RT [4].
RT for the definitive treatment of prostate can-
cer can be delivered to the prostate and, if war-
L. Valle ranted, to the surrounding pelvic lymph nodes as
Department of Radiation Oncology, David Geffen
external beam radiotherapy (EBRT) or to the
School of Medicine, University of California Los
Angeles, Los Angeles, CA, USA prostate alone with brachytherapy and is often
delivered in conjunction with hormonal therapy,
Veterans Affairs Greater Los Angeles Healthcare
System, Los Angeles, CA, USA depending on the prognostic risk stratification
e-mail: [email protected] group of the patient [5].
A. Kishan (*) For men who elect to receive definitive RT, the
Department of Radiation Oncology, David Geffen treatment of prostate cancer has improved dra-
School of Medicine, University of California Los matically due to advances in imaging and tech-
Angeles, Los Angeles, CA, USA
nology as well as improvements in our
understanding of the biology of prostate cancer,
R. Westley
particularly regarding the proposed low α/β ratio
Clinical Oncology Academic Urology Unit, Institute
of Cancer Research, The Royal Marsden Hospital, of prostate cancer cells [6] (per the linear qua-
London, UK dratic model). As a consequence, there has been a
e-mail: [email protected] trend toward both radiation dose escalation of the
F. Alongi prostate gland and simultaneously shortening the
Department of Advanced Radiation Oncology, delivery of definitive prostate cancer RT by deliv-
IRCCS Ospedale Sacro Cuore Don Calabria,
ering larger doses of radiation per fraction
Negrar, Verona, Italy
through hypofractionation.
Seminal trials have demonstrated the merits of
S. Sachdev dose escalation in prostate cancer. For example,
in MRC RT01, 74 Gy in 37 fractions delivered
Chicago, IL, USA alongside androgen deprivation therapy (ADT)
e-mail: [email protected] showed superior biochemical control rates com-
286 L. Valle et al.
pared to the standard dose fractionation of 64 Gy However, the move to delivering ultrahypofrac-
in 32 fractions [5]. Parallel data from the RTOG tionated regimens has only become possible due
0126 study in the United States established the to ongoing advances in radiation technology.
7-year biochemical control benefit to dose SBRT (Sterotactic Body Radiation Therapy) is
escalation to 79.2 Gy over 70.2 Gy (70% vs 55%, technically more challenging to deliver, requiring
p < 0.001) [7], thereby establishing the former as precise geometric targeting with in-room image
the standard of care dosing regimen for men with guidance. And while the toxicity results so far
all stages of prostate cancer. have been promising, given the favorable disease-
In parallel to these dose escalation studies, specific outcomes observed in men following
there has been similar interest in harnessing the prostate radiotherapy, patients tend to live many
differential response of prostate cancer cells to years after completing their treatment, and thus,
larger radiation doses per fraction through hypo- minimization of the long-term sequelae of treat-
fractionation. In the UK, the CHHiP trial showed ment remains a significant priority in the delivery
that hypofractioned radiotherapy using 60 Gy in of modern definitive prostate RT.
20 fractions was non-inferior to 74 Gy in 37 frac- In addition to deploying RT in the definitive
tions [8]. Additionally, RTOG 0415 demonstrated treatment setting, it is also indicated in the sal-
that 70 Gy in 28 fractions was non-inferior to vage setting in the event of a biochemical recur-
73.8 Gy in 41 fractions in low-risk prostate cancer, rence (BCR) following RP. In these scenarios, the
although a late increase in gastrointestinal and prostate bed serves as the primary target volume,
genitourinary adverse events were observed in the with the option to electively include draining pel-
hypofractionated group [9]. PROFIT then demon- vic lymph nodes in the radiation field depending
strated that the treatment of low- to intermediate- on patient- and tumor-specific features. While RT
risk prostate cancer with 60 Gy in 20 fractions was delivery with conventional fractionation remains
non-inferior to conventional radiotherapy with the mainstay of treatment for these patients, there
78 Gy in 39 fractions over 8 weeks in the absence has been parallel interest [17, 18] in shortening
of androgen deprivation therapy (ADT) [10]. RT courses in the postoperative setting as well.
Given the aforementioned level 1a evidence, Regardless of the treatment setting, the ulti-
moderately hypofractionated radiotherapy to the mate aim of RT delivery is to maximize the
prostate is acknowledged to be a standard treat- therapeutic ratio by ensuring robust tumor
ment option, yet the boundaries continue to be control by maintaining or increasing dose to
pushed. HYPO-RT-PC demonstrated that the the target while simultaneously minimizing
treatment of prostate cancer with ultrahypofrac- treatment toxicity by decreasing the radiation
tionated radiotherapy (also known as stereotactic dose to surrounding normal tissues. Further
body radiation therapy [SBRT]) is non-inferior to advances in radiotherapy technology is neces-
conventionally fractionated prostate cancer in the sary to achieve this objective.
context of failure-free survival [11] and patient- The hybridization of a magnetic resonance
reported outcomes also supported the long-term imaging (MRI) scanner and the modern linear
tolerability of SBRT for these patients [12]. PACE accelerator has resulted in a new form of prostate
B has also tested the efficacy and toxicity profile cancer treatment, termed MRI-guided RT
of ultrahypofractionated radiotherapy to the pros- (MRIgRT). MRIgRT may offer several advan-
tate with 36.35 Gy in 5 fractions over 2 weeks, tages in the treatment of prostate cancer com-
demonstrating the safety and efficacy of ultrahy- pared to the more standard modalities used to
pofractionated radiotherapy as well as similar deliver image-guided EBRT, such as with a
physician-reported GU and GI toxicity between CT-guided linear accelerator, Calypso™ tracking
SBRT and conventional fractionation [13, 14]. markers, and CyberKnife™. There are currently
The advent of hypo- and ultrahypofractionated four MRIgRT systems, two of which are avail-
courses have added considerable convenience able commercially. The Unity™ MR-Linac
[15] and cost savings [16] for healthcare systems. (Elekta AB, Stockholm, Sweden) combines a
16 MRI-Guided Radiation Therapy for Prostate Cancer: Less Is More When Technology and Evidence… 287
1.5-Tesla Philips Magnetic Resonance (MR) position of the prostate and the seminal vesicles
scanner (Philips Medical Systems, Best, the [21, 22]. Moreover, a recent systemic review
Netherlands) and a 7-MV FFF Elekta linear reported that although rectal and bladder filling
accelerator [19], while the MRIdian™ system status was the main cause for seminal vesicle
(ViewRay Inc., Mountain View, CA) combines a motion, no apparent effect on magnitude of
6-MV linear accelerator with a 0.35 T MR imag- motion was seen when different bladder and rec-
ing (Siemens, Munich, Germany) system [20]. tal preparation protocols were used [23], which
Both of these systems have been studied and can render motion challenging to predict a priori.
implemented for the treatment of prostate cancer In addition to OAR motion, the prostate itself
in both the definitive and the post-prostatectomy has also been shown to undergo changes in vol-
setting and will be the focus of this chapter. ume and dimensions during a course of radio-
therapy [24, 25], with the addition of androgen
deprivation therapy further altering prostate
Overcoming Challenges gland size and tumor characteristics [26].
with Traditional Approaches Previous image- guided techniques such as
CyberKnife, cone beam imaging, and fiducial
There are many challenges associated with tradi- matching have allowed for the treatment to be
tional CT-guided delivery of radiation for pros- matched to the position of the target, but never to
tate cancer, particularly in the era of the position and shape of both the target and
ultrahypofractionated treatments. Given the close adjacent OARs online. Even the 4D detection of
proximity of the prostate to surrounding organs at implanted electromagnetic transponders [27]
risk (OARs), particularly the bladder and rectum, (the Calypso™ system) cannot provide true
dosimetric tradeoffs between prostate target cov- “real-time” motion evaluation that is actionable
erage and sparing of OARs are frequent. The ure- in the event that surrounding OARs encroach
thra is another critical OAR that courses within upon the prostate target volume. A further limi-
the prostate gland itself and can be challenging to tation of the Calypso system is that the implanted
visualize and spare appropriately with CT-guided transponders may preclude the use of MRI in
EBRT. Additional challenges that limit the thera- posttreatment assessment [28].
peutic benefits of the more traditional approaches One approach for mitigating target motion due
to delivering prostate RT include concerns with to anatomic changes between fractions is adap-
intrafraction motion management, the inability to tive radiotherapy. The MRIgRT system allows
measure dose accumulation, exposure to ionizing for the original offline-generated treatment plan
radiation during image-guided RT (IGRT), and to be adapted through recontouring or a daily
the inability to functionally evaluate and capital- shift to account for any changes in anatomy as
ize on tumor response. Each of these consider- seen on the MRI taken on the day of treatment
ations will be discussed in turn, as well as the (Fig. 16.1). The new plan which is generated
limited scope of ultrahypofractionated RT in online in real time is then delivered to the patient,
recurrent prostate cancer and how these hurdles negating any interfraction deformations that may
may be overcome using MRIgRT. have occurred since the planning scan and
between each day of radiotherapy. The ability to
simultaneously visualize and customize treat-
Motion Management ments based on deformation of both tumor tar-
gets and OARs represents a major theoretical
It is well documented that the prostate, seminal advantage of adaptive planning with MRIgRT.
vesicles, and OARs are subject to both interfrac- Depending on the treatment workflow
tion and intrafraction motion. There are continu- adopted, there is also the potential to account for
ous changes in shape and position of the bladder, intrafraction organ motion which occurs during
bowel, and rectum which in turn influence the the delivery of a single fraction of radiotherapy
288 L. Valle et al.
a b
Fig. 16.1 Changes in anatomy as visualized on the daily vesicles anteriorly and laterally. (a) axial MR image without
MRI as the rectum fills with air and squashes the seminal rectal distention. (b) axial MR image with rectal distention
[29]. This is achieved by creating a gating margin to be shifted after the initial recontouring phase to
around the prostate CTV and assigning a toler- match to the most up-to-date position of the pros-
ance for the acceptable proportion of the CTV tate as seen on MRI taken immediately prior to
that is allowable outside of the gating boundary. treatment. This accounts for any motion that may
The intrafraction motion of the prostate and sur- have occurred during the contouring and plan
rounding OARs is then monitored by sagittal adaption phase of the workflow [31]. This is
images taken during treatment at four frames per known as adapt-to-position and is outlined in
second, which might be more aptly termed “real- greater detail in “On Table Adaption for Unity
time” monitoring. Moreover, as part of the auto- (“Unity”) and MRIdian (“ViewRay”) below.
matic beam hold feature of the MRIdian device,
treatment is paused if intrafraction motion leads
to unacceptable displacement of the prostate I nability to Measure Dose
gland from the original setup position and Accumulation
resumes when the prostate returns to the original
setup position. This provides additional reassur- Conventional radiotherapy delivery approaches
ance regarding the accuracy of treatment delivery assume that the organ motion is static during
when treating with ultrahypofractionated regi- treatment, leading to potential inaccuracies in the
mens [30]. These gating boundaries and toler- assessment of true delivered dose to targets and
ance thresholds can be established and modified OARs. Volumetric imaging acquired during the
at the discretion of the treating physician. These treatment of a patient with a fraction of radiother-
features of MRIgRT thus provide a possible apy also offers the opportunity for more accurate
advantage over platforms such as Cyberknife™, “real-time” calculation of the dose accumulation
on which “real-time” imaging and beam compen- that occurs throughout a treatment course [32].
sation occurs via kv planar images taken every This enables the opportunity to consider prostate
30–60 s depending on the threshold set. dose escalation for future fractions if the dose
While gating is primarily a feature of the received by the surrounding OARs over the pre-
ViewRay device, as of early 2023, the Unity device ceding fractions is below the allowable dose con-
received FDA 510(k) clearance for its straints. It similarly enables dose de-escalation in
“Comprehensive Motion Management with True patients with high-risk anatomy, for whom daily
3D Tracking and Automatic Gating” technology, dosimetry to critical OARs is higher than
which will allow tracking, gating, and baseline expected. A patient with small bowel adjacent to
shift correction. At the time of writing, this is not the prostate, for example, may be more safely
widely clinically implemented, and thus, the use treated with MRIgRT as the accumulated dose to
of gating to account for larger degrees of intrafrac- the bowel can be calculated throughout the
tion motion on the Unity device is more limited. treatment and the prescription dose or PTV cov-
However, Unity does allow the radiotherapy plan erage modified, if required [31].
xposure to Ionizing Radiation

E Limited Scope
During Image-Guided RT of Ultrahypofractionation
in Recurrent Prostate Cancer
MRI as a tool for image-guided radiotherapy
removes the additional ionizing radiation that As described above, ultrahypofractionation is a
patients are exposed to using other forms of well-established treatment technique for the
image-guided radiotherapy. Although the dose delivery of definitive RT to localized disease.
for daily X-ray-based image guidance may be However, RT for prostate cancer is delivered in a
considered negligible (8–18 mGy daily) in com- variety of clinical settings across the entire con-
parison to the prescription doses for therapeutic tinuum of prostate cancer, from early-stage dis-
RT, this increases if X-ray tracking is used dur- ease to oligometastatic disease to widely
ing each fraction over a whole treatment course metastatic castrate-resistant disease. Thus, there
[33]. Repeated CT-based imaging has thus has naturally been interest and enthusiasm in
sometimes been avoided due to the risks associ- bringing the benefits of ultrahypofractionation to
ated with increased integral radiation dosing. In these treatment settings as well and MRIgRT
comparison, MRIgRT allows for constant may offer the ideal platform to usher in the
motion monitoring throughout treatment and expanding scope of ultrahypofractionation in
offers the ability to visualize the position of the prostate cancer.
prostate in relation to the CTV without the con- Post-prostatectomy RT is one such setting
comitant delivery of any additional ionizing where ultrahypofractionated radiation regimens
radiation. are being explored. Post-prostatectomy RT is
delivered in either the adjuvant setting (when a
constellation of high-risk features including
Inability to Measure Functional node-positive disease, pT3/4 disease, positive
Response margins, or high genomic classifier scores are
present at the time of surgery) or as early salvage
Functional MRI of the prostate can detect RT for a rising PSA following RT. In the event of
changes in the microenvironment of the tumor biochemical recurrence, postoperative RT pro-
and surrounding stroma, but with conventional vides the sole curative option after RP. Challenges
RT delivery and image guidance, the opportunity of hypofractionation RT in this context include
to assess functional changes in prostate tumor the highly deformable “negative space” of the
biology throughout treatment becomes forfeit. postoperative bed, raising concerns related to
The 1.5 T MRI of the Elekta Unity MR-Linac elevated toxicity with hypofractionation. As a
offers the potential to guide personalized RT due result, postoperative RT (most commonly deliv-
its biomarker capabilities, and there exists the ered as early salvage RT [36–38]) is standardly
potential to establish the cumulative effect of RT delivered over 9 weeks of treatment, and thus, the
on cancer cells and surrounding normal tissue convenience of RT in the definitive setting has yet
[34]. Intuitively, the stronger the magnet, the to be fully realized in the postoperative setting.
greater the signal-to-noise ratio and the greater As will be discussed below, MRIgRT may offer a
the separation between water, fat, and other solution [39].
metabolites meaning that biomarker techniques Analogous to recurrence following RP,
become more effective (fat suppression and spec- patients who underwent definitive RT can also
troscopy) [35]. Although yet to be clinically develop recurrent disease in the prostate gland,
implemented, it is hoped that quantitative and regionally, or distantly. While less than 15% of
qualitative intrafraction visualization of the real- men treated with EBRT will develop a biochemi-
time biological impacts of prostate radiotherapy cal recurrence at 5 years, cure is still attainable
will help to tailor the consecutive dose per frac- for the patients who recur locally [4]. When
tion [34]. imaged with MRI or positron emission tomogra-
290 L. Valle et al.
phy/computed tomography (PET/CT), local tionated salvage RT might also be visualized [48]
recurrences can be detected in up to 33% of these and treated with SBRT on the MR-Linac while
men before the PSA has risen above 2.0 ng/mL sparing the dose to the surrounding organs.
[40]. Moreover, PSMA PET in combination with
mpMRI have considerably improved the detec-
tion of local-only recurrent prostate cancer [41– Contouring, PTV, and OAR
44], which invites the therapeutic dilemma of
how best to salvage these local failures, if at all. Target Volumes
A recent meta-analysis of local salvage therapies
identified reirradiation with either ultrahypofrac- Standardized guideline-concordant target volume
tionated EBRT or HDR brachytherapy as the two delineation is followed for the prostate [49] and
local salvage options that carried the lowest pelvic lymph nodes [50] in the definitive local-
aggregate risks of GU and GI toxicity [4], thus ized setting [51] as well as for the prostate bed
increasing enthusiasm for offering curative local and pelvic lymph nodes in the postoperative set-
reirradiation in appropriately selected patients ting [52]. However, because of superior image
with a low risk of distant metastases and a favor- resolution, the contouring of the prostate on MR
able toxicity course with their initial course of imaging allows for more accurate and confident
radiation. One of the larger SBRT series report- contouring which in turn lends to reduced
ing long-term outcomes appreciated the impor- interobserver variability and greater speed of
tance of sparing normal tissue with reirradiation contouring compared to CT imaging [53]. This
and therefore used no PTV margins out of con- improved accuracy results in the contouring of
cern for inciting toxicity with local reirradiation smaller prostate volumes [54], consequently
[45]. Intuitively, the risk of toxicity could be low- leading to more attainable target coverage and
ered even further when PTV margin reduction OAR avoidance, as has been observed for other
and MRIgRT are combined. Of note, intrafrac- pelvic malignancies [55].
tional planar imaging which is often used for tri- The soft tissue differentiation of MRI also
angulation of fiducial markers as a surrogate of enables the dominant intraprostatic lesion(s)
prostate location during ultrahypofractioned (DIL[s]) to be visualized and accurately targeted
treatments can prove to be challenging with during planning and treatment (Fig. 16.2). This
numerous permanently implanted LDR seeds, process has been adopted on the MRIdian system
highlighting the possibility of more benefit with using rigid propagation while the lesion can be
MRIgRT techniques in the population of patients visualized and recontoured directly on the Elekta
previously treated with LDR brachytherapy who Unity MR-Linac [56], given higher magnet
are undergoing local salvage reirradiation. strength on this device.
As described above, CT-based treatment The benefit of being able to contour and target
delivery has been employed for safe reirradiation the dominant lesion was demonstrated in the
of localized disease, either with a whole-gland FLAME trial, where it was shown that a simulta-
approach or a focal-gland approach [46]. Given neous integrated boost to the dominant intrapros-
the promise of MRIgRT in the upfront treatment tatic lesion(s) improves biochemical disease-free
setting, there has been interest in evaluating its survival (bDFS) without impacting toxicity and
role in the setting of local salvage reirradiation. quality of life in localized intermediate- and high-
One-year clinical data from 37 men who had risk patients [57]. While FLAME employed con-
intraprostatic relapsed disease re-treated with ventional fractionation, Kerkmeijer et al. are also
MRIgRT found the treatment to be well tolerated exploring focal lesion boosting with SBRT in the
with no >2 acute toxicities and only 1 grade 3 late hypo-FLAME trial, which recently demonstrated
toxicity. In this series, the GTV was treated with favorable rates of acute grade 2 GU and GI toxic-
a 3-mm isotropic margin to create the PTV [47]. ity of 34% and 5%, respectively, when treating the
Applying the same concept, gross local relapses entire prostate to 35 Gy and boosting MRI-defined
following prostatectomy and conventionally frac- gross disease to 50 Gy in 5 fractions [58]. Long-
a b
Fig. 16.2 Visualization of the dominant intraprostatic lesions seen on (a) fused diagnostic MRI and (b) 0.35 T planning
image
term oncologic outcomes associated with these strated in a prospective setting, with commensu-
dose-escalated treatments are eagerly anticipated. rate decreases in both GU and GI toxicity [60].
Further extending the reach of DIL boosting in the
treatment of definitive localized prostate cancer,
the recently published 5-year outcomes from the OAR
DELINEATE [59] trial establish the safety and
efficacy of these treatments in the moderately The advantages of MRIgRT in the pelvis are fur-
hypofractionated setting as well, where simultane- ther appreciated by the superior visualization of
ous integrated boosting to 67 Gy in 20 fractions OARs. Specifically, the boundaries of the bladder
was similarly found to be safe and effective. and the rectum are difficult to visualize on CT but
can be more easily visualized on MRI (Fig. 16.3),
further enhancing OAR avoidance during treat-
PTV ment planning.
Urethral avoidance also remains paramount to
In the treatment of prostate cancer, the whole ensure acceptable rates of late toxicity following
prostate is standardly treated as the clinical target radiotherapy [61]. However, a limitation of exter-
volume (CTV) with a surrounding PTV margin nal beam radiotherapy compared to HDR brachy-
added to account for setup error, delineation therapy is suboptimal visualization of the urethra.
error, and inter- and intrafraction motion. This The urethra can however be visualized on the
PTV margin has varied across prospective clini- Unity device (Fig. 16.4a) and its delineation has
cal trials, but usually ranges from 5 to 7 mm, with been adopted to reduce the dose it receives and
allowances for reduction in PTV margin to no hence improve GU toxicity. Furthermore, on the
less than 3 mm posteriorly, where overlap with MRIdian device, investigational 0.35 T MRI
the rectal OAR is all but assured. While MRIgRT sequences are under development that would also
does not alter CTV delineation, it does enable enable superior visualization and delineation of
considerable reductions in PTV margin. For the urethra in an effort to minimize long-term gen-
example, in the MIRAGE trial which will be dis- itourinary toxicity (Fig. 16.4b) [62].
cussed extensively below, the safety and efficacy Vascular structures can also be better visual-
of PTV margin reduction to 2 mm was demon- ized on MRI, such that treatment with MRIgRT
292 L. Valle et al.
a b
Fig. 16.3 Prostate and the surrounding normal tissue as visualized with (a) CT-based techniques vs (b) MR-based
techniques
a b
Fig. 16.4 Visualization of the urethra on (a) the Unity urethral visualization include MRIgRT 3D HASTE (top)
device as observed on the 1.5 T T2 3D Tra and (b) the and MRIgRT 3D TSE (bottom). (MRIdian images bor-
MRIdian MRIgRT platforms. MRIdian sequences with rowed from Pham et al. [63])
offers the potential to outline and limit dose to tures on the day of treatment and therefore enable
such structures [33]. For example, sparing of the maximal sparing of dose to important vascular
internal pudendal artery has been shown to structures, particularly when adaptive planning is
improve sexual dysfunction [63], which is now pursued.
the most prevalent long-term side effect of pros-
tate radiotherapy [64]. The success of the sparing
of critical vascular structures has been previously DVH Constraints
demonstrated and thus highlights a need for
MRI-based radiation planning [65], where such As with other anatomic treatment sites in radi-
structures can be reliably visualized. MRIgRT ation oncology, there exist no universally
not only allows for offline MR-guided planning accepted or standardized dose constraints for
but also the ability to visualize these vital struc- ultrahypofractionated prostate RT. However,
Table 16.1 Dosimetric objectives for target volumes and Table 16.2 Dosimetric objectives for target volumes and
OARs for definitive ultrahypofractionated treatment of OARs for definitive moderately hypofractionated treat-
intact prostate cancer on the MIRAGE clinical trial ment of intact prostate cancer on the GARUDA clinical
trial
Structure Dosimetric objective
PTV V40 Gy ≥95% Structure Dosimetric objective
PTV V42 Gy <30% PTV V60 Gy ≥95%
Rectum V20 Gy <50% Rectum V24 Gy <80%
Rectum V42 Gy <0.03 cc Rectum V52 Gy <30%
Rectum V38 Gy <2 cc Rectum V56 Gy <15%
Rectal wall V24 Gy <50% Rectum V60 Gy <3%
Posterior rectum Max <16 Gy Rectal wall V30 Gy <50%
Anal canal V30 Gy <0.03 cc Posterior rectum Max <24 Gy
Anal canal V20 Gy <5 cc Small bowel V50 Gy <17 cc
Small bowel V25 Gy <0.03 cc Bladder V40 Gy <50%
Small bowel V20 Gy <30 cc Bladder V48 Gy <25%
Sigmoid colon Max <38 Gy Bladder V59 Gy <5%
Sigmoid colon V30 Gy <1 cc Right femoral head V50 Gy <5%
Sigmoid colon V25 Gy <30 cc Left femoral head V50 Gy <5%
Bladder V42 Gy <0.03 cc Penile bulb V50 Gy <50%
Bladder V39 Gy <2 cc Penile bulb V60 Gy <10%
Bladder V20 Gy <40% Urethra Max <63 Gy
Bladder V40 Gy <5%
Right femoral head V20 Gy <5 cc
Left femoral head V20 Gy <5 cc Table 16.3 Dosimetric objectives for target volumes and
Skin V15 Gy <0.03 cc OARs for postoperative ultrahypofractionated salvage
Penile bulb V24.8 Gy <5% treatment of recurrent prostate cancer on the EXCALIBUR
clinical trial
Penile bulb Mean <16 Gy
Urethra V42 Gy <0.03 cc Structure Dosimetric objective
PTV V100% ≥95%
Rectum V27.5 Gy <45%
dose constraints employed in prospective Rectum V32.5 Gy <30%
studies where MRIgRT was employed are Rectum V33.75 Gy <25%
recapitulated in Tables 16.1, 16.2 and 16.3 Rectal wall V35.7 Gy <0.03 cc
Rectal wall V24 Gy <50%
below. The dose constraints in Table 16.1 ref-
Anal canal V30 Gy <0.03 cc
erence those used to treat intact localized Anal canal V20 Gy <5 cc
prostate cancer with ultrahypofractionated RT Small bowel V25 Gy <0.03 cc
on the MIRAGE trial. The dose constraints in Small bowel V20 Gy <30 cc
Table 16.2 reference those used to guide treat- Sigmoid colon V30 Gy <1 cc
ment of intact localized prostate cancer with Bladder V35.7 Gy <0.03 cc
moderately hypofractionated RT on the Bladder V32.5 Gy <35%
Bladder V30 Gy <45%
GARUDA trial. And the dose constraints in
Bladder wall V30 Gy <16 cc
Table 16.3 reference those used to guide ultra-
Bladder wall V25 Gy <18 cc
hypofractionated salvage RT to the prostate Bladder wall V20 Gy <20 cc
bed on the EXCALIBUR trial. Corresponding Bladder wall V20 Gy <34%
DVHs for approved plans are shown in Right femoral head V20 Gy <5 cc
Figs. 16.5, 16.6 and 16.7. Left femoral head V20 Gy <5 cc
Penile bulb V24.8 Gy <5%
294 L. Valle et al.
DVH Structures
100 Skin
90 O_Femr_rt
80
% Coverage of structure
O_Femr_lt
70
O_Bldr
60
50 O_AnalCanal
40 C_Prst
30
O_Bwel_Sm
20
O_Pblb
10
0 N_Bladder_Setup
10 20 30 40
GTV42
Dose (Gy)
C_PrxSV
C_Prst+PrxSV
Urethra
Urethra Avoidance
PTV40
O_Rctm
Fig. 16.5 Dose volume histogram (DVH) from a patient with intact prostate cancer treated with ultrahypofractionated
RT on the MIRAGE trial
DVH Structures
100 C_Prst
90 C_Prx+SV
80
C_SV
70
60
G_GTV1
50 G_GTV2
40 GTV68
30
O_AnalCanal
20
O_Bldr
10
0 O_Bwel
10 20 30 40 50 60 70
O_Femr_Lt
Dose (Gy)
O_Femr_Rt
O_Pblb
O_Rctm
O_Rctm_Pt
O_Rctm_WI
O_Sigmoid
Fig. 16.6 Dose volume histogram (DVH) from a patient with intact prostate cancer treated with moderately hypofrac-
tionated RT on the GARUDA trial
DVH Structures
100 Skin
90 O_Femr_rt
80
O_Femr_lt
70
O_Bldr
60
50 O_Rctm
40 C_PrstBed
30
O_Pblb
20
O_Anal Canal
10
0 O_Bwel_Sm
10 20 30 40
T_Rctm
Dose (Gy)
O_Skin
T_Bldr
O_Sigmoid
T_PTV25
O+Bldr_Setup
PTV_PrstBed
Fig. 16.7 Dose volume histogram (DVH) from a patient with intact prostate cancer treated with postoperative ultrahy-
pofractionated RT on the EXCALIBUR trial
Dose Escalation 4.9% of men enrolled, respectively [67]. This has

prompted caution for whole-gland dose escala-
SBRT dose escalation trials have been carried out tion above 40 Gy, at least with CT-based delivery
using CT-guided delivery with mixed success. A approaches.
phase I/II dose escalation study from UT At the same time, data from Memorial Sloan
Southwestern famously concluded that “the sky Kettering demonstrates that incremental dose
is not the limit” after observing an increase in escalation from 32.5 to 35–37.5 Gy to 40 Gy
acute and late severe (grade 3 or greater) toxicity improves biochemical failure rates, with 5-year
in patients receiving 50 Gy in 5 fx. In their study, cumulative incidence rates of PSA failure for
dose was incrementally increased from a total each of these four dose levels corresponding to
prescription dose of 45 Gy to 47.5 Gy–50 Gy 15%, 6%, 0%, and 0%, respectively. The trial
[66]. Of note, the investigators on this study took also mandated posttreatment biopsies at 2 years
particular care to minimize toxicity by requiring and found that the incidence of a positive post-
Fleet enema and rectal balloon placement for treatment biopsy was 47.6%, 19.2%, 16.7%,
each treatment, 4 mg dexamethasone for each and 7.7%, respectively, for the 4 dose arms
treatment, as well as alpha-adrenergic antagonist (p = 0.013). Moreover, dose escalation to 40 Gy
usage. Yet, in the 47 patients who received 50 Gy was well-tolerated, with no grade 3 or 4 acute
in 5 fractions, within 500 days of treatment, late GU or GI toxicities observed, only 1 late grade
grade 4 GI and GU toxicities were reported at 3 urinary toxicity (urethral stricture) in the
3.3% and 1.6%, respectively, whereas late grade 40-Gy dose arm (the stricture was corrected
3 GI and GU toxicities were reported in 6.6% and with transurethral resection), no grade 4 late GU
296 L. Valle et al.
toxicity and no late grade 3 or 4 GI toxicity ular adaptive workflows. The adaptive work-
reported. flows for both Unity and VR systems are
So in the context of CTgRT, dose escalation to described below:
40 Gy appears to strike the appropriate balance
between biochemical control and toxicity, with
dose escalation beyond 40 Gy potentially carry- Unity
ing increased risks of acute and late GI and GU
toxicity. However, given the biochemical Pretreatment Phase
recurrence- free survival benefits appreciated The Elekta Unity MR-Linac has two adaptive
from DIL boosting, future attention has under- online planning methods: adapt-to-position
standably focused on dose de-escalation of the (ATP) and adapt-to-shape (ATP), with both
prostate gland itself, as long as DILs (which have requiring an initial plan to be generated on a ref-
been shown to represent the most likely location erence image (CT or MRI) [19].
of radiorecurrent failure) can be successfully In the pretreatment phase the patient undergoes
dose escalated. The opportunities for studying bladder and rectum preparation to ensure they are
this dose modification framework with MRIgRT imaged and treated with a comfortably full bladder
are ripe, with the hope that further gains in both and an empty rectum. The patient then attends a
efficacy and toxicity reduction can be realized planning CT and/or MRI session which is used to
with MRIgRT through its ability to deliver pre- create the reference plan. The reference planning
cise and adaptive radiotherapy. scans are acquired with the patient adequately
immobilized. The contours are then outlined on
either CT or the MRI depending on the workflow
n-Table Adaption for Unity
O of the treating center. At present, it is permissible
(“Unity”) and MRIdian (“ViewRay”) to use an MR-only workflow and this is likely to
become more common in the future.
The possibility to adapt offers an important the- The MR-Linac reference plan is generated using
oretical advantage for MRIgRT in prostate can- the Monaco treatment planning system (TPS)
cer. However, it is worth noting that adapting is (Elekta AB, Stockholm, Sweden, V5.40.00) using
not essential for achieving improved outcomes the GPUMCD dose calculation algorithm [68]. The
in the definitive setting, as none of the patients secondary dose calculations are generated with the
enrolled on the MIRAGE [60] trial received Raystation TPS (Raysearch Laboratories,
adaptive treatments. Nevertheless, the adaptive Stockholm, Sweden, V8.0.0.61) with a collapsed
workflow may still offer additional benefits in cone dose calculation algorithm [31]. An example
settings with adequate resources to support reg- treatment plan is shown in Fig. 16.8.
a b
Fig. 16.8 Radiotherapy plans for prostate only SBRT from (a) Unity and (b) MRIdian. Both treatments show that the
dominant intraprostatic lesion has been contoured
Treatment Phase t omography [PET]) to the simulation MRI is per-

On the day of treatment, the patient will have formed. Target volumes and OARs are delineated
undergone adequate bowel and bladder prepara- on the primary simulation MRI. Planning, dose
tion. Once on the treatment couch, a session calculations, and QA then proceed prior to
image is acquired with a T2 3D Tra Philips treatment.
sequence (1-mm slice thickness).
In the “adapt-to-position” workflow, the refer- Treatment Phase
ence image is registered to the daily MR image On the day of treatment, a pilot image is first
and the multileaf collimator (MLC) leaves are acquired to confirm appropriate bladder filling
adapted to the translation-only rigid registration. and rectal emptying. This is followed by a high-
The adapted plan is then recalculated on the ref- resolution session MRI scan. This higher-
erence image. This adaption workflow does not resolution scan is fused to the treatment planning
account for any changes in the anatomical shape MRI based on the prostate CTV. Image verifica-
of the target or OARs [31]. tion is then performed and OARs are recontoured
In the “adapt-to-shape” workflow, the refer- if needed. Dose to the new OAR volumes is then
ence plan is deformably registered to the daily predicted, and if it is decided that adaptive plan-
MR image and the contours are propagated either ning will be pursued, then the plan is reoptimized
rigidly or deformably. The contours are then to improve dosimetry over the initially predicted
edited by the treating physician to match the dosimetry. The reoptimized plan is then verified
daily anatomy as seen on the session MRI. Based and approved and online QA is performed. A
on the updated contours, Monaco reoptimizes the repeat MR image is taken to verify positional
treatment plan using the optimization constraints accuracy, and if positioning is felt to be accept-
derived from the original plan [69]. able, the gating window is set and the treatment is
Once the online daily plan has been generated, delivered. If adaptive planning is not deployed,
checks are performed to ensure the plan parame- the workflow proceeds directly to plan verifica-
ters are as expected. A second dose calculation tion and online QA following image verification.
may then be performed, and a repeat verification Cine imaging is used for continuous motion
image is acquired. If there has been further monitoring during dose delivery. The prostate
motion of the prostate during the ATS planning and the rectum are monitored in real-time sagit-
process, then a further ATP shift may be required tal, axial, and transverse plans. If the prostate
prior to dose delivery. moves outside the motion monitoring volume,
the treatment is stopped until the prostate returns
to the correct position [69].
ViewRay
Pretreatment Phase Evaluation

As with conventional IGRT, the pretreatment
phase begins with a simulation scan in the desired Plan Evaluation
treatment position. CT and MR images are
acquired on a traditional CT scanner and the As with CTgRT, a homogenous dose distribution
MR-Linac, respectively, although derivation of within the prostate gland is preferred, though this
synthetic CT scans from MR-only images has can be more challenging if DIL boosting is
been investigated [70] as part of an MRI-only employed. Planning evaluation goals are dis-
workflow that could reduce systematic errors cussed in “DVH Constraints” above.
introduced by image fusion [71]. Registration of It is worth noting that dosimetry can also be
the diagnostic MRI and other relevant imaging impacted by the technical inability of MRIgRT
modalities (including prostate-specific mem- devices to offer modulated arc therapy, as well as
brane antigen [PSMA] positron emission the inability to deploy non-coplanar beam
298 L. Valle et al.
arrangements. While these techniques have both simulation, to the anatomy on the day of treat-
been shown to assist in improving conformality ment in comparison to cone beam CT imaging
and lower radiation doses to OARs [72], most [77]. Furthermore, since the need for surrogate
patients are likely to be adequately served by markers for prostate triangulation is abrogated
MRIgRT despite these limitations. Indeed, a with MRIgRT, this technology also removes the
direct comparison between MRI-guided IMRT need for invasive fiducial marker insertion which
and conventional VMAT planning approaches is unpleasant for patients and introduces the risk
demonstrated similar dose distributions in the of bleeding and infection in a patient population
high-dose region and the target volume DVH that may harbor an aversion to surgically invasive
parameters (D98%, D50%, D2%, and V95%) and simi- procedures.
lar homogeneity indices (HI). While MRIgRT Improved image matching on the day of treat-
resulted in a worse conformity index (CI) when ment and more accurate contouring of both treat-
compared to VMAT, MRIgRT planning demon- ment targets and organs at risk are estimated to
strated similar OAR sparing when normal PTV lead to more precise radiotherapy which could
margins were used, but significantly better spar- result in improved toxicity.
ing once reduced PTV margins were applied
[73]. Dosimetric shortfalls, however, may arise in
patients with anatomy that renders planning more Outcome
difficult, for whom fewer planning solutions may
be available with the limitations of MRIgRT. Definitive Prostate
Furthermore, because the primary mechanism
of action of RT involves activation of secondary Superior image quality and the ability to account
particles which carry an electronic charge, expo- for interfraction and intrafraction motion is ulti-
sure of these secondary charged particles to a mately of clinical relevance because it mitigates
strong magnetic field can alter their movement the errors that are factored into the CTV to PTV
and consequentially alter radiation dosimetry. margins, enabling a safe reduction in PTV mar-
The electron return effect, also known as the gin which may consequentially reduce the radio-
Lorentz force, is an example of such a phenome- therapy dose to surrounding normal tissue and in
non where electrons move in a circular rather turn reduce acute toxicity. These theoretical ben-
than linear pattern in the presence of a magnet efits have recently been borne out with phase III
[74, 75]. While this effect can, and should, be clinical data from the University of California
accounted for during radiation treatment plan- Los Angeles. The MIRAGE trial is a single-
ning, it is predominately a concern at air-tissue center trial which randomized 156 men receiving
interfaces and thus less problematic for the treat- SBRT (40 Gy in 5 fractions) to either a CT-guided
ment of internal organs such as the prostate, par- RT (CTgRT) approach employing a 4-mm CTV
ticularly with efforts to reduce rectal gas in place to PTV margin versus a MRIgRT approach which
[76]. employed a 2-mm CTV to PTV margin [60].
Final reporting demonstrated a significant reduc-
tion in acute grade ≥2 genitourinary (GU) toxic-
Evaluation of Daily Setup ity with MRI versus CT guidance (24.4% vs
43.4%, p = 0.01) as well as a similarly significant
The superior visualization and reduced interob- reduction in grade ≥2 gastrointestinal (GI) toxic-
server interpretation of pelvic anatomy when ity (0% vs 10.5%, p = 0.003). This translated to a
imaged by MR has already been discussed, but in 60% reduction in the odds of grade ≥2 GU toxic-
addition to advantages with target and OAR ity with the tighter PTV margins enabled by
delineation, MRI also enables more accurate and MRIgRT. Patient-reported urinary and bowel
reproducible matching of the anatomy on the function were also better preserved at the 1-month
treatment plan, as seen at the time of CT/MR time point. Daily treatment adaption was not per-
formed on any of the patients enrolled in EXCALIBUR (NCT04915508) is the phase II

MIRAGE, suggesting that MR guidance and gat- nonrandomized successor to SCIMITAR that
ing may be enough to significantly reduce the will exclusively treat with MRIgRT and deter-
toxicity that would have been seen with mine the change in GI and GU reported outcome.
CTgRT. MRIgRT may emerge as the preferred MRIgRT is similarly being evaluated on the
treatment modality for men with prostate cancer SHORTER trial (NCT04422132), which will
in centers that have MRIgRT capacity. It remains randomize patients to either 36.5 Gy in 5 frac-
to be seen whether the gains achieved in the tox- tions or 55 Gy in 20 fractions on the MR-Linac
icity profile of SBRT could be further improved with the aim of comparing the toxicity profile
if daily adaption was incorporated in to the between these moderate and ultrahypofraction-
workflow. ated arms.
Post-prostatectomy Potential Limitations of MRIgRT
While the majority of studies in MRIgRT have An important and practical limitation of MRIgRT
focused on targeting and treating intact prostate treatment delivery relates to the design of the
cancer in the definitive setting, the role of MR-linear accelerator. As opposed to the “open”
MRIgRT in the post-prostatectomy setting and its nature of CT-guided linear accelerator treatment,
potential improvements in convenience and effi- the 70-cm closed-bore MRIgRT delivery system
cacy have recently garnered interest. An increase requires a much tighter space, resembling that of
in understanding of the deformable and dynamic a traditional MRI. As a result, patients with a
nature of surrounding OARs in the post- large body habitus or patients with claustropho-
prostatectomy setting, including the penile bulb bia or anxiety surrounding treatment may not be
[78], has further motivated interest in the careful well-suited for treatment with MRIgRT. As diag-
implementation of MRI-guided ultrahypofrac- nostic multiparametric MRI is a critical compo-
tionated radiotherapy in the post-prostatectomy nent of the initial workup of prostate cancer, this
setting. In the SCIMITAR study, men were offers a natural “test” of which patients may be
treated with 30–34 Gy in 5 fractions to the pros- poorly suited to undergo MRIgRT-based therapy,
tate bed under either CT-guided RT or with the option to trial anxiolytics to aid in patient
MRIgRT. With a median follow-up of comfort and tolerability.
29.5 months, patient-reported outcomes showed Treatment time on the MR-Linac is longer
baseline normalization in EPIC-26 QOL scores than traditional methods of radiotherapy delivery
by 3 months post SBRT, suggesting that post- due to the more complex workflow. Adaptive
radical prostatectomy SBRT is well-tolerated on radiotherapy with repeat contouring and replan-
both platforms [39]. In total, three patients had ning in particular takes additional time, with
grade 3 toxicity, yet no patients receiving patients on the table for 45–60 min [31, 79]. This
MRIgRT reported grade 3 GU or grade ≥2 GI is particularly challenging for the treatment of
toxicity. MRIgRT was also independently associ- men with prostate cancer as full bladders have
ated with improved any-grade GI toxicity and been historically required to physically displace
improved bowel QOL. The apparent benefits of small bowel out of the treatment field, but can be
MRIgRT are most likely attributed to the reduc- difficult for patients to achieve for long periods of
tion in PTV margins with a CTV to PTV margin time.
of 3 mm compared to 5 mm used in CTgRT. This In a workflow involving gating, the size of the
reduction in PTV margins is made possible by gating margins as well as the threshold of CTV
improved soft tissue definition and contouring allowable outside of the gating margin will also
using MRI as well as consistently accurate setup influence treatment times, with more rigid gating
enabled by MR imaging. boundaries contributing to longer treatment times
300 L. Valle et al.
owing to an increased likelihood of prostate CTV Given the later-in-life presentation of most
extending outside of the allowable margin. prostate cancers, there may be a greater preva-
However, these concerns about treatment length lence of relative medical contraindications to
must be balanced against accuracy of treatment treatment on MRI-guided devices in this patient
delivery, and individual centers or physicians population, thereby theoretically reducing num-
may adopt different thresholds. Reassuringly, a bers of patients suitable for this technology.
validated patient-reported treatment question- Chief among these relative contraindications
naire which collected responses from five centers include the presence of a pacemaker or implant-
found that patient-reported experience and satis- able defibrillator, as well as metallic prostheses.
faction were consistently high despite the length While treatment of pacemaker-dependent
of treatment [79]. patients with MRIgRT is not an absolute contra-
Moreover, a recent series examined motion- indication, careful multidisciplinary collabora-
compensated dose distributions from 20 patients tion with electrophysiologists, device
for all prostate positions and only prostate posi- representatives, and radiation therapists is
tions within the gating window and found that required to ensure treatment safety and contin-
beam gating actually had a negligible dosimetric ued pacemaker function. Interrogation of the
impact in online-adaptive MRIgRT of prostate pacemaker prior to treatment, following each
cancer, albeit in a cohort with low center of mass fraction, and after completion of treatment is
translations to begin with. They also found that encouraged. Workflows have been published to
even without gating, reducing the prostate-to- ensure safe stereotactic radiation therapy deliv-
PTV margin from 5 to 3 mm maintained D95% ery in 1.5 T devices and the importance of daily
≥100%, suggesting that PTV margins might be dose monitoring with thermoluminescent
successfully reduced without the need for a dosimeters is emphasized to ensure dose toler-
treatment-prolonging gating approach in the ances for the pacemaker remain well below
majority patients [80]. 2 Gy, even when the pacemaker does not lie
To address the dilemma of lengthy treatment within proximity to the treatment field as is
times and the requirement for full bladders dur- often the case for prostate cancer patients [82].
ing treatment, research is ongoing as to whether Metal prostheses and vascular stents are other
real-time dose prediction can be deployed to examples of metallic implants that can generate
assess whether bladder constraints will be met challenges for MRIgRT. These prostheses, par-
when bladders are not consistently filled, enabling ticularly when located in the hip and low pelvic
treatment to potentially move forward with regions, can lead to geometric distortions and
incompletely filled bladders if reasonable dosim- signal loss in neighboring treatment targets or
etry can be achieved. An alternative strategy for OARs.
circumventing the requirement of a full bladder Magnet strength is likely to affect the mag-
would be to rely on online adaption for each frac- nitude of these distortions. A pilot study of
tion—a strategy which has been shown to three patients from the Netherlands reported
improve both target and OAR dosimetry in their experience with a workflow that employed
underfilled bladders in a single institution cohort B0 mapping at the time of simulation to esti-
of 22 patients [81]. mate the magnitude of distortions during
MRIgRT and guide the decision on eligibility tiation of hormone therapy can render challenges
for MRIgRT of prostate cancer patients with for DIL identification with both the MRIdian and
metallic hip implants [83]. Taken together, the Unity devices. Therefore, offline fusion of
these points suggest that metallic implants and diagnostic MRI is often required to enable accu-
devices should not be viewed as an absolute rate contouring of DIL(s). This additional step
contraindication to MRIgRT, though additional has the potential to introduce further errors into
protocols and procedures to ensure patient target delineation and hampers the verification of
safety and appropriate selection for MRIgRT boost volume alignment on the day of treatment.
are paramount. Additionally, functional sequences are more
The maximum craniocaudal field length for challenging to perform with the lower magnet
the Unity device is 22 cm, whereas the MRIdian strength on the MRIdian device.
device has a maximum length of 24 cm [84].
Many patients receiving definitive treatment for
prostate cancer will receive radiation to the pros- Ongoing Trials in MRIgRT
tate alone, and in these scenarios, field length for Prostate Cancer
concerns are unlikely to arise.
The selection of patients who benefit from Trials taking place on the Unity MR-Linac are
radiotherapy to the prostate and pelvic lymph investigating the efficacy and toxicity of treating
nodes in the definitive setting is controversial, men with ultrahypofractionated regimes in as lit-
though men with very-high-risk disease and no tle as 2 fractions as well as the ability to visualize
evidence of distant metastatic disease on molecu- and boost the dominant lesion. Furthermore, the
lar imaging will likely benefit from elective irra- ability to avoid dose delivery to OARs is being
diation of the pelvic lymph nodes [85]. In the realized, with the ERECT trial investigating the
post-prostatectomy setting, data from RTOG dose sparing of the neurovascular bundle, inter-
0534 also supports routine inclusion of pelvic nal pudendal artery, corpora cavernosa, and
lymph nodes for some patients with a rising PSA penile bulb [86] (Table 16.4).
following prostatectomy [3]. Trials investigating MRIgRT with the
In such situations where elective pelvic lymph MRIdian device are similarly interested in
node volumes are included and/or boosting of reducing total fractions and increasing dose per
gross pelvic nodal disease is desired, MRIgRT fraction, with the FORT trial as an example that
may not be suitable for all patients due to the is assessing the non-inferiority of prostate RT
need to allow for a PTV margin and penumbra delivery in 2 versus 5 fractions. Other investiga-
for dose calculation to OARs. Thus, patient selec- tional questions center around the safety and
tion on the basis of radiation field size may feasibility of conventional dose escalation to
become an important consideration, particularly DILs, as will be investigated in the SIBRT trial,
for the Elekta Unity device. as well as dose escalation in a single fraction
The lower 0.35 T magnet strength of the employing spatially fractionated GRID therapy,
MRIdian device limits direct visualization of the which will be assessed in the LEAD study
DIL during treatment planning. Additionally, ini- (Table 16.4).
302
Table 16.4 Ongoing trials in prostate cancer MRIgRT

Planned
number of Dose delivery in
Trial NCT Device Phase patients Primary outcome Dose delivered in treatment arm standard arm Accrual status
HERMES 04595019 Unity II 46 Acute grade 2+ GU toxicity 24 Gy in 2 fractions to the prostate with an 36.25 Gy to the Open to
intraprostatic boost to 27 Gy in 2 fractions prostate in 5 recruitment
fractions
UltraHypo 05183074 Unity II 50 Incidence of acute GU and GI Not stated NA Open to
toxicity recruitment
ERECT 04861194 Unity II 70 Erectile dysfunction over 36.25 Gy in 5 fractions with sparing of the NA Open to
3 years post SBRT neurovascular bundle, IPA, corpora recruitment
cavernosa, and penile bulb
Boost (MSK) 04997018 Unity II 91 A reduction in posttreatment 40 Gy in 5 fractions to the prostate with NA Open to
biopsy rates at 24 months 45 Gy to the dominant lesion recruitment
AFFIRM 05373316 Unity I 95 Acute GI and GU toxicity 35 Gy to the prostate with 50 Gy to the NA Open to
intraprostatic tumor in 5 fractions recruitment
2SMART 03588819 Unity I 30 Quality of life using EPIC 26 Gy in 2 fractions to the prostate and the NA Open to
DIL dose of up to 32 Gy in 2 fractions recruitment
delivered 1 week apart
iSMART 05600400 Unity II 144 Change in quality of life 27 Gy in 2 fractions to the prostate Five every other Open to
function day fractions of recruitment
8 Gy
LEAD 01411319 ViewRay I 25 Grade 2 or higher physician- 12–14 Gy in 1 fraction to the mpMRI N/A Completed
reported treatment-related defined GTV on day 1, followed by
adverse events standard 38 fraction IMRT
FORT 04984343 ViewRay II 136 Change in patient-reported GI 37.5 Gy in 5 fractions to the prostate 25 Gy in 2 Recruiting
symptoms using EPIC fractions to the
prostate
SIBRT 03664193 ViewRay II 30 Feasibility 35 Gy in 5 fractions to the prostate with NA Completed
prostate lesion SIB to 37.5, 40, 42.5, or
45 Gy
EXCALIBUR 04915508 ViewRay II 102 Change in patient-reported GI 30–34 Gy in 5 fractions N/A Recruiting
symptoms using EPIC
SHORTER 04422132 ViewRay II 134 Change in patient-reported GI 32.5 Gy in 5 fractions 55 Gy in 20 Recruiting
symptoms using EPIC fractions
L. Valle et al.
References 15. Kupelian P, et al. Stereotactic body radiation therapy

for prostate cancer: rational and reasonable. Pract
Radiat Oncol. 2015;5:188.
16. Hodges JC, et al. Cost-effectiveness analysis of ste-
statistics, 2022. CA Cancer J Clin. 2022;72:7–33.
reotactic body radiation therapy versus intensity-
2. Hamdy FC, et al. Fifteen-year outcomes after moni-
modulated radiation therapy: an emerging initial
toring, surgery, or radiotherapy for prostate cancer. N
radiation treatment option for organ-confined prostate
Engl J Med. 2023;388:1547. https://doi.org/10.1056/
cancer. J Oncol Pract. 2012;8:e31s–7s.
NEJMoa2214122.
17. Buyyounouski MK, et al. Primary endpoint analysis
3. Pollack A, et al. The addition of androgen depriva-
of a randomized phase III trial of hypofractionated vs.
tion therapy and pelvic lymph node treatment to
conventional post-prostatectomy radiotherapy: NRG
prostate bed salvage radiotherapy (NRG Oncology/
Oncology GU003. Int J Radiat Oncol Biol Phys.
RTOG 0534 SPPORT): an international, multicen-
2021;111:S2–3.
tre, randomised phase 3 trial. Lancet Lond Engl.
18. Tandberg DJ, et al. Postoperative radiation therapy
2022;399:1886–901.
for prostate cancer: comparison of conventional ver-
4. Valle LF, et al. A systematic review and meta-analysis
sus hypofractionated radiation regimens. Int J Radiat
of local salvage therapies after radiotherapy for
Oncol Biol Phys. 2018;101:396–405.
prostate cancer (MASTER). Eur Urol. 2020;80:280.
19. Winkel D, et al. Adaptive radiotherapy: the Elekta
https://doi.org/10.1016/j.eururo.2020.11.010.
Unity MR-linac concept. Clin Transl Radiat Oncol.
5. NCC, N. NCCN clinical practice guidelines in oncol-
2019;18:54–9.
ogy: prostate cancer 2020. 2020.
20. Mutic S, Dempsey JF. The ViewRay system: magnetic
6. Brenner DJ, Hall EJ. Fractionation and protraction
resonance-guided and controlled radiotherapy. Semin
for radiotherapy of prostate carcinoma. Int J Radiat
Radiat Oncol. 2014;24:196–9.
Oncol Biol Phys. 1999;43:1095–101.
21. Moiseenko V, Liu M, Kristensen S, Gelowitz G,
7. Michalski JM, et al. Effect of standard vs dose-
Berthelet E. Effect of bladder filling on doses to pros-
escalated radiation therapy for patients with
tate and organs at risk: a treatment planning study. J
intermediate-risk prostate cancer: the NRG oncology
Appl Clin Med Phys. 2006;8:55–68.
RTOG 0126 randomized clinical trial. JAMA Oncol.
22. Zellars RC, et al. Prostate position late in the course
2018;4:e180039.
of external beam therapy: patterns and predictors. Int
8. Dearnaley D, et al. Conventional versus hypofraction-
J Radiat Oncol Biol Phys. 2000;47:655–60.
ated high-dose intensity-modulated radiotherapy for
23. Brand VJ, Milder MTW, Christianen MEMC,
prostate cancer: 5-year outcomes of the randomised,
Hoogeman MS, Incrocci L. Seminal vesicle inter- and
non-inferiority, phase 3 CHHiP trial. Lancet Oncol.
intra-fraction motion during radiotherapy for prostate
2016;17:1047–60.
cancer: a review. Radiother Oncol. 2022;169:15–24.
9. Lee WR, et al. Randomized phase III noninferior-
24. Alexander SE, et al. Prostate volume changes during
ity study comparing two radiotherapy fractionation
extreme and moderately hypofractionated magnetic
schedules in patients with low-risk prostate cancer. J
resonance image-guided radiotherapy. Clin Oncol (R
Clin Oncol. 2016;34:2325–32.
Coll Radiol). 2022;34:e383–91.
10. Catton CN, et al. Randomized trial of a hypofraction-
25. Ma TM, et al. Dosimetric impact of interfraction pros-
ated radiation regimen for the treatment of localized
tate and seminal vesicle volume changes and rotation:
prostate cancer. J Clin Oncol. 2017;35:1884–90.
a post-hoc analysis of a phase III randomized trial
11. Widmark A, et al. Ultra-hypofractionated versus
of MRI-guided versus CT-guided stereotactic body
conventionally fractionated radiotherapy for prostate
radiotherapy. Radiother Oncol. 2022;167:203–10.
cancer: 5-year outcomes of the HYPO-RT-PC ran-
26. Hötker AM, et al. Prostate cancer: assessing the
domised, non-inferiority, phase 3 trial. Lancet Lond
effects of androgen-deprivation therapy using quan-
Engl. 2019;394:385–95.
titative diffusion-weighted and dynamic contrast-
12. Fransson P, et al. Ultra-hypofractionated versus con-
enhanced MRI. Eur Radiol. 2015;25:2665–72.
ventionally fractionated radiotherapy for prostate can-
27. Kupelian P, et al. Multi-institutional clinical experi-
cer (HYPO-RT-PC): patient-reported quality-of-life
ence with the calypso system in localization and con-
outcomes of a randomised, controlled, non-inferiority,
tinuous, real-time monitoring of the prostate gland
phase 3 trial. Lancet Oncol. 2021;22:235–45.
during external radiotherapy. Int J Radiat Oncol Biol
13. Tree AC, et al. Intensity-modulated radiotherapy ver-
Phys. 2007;67:1088–98.
sus stereotactic body radiotherapy for prostate cancer
28. Zhu X, et al. Tradeoffs of integrating real-time track-
(PACE-B): 2-year toxicity results from an open-label,
ing into IGRT for prostate cancer treatment. Phys
randomised, phase 3, non-inferiority trial. Lancet
Med Biol. 2009;54:N393–401.
Oncol. 2022;23:1308–20.
29. de Muinck Keizer DM, et al. Prostate intrafrac-
14. Brand DH, et al. Intensity-modulated fractionated
tion motion during the preparation and delivery
radiotherapy versus stereotactic body radiotherapy for
of MR-guided radiotherapy sessions on a 1.5 T
prostate cancer (PACE-B): acute toxicity findings from
MR-Linac. Radiother Oncol. 2020;151:88–94.
an international, randomised, open- label, phase 3,
30. Kontaxis C, et al. Delivered dose quantification in
non-inferiority trial. Lancet Oncol. 2019;20:1531–43.
prostate radiotherapy using online 3D cine imaging
304 L. Valle et al.
and treatment log files on a combined 1.5 T magnetic of biochemical recurrence after low dose rate brachy-
resonance imaging and linear accelerator system. therapy. Diagn Interv Radiol. 2018;24:46–53.
Phys Imaging Radiat Oncol. 2020;15:23–9. 44. Fendler WP, et al. Assessment of 68Ga-PSMA-11
31. Dunlop A, et al. Daily adaptive radiotherapy for PET accuracy in localizing recurrent prostate cancer:
patients with prostate cancer using a high field a prospective single-arm clinical trial. JAMA Oncol.
MR-linac: initial clinical experiences and assessment 2019;5:856–63.
of delivered doses compared to a C-arm linac. Clin 45. Fuller D, et al. Retreatment for local recurrence of
Transl Radiat Oncol. 2020;23:35–42. prostatic carcinoma after prior therapeutic irradiation:
32. Glitzner M, et al. On-line MR imaging for dose vali- efficacy and toxicity of HDR-like SBRT. Int J Radiat
dation of abdominal radiotherapy. Phys Med Biol. Oncol Biol Phys. 2020;106:291.
2015;60:8869–83. 46. Fuller DB, et al. High dose ‘HDR-like’ prostate SBRT:
33. Sritharan K, Tree A. MR-guided radiotherapy for PSA 10-year results from a mature, multi-institutional
prostate cancer: state of the art and future perspec- clinical trial. Front Oncol. 2022;12:935310.
tives. Br J Radiol. 2022;95:20210800. 47. Michalet M, et al. Magnetic resonance-guided reirra-
34. van Schie MA, et al. Quantitative MRI changes dur- diation for local recurrence within the prostate or in
ing weekly ultra-hypofractionated prostate cancer the prostate bed: 1-year clinical results of a prospec-
radiotherapy with integrated boost. Front Oncol. tive registry study. Cancers. 2022;14:1943.
2019;9:1264. 48. Barchetti F, Panebianco V. Multiparametric MRI
35. Datta A, Aznar MC, Dubec M, Parker GJM, O’Connor for recurrent prostate cancer post radical prostatec-
JPB. Delivering functional imaging on the MRI- tomy and postradiation therapy. Biomed Res Int.
Linac: current challenges and potential solutions. Clin 2014;2014:316272.
Oncol (R Coll Radiol). 2018;30:702–10. 49. Salembier C, et al. ESTRO ACROP consensus guide-
36. Kneebone A, et al. Adjuvant radiotherapy versus early line on CT- and MRI-based target volume delineation
salvage radiotherapy following radical prostatectomy for primary radiation therapy of localized prostate
(TROG 08.03/ANZUP RAVES): a randomised, con- cancer. Radiother Oncol. 2018;127:49–61.
trolled, phase 3, non-inferiority trial. Lancet Oncol. 50. Hall WA, et al. NRG oncology updated international
2020;21:1331–40. consensus atlas on pelvic lymph node volumes for
37. Parker CC, et al. Timing of radiotherapy after intact and postoperative prostate cancer. Int J Radiat
radical prostatectomy (RADICALS-RT): a ran- Oncol Biol Phys. 2021;109:174–85.
domised, controlled phase 3 trial. Lancet Lond Engl. 51. Harris VA, et al. Consensus guidelines and contour-
2020;396:1413–21. ing atlas for pelvic node delineation in prostate and
38. Sargos P, et al. Adjuvant radiotherapy versus early pelvic node intensity modulated radiation therapy. Int
salvage radiotherapy plus short-term androgen J Radiat Oncol Biol Phys. 2015;92:874–83.
deprivation therapy in men with localised prostate 52. Michalski JM, et al. Development of RTOG consen-
cancer after radical prostatectomy (GETUG-AFU sus guidelines for the definition of the clinical target
17): a randomised, phase 3 trial. Lancet Oncol. volume for postoperative conformal radiation therapy
2020;21:1341–52. for prostate cancer. Int J Radiat Oncol Biol Phys.
39. Ma TM, et al. Quality-of-life outcomes and toxicity 2010;76:361–8.
profile among patients with localized prostate cancer 53. Parker CC, et al. Magnetic resonance imaging in the
after radical prostatectomy treated with stereotactic radiation treatment planning of localized prostate
body radiation: the SCIMITAR multicenter phase 2 cancer using intra-prostatic fiducial markers for com-
trial. Int J Radiat Oncol Biol Phys. 2022;115(1):142– puted tomography co-registration. Radiother Oncol.
52. https://doi.org/10.1016/j.ijrobp.2022.08.041. 2003;66:217–24.
40. Jansen BHE, et al. Detection of recurrent prostate 54. Pathmanathan AU, et al. Comparison of prostate
cancer using prostate-specific membrane antigen delineation on multimodality imaging for MR-guided
positron emission tomography in patients not meeting radiotherapy. Br J Radiol. 2019;92:20180948.
the Phoenix criteria for biochemical recurrence after 55. White I, et al. Interobserver variability in target vol-
curative radiotherapy. Eur Urol Oncol. 2021;4:821–5. ume delineation for CT/MRI simulation and MRI-
41. Pfister D, et al. 68 Ga-PSMA-PET/CT helps to select guided adaptive radiotherapy in rectal cancer. Br J
patients for salvage radical prostatectomy with local Radiol. 2021;94:20210350.
recurrence after primary radiotherapy for prostate 56. Westley R, Hall E, Tree A. HERMES: delivery of a
cancer. BJU Int. 2020;126:679–83. speedy prostate cancer treatment. Clin Oncol (R Coll
42. Panebianco V, et al. Prostate magnetic resonance Radiol). 2022;34:426–9.
imaging for local recurrence reporting (PI-RR): 57. Kerkmeijer LGW, et al. Focal boost to the intra-
international consensus-based guidelines on multi- prostatic tumor in external beam radiotherapy for
parametric magnetic resonance imaging for prostate patients with localized prostate cancer: results from
cancer recurrence after radiation therapy and radical the FLAME randomized phase III trial. J Clin Oncol.
prostatectomy. Eur Urol Oncol. 2021;4:868–76. 2021;39:787–96.
43. Valle LF, et al. Multiparametric MRI for the detection 58. Draulans C, et al. Primary endpoint analysis of the
of local recurrence of prostate cancer in the setting multicentre phase II hypo-FLAME trial for interme-
diate and high risk prostate cancer. Radiother Oncol. 73. Da Silva Mendes V, et al. Dosimetric comparison of
2020;147:92–8. MR-linac-based IMRT and conventional VMAT treat-
59. Tree AC, et al. Standard and hypofractionated dose ment plans for prostate cancer. Radiat Oncol Lond
escalation to intraprostatic tumor nodules in localized Engl. 2021;16:133.
prostate cancer: 5-year efficacy and toxicity in the 74. Abraham M. Theorie der Elektrizität. Zweiter
DELINEATE trial. Int J Radiat Oncol Biol Phys. Band: Elektromagnetische Theorie der Strahlung.
2023;115:305–16. Monatshefte Für Math Phys. 1906;17:A39.
60. Kishan AU, et al. Magnetic resonance imaging- 75. Hall WA, et al. The transformation of radiation oncol-
guided vs computed tomography-guided ste- ogy using real-time magnetic resonance guidance: a
reotactic body radiotherapy for prostate cancer: review. Eur J Cancer. 2019;1990(122):42–52.
the MIRAGE randomized clinical trial. JAMA 76. Yang Y, et al. Longitudinal diffusion MRI for treat-
Oncol. 2023;365:e226558. https://doi.org/10.1001/ ment response assessment: preliminary experience
jamaoncol.2022.6558. using an MRI-guided tri-cobalt 60 radiotherapy sys-
61. Rodda S, Morris WJ, Hamm J, Duncan tem. Med Phys. 2016;43:1369–73.
G. ASCENDE-RT: an analysis of health-related qual- 77. van Herk M, McWilliam A, Dubec M, Faivre-Finn C,
ity of life for a randomized trial comparing low-dose- Choudhury A. Magnetic resonance imaging-guided
rate brachytherapy boost with dose-escalated external radiation therapy: a short strengths, weaknesses,
beam boost for high- and intermediate-risk prostate opportunities, and threats analysis. Int J Radiat Oncol
cancer. Int J Radiat Oncol Biol Phys. 2017;98:581–9. Biol Phys. 2018;101:1057–60.
62. Pham J, et al. Evaluation of T2-weighted MRI for 78. Roy A, et al. Assessing inter-fraction changes in
visualization and sparing of urethra with MR-guided the size and position of the penile bulb during daily
radiation therapy (MRgRT) on-board MRI. Cancers. MR-guided radiation therapy to the prostate bed:
2021;13:3564. do we need to adjust how we plan radiation in the
63. Spratt DE, et al. Vessel-sparing radiotherapy for local- post-radical prostatectomy setting to reduce risk
ized prostate cancer to preserve erectile function: a of erectile dysfunction? Clin Genitourin Cancer.
single-arm phase 2 trial. Eur Urol. 2017;72:617–24. 2022;20:e227–32.
64. Hamdy FC, et al. 10-Year outcomes after monitoring, 79. Barnes H, et al. Development and results of a
surgery, or radiotherapy for localized prostate cancer. patient-reported treatment experience questionnaire
N Engl J Med. 2016;375:1415–24. on a 1.5 T MR-Linac. Clin Transl Radiat Oncol.
65. Lee JY, Spratt DE, Liss AL, McLaughlin PW. Vessel- 2021;30:31–7.
sparing radiation and functional anatomy-based 80. Wahlstedt I, et al. Gating has a negligible impact
preservation for erectile function after prostate radio- on dose delivered in MRI-guided online adaptive
therapy. Lancet Oncol. 2016;17:e198–208. radiotherapy of prostate cancer. Radiother Oncol.
66. Boike TP, et al. Phase I dose-escalation study of 2022;170:205–12.
stereotactic body radiation therapy for low- and 81. Leeman JE, et al. Magnetic resonance-guided pros-
intermediate- risk prostate cancer. J Clin Oncol. tate stereotactic body radiation therapy with daily
2011;29:2020. online plan adaptation: results of a prospective phase
67. Hannan R, et al. Stereotactic body radiation therapy 1 trial and supplemental cohort. Adv Radiat Oncol.
for low and intermediate risk prostate cancer-results 2022;7:100934.
from a multi-institutional clinical trial. Eur J Cancer 82. Yang B, et al. Magnetic resonance-guided radiation
Oxf Engl. 2016;1990(59):142–51. therapy of patients with cardiovascular implantable
68. Paudel MR, et al. Experimental evaluation of a GPU- electronic device on a 1.5 T magnetic resonance-
based Monte Carlo dose calculation algorithm in the Linac. Pract Radiat Oncol. 2022;12:e56–61.
Monaco treatment planning system. J Appl Clin Med 83. Keesman R, et al. Clinical workflow for treating
Phys. 2016;17:230–41. patients with a metallic hip prosthesis using mag-
69. Menten MJ, et al. Automatic reconstruction of the netic resonance imaging-guided radiotherapy. Phys
delivered dose of the day using MR-linac treatment Imaging Radiat Oncol. 2020;15:85–90.
log files and online MR imaging. Radiother Oncol. 84. Tocco BR, Kishan AU, Ma TM, Kerkmeijer LGW,
2020;145:88–94. Tree AC. MR-guided radiotherapy for prostate cancer.
70. Chen S, Qin A, Zhou D, Yan D. Technical note: Front Oncol. 2020;10:616291.
U-net-generated synthetic CT images for magnetic 85. Murthy V, et al. Prostate-only versus whole-pelvic
resonance imaging-only prostate intensity-modulated radiation therapy in high-risk and very high-
radiation therapy treatment planning. Med Phys. risk prostate cancer (POP-RT): outcomes from
2018;45:5659–65. phase III randomized controlled trial. Clin Oncol.
71. Tenhunen M, et al. MRI-only based radiation therapy 2021;1234:JCO2003282. https://doi.org/10.1200/
of prostate cancer: workflow and early clinical experi- JCO.20.03282.
ence. Acta Oncol. 2018;57:902–7. 86. Verkooijen HM. EREctile Function Preservation for
72. Ren W, et al. Dosimetric comparison of intensity- Prostate Cancer Radiation Therapy (ERECT); a pro-
modulated radiotherapy and volumetric-modulated arc spective phase II trial. 2021. https://clinicaltrials.gov/
radiotherapy in patients with prostate cancer: a meta- ct2/show/NCT04861194.
analysis. J Appl Clin Med Phys. 2016;17:254–62.
Breast Radiotherapy Using
MR-Linac
17
Chiara De-Colle, Anna M. Kirby, Desiree van den
Bongard, and Maureen Groot Koerkamp
Introduction ing the target. In breast radiotherapy, partial

breast irradiation has been adopted in recent
Long-term treatment-related toxicity remains a years to reduce the treatment volume. However,
major concern for patients affected by breast can- correct target visualization and definition is par-
cer. The favorable prognosis found especially in ticularly important when performing partial
patients diagnosed with low-risk early-stage breast irradiation. Neoadjuvant partial breast irra-
tumors directs the current research towards mini- diation could facilitate target definition. With this
mizing toxicity while maintaining optimal onco- approach, the volume of irradiated breast tissue
logical outcomes. One way to achieve this is by can be smaller than in the postoperative setting
reducing the irradiated volumes. Technical [1]. In addition, the breast tissue receiving the
improvements in modern radiotherapy help to highest doses could be subsequently surgically
achieve this without increasing the risk of miss- removed. Furthermore, MR-guided breast radio-
therapy performed on an MR-Linac offers the
possibility to better visualize the target and
C. De-Colle (*) potentially to reduce the margins used to com-
Department of Radiation Oncology, University pensate for inter- and intrafraction variation. In
Hospital and Medical Faculty, Eberhard Karls this chapter, the rationale and potential advan-
University Tübingen, Tübingen, Germany
tages as well as specific considerations and prac-
tical suggestions for performing breast cancer
A. M. Kirby
Department of Radiotherapy, Royal Marsden NHS
treatment on an MR-Linac, with a focus on neo-
Foundation Trust and Institute of Cancer Research, adjuvant partial breast irradiation, are discussed.
Sutton, United Kingdom
D. van den Bongard Breast Cancer Overview
Department of Radiation Oncology, Amsterdam
University Medical Centers, Cancer Center
Amsterdam, Cancer Treatment and Quality of Life,
Breast cancer is the most frequent female cancer,
Cancer Biology and Immunology, with an incidence of approximately 140 per
Amsterdam, The Netherlands 100,000 women and a lifetime risk of 1 every 7
e-mail: [email protected] women [2, 3]. With more than 2.26 million of
M. G. Koerkamp new cases of breast cancer diagnosed worldwide
Department of Radiation Oncology, University in 2020 [4], it accounts for one-third of all can-
Medical Centre Utrecht, Utrecht, The Netherlands
cers diagnosed in women [5].
308 C. De-Colle et al.
Despite improvements in diagnosis and treat- better results in terms of acute and late treatment-
ment, breast cancer remains the main cause of related toxicity and cosmesis [16, 17]. In the
death in women below 65 years and the third IMPORT LOW trial, significantly better results
most common cause of death after cerebral and in terms of changed breast appearance and breast
cardiovascular diseases in women older than hardness compared to whole breast were reported
65 years [6]. With 91,826 deaths in 2020 in the [17]. However, even though smaller compared to
EU and 43,250 in 2022 in the USA, the mortality whole breast irradiation, the irradiated breast area
rates are between 20 and 30 per 100,000 women when performing external-beam adjuvant partial
per year [2, 5]. Because of its high incidence and breast irradiation often represents 1/4 of the
mortality, but at the same time high long-term breast or more [17, 18], due to the need to cover
survival rates and because of predicted further all the postoperative changes, especially in the
increase in new cases each year [7], breast cancer era of oncoplastic surgery. Additionally, planning
and long-term treatment-related toxicities repre- target volume (PTV) margins up to 10 mm are
sent a major public health burden. traditionally required to account for the inter- and
In this context, early diagnosis is essential for intrafractional variability. Because both breast
better outcome. Thanks to diagnostic improve- irradiation and high-grade subcutaneous fibrosis
ments, the majority of breast tumors are nowa- have been shown to be correlated with the breast
days detected in early-stage and the 5-year overall volume receiving a high dose [17, 19, 20], small
survival rate for these patients’ subgroup reaches volumes for partial breast irradiation are strongly
almost 100% [8]. Radiotherapy plays a main role encouraged.
in the treatment of these patients. Attempts to Another approach, investigated more recently
omit radiotherapy in patients with low-risk breast in an increasing number of studies, is neoadju-
cancer revealed a more than ten-fold higher inci- vant partial breast radiotherapy, the goal of which
dence of relapses in those treated with breast- is to further reduce the irradiated area and conse-
conserving surgery and hormonal therapy alone quently minimize treatment-related toxicities and
[9], indicating that radiotherapy remains an improve cosmetic outcomes. Neoadjuvant partial
important part of the treatment of patients under- breast irradiation allows a more precise and less
going a breast-conserving approach. observer-dependent definition of the target,
Given the high survival rates, especially for which might translate in lower treatment-related
patients diagnosed with early-stage cancer, much toxicity [21]. With neoadjuvant partial breast
clinical research in this patient group focuses on irradiation, the possibility to treat the visible
reducing treatment-related toxicities, without tumor with a margin for the clinical target vol-
compromising oncological outcomes. Risk of ume (CTV) allows for smaller target volumes
late radiotherapy-induced effects such as heart compared to the breast CTV in the adjuvant
disease, lung injury, and secondary cancers setting.
appears minimal in patients treated with modern The new MR-Linac technology improves the
techniques, not requiring nodal irradiation and tumor visualization and enables for online adap-
nonsmokers [10–12]. However, side effects in tive radiotherapy, opening the possibility to
healthy breast tissue remain an issue. reduce the CTV-PTV margin and therefore fur-
One way to try to maintain the excellent onco- ther decrease the irradiated volume.
logical outcomes and reduce the radiotherapy-
related breast tissue toxicity is to reduce the
irradiated volume. A widely investigated Overcoming Challenges
approach is external-beam adjuvant partial breast with Traditional Approaches
irradiation. Partial breast irradiation reduces the
irradiated breast volume compared to whole There are two main challenges for target volume
breast irradiation and has become a standard reduction for partial breast irradiation that
approach in selected low-risk breast cancer MR-guided breast radiotherapy, compared to tra-
patients [13–15]. This approach appears to offer ditional breast techniques, might help to over-
17 Breast Radiotherapy Using MR-Linac 309
come. The first challenge is the visualization of location, or differences in breathing pattern or
the target area in order to precisely define the arm position compared to the planning CT. In the
gross tumor volume (GTV) and CTV. MRI offers setting of standard fractionated adjuvant breast
a better soft-tissue contrast and detailed view of irradiation, a reduction of the lumpectomy cavity
breast and tumor tissue, which could help to volume to less than a half has been described,
improve target definition in neoadjuvant partial mainly due to a reduction of the seroma [33]. All
breast irradiation. The second challenge is how to such factors contribute to increase in the inter-
minimize the irradiated volume. The latter could fractional variability and have to be included as
be achieved by visualizing the target before and uncertainties in a PTV margin. In a worst-case
during dose delivery and by adapting the treat- scenario, they may lead to a replan being required.
ment plan to inter- and intrafraction variabilities Depending on the correction strategy, PTV mar-
to reduce CTV-PTV margins. Hence, the gins up to 10 mm may be necessary [34].
MR-Linac may play a role in overcoming both Treatment on an MR-Linac could facilitate online
challenges. monitoring and adaptation to interfraction vari-
Regarding the first topic, in the adjuvant setting ability [35].
radiation oncologists need to determine the loca- Additionally, the traditional image guidance
tion of the postoperative tumor bed. This could be approaches cannot compensate for intrafraction
particularly large after breast tissue mobilization motion, which therefore also has to be compen-
and displacement following oncoplastic surgery. sated for in the PTV margin. Given the increasing
Moreover, the traditional CT-based delineation of use of hypofractionated (five-fraction) schedules
the CTV might be challenging when the resection [36], compensation for intrafraction motion
cavity cannot be easily distinguished [22], e.g., becomes increasingly important because the risks
because of the absence of a well-defined seroma or of a target missing might have greater conse-
when few clips are available (cavity visualization quences due to the lack of the possibility for the
scores 1–3 [23]). MR-based CTV delineation has inaccuracies in dose delivery to be compensated
been associated with lower interobserver variabil- during other fractions. Also, fewer but longer
ity, improved accuracy, and smaller CTVs in some fractions may require larger PTV margins
investigations [24–27]. However, other studies because intrafraction drift motion increases with
demonstrated no increase in consistency of CTV longer treatment time [37]. With the MR-Linacs,
contouring when MR was used [27–29], some- a precise target visualization before and during
times ending up in larger CTVs, mainly due to the irradiation is possible. Treatments performed at
inclusion of postoperative breast tissue changes the MR-Linac might allow for (almost) no PTV
which were not visible on CT [29–31]. In the neo- margins, because the interfractional variability is
adjuvant scenario, the target delineation of a tumor reduced and the intrafractional variability might
in situ appears to be more accurate and less be tracked and corrected online [37].
affected by interobserver variability compared to After decades of experience with postopera-
the target delineation in the adjuvant setting [1]. tive whole breast irradiation with tangential fields
Additionally, MR offers clear advantages in breast encompassing the whole breast, radiation oncol-
tumor visualization compared to CT [32]. ogists performing partial breast irradiation need
Therefore, the largest benefit of MR for breast to switch to a concept where the CTV and, for
radiotherapy lies in the neoadjuvant setting. neoadjuvant breast irradiation, the GTV need to
Regarding the second topic, traditional image be precisely identified in order to avoid geo-
guidance approaches are characterized by the graphical miss. With the excellent soft tissue con-
acquisition of portal images or cone beam CTs trast of MR and the possibility of online daily
before irradiation, which are used to correct vari- MR-based plan adaptation, the MR-Linacs offer
ations in patient positioning. However, these the tools required for this concept.
image guidance methods cannot correct for In the following sections, practical recom-
changes in the breast contour, e.g., swelling and mendations for simulation, target and OAR defi-
seroma volume variations, changes in tumor bed nition, and plan adaptation for treatments
performed at the MR-Linac are provided. These Simulation: Image Acquisition

recommendations are intended for treatments of
patients with low-risk early-stage breast cancer, In the adjuvant setting, a T1-weighted sequence
according to the GEC-ESTRO and ASTRO without fat suppression is recommended for
selection criteria [38, 39]. visualization of the surgical clips, while a
T1-weighted scan with fat suppression and
especially a T2-weighted scan are preferred for
imulation, Contouring, PTV,
S the visualization of seroma [25, 28, 32, 46–49].
and OAR In the neoadjuvant setting, a T1-weighted
sequence with fat suppression (mDixon) and
Simulation: Patient Positioning gadolinium contrast enhancement offers the
best visualization of the tumor and its spiculae
Because the MR-Linac bore size is smaller than [50–53]. During treatment at an MR-Linac not
the CT bore, particular attention needs to be all preferred sequences may be available, as
paid to patient positioning for breast radiother- only a fixed set of MRI sequences is available
apy at an MR-Linac. When treatments are per- on the machines [41]. At the MRIdian, a True-
formed in the classical supine position with FISP sequence, a balanced steady-state free pre-
arms above the head, an MR-compatible immo- cession imaging sequence, is used. At the Elekta
bilization device less inclined (maximum 5°) Unity, a T1-weighted spoiled gradient echo and
compared to a CT-planning setup or a device a T2-weighted fast spin echo sequence are avail-
without inclination is usually required in order able. To aid tumor delineation, one scan should
to avoid collision between the patient’s elbows be acquired before contrast injection, and mul-
and the machine. An MR-compatible knee sup- tiple scans should be acquired at different time
port or feet immobilization device is recom- points in the first minutes after contrast injec-
mended. The target area should be as close as tion [54]. An example of tumor visualization in
possible to the isocenter to minimize geometric a breast cancer patient treated in the neoadju-
distortions in the target area, which may be vant setting at an Elekta Unity MR-Linac is dis-
achieved by shifting the patient to the contralat- played in Fig. 17.1. Diffusion-weighted imaging
eral side of the table. The anterior receiver coil (DWI) sequences allow for differentiation
can be placed on coil bridges to avoid breast between malignant and benign lesions, but they
image deformation [40, 41]. Also, application of do not offer anatomical references, so that in the
flex coils placed over the breasts has been context of radiotherapy they might be more
described [42]. interesting for evaluation of tumor response
Treatments in prone position have been inves- after irradiation than for target definition [50].
tigated [43–45] and appear to be feasible, but Regarding the image quality, planning studies
may be challenging, due to the smaller bore size demonstrated that images acquired at the
leaving limited space for pendulous breasts to MR-Linac are mostly hampered neither by the
hang freely without touching the table, especially use of positioning devices nor by breath-related
combined with a receiver coil on the back of the motion [47, 48]. Nevertheless, if needed,
patients [41]. MR-navigated or gated sequences might be
When a planning CT is still required, the adopted. To be able to evaluate the impact of the
planning CT and planning MR should be per- electron stream effect (“Considerations for
formed with the same immobilization device Breast Treatment Planning at the MR-Linac
and, if possible, on the same day in order to Global Web Icon”), the simulation scan should
improve the consistency of patient and target cranially include the region of the chin.
position and shape between the two imaging Caudally, it should include the bottom of the
modalities. diaphragm. Recommended image slice thick-
ness is 1–2 mm. In the adjuvant setting a 2-mm
a b
c d
Fig. 17.1 Example of a patient with a T1b N0 M0 (Stage T2-weighted FLAIR sequence (c, d, axial and sagittal)
IA) invasive ductal carcinoma of the left breast treated at acquired at the MR-Linac (Courtesy of Adam Currey,
an Elekta Unity MR-Linac. The tumor is visible on a 3D Department of Radiation Oncology, Medical College of
T1-weighted sequence (a, b, axial and sagittal) and a 3D Wisconsin, Milwaukee, WI, United States)
slice thickness appears appropriate, while in the tumor bed should include the lumpectomy cav-
neoadjuvant setting thinner slices are ity, the seroma if present, the clips if present,
recommended. and the postoperative changes visible on the
T1-weighted and T2-weighted images. Based
on the IMPORT protocols [17, 55], the CTV is
Contouring: Target Volumes defined as an isotropic expansion of 15 mm of
in the Adjuvant Context the tumor bed, limited by the anatomical barri-
ers of the ipsilateral breast. Regarding the PTV,
As previously discussed, the added value of MR as discussed in the previous section, a margin of
images for target contouring in the context of 5–10 mm is traditionally required for breast
adjuvant partial breast irradiation is arguable. treatments. At present, the impact of MR-based
Nevertheless, in situations where the tumor bed plan adaptation on PTV margin reduction has
delineation is challenging because few clips are not been clearly quantified yet. In a study con-
placed, no seroma is visible, and postoperative ducted at the ViewRay MR 60Co system, a PTV
tissue changes after oncoplastic surgery are dif- reduction of more than 50% could be reached
ficult to be recognized on CT images, an through an MR-based setup correction used to
MR-based contouring might be helpful. The align the surgical cavity [40]. However, other
a b
c d
Fig. 17.2 Example of target contouring for partial breast and the postoperative changes; the CTV (violet) is an iso-
irradiation on an Elekta Unity MR-Linac in the adjuvant tropic expansion of 15 mm, limited to the breast tissue;
setting. Displayed is a T2-weighted sequence, where the the PTV (light green) is in this case an isotropic expansion
seroma appears bright. The seroma is contoured in dark of 10 mm, limited to 5 mm below the patient’s surface. (a)
blue; one clip lays lateral to the seroma (yellow); the Axial, (b) 3D reconstruction, (c) coronal, (d) sagittal
tumor bed (emerald green) includes the seroma, the clips,
investigations suggest that margins of 6–8 mm contouring guidelines are available [54]. The
are still required when p erforming adjuvant par- GTV should include the fiducial marker and
tial breast irradiation at the MR-Linac [56]. the tumor as visible in the T1-weighted
Here it should also be mentioned that, compared gadolinium-enhanced sequence with fat sup-
to traditional treatments, at the MR-Linac the pression, where it appears hyperintense.
geometric image accuracy represents a potential Spiculae are included and the GTV should be
source of error eventually needed to be compen- cross-checked on coronal images. The CTV
sated within the PTV. Distortions up to 2 mm should be defined as a GTV isotropic expan-
have been described, especially when the target sion of 15–20 mm, limited by the anatomical
is distant from the isocenter [57–59]. For breast barriers of the ipsilateral breast [60]. In the
treatments, this should be considered for targets setting of neoadjuvant treatments performed at
located in the outer quadrants of big breasts. the MR-Linac, smaller PTV margins com-
Finally, a structure for online motion monitor- pared to the adjuvant setting might be used,
ing, e.g., a copy of the PTV, should be defined. since it could be assumed that some of the
An example of target delineation in the adjuvant errors contributing to the PTV margins in the
setting is shown in Fig. 17.2. adjuvant scenario play here a minor role, as
follows. First, due to a better visualization of
the target, delineation errors as well as match-
Contouring: Target Volumes ing errors are minimized. Second, a target
in the Neoadjuvant Context deformation, i.e., surgical clip displacement in
the adjuvant setting, might be also reduced
Clear advantages of MR compared to CT have compared to the adjuvant setting. Other fac-
been described in the neoadjuvant setting and tors contributing to possible errors and vari-
a b c
Fig. 17.3 Example of target contouring for partial breast CTV (orange). (a) Axial, (b) coronal, (c) sagittal (Courtesy
irradiation on an Elekta Unity MR-Linac in the neoadju- of Adam Currey, Department of Radiation Oncology,
vant setting. The GTV includes the tumor and its spiculae Medical College of Wisconsin, Milwaukee, WI, United
(magenta), the CTV is a 15-mm expansion from the GTV States)
(light blue) and the PTV is a 3-mm expansion from the
ability, such as intrafraction target motion the breast tissue/target volume) and a 3–5-mm-
according to the breathing, geometrical distor- thick structure at the lung/tissue (ribs, inter-
tion, technical delivery accuracy, and beam costal muscles) interface.
penumbra, need to be considered in the neoad-
juvant scenario in the same way as in the adju-
vant setting. As before, a structure for the ose Prescription and DVH
D
motion monitoring, e.g., a copy of the PTV, Constraints
should be defined. An example of target delin-
eation in the neoadjuvant setting is shown in In the adjuvant setting, there are three preferred
Fig. 17.3. fractionation schedules for external beam partial
breast irradiation [62]: the 3-week hypofraction-
ated schedule as per IMPORT-LOW, namely,
Contouring: Organs at Risk 40 Gy in 15 fractions [17]; the 1-week highly
hypofractionated schedule as per FAST-Forward,
The following structures are recommended to namely, 26 Gy in 5 fractions [36]; and a 2-week
be identified as organs at risk when perform- highly hypofractionated schedule based on the
ing partial breast irradiation, either neoadju- Florence protocol, namely, 30 Gy in 5 fractions
vant or adjuvant, at the MR-Linac: ipsilateral delivered in nonconsecutive days [63].
and contralateral breasts, heart, lungs, and, In the neoadjuvant setting, different schedules
especially for neoadjuvant partial breast irra- have been applied in clinical trials. Preferred
diation, chest wall. Additionally, some dummy fractionation schedules appear to be 20–24 Gy in
structures might be required for MR-Linac- one fraction or 30–40 Gy in 3–5 fractions. These
specific planning purposes. For the breasts, fractionations are based on published studies and
the breast tissue is best visualized on protocols [51, 52, 60, 64–71] and ongoing trials
T1-weighted sequences. T1-weighted mDixon [64, 70–81]. Depending on the fractionation
sequences are also preferred for other thoracic schedule, different constraints might be used. An
organs at risk [61]. To evaluate the electron example of an MR-Linac treatment plan in the
return effect (“Considerations for Breast neoadjuvant setting is shown in Fig. 17.4.
Treatment Planning at the MR-Linac Global A summary of constraints applied in clinical
Web Icon”), two dummy structures might be trials of partial breast irradiation, which could be
delineated, namely, a 3–5-mm-thick structure used when performing partial breast irradiation at
under the skin (between the skin surface and the MR-Linac, is reported in Table 17.1.
a b
Fig. 17.4 Treatment plan for a partial breast irradiation fifth and last treatment fraction (Courtesy of Adam Currey,
on an Elekta Unity MR-Linac in the neoadjuvant setting. Department of Radiation Oncology, Medical College of
The patient was treated with 30 Gy in 5 fractions. (a) Base Wisconsin, Milwaukee, WI, United States)
treatment plan, (b) the match with the daily MRI at the
Table 17.1 Summary of target coverage and constraints used in different clinical trials in the adjuvant or neoadjuvant
setting
Single-dose neoadjuvant APBI
Target/OAR 30 Gy in 5 fractions [63] 26 Gy in 5 fractions [36] [82]
PTV coverage V28.5 Gy = 100% (100% of V24.7 Gy = 95% (95% of ≥95% dose in ≥97% volume with
PTV covered by 95% of the PTV covered by 95% of the ≥95% dose in ≥99% of the GTV
PD) PD) Dmax <110%
Dmax = 31.5 Gy (<105%) V27.3 Gy <5% (<5% of PTV
Dmin = 28 Gy should receive ≥105% of the
PD)
V27.8 Gy <2% (<2% of PTV
should receive ≥107% of the
PD)
Dmax = 28.6 Gy (<110%)
Ipsilateral V15 Gy < 50% – Dmean ≤4.8 Gy
breast
Skin – – Dmax ≤100% of the PD
3 mm: D 1 cc ≤13.2 Gy
3 mm: D 10 cc ≤9.5 Gy
5 mm: D 1 cc ≤16.3 Gy
5 mm: D 10 cc ≤12.4 Gy
Chest wall – – D 20 cc ≤10.2 Gy
Heart V3 Gy <10% V1.5 Gy <30% Dmean ≤0.6 Gy
V7 Gy <5% Dmax ≤3.4 Gy
Ipsilateral lung V10 Gy <20% V8 Gy <15% Dmean ≤1.0 Gy
Contralateral Dmax <1 Gy – Dmax ≤1.6 Gy
breast
Contralateral V5 Gy <10% – Dmean ≤0.4 Gy
lung
LST left-sided tumor, RST right-sided tumor, PD prescribed dose, Dmean mean dose, Dmax maximal dose, Dmin mini-
mal dose
Considerations for Breast Treatment electrons in and around the patient. Compared
Planning at the MR-Linac to the other tumor sites, this is particularly
important for patients receiving breast radio-
When performing treatments at the MR-Linac, therapy because of the superficial target loca-
it must be considered that the presence of the tion. Two effects induced by the magnetic field
magnetic field impacts the motion of secondary have to be considered. The electron stream
effect may cause an unwanted out-of-field dose Another aspect to consider for breast radio-
deposition to the chin or to the arm of the patient therapy at the MR-Linac is the possible impact of
[18, 41, 83–85]. The electron return effect might geometric distortions on treatment planning.
induce an increased dose to the skin and at the Dosimetric variation due to the geometric distor-
lung/chest wall interface, both air/tissue inter- tions has been described [58, 89, 90], but this
faces [83, 86–88]. It has been shown that both topic needs further investigation in the field of
effects are correctly calculated by the modern partial breast irradiation. As previously described,
treatment planning systems [18, 83]. Evaluation the geometric distortions should be taken into
in the first patients treated with adjuvant partial account as possible source of errors in the defini-
breast irradiation at the Elekta Unity MR-Linac tion of the PTV.
showed that the maximal dose to the skin and Finally, at the MR-Linacs, volumetric arc
the interface lung/chest wall was low and that techniques are not available and treatments are
the use of a 1-cm bolus can effectively minimize planned with a step-and-shoot technique.
the dose on the chin or on the arm of the patient Therefore, the choice of the beam angles is
[18, 84], as shown in Fig. 17.5. Another possi- important. It should be considered that some
bility might be to shield the upper part of the angles are not available or should be avoided in
patient’s body with a “bolus curtain” placed treatment planning at the MR-Linac. Due to tech-
approximately in correspondence of the patient’s nical limitations, angles between 30° and 33° and
neck (Fig. 17.5). Until now no increased acute between 8° and 18° cannot be used at the
skin toxicities have been reported for breast ViewRay system and at the Elekta system,
patients treated at the MR-Linac. respectively [91, 92]. Particularly important for
a b
c d e f
Fig. 17.5 Setup possibilities to shield the patient from port appositely constructed (c: original design) holds a
the electron stream effect. (a, b) A 1-cm bolus can be bolus (d: original design, e: finalized product, f: schematic
placed on the patient’s arm or chin, depending on the reproduction) (Courtesy of Joan Chick and Jim Sullivan,
planned out-of-field dose distribution due to the electron Royal Marsden NHS Foundation Trust and Institute of
stream effect. (c–f) A “bolus curtain” can be placed Cancer Research, Sutton, United Kingdom)
between the coil and patient’s chin and arms, where a sup-
breast treatments, angles between 130°–150° and However, currently, contrast agent injection is
210°–230°, depending on the tumor location and not recommended for imaging before irradia-
patient’s anatomy, should be avoided because of tion, because of potential radiosensitizing prop-
the presence of the couch edges with high-density erties [93] and the lack of data regarding its use
components [41]. in combination with radiotherapy. In the neoad-
juvant setting, as for adjuvant partial breast irra-
diation, a rigid registration with a simple plan
On-Table Adaptation adaptation based on the patient’s position of the
day appears reasonable, for different reasons.
The first MR-based breast treatments were per- First, tumors are small, and without contrast
formed at the MR 60Co system, a previous ver- agent, it might be difficult to exactly identify the
sion of the ViewRay MRIdian system which lesion edges and its spiculae for precise recon-
used three 60Co heads instead of a Linac [40]. In touring. Second, as described in the previous
these initial treatments, MR was only used for section, neoadjuvant treatments are performed in
setup correction; no online adaptation or online very few fractions, so that significant changes in
motion management was performed. The first the target and organs at risk anatomy (e.g., tumor
experiences with a simple online adaptation response, breast swelling) are not to be expected.
based on the patient’s position were performed a Nevertheless, in case of significant changes, a
few years ago at the Elekta MR-Linac [18, 83]. full plan adaptation is recommended.
The adapt- to-
position method allowed for a Especially in the context of neoadjuvant par-
quicker adaptation compared to a full adaptation tial breast irradiation, the management of the
where the target and the organs at risk are recon- intrafraction motion is particularly important.
toured and a completely new plan is generated. First, as previously mentioned, the PTVs are
These initial experiences indicate that, in the expected to be smaller. Second, if using fewer
context of adjuvant partial breast irradiation, a fractions and higher dose per fraction, the beam-
plan adaptation based on the patient’s position of on time is longer. Third, at the MR-Linac no
the day is sufficient for adequate target coverage volumetric arc techniques are available. It has
and allows for the treatment to be completed been documented that the intrafraction breast
with an in-room time of approximately 25 min. target motion, mostly due to breathing, is gener-
A short treatment time appears particularly ally regular and limited to a median of 2–3 mm
important for breast patients, because they are [37, 40, 94, 95]. Larger variations due to deep
commonly treated with arms above the head and inspiration or whole-body displacements were
a longer treatment time might result in discom- observed but rare, and therefore, their impact on
fort. More recently, the first experiences with target coverage is expected to be limited [40, 94,
online adaptation on the MRIdian MR-Linac 95]. However, it has been shown that margins
were reported for partial breast irradiation in required to account for intrafraction motion
prone position [45]. Use of daily adaptive treat- without motion mitigation strategies increase
ment planning was feasible and online MRI with longer treatment durations [37]. Motion
guidance led to smaller target volumes in com- monitoring with online beam-on MR performed
parison to conventional CT simulation for the during treatment delivery is possible at both the
treated patients. The preliminary results indi- ViewRay and Elekta MR-Linacs. Since the
cated that there may be potential to decrease recent implementation on the Unity system, gat-
CTV and PTV with the use of online MRI guiding techniques are available at both systems as
ance. It should be noticed that no contrast agent well. It should be noted that gating techniques
was used in aforementioned experiences. In the might prolong the delivery time. Cine gating
context of neoadjuvant partial breast irradiation, appeared to be an important reason for selecting
the use of gadolinium is recommended for target MR-guided radiotherapy for accelerated partial
delineation, as described in the previous section. breast irradiation [96].
Outcome Ongoing studies are currently investigating

neoadjuvant partial breast irradiation at the
Adjuvant partial breast irradiation treatments MR-Linac [72, 75, 76]. The advantages of
have been successfully performed at the ViewRay MR-guided partial breast radiotherapy have been
0.35 T MR 60Co system [40, 96, 97], at the discussed elsewhere in this chapter and are repre-
ViewRay 0.35 T MR-Linac [45], and at the sented by more precise target definition and mar-
Elekta 1.5 T Unity MR-Linac [18, 83]. At gin reduction due to the possibility of a MR-based
Washington University, almost 200 breast online plan adaptation and online motion moni-
patients were treated at the MR 60Co system toring. Additional advantages may be the use of
[96], but oncological outcomes and toxicity were DCE- and DWI-MR to provide (semi)quantita-
not described in detail. At Seoul University, 173 tive parameters, which could be investigated as
patients received adjuvant partial breast irradia- biomarkers for response assessment after neoad-
tion at the MR 60Co system, with very good juvant radiotherapy [21, 64, 99–101].
results in terms of both toxicity, with no patients
with more than grade 3 toxicity, and oncological
outcome, with recurrence-free survival rates Summary and Conclusion
close to 100% after a 3-year follow-up. A smaller
cohort of 11 patients was treated with adjuvant MR-guided radiotherapy for breast cancer
partial breast irradiation at the 1.5 T Elekta patients has no widespread clinical application
MR-Linac with excellent results, i.e., no patients yet. MR-guided breast radiotherapy may help to
with more than grade 2 toxicity and good or minimize the irradiated breast volume in partial
excellent cosmetic results. There were no breast irradiation by aiding target definition for
relapses, but the number of patients was limited neoadjuvant partial breast irradiation and by
and the follow-up short [18, 98]. Also, the small potential margin reduction for both adjuvant and
cohort of 11 patients treated with adjuvant partial neoadjuvant partial breast irradiation through
breast irradiation in prone position at an ViewRay online adaptation and monitoring when perform-
MR-Linac showed no grade 2 or higher toxici- ing treatment on an MR-Linac. MR-Linac-
ties, though reported results only covered a delivered adjuvant partial breast irradiation has
median follow-up of 6 months [45]. been demonstrated feasible and studies evaluat-
Different groups have published their experi- ing MR-Linac treatment in the neoadjuvant set-
ence with external beam neoadjuvant partial ting are ongoing.
breast radiotherapy performed at a CT-Linac or
CyberKnife [51, 52, 64–68, 70]. MR-based
radiological and pathological responses ranged References
between 17–42% [64, 65, 68] and 9–42% [65, 66,
1. van der Leij F, et al. Target volume delineation in
68, 70], respectively. Delayed surgery performed
external beam partial breast irradiation: less inter-
6–8 months instead of few weeks after radiother- observer variation with preoperative-compared
apy appears to correlate with a higher radiologi- to postoperative delineation. Radiother Oncol.
cal and pathological complete response. These 2014;110(3):467–70.
2. E.E.C.I. System. J.R. Centre, Editor. 2020. https://
studies could also document the safe profile of
ecis.jrc.ec.europa.eu/pdf/Breast_cancer_factsheet-
high-dose neoadjuvant partial breast irradiation, Oct_2020.pdf.
with only very few cases of grade 3 toxicities, the 3. U.S. Cancer Statistics, T.O.F.C.S.C.f.D.C.a.P. Cancer
large majority of patients reporting excellent or statistics at a glance, 2019. https://gis.cdc.gov/
Cancer/USCS/#/AtAGlance/.
good cosmetic results, and very few local
4. World Cancer Research. 2022. https://www.wcrf.
relapses. Until now, no reports have been pub- org/cancer-trends/breast-cancer-statistics.
lished on neoadjuvant partial breast irradiation 5. Siegel RL, et al. Cancer statistics, 2022. CA Cancer
performed at an MR-Linac. J Clin. 2022;72(1):7–33.
6. Eurostat. 2023. https://ec.europa.eu/eurostat/ 22. van Mourik AM, et al. Multiinstitutional study on
statistics-explained/index.php?title=Causes_of_ target volume delineation variation in breast radio-
death_statistics#Major_causes_of_death_in_the_ therapy in the presence of guidelines. Radiother
EU_in_2017. Oncol. 2010;94(3):286–91.
7. Arnold M, et al. Current and future burden of breast 23. Landis DM, et al. Variability among breast radia-
cancer: global statistics for 2020 and 2040. Breast. tion oncologists in delineation of the postsurgical
2022;66:15–23. lumpectomy cavity. Int J Radiat Oncol Biol Phys.
8. DeSantis CE, et al. Breast cancer statistics, 2019. 2007;67(5):1299–308.
CA Cancer J Clin. 2019;69(6):438–51. 24. Al-Hammadi N, et al. MRI reduces variation of con-
9. Kunkler IH, et al. Breast-conserving surgery with or touring for boost clinical target volume in breast can-
without irradiation in early breast cancer. N Engl J cer patients without surgical clips in the tumour bed.
Med. 2023;388(7):585–94. Radiol Oncol. 2017;51(2):160–8.
10. Lee BM, et al. Hypofractionated radiotherapy dose 25. Jolicoeur M, et al. Localization of the surgical
scheme and application of new techniques are asso- bed using supine magnetic resonance and com-
ciated to a lower incidence of radiation pneumonitis puted tomography scan fusion for planification of
in breast cancer patients. Front Oncol. 2020;10:124. breast interstitial brachytherapy. Radiother Oncol.
11. Yeboa DN, Evans SB. Contemporary breast radio- 2011;100(3):480–4.
therapy and cardiac toxicity. Semin Radiat Oncol. 26. Mast M, et al. Target volume delineation in breast
2016;26(1):71–8. conserving radiotherapy: are co-registered CT
12. Taylor C, et al. Estimating the risks of breast cancer and MR images of added value? Radiat Oncol.
radiotherapy: evidence from modern radiation doses 2014;9:65.
to the lungs and heart and from previous randomized 27. Pogson EM, et al. Comparison of magnetic resonance
trials. J Clin Oncol. 2017;35(15):1641–9. imaging and computed tomography for breast target
13. NCCN Clinical Practice Guidelines in Oncology, volume delineation in prone and supine positions.
Breast Cancer, N.C.C. Network, Editor. 2023. p. Int J Radiat Oncol Biol Phys. 2016;96(4):905–12.
BINV-I. 28. Giezen M, et al. MRI- versus CT-based volume
14. Interdisziplinäre S3-Leitlinie für die Früherkennung, delineation of lumpectomy cavity in supine position
Diagnostik, Therapie und Nachsorge des in breast-conserving therapy: an exploratory study.
Mammakarzinoms. 2021. p. 150. Int J Radiat Oncol Biol Phys. 2012;82(4):1332–40.
15. Radiologists TRCo. Postoperative radiotherapy 29. Den Hartogh MD, et al. Post-lumpectomy CT-guided
for breast cancer: UK consensus statements. 2016, tumor bed delineation for breast boost and partial
update 2021. breast irradiation: can additional pre- and postopera-
16. Meattini I, et al. Accelerated partial-breast irra- tive imaging reduce interobserver variability? Oncol
diation compared with whole-breast irradiation for Lett. 2015;10(5):2795–801.
early breast cancer: long-term results of the random- 30. Sabine B, Giovanna D, Peter P, Clara J, Bert P, John
ized phase III APBI-IMRT-Florence Trial. J Clin K. Open low-field magnetic resonance (MR) versus
Oncol. 2020;38(35):4175–83. CT scanner (CT) imaging in breast radiotherapy
17. Coles CE, et al. Partial-breast radiotherapy after treatment planning. Int J Radiat Oncol Biol Phys.
breast conservation surgery for patients with 2005;63:S232–3.
early breast cancer (UK IMPORT LOW trial): 31. Kirby AM, et al. Tumor bed delineation for partial
5-year results from a multicentre, randomised, breast and breast boost radiotherapy planned in the
controlled, phase 3, non-inferiority trial. Lancet. prone position: what does MRI add to X-ray CT
2017;390(10099):1048–60. localization of titanium clips placed in the exci-
18. De-Colle C, et al. Analysis of the electron-stream sion cavity wall? Int J Radiat Oncol Biol Phys.
effect in patients treated with partial breast irra- 2009;74(4):1276–82.
diation using the 1.5 T MR-linear accelerator. Clin 32. den Hartogh MD, et al. MRI and CT imaging for
Transl Radiat Oncol. 2021;27:103–8. preoperative target volume delineation in breast-
19. Hepel JT, et al. Toxicity of three-dimensional conserving therapy. Radiat Oncol. 2014;9:63.
conformal radiotherapy for accelerated partial 33. Tersteeg RJ, et al. Changes in excision cavity vol-
breast irradiation. Int J Radiat Oncol Biol Phys. ume: prediction of the reduction in absolute volume
2009;75(5):1290–6. during breast irradiation. Int J Radiat Oncol Biol
20. Jagsi R, et al. Unacceptable cosmesis in a proto- Phys. 2009;74(4):1181–5.
col investigating intensity-modulated radiotherapy 34. Topolnjak R, et al. Breast-conserving therapy: radio-
with active breathing control for accelerated partial- therapy margins for breast tumor bed boost. Int J
breast irradiation. Int J Radiat Oncol Biol Phys. Radiat Oncol Biol Phys. 2008;72(3):941–8.
2010;76(1):71–8. 35. Jeon SH, et al. Seroma change during magnetic
21. Civil YA, et al. Preoperative partial breast irradiation resonance imaging-guided partial breast irradia-
in patients with low-risk breast cancer: a systematic tion and its clinical implications. Radiat Oncol.
review of literature. Ann Surg Oncol. 2023;30:3263. 2017;12(1):103.
36. Murray Brunt A, et al. Hypofractionated breast radio- based guidelines for contouring tumors for preop-
therapy for 1 week versus 3 weeks (FAST-forward): erative partial breast irradiation. Radiother Oncol.
5-year efficacy and late normal tissue effects results 2019;133:S719.
from a multicentre, non-inferiority, randomised, 51. Horton JK, et al. Preoperative single-fraction
phase 3 trial. Lancet. 2020;395(10237):1613–26. partial breast radiation therapy: a novel phase 1,
37. Hoekstra N, et al. Intrafraction motion during par- dose-escalation protocol with radiation response
tial breast irradiation depends on treatment time. biomarkers. Int J Radiat Oncol Biol Phys.
Radiother Oncol. 2021;159:176–82. 2015;92(4):846–55.
38. Polgár C, et al. Patient selection for accelerated 52. Guidolin K, et al. Stereotactic image-guided neo-
partial-
breast irradiation (APBI) after breast- adjuvant ablative single-dose radiation, then
conserving surgery: recommendations of the lumpectomy, for early breast cancer: the SIGNAL
Groupe Européen de Curiethérapie-European prospective single-arm trial of single-dose radiation
Society for Therapeutic Radiology and Oncology therapy. Curr Oncol. 2019;26(3):e334–40.
(GEC-ESTRO) breast cancer working group based 53. Mouawad M, et al. Reducing the dose of gadolinium-
on clinical evidence (2009). Radiother Oncol. based contrast agents for DCE-MRI guided SBRT:
2010;94(3):264–73. the effects on inter and intra observer variability
39. Correa C, et al. Accelerated partial breast irradiation: for preoperative target volume delineation in early
executive summary for the update of an ASTRO stage breast cancer patients. Radiother Oncol.
evidence-based consensus statement. Pract Radiat 2019;131:60–5.
Oncol. 2017;7(2):73–9. 54. Vasmel JE, et al. Consensus on contouring primary
40. Acharya S, et al. Magnetic resonance image guided breast tumors on MRI in the setting of neoadjuvant
radiation therapy for external beam accelerated partial breast irradiation in trials. Pract Radiat Oncol.
partial-
breast irradiation: evaluation of delivered 2020;10(6):e466–74.
dose and intrafractional cavity motion. Int J Radiat 55. The Institute of Cancer Research, London.
Oncol Biol Phys. 2016;96(4):785–92. 2013. https://www.icr.ac.uk/media/docs/default-
41. Groot Koerkamp ML, et al. Optimizing MR-guided source/default-d ocument-l ibrary/import-h igh-
radiotherapy for breast cancer patients. Front Oncol. planning-pack-version-4-0-final-23052013-edited.
2020;10:1107. pdf?sfvrsn=6d2c2169_0.
42. Moore-Palhares D, et al. Clinical implementation of 56. Tan IZ, Mitchell A, McNair H, Dunlop A, Herbert T,
magnetic resonance imaging simulation for radiation Nartey J, Lawes R, O’Connell N, Cooper D, De-Colle
oncology planning: 5 year experience. Radiat Oncol. C, Diaz J, Donovan EK, Han K, Hahn E, Nelms B,
2023;18(1):27. Russell N, Kirby AM, on behalf of the MR-Linac
43. Charaghvandi KR, et al. Single dose partial breast Consortium Breast Tumour Working Group. A mul-
irradiation using an MRI linear accelerator in the ticentre study of clinical to planning target volume
supine and prone treatment position. Clin Transl margins for adjuvant partial breast irradiation deliv-
Radiat Oncol. 2019;14:1–7. ered on the 1.5 T MR-Linear accelerator. Int J Radiat
44. Groot Koerkamp ML, et al. Prone vs. supine accel- Oncol Biol Phys. 2023;117(2):e725.
erated partial breast irradiation on an MR-Linac: a 57. Ginn JS, et al. Characterization of spatial distortion
planning study. Radiother Oncol. 2021;165:193–9. in a 0.35 T MRI-guided radiotherapy system. Phys
45. Ng J, Pennell R, Formenti SC. The initial experi- Med Biol. 2017;62(11):4525–40.
ence of MRI-guided precision prone breast irra- 58. Fransson A, Andreo P, Pötter R. Aspects of MR
diation with daily adaptive planning in treating image distortions in radiotherapy treatment plan-
early stage breast cancer patients. Front Oncol. ning. Strahlenther Onkol. 2001;177(2):59–73.
2022;12:1048512. 59. Walker A, et al. MRI distortion: considerations
46. Paulson ES, et al. Comprehensive MRI simulation for MRI based radiotherapy treatment planning.
methodology using a dedicated MRI scanner in radi- Australas Phys Eng Sci Med. 2014;37(1):103–13.
ation oncology for external beam radiation treatment 60. Charaghvandi RK, et al. Redefining radiotherapy
planning. Med Phys. 2015;42(1):28–39. for early-stage breast cancer with single dose abla-
47. Ahn KH, et al. MRI guidance for accelerated partial tive treatment: a study protocol. BMC Cancer.
breast irradiation in prone position: imaging proto- 2017;17(1):181.
col design and evaluation. Int J Radiat Oncol Biol 61. Whiteside L, et al. To see or not to see: evaluation
Phys. 2009;75(1):285–93. of magnetic resonance imaging sequences for use
48. Batumalai V, et al. Assessment of MRI image quality in MR-Linac-based radiotherapy treatment. J Med
for various setup positions used in breast radiother- Imaging Radiat Sci. 2022;53(3):362–73.
apy planning. Radiother Oncol. 2016;119(1):57–60. 62. Meattini I, et al. European Society for Radiotherapy
49. Liney GP, Moerland MA. Magnetic resonance imag- and Oncology Advisory Committee in Radiation
ing acquisition techniques for radiotherapy plan- Oncology Practice consensus recommendations
ning. Semin Radiat Oncol. 2014;24(3):160–8. on patient selection and dose and fractionation for
50. Vasmel JE, Koerkamp G, Kirby A, Russell N, external beam radiotherapy in early breast cancer.
Shaitelman S, Vesprini D, et al. Evaluation of MRI- Lancet Oncol. 2022;23(1):e21–31.
63. Livi L, et al. Accelerated partial breast irradia- 78. Rahimi A. A phase I dose escalation study of single
tion using intensity-modulated radiotherapy versus fraction ablative pre-operative partial breast (S-PBI)
whole breast irradiation: 5-year survival analysis of for early stage breast cancer. ClinicalTrials.gov
a phase 3 randomised controlled trial. Eur J Cancer. Identifier: NCT04040569.
2015;51(4):451–63. 79. Takita C. Study for selected early stage breast
64. Weinfurtner RJ, et al. Quantitative changes in cancer (SABER). ClinicalTrials.gov Identifier:
intratumoral habitats on MRI correlate with NCT04360330.
pathologic response in early-stage ER/PR+ 80. Cedrone L. SBRT in early breast cancer in
HER2− breast cancer treated with preoperative elderly women. ClinicalTrials.gov Identifier:
stereotactic ablative body radiotherapy. J Breast NCT05071105.
Imaging. 2022;4(3):273–84. 81. Blitzblau R. Preoperative single-fraction radiother-
65. Bosma SCJ, et al. Five-year results of the preoperative apy in early stage breast cancer. ClinicalTrials.gov
accelerated partial breast irradiation (PAPBI) trial. Identifier: NCT02482376.
Int J Radiat Oncol Biol Phys. 2020;106(5):958–67. 82. Charaghvandi RK, et al. Treatment constraints for
66. Nichols E, et al. Preoperative accelerated partial single dose external beam preoperative partial breast
breast irradiation for early-stage breast cancer: pre- irradiation in early-stage breast cancer. Clin Transl
liminary results of a prospective, phase 2 trial. Int J Radiat Oncol. 2017;6:7–14.
Radiat Oncol Biol Phys. 2017;97(4):747–53. 83. Nachbar M, et al. Partial breast irradiation with the
67. Tiberi D, et al. Tumour response 3 months after neo- 1.5 T MR-Linac: first patient treatment and analy-
adjuvant single-fraction radiotherapy for low-risk sis of electron return and stream effects. Radiother
breast cancer. Curr Oncol. 2020;27(3):155–8. Oncol. 2020;145:30–5.
68. Vasmel JE, et al. Tumor response after neoadjuvant 84. Park JM, et al. Air-electron stream interactions during
magnetic resonance guided single ablative dose par- magnetic resonance IGRT: skin irradiation outside
tial breast irradiation. Int J Radiat Oncol Biol Phys. the treatment field during accelerated partial breast
2020;106(4):821–9. irradiation. Strahlenther Onkol. 2018;194(1):50–9.
69. Vanderwalde NA, et al. Phase 2 study of pre- 85. Hackett SL, et al. Spiraling contaminant electrons
excision single-dose intraoperative radiation therapy increase doses to surfaces outside the photon beam
for early-stage breast cancers: 6-year update with of an MRI-linac with a perpendicular magnetic field.
application of the ASTRO accelerated partial breast Phys Med Biol. 2018;63(9):095001.
irradiation consensus statement criteria. Cancer. 86. van Heijst TC, et al. MR-guided breast radiotherapy:
2013;119(9):1736–43. feasibility and magnetic-field impact on skin dose.
70. Meattini I, et al. Preoperative robotic radiosurgery Phys Med Biol. 2013;58(17):5917–30.
for early breast cancer: results of the phase II ROCK 87. Raaijmakers AJ, Raaymakers BW, Lagendijk
trial (NCT03520894). Clin Transl Radiat Oncol. JJ. Integrating a MRI scanner with a 6 MV radiother-
2022;37:94–100. apy accelerator: dose increase at tissue-air interfaces
71. Zerella MA, et al. Single fraction ablative preopera- in a lateral magnetic field due to returning electrons.
tive radiation treatment for early-stage breast cancer: Phys Med Biol. 2005;50(7):1363–76.
the CRYSTAL study—a phase I/II clinical trial pro- 88. Kim A, et al. Magnetic field dose effects on differ-
tocol. BMC Cancer. 2022;22(1):358. ent radiation beam geometries for hypofractionated
72. van den Bongard HJGD. Single-dose preoperative partial breast irradiation. J Appl Clin Med Phys.
partial breast irradiation in low-risk breast can- 2017;18(6):62–70.
cer patients (ABLATIVE-2). ClinicalTrials.gov 89. Prott FJ, et al. Effect of distortions and asymmetry in
Identifier: NCT05350722. MR images on radiotherapeutic treatment planning.
73. Brackstone M. Stereotactic image-guided neoadju- Int J Cancer. 2000;90(1):46–50.
vant ablative radiation then lumpectomy (SIGNAL 90. Walker A, et al. MRI geometric distortion: impact
2). ClinicalTrials.gov Identifier: NCT02212860. on tangential whole-breast IMRT. J Appl Clin Med
74. Yassa M. Single pre-operative radiation ther- Phys. 2016;17(5):7–19.
apy—with delayed or no surgery (SPORT-DNS). 91. Klüter S. Technical design and concept of a
ClinicalTrials.gov Identifier: NCT05217966. 0.35 T MR-Linac. Clin Transl Radiat Oncol.
75. Currey A. MRI-based preoperative accelerated par- 2019;18:98–101.
tial breast irradiation. ClinicalTrials.gov Identifier: 92. Bijman R, et al. MR-Linac radiotherapy—the
NCT02728076. beam angle selection problem. Front Oncol.
76. Shaitelman SF. Radiation therapy before surgery 2021;11:717681.
in treating patients with hormone receptor positive, 93. Taupin F, et al. Gadolinium nanoparticles and con-
HER2 negative breast cancer. ClinicalTrials.gov trast agent as radiation sensitizers. Phys Med Biol.
Identifier: NCT03359954. 2015;60(11):4449–64.
77. Wapnir I. Phase 2 surgical excision vs neoadjuvant 94. van Heijst TC, et al. Quantification of intra-
radiotherapy + delayed surgical excision of ductal fraction motion in breast radiotherapy using supine
carcinoma (NORDIS). ClinicalTrials.gov Identifier: magnetic resonance imaging. Phys Med Biol.
NCT03909282. 2016;61(3):1352–70.
95. Groot Koerkamp ML, et al. Intrafraction motion 99. Mouawad M, et al. DCE-MRI assessment of
during radiotherapy of breast tumor, breast tumor response to neoadjuvant SABR in early stage breast
bed, and individual axillary lymph nodes on cine cancer: comparisons of single versus three fraction
magnetic resonance imaging. Phys Imaging Radiat schemes and two different imaging time delays post-
Oncol. 2022;23:74–9. SABR. Clin Transl Radiat Oncol. 2020;21:25–31.
96. Henke LE, et al. Magnetic resonance image- 100. Wang C, et al. Assessment of treatment response
guided radiotherapy (MRIgRT): a 4.5-year clini- with diffusion-weighted MRI and dynamic contrast-
cal experience. Clin Oncol (R Coll Radiol). enhanced MRI in patients with early-stage breast
2018;30(11):720–7. Cancer treated with single-dose preoperative radio-
97. Lee HI, et al. The acute and late toxicities of MRI- therapy: initial results. Technol Cancer Res Treat.
guided external beam partial breast irradiation deliv- 2016;15(5):651–60.
ered using a once-per-day regimen. Front Oncol. 101. Vasmel JE, et al. Dynamic contrast-enhanced and
2021;11:649301. diffusion-weighted magnetic resonance imaging for
98. De-Colle C, et al. Estimation of secondary can- response evaluation after single-dose ablative neoad-
cer projected risk after partial breast irradia- juvant partial breast irradiation. Adv Radiat Oncol.
tion at the 1.5 T MR-linac. Strahlenther Onkol. 2022;7(2):100854.
2022;198(7):622–9.
Considerations for Using MR Linac
for the Treatment of Patients
18
with Gynecologic Cancer:
A Practical Guide and Early Clinical
Experience
Lorraine Portelance, Ina Jürgenliemk-Schulz,

Kyle R. Padgett, Rosa Patricia Castillo,
and Astrid van Lier
Introduction of locally advanced cervical cancer (LACC) is a

combination of chemotherapy and radiation
Radiation therapy (RT) plays a major role in the therapy (external beam and brachytherapy
management of patients with gynecologic can- [BT]). Vaginal cancers are treated with a similar
cer. Cervical cancer is the fourth most frequent approach. Both pathologies being relatively rare
cancer in women worldwide, with an estimated in developed countries, these cancers are ideally
604,000 new cases in 2020. However, the inci- treated in centers with expertise in BT.
dence of cervical cancer is much higher in low- MR-Linac is a technology which allows for the
and middle-income countries [1]. According to delivery of external beam radiation with higher
the American Cancer Society, cervical cancer precision, and patients diagnosed with LACC
incidence in the United States has decreased by might benefit from this technology. This chapter
more than half since the mid-1970s and other will present an overview of these potential
Western countries have observed similar trends opportunities.
[2]. The decline in the incidence rate of cervical Endometrial cancer is by far the most frequent
cancer seen in developed countries is attributed gynecologic malignancy in Western countries.
to successful screening and HPV vaccination While surgery is the primary treatment for uter-
programs. The standard of care for the treatment ine cancer, radiation is key for adjuvant treatment
for patients with a higher risk of recurrence
L. Portelance (*) · K. R. Padgett (higher-grade disease, deep myometrial invasion,
Radiation Oncology Department, Sylvester lymphovascular invasion, higher FIGO stage).
Comprehensive Cancer Center, Miller School of
The dose of radiation delivered in the adjuvant
Medicine, University of Miami, Miami, FL, USA
e-mail: [email protected] setting, however, is relatively low, and the con-
formality provided by endocavitary BT and pel-
I. Jürgenliemk-Schulz · A. van Lier
Department of Radiation Oncology, University vic intensity-modulated radiation therapy (IMRT)
Medical Center Utrecht, University Utrecht, is associated with good treatment outcomes.
Utrecht, The Netherlands However, the management of patients with inop-
erable endometrial cancer requires a higher dose
R. P. Castillo of radiation, and MR-Linacs might become an
Radiology Department, Miller School of Medicine,
excellent treatment option for these patients.
University of Miami, Miami, FL, USA
324 L. Portelance et al.
Radiation is also frequently used as a salvage tial for online replanning for certain body sites,
treatment of recurrent gynecologic cancer, either like the pelvis.
alone or in combination with chemotherapy or MRI is now commonly used at initial diagno-
surgery. Additionally, there is a fast-growing lit- sis for identifying tumors and differentiating
erature demonstrating that radiation has a role to them from adjacent organs in the management of
play in the management of oligometastatic dis- patients with gynecologic cancer. Gynecologic
ease for many disease sites, including gyneco- cancers often have poor contrast resolution com-
logic malignancies [3]. pared to the surrounding soft tissues on CT scans,
During the last three decades, research in and consequently, MRI is the diagnostic imaging
oncology focused on developing systemic agents modality of choice. Beyond its superiority for
offering personalized medicine options and diagnostic purposes, MRI plays a role throughout
adapting the systemic treatment to specific tumor all stages of care, including EBRT planning, out-
characteristics (targeted agents and immunother- come evaluation, and detecting tumor
apy, preferably directed to specific genetic signa- recurrence.
tures). In parallel, RT has significantly benefited Cervical, vaginal, endometrial, and oligomet-
from technological advances that lead to astatic gynecologic cancers are being treated on
improved treatment precision. Breakthroughs in MR-Linacs in increasing numbers worldwide.
dynamic multi-leaf collimator design and treat- This chapter provides an overview of the princi-
ment planning systems have improved treatment ples of MR imaging for gynecologic cancer, fol-
delivery conformality, translating into decreased lowed by the current state and potential future
treatment toxicity and a better outcome for directions for the clinical use of MR-Linacs in
patients with gynecologic cancer [4]. the management of gynecologic cancers.
The introduction of online MRI-guided exter-
nal beam RT (MRgRT) in 2016 marked a signifi-
cant milestone and effort to increase the precision rinciples of Magnetic Resonance
P
of RT delivery. The MR-Linacs provide the Imaging for Gynecologic Cancer
advantage of onboard MRI combined with the
possibility of performing onboard adaptive plan- MRI scanners were first introduced for clinical
ning when necessary. In parallel with the imple- use just 40 years ago. Significant technology
mentation of MRgRT, numerous publications development followed, and today, MRI is a diag-
showed the benefits of MRI-guided BT in the nostic tool widely used in oncology. As early as
same years [5, 6]. 1990, MRI was described as a promising tool in
The ability to perform online adaptive radio- the management of gynecologic cancers provid-
therapy requires daily imaging datasets with ing superior visualization of tumors and adjacent
enough quality to delineate both targets and pelvic organs compared to CT scan [7]. A review
organs at risk (OARs) effectively. These datasets published in 1992 by Russell highlighted the
need to be artifact resistant and spatially accu- potential for MRI guidance to avoid marginal
rate. Additionally, these high-fidelity onboard tumor misses in EBRT treatment for gynecologic
imaging systems must be coupled with a fast and cancers [8]. In the early 2000s, Potter et al. devel-
robust adaptive treatment planning system and a oped a cervical BT planning approach for patients
modern radiation delivery platform. With these with LACC, integrating the use of pelvic MRI to
requirements in place, onboard imaging and delineate tumor targets according to possible
online replanning based on the daily anatomy remaining cell density at the time of BT planning
allows for dose escalation and better sparing of [9]. Subsequently, efforts of the American
OARs. It is important to note that onboard MRI Brachytherapy Society (ABS) [10] and the
provides superior soft tissue contrast resolution Groupe Europeen de Curietherapie European
compared to cone beam computed tomography Society for Radiation Oncology (GEC ESTRO)
(CBCT). This increased image quality is essen- [11] led to further development of the concepts.
18 Considerations for Using MR Linac for the Treatment of Patients with Gynecologic Cancer… 325
Twenty years later, there is a wealth of literature of collecting images of sufficient quality to prop-
confirming that MRI-guided adaptive BT leads to erly diagnose and stage cervical cancer, 3.0 tesla
better treatment outcomes, particularly in LACC systems have some distinct advantages with a
cases [12], paving the way for the use of MRI in few potential drawbacks as well. The higher field
treatment planning for EBRT as well. strength of 3.0 tesla systems provides a higher
signal-to-noise ratio which is primarily used to
increase resolution, improve contrast, and speed
elvic MRI for Initial Staging
P up acquisition. On the other hand, the higher field
of Gynecologic Malignancies strengths of 3.0 tesla systems come with increased
image artifacts when encountering bowel or rec-
I nitial Staging of Cervical and Vaginal tal gas and when hip prostheses or other metal
Cancer implants are found in the pelvis: surgical clips,
MRI plays a vital role in the initial staging of cer- etc. These artifacts are more pronounced on cer-
vix cancer [13], and its findings are now included tain pulse sequences (gradient recalled echo,
in the International Federation of Gynecology echo planar imaging, diffusion-weighted imag-
and Obstetrics (FIGO) and American Joint ing [DWI], among others), many of which are
Committee on Cancer (AJCC) cervical cancer used for functional imaging. To summarize, both
staging system [14]. MRI’s high soft tissue con- 1.5 tesla and 3.0 tesla produce good-quality
trast resolution and ability to detect disease extent images, making either choice a suitable option.
in more than one plane make it the preferred Hence, 1.5 tesla systems are significantly more
imaging modality for detecting cervical and vagi- common than 3.0 tesla systems and provide
nal cancer [15, 16]. images of sufficient quality for diagnosis and
Although gadolinium is not nephrotoxic, a staging of cervical cancer, and 3.0 tesla systems
review of recent laboratories should always be generally provide images with superior resolu-
done before ordering a contrast-enhanced (CE) tion, albeit with a significantly higher cost, and
MRI; gadolinium should be used cautiously in may produce inferior results in certain patients
patients with severe renal insufficiency since with excessive bowel gas or with exogenous
these patients are at higher risk of nephrogenic implanted objects. The key sequences for the
systemic fibrosis. A careful review of creatinine evaluation of cervical cancer are the multiplanar
levels should be done when prescribing a CE pel- fast spin-echo T2-weighted images without fat
vic MRI for patients with LACC since hydrone- saturation. Examples of image acquisition proto-
phrosis and increased creatinine level are cols used for a diagnostic study and treatment
common findings in patients with FIGO stage planning simulation pelvic MRI for cervical can-
IIIB or IVA LACC. cer are presented in Tables 18.1 and 18.2.
To minimize bowel motion during pelvic During image acquisition for diagnostic pur-
image acquisition, patients may be advised to fast poses, a large field of view is first used to screen
for 4 hours before the examination. Some medi- the lower abdomen and pelvis for lymphadenop-
cal facilities may also administer medication athy, bone metastasis, or any other incidental
such as scopolamine butyl bromide (Buscopan) findings (5–6 mm thickness through the entire
or glucagon to decrease peristalsis, although this low abdomen and pelvis). In the localizer, sagittal
may cause short-time nausea. Additionally, a T2-weighted images are used to locate the tumor
low-fiber diet to reduce bowel gas may be recom- in the cervix and plan the high-spatial-resolution
mended to begin the day before the diagnostic T2-weighted images in axial and coronal oblique
MRI. planes. The possibility that MRI offers to per-
The two most common field strengths form image acquisition in two orthogonal planes
employed when performing diagnostic imaging along the tumor axis provides important informa-
for cervical cancer are 1.5 tesla and 3.0 tesla. tion on disease extent for cervical cancer staging.
While both field strengths are more than capable The cervical tumor size and information showing
Table 18.1 MRI protocols for image acquisition in the management of cervical cancer: (a) University of Miami: diag-
nostic image acquisition protocol
Planes Sequences
Non-contrast Localizer (three planes) T2, TSE, TSE FS 6 mm (whole pelvis)
(localizer) Coronal T2, TSE 6 mm (from renal hilum to pubic
Axial oblique (perpendicular to the T2, TSE symphysis)
cervical canal) T2, TSE 3.5 mm (small FOV)
Sagittal 4 mm (small FOV)
Contrast Axial T1, TSE, TSE FS 6 mm (whole pelvis)
gadolinium Coronal T1, TSE 3.5 mm (small FOV)
Axial T1, TSE 3.5 mm (small FOV)
DWI Axial oblique B value 0–750
4 mm
Small FOV recommended is 240 × 240 mm
TSE turbo spin echo, FS fat saturated, FOV field of view, DWI diffusion-weighted imaging
Table 18.2 MRI protocols for image acquisition in the management of cervical cancer: (b) University Medical Center
Utrecht: MRI protocol for treatment planning simulation
Sequence Plane Details Bladder filling
T2 TSE Transverse 0.48 × 0.48 × 3 mm (whole pelvis) Medium
T2 TSE 3D (transverse) 0.63 × 0.63 × 2 mm3 (whole pelvis) Medium
pseudo CT 3D (transverse) 1 × 1 × 1 mm3 (whole pelvis) Medium
B0 mapa 3D (transverse) 1.6 × 1.6 × 5 mm3 (whole pelvis) Medium
PD TSE Transverse 0.88 × 0.88 × 6.6 (very large FOV Medium
including PAO lymph nodes)
DWI SPAIR Transverse 1.5 × 1.5 × 5.0 (small FOV) Medium
Wait for extra bladder filling
T2 TSE 3D (transverse) 0.63 × 0.63 × 2 mm (whole pelvis) Full
Patient empties bladder
T2 TSE 3D (transverse) 0.63 × 0.63 × 2 mm (whole pelvis) Empty
The image acquisition protocol includes specific instructions to assess the impact of bladder filling on the position of
the uterus
PD proton density
a
B0 map to verify that the main magnetic field is homogeneous and thus that images are geometrically correct
the relationship between the tumor and adjacent mass with increased signal intensity compared
structures (e.g., parametrial tissue involvement with the surrounding normal cervical stroma. The
and pelvic sidewall invasion) are more accurately use of intravenous (IV) contrast (gadolinium) in
assessed on the oblique planes. the evaluation of cervical cancer is controversial.
In a multicenter study, Zhang et al. [17] dem- Some do not recommend routine administration
onstrated that the overall accuracy of staging for of IV contrast material for staging protocols,
cervical cancer increased from 61 to 81.4% when given the various degrees of enhancement of cer-
MRI findings were combined with clinical find- vical tumors. However, dynamic contrast-
ings. Clinical staging particularly underestimates enhanced (DCE) MRI is reported to be of value
tumor size if there is endocervical tumor growth for small lesions, for assessment of residual
towards the lower uterine wall. In these situa- tumor post-RT, for surgical planning in patients
tions, MRI measurement of the tumor extent will who wish to undergo fertility-sparing surgery,
provide a more appropriate initial evaluation than and to distinguish between cervical and endome-
clinical evaluation. trial tumors when the primary tumor is a locally
The cervical tumor is best depicted on extensive adenocarcinoma [18–21].
T2-weighted images without fat suppression, on Although MRI has a high negative predictive
which it appears as an expansile or infiltrative value (94–100%) for parametrial invasion,
reported sensitivities vary between 38 and 100%

[22, 23], and clinical findings will always rule
over MRI findings for parametrial tissue involve-
ment. Disruption of the stromal ring, obliteration
of the peri-uterine fat planes, encasement of the
ureters, and involvement of peri-uterine vascula-
ture are signs consistent with parametrial inva-
sion. Diffusion-weighted image (DWI) sequences
add value to pelvic MRI for evaluation of para-
metrial involvement, increasing the sensitivity
and specificity (82% and 97%, respectively)
compared with MRI without DWI sequences
(72% and 91%, respectively) [24]. Pelvic side-
wall invasion is defined as a tumor that is less
than 3 mm from the pelvic muscles (internal
obturator, levator ani, pyriformis muscle) or the
iliac vessels. Gross invasion is easy to recognize
and could be associated with an enlargement of
the pelvic wall muscles, encasement or narrow-
ing of the iliac vessels, or extension to pelvic Fig. 18.1 Siemens 3 T MRI T2 FSE sagittal view of a
bones [25]. pelvic MRI of a patient with LACC stage IVA; endovagi-
nal gel was used to fill the vaginal canal, thus allowing for
To evaluate the extent of vaginal involvement
better visualization of disease extent in the upper vaginal
in patients with cervical cancer, combining find- wall. Note: Patient had bilateral external nephrostomy
ings from physical examination with MRI leads tubes; absence of urine in the bladder leads to suboptimal
to an accuracy as high as 86–93% [26]. Vaginal condition to assess bladder involvement
wall invasion is detected when a high-signal-
intensity tumor replaces the low signal intensity
of the vaginal wall. Evaluation of vaginal wall Finally, DWI is obtained in the axial oblique
integrity may be difficult to assess, with large plane with a 3D gradient-echo T1-weighted
exophytic cervical tumors protruding into the sequence after IV gadolinium administration.
vagina. The assessment of vaginal wall invasion Sequential images are acquired before contrast
by cervical tumor can be easier if medical gel injection and then during multiple sequential
used for ultrasound is injected into the vaginal phases of enhancement and a delayed CE
canal before image acquisition [27], as shown in T1-weighted image in the axial oblique plane.
Fig. 18.1. CE images help with the assessment of invasion
DWI may improve lesion detection and is of adjacent structures (bladder and rectum) and
obtained in the same two oblique planes (coronal evaluation of lymph node involvement which is
and axial) as tumor extension assessment [28– key information for cervical cancer staging. The
30]. During DWI, random Brownian motion of specificity of MRI for metastatic lymph nodes in
water within tissues is detected, which is hin- cervical cancer has been reported to be as high as
dered by the tissue architecture. Information 71–100%. MRI criteria indicating lymph node
obtained from the DWI study could be used to involvement include lymph nodes larger than
distinguish malignant from benign tumors by 1 cm in the short axis, lobulated or spiculated
measuring apparent diffusion coefficient (ADC) contours, signal intensity like the primary tumor,
values. While malignant tumors exhibit a low and heterogeneity of the internal architecture
ADC value, DWI is highly sensitive in differenti- [31]. The presence of lymph nodes with spicu-
ating tumors from surrounding tissues. lated contours, the same signal intensity as the
primary tumor, and heterogeneous signal are also During the EBRT planning process for LACC,
indicative of metastatic disease in lymph nodes the diagnostic MRI should be fused with the
that are less than 1 cm in the short axis, even in planning studies (planning CT or planning MRI).
case of negative PET-CT [32]. Although PET-CT MR-Linacs commercially available do not pro-
has a role in staging cervical cancer for detecting vide the possibility of image acquisition in an
the location of disease (primary tumor, metastatic oblique plane perpendicular to the cervical canal,
nodes, and possible distant metastasis), MRI is and for now, onboard MRI cannot be used to
superior for precisely defining tumor extension. replace a good-quality diagnostic or treatment
Local extension of the primary tumor into the planning simulation MRI for the planning
bladder can be evaluated using MRI with a sensi- process.
tivity of 83% and specificity of 100%. Per FIGO
staging, mucosal invasion of the bladder must be I nitial Staging of Endometrial Cancer
present for the tumor to be classified as FIGO Although the staging of endometrial cancer is
stage IVA; mucosal invasion carries a significantly primarily based on pathologic findings, pelvic
worse prognosis compared with only serosal or MRI at initial diagnosis has real value in certain
muscular invasion [33]. MRI findings that suggest clinical situations. For patients with inoperable
bladder invasion include nodular and irregular endometrial cancer, pelvic MRI is the best tool
bladder wall, tumor protruding into the bladder to determine the extent of the primary disease
lumen, and high signal intensity of the anterior (depth of myometrial involvement, cervix stro-
aspect of the posterior wall of the bladder on mal invasion, invasion into the surrounding tis-
T2-weighted images. The presence of vesicovagi- sue) and the presence of pelvic lymph node
nal fistula is diagnostic for bladder wall invasion. metastasis [35]. For patients with inoperable
Rectal wall invasion by cervical cancer is less fre- endometrial cancer, when there is no evidence
quent than bladder invasion; rectal wall invasion of nodal involvement on imaging, it is common
carries a worse prognosis [34]. When bladder or practice to assess the risk of microscopic dis-
rectal invasion is suspected, on MRI, cystoscopy ease spread in the pelvic nodes by combining
and proctoscopy can be performed to confirm the tumor grade with the depth of myometrial
mucosal involvement (Fig. 18.2). invasion seen on MRI using the Creasman rules
[36]. When the risk of nodal involvement is low,
treating patients with inoperable endometrial
cancer with endocavitary BT only is reasonable.
In the adjuvant setting for patients who did not
undergo a surgical nodal staging procedure, the
risk of pelvic node disease added to the tumor
grade might be used to formulate individual
treatment recommendations. This includes
deciding if a plan for postoperative treatment
could be limited to endocavitary BT or if pelvic
EBRT (for patients with a higher risk of nodal
involvement) should be applied with or without
additional BT [37].
Fig. 18.2 Siemens 3 T MRI T2 FSE axial view of a pel- Pelvic MRI can also provide key information
vic MRI of a patient with LACC stage IVA. The blue to distinguish between a primary cervical cancer
arrow indicates the absence of a plane between the cervi- versus a primary endometrial cancer in situations
cal gross tumor and the rectal wall, a situation where a
rectoscopy could be indicated to rule out rectal
where adenocarcinoma is found on cervical and
involvement endometrial biopsies.
elvic MRI During Radiation

P serial functional imaging on MR-Linac systems
Treatment through the course of EBRT brings information
that could be used to guide decisions on boosting
Sequential diagnostic MRIs obtained during poorly responding targets or reaching for addi-
EBRT can provide predictive information on tional therapies as the treatment course
tumor response [38–41]. The integration of an progresses.
MRI in a linear accelerator (MR-Linac) treat- Towards the end of a course of pelvic radia-
ment unit (Unity, Elekta, Sweden; MRIdian, tion for patients with LACC or vaginal cancer,
View Ray, Cleveland, OH, USA) constitutes a performing a diagnostic pelvic MRI in prepara-
breakthrough in the management of gynecologic tion for the BT treatment is standard of care.
cancer. The onboard MRI in the Unity system has This study provides essential information on
a field strength of 1.5 tesla, and the onboard MRI residual tumor size and extension, which is
in the MRIdian system has a lower field strength taken into consideration in addition to the physi-
of 0.35 tesla. While a higher-strength magnet cal examination findings to guide the BT appli-
might produce better diagnostic image quality, cator selection. With improved onboard image
there are several advantages to a low-field- quality, this end of EBRT MRI may become
strength magnet in RT. First, low field strength obsolete, and the onboard MRI will provide the
allows the linear accelerator and the MRI to be in needed information.
closer proximity. This can help increase the dose Radiomics is a powerful technique where
rate and reduce the penumbra. The second key many imaging features are extracted from the
benefit is the reduction in both image distortion datasets, and automated analysis of these features
and radiation dose distortion. Finally, for imag- is performed to correlate imaging findings with
ing sequence speed, at lower field strength, impactful clinical endpoints. Following image
images can be acquired more quickly, and field acquisition, the radiomics process is initiated by
strength enables real-time imaging, which segmenting the portion of the image that is to be
MRIdian uses for tissue tracking (providing eight analyzed (i.e., areas of suspected tumor, cervix,
images per second) and beam gating. On board, uterus, etc.). Imaging features are then extracted,
MRI provides valuable information on changes but the number of features that are extracted can
in tumor volume and position of the tumor and be quite large, up to 5000 [44]. These features are
surrounding OAR. Since the onboard image is then significantly reduced to a manageable num-
acquired with the same protocol daily, valuable ber of features to perform analysis; ideally, fea-
radio genomics data can be extracted from tures that are useful and reproducible are utilized,
onboard imaging for further study. and the remainder are excluded. Once the fea-
As MRgRT use continues to grow, tools are tures to be included have been chosen, a statisti-
adapted from diagnostic MRI to the treatment cal framework can be developed with the goal of
delivery environment. For example, studies have aiding with diagnosis, treatment response, toxic-
demonstrated that serial ADC measurements dur- ity, etc. While many of the initial radiomics stud-
ing the treatment course can be used as an inde- ies were developed to aid in the diagnostic
pendent prognostic factor for treatment response, setting, these powerful techniques have also been
where an increase in ADC values during treat- applied to the treatment and follow-up phases.
ment represents tumor response, thus aiding in Some of the common applications include pre-
identifying good responders [42, 43]. DWI could dicting tumor response and predicting patient
be used to demonstrate heterogeneity within the outcome, but other applications have been devel-
tumor, indicating areas of resistant clones and oped or are possible such as dose painting and
areas of disease regression. Acquiring multiple prediction of side effects [45].
MRI datasets in the traditional clinical setting is Many of these radiomics studies have been
challenging due to patient burden and limited demonstrated on body sites other than the female
diagnostic MRI resources. The ability to perform pelvis. One powerful example of radiomics
potential for gynecologic malignancies was pro- let the tissue heal from the acute inflammatory
vided by Takada et al. [46]. Their research uti- reaction seen post RT; therefore, the first follow-
lized machine learning and radiomics to predict up pelvic MRI is usually done about 3 months
in-field tumor control after definitive RT using after treatment completion. In addition to find-
two MRI scans, T2-weighted imaging, and ADC ings from physical examination, important infor-
datasets. Their research was revealing on several mation is obtained from this imaging study:
levels; one main finding was that utilizing a vol- residual tumor size, distinguished between resid-
ume of interest larger than the GTV resulted in a ual tumors and fibrosis. Fibrosis could be distin-
stronger prediction of in-field tumor control than guished from residual disease, which has an
when the GTV was used without an expansion. intermediate signal intensity on T2-weighted
Another significant finding was that radiomics imaging [47]. Differentiating residual viable
analysis on the ADC datasets was more predic- tumor or tumor recurrence from posttreatment
tive of in-field tumor control than T2-weighted changes can be challenging; DWI sequences may
imaging. Other radiomics applications—delta help to differentiate persistent tumors from fibro-
radiomics, dose-painting applications, and pre- sis. Residual viable tumors usually show restric-
diction of side effects—have not been robustly tion on DWI sequences. Multiple studies have
investigated in the gynecologic cancer setting but used ADC as a quantitative metric for treatment
are ripe for future investigations. response [42, 48]. Interval increase in ADC value
(attributed to apoptotic-induced cell death or
necrosis) has been seen in patients with a com-
elvic MRI for Assessment
P plete response but not in patients with partial or
of Treatment Response After no treatment response [49–52]. Figure 18.3 is an
Treatment excellent example of a pelvic MRI performed at
diagnosis and after treatment, with dramatic
It is standard of care to use pelvic MRI to assess changes in the cervical tissue showing an excel-
treatment response after completion of treatment lent treatment response. FDG positron emission
for LACC, vaginal cancer, or inoperable endome- tomography (PET/CT) is frequently used in early
trial cancer. It is essential to wait long enough to follow-up of patients with LACC.
a b
Fig. 18.3 Siemens 3.0 tesla pelvic MRI, axial T2 view showing the extent of a cervical squamous cell carcinoma at
diagnosis (a) and after 5 weeks of chemoradiation pre-brachytherapy (b)
elvic MRI for the Diagnosis

P cervical tumor extension, which could be
of Disease Recurrence associated with dense fibrosis and decreased cer-
vix volume. DWI and contrast imaging are cru-
MRI also plays an essential role in diagnosing cial to distinguish fibrosis from residual tumor;
pelvic recurrence of gynecologic cancer. The fibrosis will demonstrate delayed enhancement
recurrence can be local in the region of the pri- and no restriction in DWI, as opposed to restric-
mary tumor, locoregional (in pelvic or para-aortic tive hypovascular residual masses, which will
nodes), or distant. In addition to findings from display restriction in DWI.
physical examination and tissue biopsy, a diag- For patients treated with surgery (total hyster-
nostic MRI should be done to document the ectomy, modified radical hysterectomy, or radical
extent of a recurrence. While any imaging modal- hysterectomy) at diagnosis, a nodular appearance
ity can be used (CT scan, PET/CT, or MRI), pel- of the vaginal vault with thickening of the vaginal
vic MRI is superior to CT scan, especially in the wall will be best detected by a pelvic MRI.
detection of small recurrences. Locoregional
relapse or disease progression for cervical and
vaginal cancer usually occurs within the first 2 Role of MR-Linac
years after treatment [53]; however, the most fre- in the Management of Cervical
quent sites of recurrence are distant metastases Cancer
(including distant nodes and visceral metastases)
[54]. Tumor recurrence for endometrial cancer Surgery is the standard of care for patients diag-
usually occurs within 3 years after surgery; the nosed with early cervical cancer. Very early dis-
most frequent sites of recurrence are in the vagi- ease might even be managed by fertility-sparing
nal vault and/or in regional pelvic nodes [55]. surgery, while patients with FIGO stage IB1 or
Disease recurrence should be distinguished IB2 cancer will be treated by a more radical sur-
from residual disease. The general rule is that dis- gical approach (modified radical hysterectomy or
ease recurrence is defined as the persistent visu- radical hysterectomy with bilateral salpingo-
alization of a tumor in the same locations as the oophorectomy and lymph node surgery of vari-
primary tumor, possible metastatic nodes, or dis- ous extent—from sentinel node biopsy to full
tant metastasis after complete remission and lymphadenectomy). While surgery is the pre-
occurring at least 6 months after treatment. ferred treatment for early-stage disease, the stan-
Residual disease is defined as incomplete remis- dard of care for patients diagnosed with LACC is
sion or reappearance of the tumor in the original a combined modality approach of EBRT (either
locations in a period shorter than 6 months. For limited to the pelvis or extended field) combined
local disease recurrence, a thorough clinical with weekly cisplatin and intracavitary/intersti-
examination and MRI should be performed [56]. tial image-guided BT (IGBT). The EBRT vol-
The MRI scanning protocol during the follow- umes cover the central disease (primary tumor,
up phase of patients with LACC is the same as the entire cervix, uterus, parametria, and upper
for the pretreatment MRI. The use of medical gel vagina), draining lymphatic regions and nodal
in the vaginal canal is important to help distin- boosts to positive nodes. Adding a margin to the
guish vaginal fibrosis from residual/new cervical central disease to account for inter- and intra-
tumor extension. The sensitivity and specificity fraction motion is recommended.
to detect recurrent or residual tumors using MRI Elective para-aortic (PAO) node irradiation
have been described as 82–100% and 78–100%, may be indicated in selected patients. Treating
respectively [57]. patients with LACC with prophylactic PAO node
Changes seen in the cervix and surrounding irradiation fell out of favor after an intergroup
tissues on pelvic MRI after RT are characterized clinical trial showed a survival benefit for pelvic
by the recovery of low signal intensity in the cer- RT combined with chemotherapy over extended
vical stroma and in locations of initial extra- field RT in 1999 [58]. However, recent data indi-
cate that patients with high-risk cervical cancer using MRI is that it is not associated with addi-
(≥one pathologic node at common iliac or above tional ionizing radiation exposure.
or ≥three pathologic nodes in the pelvis) might Online MRgRT is a strategy that allows for
benefit from the inclusion of PAO nodes up to the increased EBRT conformality and more individ-
renal vein even if no suspicious nodes are seen in ualized treatment plans. Treating cervix cancer
the lower PAO region below the renal vein on ini- with an MRgRT approach can allow for a further
tial workup imaging [59, 60]. reduction in treated volume, which should lead to
MR-Linacs (Unity, ViewRay) allow physi- reduced treatment toxicity. MRgRT could also
cians to perform online adaptive radiation ther- allow for dose escalation where needed, thus
apy (ART) based on the onboard MRIs of the improving tumor control since IGABT results
day. Onboard MRIs provide valuable information clearly demonstrate the existence of a dose-
on changes in tumor position or volume and the response curve in patients with LACC [5, 61, 64].
daily position of the surrounding OAR. Patient selection for MRgRT depends on
Based on the improvement in treatment out- patients and disease characteristics. The presence
comes observed with the development of image- of non-MRI-compatible implantable cardiac
guided adaptive BT (IGABT) led by the GEC electronic devices or other types of incompatible
ESTRO gynecology working group, we have evi- metallic implants/foreign bodies constitutes con-
dence that integrating MRI information in treat- traindication to MRgRT. Nephrostomy catheters
ment planning adds value to the management of (internal or external) are usually MRI-compatible,
patients with LACC. There is now a bulk of lit- but verifying on the manufacturer’s website is
erature demonstrating that IGABT leads to better always safer before bringing a patient to the
tumor control, increased survival, and decreased MR-Linac room. Clinical conditions which con-
treatment toxicity [5, 12, 61–64]. When perform- stitute a formal or relative contraindication to
ing MRI-guided (MRg) BT, the treatment vol- MRgRT also include claustrophobia, severe anxi-
ume accounts for the topography of the primary ety, and pain preventing the patient from holding
tumor at diagnosis and the regression observed the same position for a long time, given the fact
during EBRT [11] instead of using a fixed-point that the whole replanning and treatment delivery
A for treatment planning. IGABT is supported by process might take up to 60 min, and morbid
both the GEC ESTRO and the ABS and several obesity.
guidelines have been published [11, 12, 65, 66]. In the curative treatment of LACC, MRgRT
It is well-established that both IGABT and may also be used for nodal boosts or primary
EBRT benefit from MRI guidance [67]. Since the tumor boosts if first-line BT is not feasible, and
1990s, EBRT treatment delivery has evolved MRg stereotactic body radiation therapy (SBRT)
from the use of port films and skin marks for could be applied to both nodal and soft tissue
position verification to the use of cone beam CT metastasis to achieve target tumor control with
(CBCT) with or without fiducial markers for limited morbidity. The use of MRgRT for patients
more precise targeting of soft tissue lesions trans- who cannot undergo BT is discussed in further
lating in a reduction in treatment margins and detail at the end of this section. Strategies for the
irradiated volumes. For the curative treatment of use of MRgRT for the delivery of pelvic radiation
most gynecologic cancers, the use of daily are described in the next section.
onboard image guidance is the standard of care,
but the suboptimal soft tissue contrast provided
by CBCT makes it challenging at times to distin- daptive MRgRT in the Management
A
guish soft tissue tumors from surrounding normal of LACC
tissues. Given the fact that MRI provides superior
soft tissue contrast than CT, it is an imaging In the management of patients diagnosed with
modality of choice to verify the daily delivery of LACC, it is well known that large inter-fraction
radiation. Another advantage of onboard imaging motion occurs in the target volume localized in
the center of the pelvis due to variations in the extent of a cervical tumor in the uterine wall, the
volume and shape of the surrounding pelvic consensus was to include the whole uterus in the
organs (bladder, rectum, and sigmoid) seen dur- CTV and to add a large margin on this volume to
ing the delivery of pelvic EBRT. Haripotepornkul account for inter- and intra-fraction motion which
et al. [68] calculated the inter-fraction movement could be significant when attempting to cover
of the cervix during IMRT in different axes and potential daily fundus motion. The ability of MRI
demonstrated that during a course of treatment, to depict tumor extension in the uterine wall
the cervix could move: 1.9 mm in the lateral introduces the possibility of targeting only the
direction, 4.1 in the vertical direction, and 4.2 mm primary tumor, whole cervix parametrial tissue,
in the anterior-posterior axis. The simplest strat- with a security margin above the disease exten-
egy to deal with target inter-fraction and intra- sion in the uterine wall rather than including the
fraction motion has been to add an internal target entire uterus in the central tumor volume.
motion margin of 1.5–2.0 cm into the planning Preliminary modeling suggests this is a feasible
target volume (PTV) [69]. This security margin is approach that could reduce OAR doses. Based on
necessary to ensure a full dose is delivered to the CTV to PTV margins of 3–5 mm, an online adap-
target during every fraction. The cost of this tive planning strategy can reduce the dose to the
approach is that a large volume of the surround- rectum V40Gy by 36–47%, the dose to the blad-
ing normal organs receives the dose prescribed to der V40Gy by 43–59%, and the dose to the bowel
the target volume. Inter-fraction motion uncer- V40Gy by 13–30% compared to a nonadaptive
tainties are eliminated when onboard adaptive approach [70].
radiation is used during MRgRT. However, a Kozak published a single-institution retro-
margin for intra-fraction motion should always spective study of 53 patients with LACC treated
be used as treatment time per fraction increases per institution policy with less than whole uterus
with MRgRT compared to traditional RT frac- irradiation volume. The study found that the
tions, and longer treatment time could lead to patients had comparable locoregional control and
changes in bladder filling and central PTV posi- a reduced bowel dose (V40 and D200cc) com-
tion during the delivery process. pared to historical series [71]. These results are
When using an online adaptive approach, the encouraging since the incidence of cervical can-
contours of the central target volume could be cer in developed countries where this technology
adjusted daily to correct for inter-fraction is available is decreasing, and preliminary results
motion during EBRT, resulting in reduced cen- of MRgRT should be tested in a large multicenter
tral PTV margins and decreased volume of prospective study to investigate the safety of
OARs that receive the prescribed dose. Studies reducing treatment volumes as described above
on the use of MRgRT demonstrated that daily (Fig. 18.4).
MRI permits the adaptation of EBRT plans to Another caveat associated with the MRgRT
daily tumor and OAR positions [40]. The use of approach for a course of pelvic radiation is that
ART potentially leads to a considerable reduc- the online adaptive process is time-consuming.
tion in OAR dose by facilitating improved accu- Until the MR-Linac technology offers a faster
racy of treatment delivery and enabling margin replanning process with robust auto-
reduction. A recent study comparing various segmentation tools, ART will remain resource
ART techniques using CBCT with standard intensive and would be challenging to integrate
margins, reduced margins, and MRgRT found into the clinic in nonacademic centers. Since the
that significant OAR sparing can be achieved first clinical use of MR-Linac technology in
using MRgRT [70]. 2014, significant improvements have been made
Consensus contouring guidelines for IMRT in the treatment replanning process. The latest
for cervical cancer advise including the whole MRIdian upgrade (Ai3) enables the completion
uterus [69] in the target volume. Since, when of multiple tasks happening in parallel (for
using CT scan, it is challenging to outline the example, target contouring could be done by the
a b
Fig. 18.4 Sagittal and axial view of a 0.35 T MRI show- where a 2-cm margin within the uterine wall is added to
ing the PTV in a situation where the whole uterus is being the GTV (b)
included in the treatment volume (a) versus an approach
treating physician while a trained technologist A practical approach to mitigate the large
simultaneously contours the OARs, and the treatment margins necessary for accounting for
physicist performs the electron density calcula- inter-fraction motion without spending time on a
tion), which translates into a reduction in the daily adaptive plan is to utilize a plan-of-the-day
time required to complete the whole on-board (POTD) approach [72]. The POTD technique
adaptive process. As MRgRT and auto-segmen- uses an individualized IMRT plan library that is
tation technology continue to improve, the bur- selected based on the patient’s internal anatomy
den of daily adaptive planning will likely be at the time of the daily setup. The POTD tech-
significantly reduced, and daily adaptive treat- nique has the potential to reduce the treatment
ment for patients with LACC will become margins compared to conventional treatments,
accessible. and it has a more manageable workload and
Similarity Similarity
Daily 3D MRI Planning Library Import Plan Treatment
Evaluation Evaluation
START
HERE
Similarity
Evaluation
On-board Adaptive Workflow

Changes in Plan Plan Approval Add to Planning
Daily Contouring Dose Calculation
target/OAR Re-Optimization & QA Library
Fig. 18.5 Illustration of the plan-of-the-day approach
faster treatment time compared to daily since extending the treatment volume to include
ART. Buschmann et al. [72] published their expe- the common iliac nodes or PAO will lead to a vol-
rience with 16 patients using a volumetric modu- ume that would not be compatible with the field
lated arc therapy (VMAT) plan library for bladder constraints of the MR-Linac. A multiple isocen-
full, bladder empty, and a motion robust backup ter approach may solve this; however, this would
plan, where plan selection is based upon daily translate into long treatment times, added com-
setup CBCT. MR-Linac systems could use a sim- plexity, and the risk of irradiating the same bowel
ilar methodology. MR-Linac offers the added volume twice. Until the MR-Linac systems can
benefit of allowing the users to create a plan treat longer volumes, patients with intermediate-
library as needed for fractions that do not have a or high-risk LACC are not good candidates for
suitable match in the current library for the this technology.
patient, resulting in an adapted plan for the day Studies exploring the benefits of integrated
and an additional entry into the plan library. This MR-Linacs for enhanced target and OAR visual-
process used in the University of Miami for ization and online adaptation to treat LACC with
selected cases is described in Fig. 18.5. hypofractionated schedules are in progress [73].
The improved image quality of onboard MRI If successful, these studies will facilitate the
compared to CBCT eases plan selection for those broader adoption of daily replanning for cervical
fractions where a predefined plan will suffice. cancer.
Although using MRgRT for LACC has been seen
as one of the key examples for using the ART
approach with MR-Linacs, to date, clinical RgRT Boost When Brachytherapy Is
M
implementation of this treatment is limited. In Not Feasible
addition to the time constraint, another drawback
of the currently available MR-Linac systems is MRgRT can be used to substitute the final BT
the limited treatment field size (feet/head extent: boost for patients who cannot be treated with
22 cm [Unity/Elekta], 24.1 cm [MRIdian]). With BT for different reasons (e.g., patients with
this field length limitation, only patients with comorbidities limiting their capacity to undergo
low-risk LACC could be treated with MRgRT invasive procedures, patients unfit for any type
of anesthesia, patients receiving therapeutic Gynecologic Oncology and the ABS have
anticoagulation therapy, BT implantation tech- recently published a review to reinforce that BT
nically not feasible). The first preliminary expe- is a critical component of primary radiation ther-
rience with this novel treatment approach was apy for LACC [81].
presented at the ESTRO annual meeting in 2020 MR-Linac treatments are not considered
[74], and a single-institution series has been replacements for BT but should be seen as an
published [75]. alternative for selected cases where BT is not
Many series on the use of SBRT to substitute feasible. In these cases, MRgRT might be prefer-
for BT have been published [76–78]. When using able over CBCT-guided SBRT as the better soft
SBRT to deliver the final boost, instead of BT, the tissue contrast of MRI allows to better tailor the
target dose is more limited since the dose to OAR dose to the tumor according to the daily online
must be maintained within safe constraints. anatomy and away from the surrounding organs.
Focus on the OAR constraints is important for Currently, the clinical experience with MR-Linac
finding a balance between treatment-related mor- treatments in the above-described situation is
bidity and quality of life against tumor control. slowly growing, as is the number of centers that
The high toxicity reported in a single-institution use them. We regularly notice that the delivered
phase II clinical trial on the use of SBRT for target dose is somewhat less than the planned
patients with LACC clearly demonstrates that the one. This is a notable difference from the BT
use of SBRT boost as a replacement for BT has approaches. The reason is that during BT deliv-
been associated with increased toxicity and ery, the applicator and, therefore, the radiation
decreased efficacy and should be used with cau- source positions are inside or near the tumor and
tion in the context of clinical trials [78]. therefore fixed in relation to the treatment targets
Using the recommended BT, OAR dose con- during dose delivery. If the target moves, the
straints from ABS or GEC ESTRO guidelines, applicator will move with it; OARs, however, can
and an isotoxic approach seems the way to move move independently. This fixed relation of target
forward when using SBRT to deliver the boost. In and radiation source is not reproducible when
such an approach, the OAR dose drives the delivering EBRT with a steady source at a dis-
choices in treatment planning. It can be expected tance and independently changing tumor and
that daily online replanning with MR-Linacs may OAR position during radiation delivery.
allow for the safe delivery of a higher dose to the MR-Linac centers are still in their learning curves
target while maintaining D2cc to the bladder, rec- for these specific treatments, experience techni-
tum, and sigmoid within recommended guide- cal challenges, and continuously adapt
lines. Using MR-Linacs could allow for the safe approaches according to the slowly growing
delivery of a higher dose to the target compared experience. Centers are collaborating with others
to nonadaptive CBCT-guided treatments [74]. It to discuss treatment planning approaches, frac-
was demonstrated in the BT literature that adher- tionation schedules, dose-volume parameters,
ing to high-dose levels to the HR-CTV increases and uncertainties during dose delivery. The feasi-
the chance of obtaining local control [12, 79]. bility of the new treatment approaches can be
Recent preliminary studies show that the target investigated in longitudinal prospective cohort
dose is reduced using SBRT schedules on both studies. One of them is the Momentum cohort
MR-Linacs and standard CBCT Linacs com- within the MR-Linac consortium (The
pared to BT [74, 78]. MOMENTUM Study: The Multiple Outcome
Until proven otherwise, BT should remain the Evaluation of Radiation Therapy Using the
first choice to deliver the final boost to the tumor. MR-Linac Study—[ClinicalTrials.gov]), which
A SEER review published by Han et al. has provides the possibility of cooperation for pro-
clearly demonstrated that the use of BT in the spective data registration including treatment
management of patients with LACC is associated parameters and clinical outcome data in terms of
with improved survival [80]. The Society of tumor control, morbidity, and quality of life.
Finally, the availability of MR-Linac treat- isotoxic approach with priority given to hard
ment units in the radiation oncology clinic has OAR dose constraints over target coverage. The
the additional benefit of providing easy access to HR-CTV and relevant OAR are being re-
MRI datasets with applicators in place to aid in contoured for the pretreatment online plan and
MRg BT planning. This can significantly sim- before each fraction based on the actual anatomy.
plify the logistics of doing IGABT for many For each fraction, daily MRI and online ART are
institutions that until now had relied on the lim- performed. The pelvic EBRT dose (45 Gy or
ited availability of MRI scanners in the diagnos- 44.25 Gy, EQD2α/β = 10) is added to the dose
tic radiology department [82]. from the four online adaptive plans to calculate
the cumulative dose, which sums up as follows:
Case Example 1 Primary cervix cancer, stage D90 HR-CTV, 76.4 Gy EQD2α/β = 10 (i.e.,
IV; BT boost to primary tumor not applicable; 6.0 Gy, or 8.1 Gy, EQD2α/β = 10 per fraction).
alternatively, MR-Linac boost is applied, and OAR doses are for bladder D2cc, 90.9; rectum
treatment intention is pelvic control. D2cc, 70.0; sigmoid D2cc, 47.3; and bowel D2cc,
74.9 Gy EQD2α/β = 3. Although D90 HR-CTV
A 54-year-old patient presents with is below the recommended dose (D90 ≥85–
hemodynamic-relevant vaginal bleeding, abdom- 90 Gy EQD2α/β = 10), using this stereotactic
inal pain, and bladder cramps. The diagnostic planning approach at least part of the HR-CTV
workup reveals a FIGO stage IVA cervical cancer receives this dose aim with V90 Gy
with infiltration into distant parametrial tissue EQD2α/β = 10 = 19% and V85 Gy
bilaterally, into the upper vagina, rectovaginal EQD2α/β = 10 = 64%. Our institutional approach
septum, and rectal wall. In addition, there is infil- (UMCU) for CBCT Linacs would allow for a
tration of the bladder wall and both ureters, asso- total D90 HR-CTV of 70 Gy EQD2α/β = 10
ciated with bilateral hydronephrosis. The using VMAT with uniform target dose
patient’s history includes brainstem infarction distribution.
with persistent hemiplegia and the need for anti- Treatment is overall well tolerated without
coagulation. The multidisciplinary recommenda- unexpected early toxicity. At first follow-up
tion is to offer a curative treatment approach. 3 months after treatment, ongoing tumor regres-
EBRT is delivered with VMAT (45 Gy elective sion is visible on MRI. Vaginal bleeding and pain
dose in 25 fractions to tumor and lymphatic have resolved completely. The patient experi-
drainage) with bilateral simultaneous integrated ences no major side effects, especially no organ
boosts (2.35 Gy per fraction) for two metastatic fistulation (Fig. 18.6).
obturator nodes. Concurrent chemotherapy can-
not be administered due to severely impaired kid-
ney function and comorbidity. The initial plan ole of MR-Linac in the Management
R
includes a BT boost (four HDR fractions, aiming of Recurrent Cervical Cancer
at a total D90 HR-CTV of 85–90 Gy
EQD2α/β = 10). Cervical cancer will most frequently recur in the
MRI is repeated after 18 EBRT fractions but pelvis, either in the vaginal vault for patients who
unfortunately shows only minor tumor regression were initially treated by a surgical approach or in
with persistent invasion of the rectum and blad- the cervix, parametrium, pelvic, and PAO nodes
der. Tumor volume is reduced from 174 to for patients treated with chemoradiation [83].
118 mm3, including a decrease of the largest After ruling out the presence of any metastatic
tumor extension from 93 to 80 mm. BT is not spread outside of the pelvis, patients who did not
considered feasible, and therefore, a boost to the receive a prior course of radiation should be
primary tumor is delivered on a 1.5 T MR-Linac. treated with pelvic radiation combined with che-
Four IMRT fractions are given, using an 11-field motherapy and BT in case of local relapse. For
beam arrangement. Planning is done using an the reasons mentioned above, delivering a whole
Fig. 18.6 Cervix cancer; MR-Linac boost of CTV-HR boost HR-CTV contours are shown on the MRI scans
after elective radiotherapy. Cervix cancer; MR-Linac (pink at the time of treatment planning and yellow after
boost to HR-CTV after pelvic EBRT. Left column (top to 32.4 Gy EBRT). On the CBCT scans, initial HR-CTV is
bottom), transversal and sagittal T2-weighted MRI at the shown. For the MR-Linac boost plan, the online delinea-
time of treatment planning and onboard CBCT scans in tion of the first fraction for HR-CTV and rectum is shown.
the first week of elective EBRT. Right column (top to bot- The images show the improved visualization of MRI com-
tom), MRI scans after 32.4 Gy of elective EBRT and first. pared to CBCT for online position verification
MR-Linac boost plans. For comparison, the initial and
course of pelvic radiation on the MR-Linac sys- surgery is omitted, leaving the uterus and cervix
tem might be tedious. However, a boost to the site in situ. Decision for chemoradiation is made and
of recurrence using MRgRT to bring the tumor started 3 weeks after surgery. After four EBRT
volume to an ablative dose might be of interest if fractions and one cycle of cisplatin, patient
BT is not possible or in case of nodal disease develops a massive thromboembolic event in the
recurrence. Hadi et al. published early clinical superior mesenteric artery A (outside irradiated
results of MR-g SBRT boost for patients with volume). Emergency surgery follows, including
locally advanced or recurrent gynecological can- removal of necrotic bowel parts (nearly the entire
cers ineligible for BT. Five patients in this series small bowel and part of the right hemicolon) and
had recurrent cervical cancer and were success- colostomy. The patient survives with dependency
fully treated to a median cumulative dose of on total parenteral nutrition. Two months later,
73.6 Gy with a combination of pelvic IMRT and cervical cancer restaging is performed, showing
MRgRT boost of 21 Gy in 4 fractions [75]. only minor progression. Therefore, radical hys-
For patients treated primarily with combined terectomy and lymphadenectomy are performed
chemoradiation as a curative intent who present a with curative intent. The pathology report reveals
central pelvic recurrence, surgery might be a treat- adenocarcinoma of the cervix with lympho-
ment option [84]. For all other pelvic recurrence or vascular space invasion component, tumor size
pelvic progression, chemotherapy will generally 27 mm, R0 resection, and no lymph node involve-
be offered. For patients who reach a plateau on ment. The patient recovers slowly but remains
systemic treatment, MRgRT is a promising tech- dependent on parenteral nutrition. Fifteen months
nology to deliver an ablative dose to a site of recur- later, a solitary vaginal vault recurrence is detected
rent disease. When using MRgRT in this clinical during a routine follow-up. On MRI and PET-CT,
situation, typically a hypofractionation approach there was no indication of nodal or distant spread,
will be used (40–50 Gy in 5 fractions), and the and the multidisciplinary tumor board advised
advantage of performing ART will allow for safer for second salvage treatment using radiotherapy
dose delivery, sparing the surrounding organs from consisting of limited elective field EBRT and
receiving a high dose. In a phase II prospective boost to the vaginal tumor. BT is considered not
clinical trial of treating inoperable pancreatic can- feasible due to tumor extension into the bladder
cer, Parikh et al. have demonstrated that using wall, its close relation to remaining large bowel
MRgRT and an isotoxic approach, a dose of 50 Gy loops, and the need for anticoagulation treatment.
in 5 fractions could be delivered to a target volume Alternatively, radiotherapy salvage treatment is
adjacent to surrounding organs without any grade considered, including small-field elective irradia-
3 or higher acute toxicity [85]. These results tion and MR-Linac boost. The elective EBRT field
achieved treating upper abdomen targets are likely includes some margin into the surrounding of the
to be reproducible when treating any pelvic or vaginal tumor in all directions (respecting bound-
abdominal target near luminal organs. aries of surrounding, not affected organs) and the
entire vagina, with 45 Gy being administered in
Case Example 2 Recurrent cervical cancer 25 fractions. This is immediately followed by an
after complicated course of previous curative SBRT boost using a 1.5 T MR-Linac. A sched-
treatment approaches, second salvage ule of 6 fractions is delivered with 3 fractions per
approach with curative intention including week and respecting an overall treatment time
MR-Linac boost to the recurrent tumor in the of 6 weeks aiming at a total dose of 85–90 Gy
vaginal vault. (EQD2α/β 10) to the vaginal tumor remnant
A 47-year-old patient with recurrent cervi- without violation of the OAR constraints. For the
cal cancer in the vaginal vault. Two years ago, MR-Linac boost, an isotoxic approach is used,
the patient is diagnosed with FIGO stage IB2 giving priority to OAR dose constraints during
adenocarcinoma of the cervix, curative surgery the planning process. For the offline pretreatment
is intended, but intraoperatively, nodal pathology planning situation, the prescribed dose to the vault
is detected in the right obturator region. Further recurrence (D90 HR-CTV), calculated as the sum
of elective EBRT and the six boost fractions, is are mainly caused by changes in intra-fraction
98.7 Gy (EQD2α/β 10) with all OAR constraints bladder filling, causing a minor position shift of
being tolerated. In the online and actual treatment the target.
situation, the total D90 HR-CTV based on the sum Three months after salvage treatment, clinical
of daily pre-fraction plans is somewhat lower with investigation and MRI reveals complete remis-
94.0 Gy (EQD2α/β 10) but well in the range of the sion of the vaginal vault tumor and no signs of
intended target dose. The total D90 HR-CTV, as progression elsewhere. The patient is recovering
calculated on the anatomy on post-fraction MRI, well without signs of additional damage to the
is 91.7 Gy (EQD2α/β 10). For details regarding remaining bowel parts. She reports G2 urinary
OAR dose levels and differences calculated for urgency and frequency and G1 vaginal stenosis
moments in time, see the table below. Differences and mucositis (Fig. 18.7, Table 18.3).
Fig. 18.7 Recurrent cervical cancer after complicated treatment planning of the MRL boost showing only minor
course of previous curative treatment approaches, second tumor regression; lower row shows images acquired on
salvage approach with curative intent including MR-Linac 1.5 T MR-Linac for online contouring and treatment plan-
boost to the vaginal vault recurrence. Upper row shows ning, treatment targets are contoured with GTV tumor
images acquired on 1.5 T MR sim for elective EBRT treat- dark blue and PTV 3 mm orange, and dose distribution of
ment planning, CTV including entire vagina in red, GTV the online plan is shown with D98% in red, D90% in yel-
of recurrent tumor in vagina top in yellow; middle row low, D50% in light blue, and D20% in dark blue
shows images acquired on 1.5 T MR sim for offline pre-
Table 18.3 Dose level differences based on changing anatomical relations of target and organs at risk during MR-Linac
boost delivery
Dose in Gy
Dose Constraints Differences
Offline Online Post-recontoured Post-recontoured—online
HR-CTV D98% 95.6 91.0 88.1 −3.0
D90% 98.7 94.0 91.7 >90 −2.2
PTV+3 mm D98% 85.4 83.0 78.4 <80 −4.5
OAR Soft Hard
Bladder D2cc 78.3 79.4 85.7 <80 <90 6.2
D10cc 67.6 66.0 68.6 <75 2.6
Rectum D2cc 74.6 73.7 70.1 <65 <75 −3.7
D5cc 62.7 61.1 58.8 <66 −2.3
Sigmoid D2cc 59.5 55.8 56.5 <70 <75 0.7
D5cc 54.3 52.1 52.2 0.1
Bowel D2cc 47.7 47.9 46.8 <70 <75 −1.0
D5cc 45.7 46.2 45.4 −0.8
Urethra D0.01cc 44.0 44.0 44.0 <90 0.0
Ureter right D0.01cc 102.3 98.4 102.2 <90 3.8
Ureter left D0.01cc 100.1 99.0 101.1 <90 2.1
In the management of cervical cancer, an iso- therapeutic window with better tailoring of the
lated PAO node recurrence is considered a locore- dose to the metastatic nodes and away from the
gional recurrence and should be treated with a surrounding bowel (duodenum) loops.
curative approach. If the patient did not receive The management of cervical cancer recur-
prior radiation to PAO nodes, the whole PAO area rence to metastatic sites is discussed in the last
(from the superior border of the previous field to section of this chapter on the management of
2–3 cm above the involved node) should be irra- oligometastatic disease. It can also be seen in
diated using fractionated RT with chemotherapy Chap. 12 of this book.
with a boost (to 55–59 Gy) to the involved node.
Here again, MRgRT is the technology of choice
to deliver the boost since surrounding bowel Role of MR-Linac
loops could be spared from high doses. An alter- in the Management of Endometrial
native would be resecting the involved node; Cancer
however, radiation to an ablative dose is a nice
noninvasive alternative. reatment of Inoperable Endometrial
T
The dose levels recommended to achieve dis- Cancer
ease control are 45–50 Gy in 25–28 fractions for
the elective nodal volume and a dose range of The standard treatment for nonmetastatic
50–65 Gy for macroscopic disease [86]. The endometrial cancer is surgery consisting of
proximity of these nodes to the bowel is a dose- hysterectomy
with bilateral salpingo-
limiting factor. Severe duodenal morbidity was oophorectomy with or without regional lymph
reported after PAO irradiation using simultane- node dissection or sentinel lymph node map-
ous integrated boost to nodes in the upper abdo- ping. Surgery may need to be followed by
men [85, 87]. MR-Linacs might be the radiotherapy and/or systemic treatment
technology of choice in these situations as daily depending on histopathologic risk factors.
visualization of the anatomy, together with the A minority of patients diagnosed with early
possibility for online ART, allows for a broader endometrial cancer cannot undergo surgery due
to advanced age, poor performance status, or Case Example 3 Primary endometrial can-
medical contraindications to anesthesia. These cer; coronary bypass surgery is indicated
patients can be treated with definitive radiother- prior to uterine surgery, MR-Linac boost is
apy consisting of BT with or without EBRT (+/− applied to control vaginal bleeding prior to
chemotherapy). Depending on the tumor stage, any surgery, and treatment intent is curative.
good disease control and long-term survival
could be achieved [86–89]. In a cohort of 1322 A 68-year-old patient presents with postmeno-
patients with endometrial cancer treated with pausal bleeding. The diagnostic workup (including
radiotherapy alone (EBRT and/or BT) for various a CT scan) is suggestive of low-stage, high-grade
reasons, the disease-specific survival at 5 years endometrial cancer (endometroid adenocarci-
was 78.5%, and severe late morbidity (grade ≥ or noma). Gynecological surgery with curative intent
3) was as low as 3.7% for the combined treatment is indicated, consisting of laparoscopic staging
approach [90]. In a smaller retrospective study, with hysterectomy, bilateral salpingo-ovariectomy,
74 patients with stage I and II endometrial cancer and pelvic/para-aortic nodal dissection. During
received a combination of pelvic EBRT and BT diagnostic workup, severe coronary artery disease
with curative intent, resulting in a 3-year is diagnosed with an indication for coronary artery
progression-free survival of 68% with a median bypass, resulting in an absolute contraindication
interval of 43.5 months [87]. In selected patients, for the intended gynecological surgery. However,
BT alone could be used with curative intent, with clinically relevant vaginal bleeding persists and
excellent LC up to 100% in well-selected patients interferes with the necessary cardiac interventions.
[37, 90, 91]. Therefore, the decision to offer a short course of
In this population of women who are not sur- hypofractionated radiotherapy prior to any other
gical candidates, invasive BT procedures requir- treatment is taken. The treatment plan in this situa-
ing anesthesia might not be an option either, and tion is (1) short-course SBRT to the entire uterus to
treatment alternatives to deliver an ablative dose control vaginal bleeding, (2) coronary artery
to the tumor are valuable. For patients not able to bypass, (3) and radical uterine surgery.
undergo surgery and BT, MRgRT can be an alter- The patient is treated on a 1.5 T MR-Linac
native treatment technique to deliver an ablative with a total of ten online adaptive SBRT frac-
dose to the tumor. tions (4 fractions per week). Total planning dose
Daily changes in the position of the uterus aim of ≥30 Gy (32.5 Gy EQD2) to the entire
due to bladder or rectal filling can be observed uterus, including the cervix, with at least
and adjusted for. To save time in the replanning ≥40 Gy (46.7 Gy EQD2) to the GTV as depicted
process, a POTD method could be imple- on MRI is easily achievable and not violating
mented. As described earlier, for patients with the clinical OAR dose constraints. The final
LACC, the MR-Linac treatment fields are lim- online delivered total dose levels are within the
ited in a cranial-caudal direction. The currently planning aims for both GTV and the entire
available field lengths (Unity/Elekta, 22 cm; uterus with doses as high as 43 Gy (51.2 Gy
MRIdian/ViewRay, 24.1 cm) can be too lim- EQD2) reached in GTV.
ited for pelvic fields in tall patients. If The entire RT series is delivered as intended.
MR-Linac field length limits its use for the Vaginal bleeding resolves entirely, and treatment-
entire course of treatment, MR-Linacs consti- related morbidity is mild with G1 diarrhea and
tute a nice option for boosting the uterus and urinary frequency at the end of treatment. Three
any metastatic nodes in addition to elective weeks after finalizing RT, a cardiac artery bypass
EBRT when a BT boost to the uterus is not fea- is performed as intended. A radical hysterectomy
sible. Daily MRI and replanning allow for bet- is done 6 weeks later revealing necrosis but no
ter targeting of the HRCTV while adapting for remaining vital tumor on pathology. At her 2-year
the variable positions of the sigmoid, small follow-up, the patient is alive without signs of
bowel, and bladder. tumor relapse (Fig. 18.8).
Fig. 18.8 Endometrial cancer; MR-Linac treatment to on 1.5 T MR-Linac for online contouring and treatment
prevent vaginal bleeding and further tumor progression planning; treatment targets are contoured with GTV
while waiting for surgery. Upper row shows images tumor light blue, CTV entire uterus dark blue, and PTV
acquired on 1.5 T MR sim for offline pretreatment plan- 3 mm green; dose distribution of the online plan is shown
ning; sagittal and transversal views on T2 3D image and with GTV D98% in red, CTV D98% in green, and D50%
transversal DWI spair. Lower row shows images acquired in blue
reatment of Recurrent Endometrial

T ing damage to the surrounding organs during BT
Cancer needle insertion.
Although BT offers the most definitive boost
Vaginal Vault Recurrences treatment and high LC rates in this clinical situa-
Depending on the initial management, treatment tion [95, 96], it is associated with significant risk.
of vaginal recurrence of endometrial cancer may In an overview of 28 patients described by Fokdal
consist of surgery or RT [92, 93]. In patients who et al. the 2-year LC rate was 92% [96]. In addi-
did not receive a prior course of pelvic radiation, tion, not all recurrences are amenable to BT. An
EBRT combined with BT is usually the preferred example may be a rectovaginal septum recur-
treatment option. While BT is the modality of rence in close proximity to the rectal wall. In
choice to deliver a boost to vaginal recurrence, many situations, interstitial implants might be
this approach might be challenging at times. needed but are not achievable or are associated
Endocavitary BT using single-channel vaginal with a high risk of fistula and other complica-
applicators can only be used if the thickness of tions. An external beam boost using MRgRT and
the tumor volume to be treated is less than 5 mm adjusting the plan to the daily anatomy using an
[94]. Multichannel endocavitary vaginal applica- isotoxic approach may provide a good
tors can be used for focal lesions up to 1 cm thick, alternative.
but interstitial BT is required for a larger volume Using the BT recommended dose constraints
of disease. After a hysterectomy, small bowel to the OARs (D2cc to bladder <75 Gy, rectum
loops and other pelvic organs are near the vaginal and sigmoid <65 Gy EQD2), isotoxic ART allows
vault. Unless real-time image guidance or a lapa- to keep the dose away from the uninvolved rec-
roscopic approach is used, there is a risk of caus- tum and other surrounding organs. Avoiding
extremely high doses around interstitial BT nee- being 91.9 Gy (EQD2α/β 10). In the online and
dles might result in less normal organ damage actual treatment situation, however, the total D90
and necrosis. HR-CTV based on the sum of daily online plan-
ning is 82 Gy (EQD2α/β 10). For D2 cc bladder
Case Example 4 Primary endometrial can- and rectum, the pretreatment and online doses are
cer; limited early recurrence in the vaginal 84.9 Gy/73.5 Gy and 68.3 Gy/73.7 Gy (EQD2,
vault and abdominal wall; salvage radiother- α/β 3), respectively. The differences in pretreat-
apy is intended to include BT boost to the vag- ment and online doses are mainly caused by vari-
inal tumor; BT however, is not applicable, and ations in rectum positions and filling status,
alternatively, MR-Linac boost is applied. which in our isotoxic planning approach results
A 68-year-old patient presents with vaginal in a reduced target dose. For the abdominal wall
bleeding. She is diagnosed with low-grade/high- metastasis, the SBRT planning aim is 35 Gy in
stage cancer of the uterus and is treated with five fractions to 95% of the target. For both the
laparoscopic hysterectomy and bilateral salpingo- pretreatment and online plan, the GTV35 Gy
oophorectomy. The final pathological report (EQD2α/β 10 = 50 Gy) has a median value of
reveals a FIGO stage IIIB grade 2 endometroid 100%.
adenocarcinoma, which qualifies for postopera- MRI performed at 3 months after treatment
tive EBRT. On treatment planning CT scan and shows no residual tumor in both locations and no
MRI, an early recurrent tumor is detected in the evidence of disease progression. At this moment
vaginal vault, with a second lesion seen in the patient does not report any unexpected or grade
anterior abdominal wall (most likely a laparo- ≥3 treatment toxicity (Fig. 18.9).
scopic port site recurrence). A biopsy of both
lesions confirms metastases from endometrial Case Example 5 Vaginal vault recurrence;
cancer. The patient is considered ineligible for MR-Linac boost after a course of pelvic IMRT
additional surgery due to comorbidities, includ- curative treatment intent.
ing obesity. Definitive radiotherapy is planned, A 60-year-old patient is diagnosed with a
consisting of sequentially 1.5 T MR-Linac SBRT vaginal vault recurrence of endometrial cancer.
for the abdominal wall metastasis (35 Gy in five The tumor extends to the right pelvic wall caus-
fractions), followed by pelvic EBRT with VMAT ing hydronephrosis (A). After completing a
on a conventional Linac (45 Gy in 25 fractions) course of pelvic IMRT to a dose of 45 Gy in 25
and finally a sequential MR-Linac boost to the fractions, the patient receives a boost of 30 Gy in
vaginal vault recurrence (28 Gy in four fractions). 6 fractions to the PTV using MRgRT. The
Vaginal vault BT is considered but not deemed HRCTV is expanded by 3 mm to generate the
feasible since a complex interstitial approach PTV. Then the optimized PTV (PTVopt) is
under anesthesia would be necessary but not fea- obtained by subtracting 3 mm from the sur-
sible due to comorbidity. A composite approach rounding OARs (B). The dose constraints for the
is done for each of the three radiotherapy courses composite (pelvic IMRT and MR-Linac boost)
(see Fig. 18.9 below). are the following: dose to D2cc to bladder,
For the MR-Linac boosts to the vaginal vault 77 Gy; rectum, 65 Gy; sigmoid, 65 Gy; and
recurrence, an isotoxic approach is used, giving small bowel, 65 Gy. An isotoxic plan is gener-
priority to hard OAR dose constraints during the ated, meeting all these dose constraints, and a
planning process, calculating the sum of elective daily adaptive approach is used. More than 2
EBRT and the four boost fractions. For the offline years posttreatment, imaging shows a complete
pretreatment planning situation, the total dose to response, and the patient has no treatment toxic-
the vault recurrence (D90 HR-CTV) is calculated ity (Fig. 18.10, Table 18.4).
a b
c d
e f
Fig. 18.9 Endometrial cancer; MR-Linac SBRT for con- rence in blue. Target and OAR are clearly visualized on
current local recurrent abdominal wall metastasis. the MR images allowing for daily adaptation. (c, d)
Endometrial cancer; MR-Linac SBRT for concurrent vag- Typical daily MR-Linac plans for both lesions (isodoses
inal vault recurrence and anterior abdominal wall metasta- red 110%, orange 100%, blue 25%). (e) Elective EBRT
sis. (a) T2 sagittal, (b) ADC map derived from plan on planning CT (yellow 95%, green 82%, blue 52%).
diffusion-weighted image acquired with MR-Linac. (f) Overlay of elective and boost plans (range 0–70 Gy
Abdominal metastasis in pink and vaginal vault recur- physical dose)
Fig. 18.10 A coronal view of 3.0 tesla diagnostic MRI

showing the residual disease at the vaginal vault
extending to the right pelvic wall and causing a
hydroureter. (b) Isodose line achieved for the MR-Linac
boost, the red isodose line being the 30 Gy isodose
Table 18.4 Prescription of the MR-Linac boost for a vaginal vault recurrence of endometrial cancer
Treatment of Oligometastatic even be approached with curative intent. SBRT

Gynecologic Cancer of oligometastatic disease has been reported to
increase survival while preserving the quality of
The concept of oligometastatic disease was first life. Chapter 12 reviews the rationale for using
introduced in 1995 by Hellman and Weiselbaum MRgRT to deliver an ablative dose to oligometa-
[97], who described a subcategory of patients static disease, and the reader should refer to this
with metastatic disease whose prognosis is better section of the book to get a full scope of the sub-
than widespread metastatic disease and might ject. This section will briefly present a review of
Table 18.5 Published series on the use of SBRT for oligometastatic gynecologic cancer
Patients Dose LC Toxicity
Mixed Gyn cancer
Kunos et al. (2012) 50 pts. 24 Gy/3 fx LC 100% Gde 1 and 2:
[98] Metastatic GYN ca (Rx at 70% isodose) (CR + PR) 76%
all sites BED10: 43 At 2 years Gde 3: 4%
Only 1 CR Gde 4: 2%
Mesko et al. (2017) 28 pts. Median total dose (Gy) 40 LC 83% Gde 3: 1 pt.
[99] (47 lesions) Median Fx: 5 34% (SD)
Metastatic GYN BED10: >72 associated with 32% (PR)
Ca all sites better outcome 17% (CR)
17% (PD)
Median FU: 12
months
Ovarian cancer
Lazzari et al. (2018) 82 pts 24 Gy/3 fx LC 70% all sites No Gde 3–4
[100] Ovarian cancer BED10: 43 (81% if LN) toxicity
At 17.4 months
Macchia et al. 449 pts Median total dose (Gy): 25 LC 81.9% No Gde 3–4
(2020) [101] Ovarian cancer (5–75) Median FU: 22 toxicity
Median no. of fx 4 (1–13) months Gde 1–2:
Median dose/fx, (Gy): 8 (3–30) 20.7%
Note: BED 10 >70 Gy = higher
CR in MVA
the literature (Table 18.5) supporting the use of tracking and automatic beam stopping. For oligo-
SBRT for oligometastatic gynecologic cancer metastatic targets localized in the chest or upper
with a discussion on guidelines to select an abdomen, intra-fraction target motion is impor-
appropriate dose fractionation for oligometastatic tant to consider, and having the ability to perform
gynecologic malignancies. real-time tracking with automatic beam stopping
Surgery is rarely possible in the management when motion is detected allows treatment with
of oligometastatic gynecologic cancer, either due reduced margins and increased safety.
to unfavorable locations and/or patients unfit for Patients are often asymptomatic from the site
surgery [102]. Combined with systemic treat- of oligometastatic disease when the disease bur-
ment, salvage irradiation can be used to treat den is low. However, the paradigm that radiation
these disease sites. The fact that SBRT is deliv- should only be offered to palliate symptoms in
ered in only a few treatment sessions and is a the metastatic status is now being challenged,
noninvasive ablative treatment alternative makes and data support that there is a potential survival
it a great local therapy approach to integrate with benefit in treating these small sites of metastatic
systemic treatment with minimal interference deposits [3]. SBRT for small target volumes pro-
with the systemic treatment schedule. Radiation vides good tumor control rates with a low toxic-
can also be offered to patients with oligometa- ity profile [104–106].
static gynecologic cancer to allow for a pause in Focusing only on the site of recurrence car-
systemic treatment or postpone resuming sys- ries the risk of out-of-field local progression in
temic treatment [103]. MRI guidance and ART other regional lymph nodes. Therefore, it is
offer the advantage of better tumor localization essential to carefully assess the clinical situation
and high precision in dose delivery, which is at the time of disease recurrence. Treatment of a
especially important when planning to deliver a patient with cervical cancer presenting a PAO
high dose per fraction for a target surrounded by recurrence should be managed as a locoregional
luminal organs. Another important feature of progression. These patients should be offered a
MR-Linac systems is the advantage of real-time curative treatment with a comprehensive chemo-
radiation approach, as discussed in this chapter A large retrospective multicenter analysis of

in the section on the management of cervical SBRT of 449 ovarian cancer lesions found that
cancer recurrence. Out-of-field progression patient’s age ≤60-year-old, a PTV size of
could be prevented by using larger EBRT fields; ≤18 cm3, lymph node disease, and a BED (a/b10)
however, in some situations, overlapping vol- of >70 Gy were independent predictive factors of
umes of treatment fields with previous radio- complete response on multivariate analysis [101].
therapy areas lead to higher morbidity rates This analysis led to the development of the ongo-
[104], and in some situations, patients would ing MITO-RT3/RAD prospective clinical trial, a
not tolerate longer fractionation and larger treat- multicenter phase II study aimed at identifying
ment volumes. potential predictors of response and clinical out-
There is a clear dose-response correlation comes after SBRT treatment for patients with
between higher biologically equivalent doses oligometastatic/persistent/recurrent ovarian
(BEDs) and LC rates [98, 99, 107, 108]. For cancer.
lymph node metastases, 5-year LC rates of uter- The study hypothesis is that SBRT delivered
ine, cervical, and ovarian cancer range from 70 to by predefined treatment schedules and shared
97%, and good rates of disease-free survival and constraints could improve the complete response
overall survival are reported in retrospective rate, and the primary endpoint is the complete
series [100, 101, 105, 108–113]. Lymph node clinical response rate. All patients accrued will be
metastases of gynecologic malignancies are often treated with a radiotherapy dose in the range of
located in the pelvis or abdomen, where conven- 30–50 Gy by 1, 3, or 5 fractions to all sites of
tional SBRT using CBCT provides poor soft tis- active metastatic disease. The best response for
sue contrast. When delivering treatment on each lesion basis will be evaluated by CT scan,
standard Linac, dose escalation will be limited in positron emission tomography/CT, or MRI in
these clinical situations since the dose needs to be case of brain lesions every 3 months.
maintained below the dose tolerance of the sur- Approximately 205 lesions will be treated (90
rounding bowel. In this clinical situation, MRgRT lymph nodes and 115 parenchyma lesions). The
offers significant advantages, as it allows direct investigators plan to complete enrollment by
visualization of the metastases on MRI and spring 2023, and the analysis will be completed
enables margin reduction or dose escalation strat- in the same year [115].
egies by using online ART and automated gating The fractionation of SBRT varies widely
systems [114]. based on tumor-related parameters (size, prox-
Lazzari reported on the outcome of SBRT in imity to vulnerable OAR, tumor motion) and if
oligometastatic ovarian cancer [100]. In this the target lies in a previously irradiated field.
series, SBRT was used for oligorecurrent or oli- Breathing motion for oligometastasis located in
goprogressive disease in heavily pretreated the chest or upper abdomen and changes in the
patients (median of three prior systemic therapy filling status of surrounding OARs in the pelvis
regimens). Excellent LC was achieved without can present a real challenge. MRgRT can
any ≥ grade 3 acute or late toxicity. The median improve the feasibility of dose escalation to sites
systemic treatment-free interval after SBRT in of oligometastatic in the abdomen and pelvis. In
this cohort was 7.4 months, and over one-third of addition, the fact that the MR-Linac system can
patients were still disease-free 1 year after automatically gate the beam by using real-time
SBRT. SBRT was able to postpone the need to tracking eliminates the need for invasive implan-
reinitiate systemic therapy allowing patients to tation of fiducial markers and the addition of an
recover from systemic treatment toxicity for ITV to account for inter-fraction motion, leading
many months before resuming treatment. If to reduced healthy surrounding tissue
focused treatment is being used, when needed, irradiation.
multiple SBRT courses could be used in case of For many tumor sites, the combination of
subsequent recurrence. SBRT with immunotherapy has been shown to
enhance the immune response. With the emer- 4. Gregoire V, Guckenberger M, Haustermans K, et al.
gent use of immunotherapy in managing gyneco- Image guidance in radiation therapy for better cure
of cancer. Mol Oncol. 2020;14:1470–91.
logic cancer, MRg SBRT might play a dominant 5. Sturdza A, Pötter R, Fokdal LU, Haie-Meder
role in the multidisciplinary treatment strategy C, et al. Image-guided brachytherapy in locally
for oligometastatic gynecologic cancer. advanced cervical cancer: improved pelvic control
and survival in RetroEMBRACE, a multicenter
cohort study. Radiother Oncol. 2016;120(3):428–33.
https://doi.org/10.1016/j.radonc.2016.03.011.
Conclusion 6. Westerveld H, Schmid MP, Nout RA, et al. Image-
guided adaptive brachytherapy (IGABT) for pri-
MR-Linac is an emerging technology that broad- mary vaginal cancer: results of the international
multicenter RetroEMBRAVE cohort study. Cancers
ens the radiation treatment options for patients (Basel). 2021;13(6):1459. https://doi.org/10.3390/
with gynecologic cancers. This promising tech- cancers13061459.
nology is still very recent, and tremendous work 7. Scoutt LM, McCarthy SM. Applications of magnetic
needs to be done to explore its clinical relevance resonance imaging to gynecology. Top Magn Reson
Imaging. 1990;2(3):37–49.
and possible superiority over treatment systems 8. Russell AH, Anderson M, Walter J, Kinney W,
with CT scanners on board. Gynecologic cancers Smith L, Scudder S. The integration of computed
develop in organs associated with large inter- tomography and magnetic resonance imaging in
fraction motion and significant changes in shape treatment planning for gynecologic cancer. Clin
Obstet Gynecol. 1992;35(1):55–72. https://doi.
and extension through the course of external org/10.1097/00003081-199203000-0001.
beam treatment. Therefore, MRgRT, with the 9. Pötter R, Dimopoulos J, Georg P, et al. Clinical
potential of daily online treatment planning impact of MRI assisted dose volume adaptation and
according to the actual anatomical situation, dose escalation in brachytherapy of locally advanced
cervix cancer. Radiother Oncol. 2007;83(2):148–55.
could significantly improve the treatment out- https://doi.org/10.1016/j.radonc.2007.04.012.
come. For boosts to primary tumor lesions in 10. Nag S, Chao C, Erickson B, et al. The American
LACC and vaginal cancer, image-guided BT Brachytherapy Society recommendations for low-
remains the treatment of choice, and MRgRT dose-rate brachytherapy for carcinoma of the cervix.
Int J Radiat Oncol Biol Phys. 2002;52(1):33–48.
should only be considered if BT is not feasible. https://doi.org/10.1016/s0360-3016(01)01755-2.
Through joint efforts and collaboration, clinical 11. Haie-Meder C, Pötter R, Van Limbergen E, et al.
evidence to explore the full potential of this new Gynaecological (GYN) GEC-ESTRO Working
technology for the management of women with Group Recommendations from Gynaecological
(GYN) GEC-ESTRO Working Group (I): concepts
gynecologic cancer should be built during the and terms in 3D image-based 3D treatment planning
next decade. in cervix cancer brachytherapy with emphasis on
MRI assessment of GTV and CTV. Radiother Oncol.
2005;74(3):235–45. https://doi.org/10.1016/j.
radonc.2004.12.015.
References 12. Pötter R, Tanderup K, Schmid MP, EMBRACE
Collaborative Group, et al. MRI-guided adaptive
1. Sung H, Ferlay J, Siegel RL, Jemal A, et al. Global brachytherapy in locally advanced cervical cancer
cancer statistics 2020; GLOBOCAN estimates of (EMBRACE-I): a multicenter prospective cohort
incidence and mortality worldwide for 36 cancers in study. Lancet Oncol. 2021;22(4):538–47. https://doi.
185 countries. CA Cancer J Clin. 2021;71(3):209– org/10.1016/S1470-2045(20)30753-1.
49. https://doi.org/10.3322/caac.21660. 13. Balleyguier C, Sala E, Da Cunha T, et al. Staging
2. Siegel RL, Miller KD, Wagle NS, Jemal A. Cancer of uterine cervical cancer with MRI: guidelines of
statistics 2023. CA Cancer J Clin. 2023;73(1):17– the European Society of Urogenital Radiology.
48. https://doi.org/10.3322/caac.21763. Eur Radiol. 2011;21(5):1102–10. https://doi.
3. Palma DA, Olson R, Harrow S, et al. Stereotactic org/10.1007/s00330-010-1998-x.
ablative radiotherapy for the comprehensive treat- 14. AJCC cancer staging manual. 8th ed. Cervical can-
ment of oligometastatic cancers: long-term results cer staging. Springer; 2017.
of the SABR-COMET phase II randomized trial. 15. Sala E, Rockall A, Rangarajan D, Kubik-Huch
J Clin Oncol. 2020;38(25):2038–838. https://doi. RA. The role of dynamic contrast enhanced and
org/10.1200/JCO.20.00818. diffusion-weighted magnetic resonance imaging in
the female pelvis. Eur J Radiol. 2010;76(3):367–85. 28. Lin G, Ho KC, Wang JJ, et al. Detection of lymph
https://doi.org/10.1016/j.ejrad.2010.01.026. node metastasis in cervical and uterine cancers by
16. Otero-Garcia MM, Mesa-Alvarez A, Nikolic O, et al. diffusion-weighted magnetic resonance imaging
Role of MRI in staging and follow up of endometrial at 3T. J Magn Reson Imaging. 2008;28(1):128–35.
and cervical cancer: pitfalls and mimickers. Insights https://doi.org/10.1002/jmri.21412.
Imaging. 2019;10(19):1–22. https://doi.org/10.1186/ 29. Liu Y. Benign ovarian and endometrial uptake
s13244-019-0696-8. on FDG PET-CT: patterns and pitfalls. Ann Nucl
17. Zhang W, Chen C, Liu P, et al. Impact of pelvic MRI Med. 2009;23(2):107–12. https://doi.org/10.1007/
in routine clinical practice on staging of IB1-IIA2 s12149-008-0227-z.
cervical cancer. Cancer Manag Res. 2019;11:3603– 30. Nakai G, Matsuki M, Inada Y, et al. Detection and
9. https://doi.org/10.2147/CMAR.S197496. evaluation of pelvic lymph nodes in patients with
18. Akita A, Shinmoto H, Hayashi S, et al. Comparison gynecologic malignancies using body diffusion-
of T2-weighted and contrast-enhanced T1-weighted weighted magnetic resonance imaging. J Comput
MR imaging at 1.5 T for assessing the local extent of Assist Tomogr. 2008;32(5):764–8. https://doi.
cervical carcinoma. Eur Radiol. 2011;21(9):1850–7. org/10.1097/RCT.0b013e318153fd43.
https://doi.org/10.1007/s00330-011-2122-6. 31. Choi HJ, Kim SH, Seo SS, et al. MRI for pretreat-
19. Meissnitzer M, Forstner R. MRI of endometrium ment lymph node staging in uterine cervical cancer.
cancer—how we do it. Cancer Imaging. 2016;16:11. Am J Roentgenol. 2006;187(5):W538–43. https://
https://doi.org/10.1186/s40644-016-0069-1. doi.org/10.2214/AJR.05.0263.
20. Seki H, Azumi R, Kimura M, Sakai K. Stromal 32. Bates DB, Homsi ME, Chang KJ, et al. MRI for rec-
invasion by carcinoma of the cervix: assessment tal cancer: staging, MR-CRM, EMVI, lymph node
with dynamic MR imaging. Am J Roentgenol. staging and post-treatment response. Clin Colorectal
1997;168(6):1579–85. https://doi.org/10.2214/ Cancer. 2022;21(1):10–8. https://doi.org/10.1016/j.
ajr.168.6.9168730. clcc.2021.10.007.
21. Vargas HA, Akin O, Zheng J, et al. The value of MR 33. Nam H, Huh SJ, Park W, et al. Prognostic signifi-
imaging when the site of uterine cancer origin is cance of MRI-detected bladder muscle and/or sero-
uncertain. Radiology. 2011;258(3):785–92. https:// sal invasion in patients with cervical cancer treated
doi.org/10.1148/radiol.10101147. with radiotherapy. Br J Radiol. 2010;83:868–73.
22. Shweel MA, Abdel-Gawad EA, Abdel-Gawad EA, https://doi.org/10.1259/bjr/66646798.
Abdelghany HS, Abdel-Rahman AM, Ibrahim 34. Wakatsuki M, Kato S, Kiyohara H, et al. The prog-
EM. Uterine cervical malignancy: diagnostic nostic value of rectal invasion for stage IVA uter-
accuracy of MRI with histopathologic correla- ine cervical cancer treated with radiation therapy.
tion. J Clin Imaging Sci. 2012;2:42. https://doi. BMC Cancer. 2016;16:244. https://doi.org/10.1186/
org/10.4103/2156-7514.99175. s12885-016-2268-3.
23. Thomeer MG, Gerestein C, Spronk S, van Doorn 35. Nougaret S, Horta M, Sala E, et al. Endometrial can-
HC, van der Ham E, Hunink MG. Clinical exami- cer MRI staging: update guidelines of the European
nation versus magnetic resonance imaging in the Society of Urogenital Radiology. Eur Radiol.
pretreatment staging of cervical carcinoma: sys- 2019;29(2):792–805. https://doi.org/10.1007/
tematic review and meta-analysis. Eur Radiol. s00330-018-5515.
2013;23(7):2005–18. https://doi.org/10.1007/ 36. Creasman WT, Morrow CP, Bundy BN, Homesly
s00330-013-2783-4. HD, Graham JE, Heller PB. Surgical patho-
24. Woo S, Suh CH, Kim SY, Cho JY, Kim SH. Magnetic logic spread patterns of endometrial can-
resonance imaging for detection of parametrial inva- cer. A Gynecologic Oncology Group Study.
sion in cervical cancer: an updated systematic review Cancer. 1987;60(8):2035–41. https://doi.
and meta-analysis of the literature between 2012 and org/10.1002/1097-0142(19901015)60:8.
2016. Eur Radiol. 2018;28(2):530–41. https://doi. 37. Schwarz JK, Beriwal S, Esthappan J, et al.
org/10.1007/s00330-017-4958-x. Consensus statement for brachytherapy for the treat-
25. Bourgioti C, Chatoupis K, Moulopoulos LA. Current ment of medically inoperable endometrial cancer.
imaging strategies for the evaluation of uterine cer- Brachytherapy. 2015;14(5):587–99. https://doi.
vical cancer. World J Radiol. 2016;8(4):342–54. org/10.1016/j.brachy.2015.06.002.
https://doi.org/10.4329/wjr.v8.i4.342. 38. van de Bunt L, van der Heide UA, Ketelaars M, de
26. Sala E, Wakely S, Senior E, Lomas D. MRI of malig- Kort GAP, Jugenliemk-Schulz IM. Conventional,
nant neoplasms of the uterine corpus and cervix. Am conformal, and intensity-modulated radiation ther-
J Roentgenol. 2007;188(6):1577–87. https://doi. apy treatment planning of external beam radiotherapy
org/10.2214/AJR.06.1196. for cervical cancer: the impact of tumor regression.
27. Atcı N, Özgür T, Öztürk F, Dolapçıoğlu KS. Utility Int J Radiat Oncol Biol Phys. 2006;64(1):189–96.
of intravaginal ultrasound gel for local stag- https://doi.org/10.1016/j.ijrobp.2005.04.025.
ing of cervical carcinoma on MRI. Clin Imaging. 39. van de Bunt L, Jürgenliemk-Schulz IM, de Kort
2016;40(6):1104–7. https://doi.org/10.1016/j. GAP, Roesink JM, Tersteeg RJHA, van der Heide
clinimag.2016.07.004. UA. Motion and deformation of the target volumes
during IMRT for cervical cancer: what margins do 51. Liu Y, Sun H, Bai R, Ye Z. Time-window of early
we need? Radiother Oncol. 2008;88(2):233–40. detection of response to concurrent chemora-
https://doi.org/10.1016/j.radonc.2007.12.017. diation in cervical cancer by using diffusion-
40. Kerkhof EM, Raaymakers BW, van der Heide UA, weighted MR imaging: a pilot study. Radiat
van de Bunt L, Jürgenliemk-Schulz IM, Lagendijk Oncol. 2015;10:185.
JJW. Online MRI guidance for healthy tissue sparing 52. Somoye G, Harry V, Semple S, et al. Early diffusion-
in patients with cervical cancer: an IMRT planning weighted magnetic resonance imaging can predict
study. Radiother Oncol. 2008;88(2):241–9. https:// survival in women with locally advanced cancer of
doi.org/10.1016/j.radonc.2008.04.009. the cervix treated with combined chemo-radiation.
41. Bowen SR, Yuh WTC, Hippe DS, et al. Tumor Eur Radiol. 2012;22:2319–27.
radiomic heterogeneity: multiparametric functional 53. de Foucher T, Bendifallah S, Ouldamer L, et al.
imaging to characterize variability and predict Patterns of recurrence and prognosis in locally
response following cervical cancer radiation ther- advanced FIGO stage IB2 to IIB cervical cancer: ret-
apy. J Magn Reson Imaging. 2018;47(5):1388–96. rospective multicenter study from the FRANCOGYN
https://doi.org/10.1002/jmri.25874. group France. Eur J Surg Oncol. 2019;45(4):659–65.
42. De Boer P, Mandija S, Werensteijn-Honingh AM, https://doi.org/10.1016/j.ejso.2018.11.014.
van den Berg CAT, de Leeuw AAC, Jürgenliemk 54. Schmid MP, Franckena M, Kirchheiner K, et al.
Schulz IM. Cervical cancer apparent diffusion coef- Distant metastasis in patients with cervical cancer
ficient values during external beam radiotherapy. after primary radiotherapy with or without chemo-
Phys Imaging Radiat Oncol. 2019;9:77–82. https:// therapy and image-guided adaptive brachytherapy.
doi.org/10.1016/j.phro.2019.03.001. Gynecol Oncol. 2014;133(2):256–62. https://doi.
43. Han K, Croke J, Foltz W, et al. A prospective study org/10.1016/j.ygyno.2014.02.004.
of DWI, DCE-MRI and FDG PET imaging for tar- 55. Sohaib SA, Houghton SL, Meroni R, Rockall AG,
get delineation in brachytherapy for cervical cancer. Blake P, Reznek RH. Recurrent endometrial cancer:
Radiother Oncol. 2016;120(3):519–25. https://doi. patterns of recurrent disease and assessment of prog-
org/10.1016/j.radonc.2016.08.002. nosis. Clin Radiol. 2007;62(1):28–34. https://doi.
44. Nougaret S, McCague C, Tibermacine H, Vargas org/10.1016/j.crad.2006.06.015.
HA, Rizzo S, Sala E. Radiomics and radiogenomics 56. Lucas R, Dias JL, Cunha TM. Added value of
in ovarian cancer: a literature review. Abdom Radiol diffusion-weighted MRI in detection of cervical
(NY). 2021;46(6):2308–22. https://doi.org/10.1007/ cancer recurrence: comparison with morphologic
s00261-020-02820-z. and dynamic contrast-enhanced MRI sequences.
45. Michalet M, Azria D, Tardieu M, Tibermacine H, Diagn Interv Radiol. 2015;21(5):368–75. https://doi.
Nougaret S. Radiomics in radiation oncology for org/10.5152/dir.2015.14427.
gynecological malignancies: a review of literature. 57. Meads C, Davenport C, Małysiak S, et al. Evaluating
Br J Radiol. 2021;94(1125):20210032. PET-CT in the detection and management of
46. Takada A, Yokota H, Watanabe Nemoto M, recurrence cervical cancer: systematic reviews
Horikoshi T, Matsushima J, Uno T. A multi-scanner of diagnostic accuracy and subjective elicita-
study of MRI radiomics in uterine cervical cancer: tion. BJOG. 2014;121(4):398–407. https://doi.
prediction of infield tumor control after definitive org/10.1111/1471-0528.12488.
radiotherapy based on a machine learning method 58. Morris M, Eifel PJ, Lu J, et al. Pelvic radiation with
including peritumoral regions. Jpn J Radiol. concurrent chemotherapy compared with pelvic and
2020;38(3):265–73. https://doi.org/10.1007/ para-aortic radiation for high-risk cervical cancer.
s11604-019-00917-0. N Engl J Med. 1999;340(15):1137–43. https://doi.
47. Engin G. Cervical cancer: MR imaging findings org/10.1056/NEJM199904153401501.
before, during, and after radiation therapy. Eur 59. Peters M, de Leeuw AAC, Nomden CN, et al. Risk
Radiol. 2006;16(2):313–24. https://doi.org/10.1007/ factors for nodal failure after radiochemotherapy and
s00330-005-2804-z. image-guided brachytherapy in locally advanced
48. Dappa E, Elger T, Hasenburg A, Düber C, Battista cervical cancer: an EMBRACE analysis. Radiother
MJ, Hötker AM. The value of advanced MRI tech- Oncol. 2021;163:150–8. https://doi.org/10.1016/j.
niques in the assessment of cervical cancer: a review. radonc.2021.08.020.
Insights Imaging. 2017;8:471–81. 60. Bukkems LJH, Jurgenliemk-Schulz IM, van der Leij
49. Fu C, Feng X, Bian D, et al. Simultaneous changes F, Peters M, Gerestein CG, Zweemer RP. The impact
of magnetic resonance diffusion-weighted imaging of para-aortic lymph node irradiation on disease-free
and pathological microstructure in locally advanced survival in patients with cervical cancer: a system-
cervical cancer caused by neoadjuvant chemother- atic review and meta-analysis. Clin Transl Radiat
apy. J Magn Reson Imaging. 2015;42:427–35. Oncol. 2022;35:97–103. https://doi.org/10.1016/j.
50. Kuang F, Yan Z, Wang J, Rao Z. The value of ctro.2022.05.006.
diffusion-weighted MRI to evaluate the response to 61. Pötter R, Tanderup K, Kirisits C, The EMBRACE
radiochemotherapy for cervical cancer. Magn Reson Collaborative Group, et al. The EMBRACE II study:
Imaging. 2014;32:342–9. the outcome and prospect of two decades of evolu-
tion within the GEC-ESTRO GYN working group ning for cervical cancer—a comparison of offline,
and the EMBRACE studies. Clin Transl Radiat online and MRI-guided techniques to reduce organ
Oncol. 2018;11(9):48–60. https://doi.org/10.1016/j. at risk (OAR) dose using a dose accumulation
ctro.2018.01.001. model. Int J Radiat Oncol Biol Phys. 2019;105(1):
62. Tanderup K, Lindegaard JC, Kirisits C, et al. Image S250–1.
guided adaptive brachytherapy in cervix cancer: a 71. Kozak M, Koenig JL, von Eyben R, Kidd EA. Less
new paradigm changing clinical practice and out- than whole uterus irradiation for locally advanced
come. Radiother Oncol. 2016;120(3):365–9. https:// cervical cancer maintains locoregional con-
doi.org/10.1016/j.radonc.2016.08.007. trol and decreases radiation dose to bowel. Pract
63. Grover S, Harkenrider MM, Cho LP, et al. Image Oncol Radiother. 2018;9(2):e164–71. https://doi.
guided cervical brachytherapy: 2014 survey of org/10.1016/j.prro.2018.10.009.
the American Brachytherapy Society. Int J Radiat 72. Buschmann M, Majercakova K, Sturdza A, et al.
Oncol Biol Phys. 2014;94:598–60. https://doi. Image guided adaptive external beam radiation
org/10.1016/j.ijrobp.2015.11.024. therapy for cervix cancer: evaluation of a clini-
64. Pötter R, Georg P, Dimopoulos JCA, et al. Clinical cally implemented plan-of-the-day technique. Med
outcome of protocol-based image (MRI) guided Phys. 2018;28(3):184–95. https://doi.org/10.1016/j.
adaptive brachytherapy combined with 3D confor- zemedi.2017.09.004.
mal radiotherapy with or without chemotherapy 73. Mendez LC, Raziee H, Davidson M, et al. Should
in patients with locally advanced cervical cancer. we embrace hypofractionated radiotherapy for
Radiother Oncol. 2011;100:116–23. https://doi. cervical cancer? A technical note on manage-
org/10.1016/j.radonc.2011.07.012. ment during COVID-19 pandemic. Radiother
65. Potter R, Haie-Meder C, Van Limbergen E, et al. Oncol. 2020;148:270–3. https://doi.org/10.1016/j.
GEC ESTRO Working Group. Recommendations radonc.2020.05.032.
from Gynaecological (GYN) GEC ESTO Working 74. Hes J, Van Lier A, De Leeuw A, Kroon P, DeGroot-
Group (II): concepts and terms in 3D image-based van Breugel E, Jurgenkiemk-Schulz I. OC-0713:
treatment planning in cervix cancer brachytherapy- MR-Linac boosts for patients with cervix cancer
3D dose volume parameters and aspects of 3D
ineligible for brachytherapy; preliminary experi-
image-based anatomy, radiation physics, radiobiol- ence. Radiother Oncol. 2020;152:S402–3. https://
ogy. Radiother Oncol. 2006;78(1):67–77. https:// doi.org/10.1016/S0167-8140(21)00735-09.
doi.org/10.1016/j.radonc.2005.11.014. 75. Hadi I, Eze C, Schönecker S, von Bestenbostel
66. Nag S, Cardenes H, Chang S, et al. Image-guided R, et al. MR-guided SBRT boost for patients with
brachytherapy working group. Proposed guidelines locally advanced or recurrent gynecological cancers
for image-based intracavitary brachytherapy for cer- ineligible for brachytherapy: feasibility and early
vical carcinoma: report from image-guided brachy- clinical experience. Radiat Oncol. 2022;17(8):1–9.
therapy working group. Int J Radiat Oncol Biol Phys. https://doi.org/10.1186/s13014-022-01981-z.
2004;60(4):1160–72. https://doi.org/10.1016/j. 76. Mahmoud O, Kilic S, Khan AJ, Beriwal S, Small
ijrobp.2004.04.032. W. External beam techniques to boost cervical cancer
67. Berger T, Seppenwoolde Y, Pötter R, EMBRACE when brachytherapy is not an option—theories and
Collaborative Group, et al. Importance of technique, applications. Ann Transl Med. 2017;5(10):5. https://
target selection, contouring, dose prescription, and doi.org/10.21037/atm.2017.03.102.
dose-planning in external beam radiation therapy 77. Kilic S, Cracchiolo B, Mahmoud O. Non-
for cervical cancer: evolution of practice from brachytherapy alternatives in cervical cancer radio-
EMBRACE-I to II. Int J Radiat Oncol Biol Phys. therapy: why not? Appl Radiat Oncol. 2015;4:11–7.
2019;104(4):885–94. https://doi.org/10.1016/j. 78. Albuquerque K, Tumati V, Lea J, Ahn C, Richardson
ijrobp.2019.03.020. D, Miller D, Timmerman R. A phase II trial of
68. Hariteportunkel NH, Nath SK, Scanderbeg D, stereotactic ablative radiation therapy as a boost
Saenz C, Yashar CM. Evaluation of intra-and inter- for locally advanced cervical cancer. Int J Radiat
fraction movement of the cervix during intensity Oncol Biol Phys. 2020;106(3):464–71. https://doi.
modulated radiation therapy. Radiother Oncol. org/10.1016/j.ijrobp.2019.10.042.
2011;98(3):347–51. https://doi.org/10.1016/j. 79. Tanderup K, Fokdal LY, Sturdza A, et al. Effect of
radonc.2010.11.015. tumor dose, volume and overall treatment time on
69. Lim K, Small W, Portelance L, et al. For the GYN local control after radiochemotherapy including
IMRT consortium consensus guidelines for delin- MRI guided brachytherapy of locally advanced cer-
eation of clinical target volume for intensity- vical cancer. Radiother Oncol. 2016;120(3):441–6.
modulated pelvic radiotherapy for the definitive https://doi.org/10.1016/j.radonc.2017.01.022.
treatment of cervix cancer. Int J Rad Oncol Biol 80. Han K, Milosevic M, Fyles A, Pintilie M,
Phys. 2011;79(2):348–55. https://doi.org/10.1016/ Viswanathan AN. Trends in the utilization of brachy-
ijrobp.2009.10.075. therapy in cervical cancer in the United States. Int J
70. White I, Mcquaid E, Godwin E, et al. Image-guided Radiat Oncol Biol Phys. 2013;87(1):111–9. https://
adaptive external beam radiotherapy (EBRT) plan- doi.org/10.1016/j.ijrobp.2013.05.033.
81. Holschneider CH, Petereit DG, Chu C, et al. Phys. 2005;62(2):468–78. https://doi.org/10.1016/j.
Brachytherapy: a critical component of primary radi- ijrobp.2004.10.013.
ation therapy for cervical cancer: from the Society 92. Hardarson HA, Heidemann LN, dePont CR,
of Gynecologic Oncology (SGO) and the American Mogensen O, Jochumsen KM. Vaginal vault recur-
Brachytherapy Society (ABS). Gynecol Oncol. rences of endometrial cancer in non-irradiated
2019;152(3):540–7. https://doi.org/10.1016/j. patients—radiotherapy or surgery. Gynecol Oncol
ygyno.2018.10.016. Rep. 2015;11:26–30. https://doi.org/10.1016/j.
82. Ko HC, Huang JY, Miller JR, et al. Novel use gore.2015.01.002.
of ViewRay MRI guidance for high-dose-rate 93. Pai HH, Souhami L, Clark BG, Roman T. Isolated
brachytherapy in the treatment of cervical can- vaginal recurrences in endometrial carcinoma:
cer. Brachytherapy. 2018;17:680–8. https://doi. treatment results using high-dose-rate intracavi-
org/10.1016/j.brachy.2018.04.005. tary brachytherapy and external beam radiotherapy.
83. Babar S, Rockall A, Goode A, Shepherd J, Gynecol Oncol. 1997;66(2):300–7. https://doi.
Reznek R. Magnetic resonance imaging appear- org/10.1006/gyno.1997.4752.
ances of recurrent cervical carcinoma. Int J 94. Kamrava M, Sushil Beriwal S, Erickson B, et al.
Gynecol Cancer. 2007;17(3):637–45. https://doi. American Brachytherapy Society recurrent carci-
org/10.1111/j.1525-1438.2007.00849.x. noma of the endometrium task force patterns of
84. Pervin S, Ruma FI, Rahman K, et al. Adjuvant hys- care and review of the literature. Brachytherapy.
terectomy in patients with residual disease after 2017;16(6):1129–43. https://doi.org/10.1016/j.
radiation for locally advanced cervical cancer: brachy.2017.07.012.
a prospective longitudinal study. J Glob Oncol. 95. Weitmann HD, Knocke TH, Waldhäusl C, Pötter
2019;4:157. https://doi.org/10.1200/JGO.18.00157. R. Ultrasound-guided interstitial brachytherapy in
85. Parikh PJ, Lee P, Low DA, et al. Multi-institutional the treatment of advanced vaginal recurrences from
phase 2 trial of ablative 5-fraction stereotactic cervical and endometrial carcinoma. Strahlenther
MR-guided on 2 table adaptive radiation therapy Onkol. 2006;182(2):86–95. https://doi.org/10.1007/
(SMART) for inoperable pancreatic cancer. Int J s00066-006-1420-4.
Radiat Biol Oncol Phys. 2023;117:799–808. https:// 96. Fokdal L, Tanderup K, Nielsen SK, et al. Image
doi.org/10.1016/j.ijrobp.2023.05.023. and laparoscopic guided interstitial brachytherapy
86. Yoo S, Hegarty SE, Mishra MV, Patel N, Cantrell for locally advanced primary or recurrent gynaeco-
LA, Showalter TN. Definitive radiation therapy logical cancer using the adaptive GEC ESTRO tar-
for stage I–II endometrial cancer: an observational get concept. Radiother Oncol. 2011;100(3):473–9.
study of nonoperative management. Am J Clin https://doi.org/10.1016/j.radonc.2011.08.016.
Oncol. 2017;40(6):582–9. https://doi.org/10.1097/ 97. Hellman S, Weichselbaum RR. Oligometastasis. J
COC.0000000000000204. Clin Oncol. 1995;13:8–10. https://doi.org/10.1200/
87. Podzielinski I, Randall ME, Breheny PJ, et al. JCO.1995.13.1.8.
Primary radiation therapy for medically inoperable 98. Kunos CA, Brindle J, Waggoner S, et al. Phase II
patients with clinical stage I and II endometrial car- clinical trial of robotic stereotactic body radiosurgery
cinoma. Gynecol Oncol. 2012;124(1):36–41. https:// for metastatic gynecologic malignancies. Front
doi.org/10.1016/j.ygyno.2011.09.022. Oncol. 2012;2(181):1–6. https://doi.org/10.3389/
88. Nag S, Erickson B, Parikh S, Gupta N, Varia M, fonc.2012.00181.
Glasgow G. The American Brachytherapy Society 99. Mesko S, Sandler K, Cohen J, Konecny G, Steinberg
recommendations for high-dose-rate brachytherapy M, Kamrava M. Clinical outcomes for stereotac-
for carcinoma of the endometrium. Int J Radiat tic ablative radiotherapy in oligometastatic and
Oncol Biol Phys. 2000;48(3):779–90. https://doi. oligoprogressive gynecological malignancies. Int
org/10.1016/s0360-3016(00)00689-1. J Gynecol Cancer. 2017;27:403–8. https://doi.
89. van der Steen-Banasik E, Christiaens M, Shash E, org/10.1097/IGC.0000000000000869.
et al. Systemic review: radiation therapy alone in 100. Lazzari R, Ronchi S, Gandini S, et al. Stereotactic
medical non-operable endometrial carcinoma. Eur body radiation therapy for oligometastatic ovarian
J Cancer. 2016;65:172–81. https://doi.org/10.1016/j. cancer: a step toward a drug holiday. Int J Radiat
ejca.2016.07.005. Oncol Biol Phys. 2018;101:650–60. https://doi.
90. van der Steen-Banasik E. Primary brachytherapy org/10.1016/j.ijrobp.2018.03.058.
as a radical treatment for endometrial carcinoma. J 101. Macchia G, Lazzari R, Colombo N, et al. A large,
Contemp Brachytherapy. 2014;6(1):106–12. https:// multicenter, retrospective study on efficacy and
doi.org/10.5114/jcb.2014.42028. safety of stereotactic body radiotherapy (SBRT) in
91. Weitmann HD, Pötter R, Waldhäusl C, Nechvile oligometastatic ovarian cancer (MITO RT1 study):
E, Kirisits C, Knocke TH. Pilot study in the treat- a collaboration of MITO, AIRO GYN, and MaNGO
ment of endometrial carcinoma with 3D image- groups. Oncologist. 2020;25(2):e311–20. https://
based high-dose-rate brachytherapy using modified doi.org/10.1634/theoncologist.2019-0309.
Heyman packing: clinical experience and dose- 102. Chiva LM, Lapuente F, González-Cortijo L, et al.
volume histogram analysis. Int J Radiat Oncol Biol Surgical treatment of recurrent cervical cancer:
state of the art and new achievements. Gynecol a review. Technol Cancer Res Treat. 2018;17:1–8.
Oncol. 2008;110(3 Suppl 2):S60–6. https://doi. https://doi.org/10.1177/1533033818803597.
org/10.1016/j.ygyno.2008.05.024. 110. Loi M, Frelinghuysen M, Klass ND, et al.
103. Kubota H, Tsujino K, Sulaiman NS, et al. Locoregional control and survival after lymph
Comparison of salvage therapies for isolated para- node SBRT in oligometastatic disease. Clin
aortic lymph node recurrence in patients with Exp Metastasis. 2018;35:625–33. https://doi.
uterine cervical cancer after definitive treatment. org/10.1007/s10585-018-9922-x.
Radiat Oncol. 2019;14:236. https://doi.org/10.1186/ 111. Iftode C, DʼAgostino GR, Tozzi A, et al. Stereotactic
s13014-019-1442-6. body radiation therapy in oligometastatic ovar-
104. Jereczek-Fossa BA, Ronchi S, Orecchia R. Is ste- ian cancer: a promising therapeutic approach. Int
reotactic body radiotherapy (SBRT) in lymph node J Gynecol Cancer. 2018;28:1507–13. https://doi.
oligometastatic patients feasible and effective? Rep org/10.1097/IGC.0000000000001324.
Pract Oncol Radiother. 2015;20:472–83. https://doi. 112. Kowalchuk RO, Waters MR, Richardson KM, et al.
org/10.1016/j.rpor.2014.10.004. Stereotactic body radiation therapy in the treatment
105. Ning MS, Ahobila V, Jhingran A, et al. Outcomes of ovarian cancer. Radiat Oncol. 2020;15:1–10.
and patterns of relapse after definitive radiation https://doi.org/10.1186/s13014-020-01564-w.
therapy for oligometastatic cervical cancer. Gynecol 113. Trippa F, Draghini L, di Marzo A, et al. Long-term
Oncol. 2018;148:132–8. https://doi.org/10.1016/j. palliation of lymph node oligometastatic ovar-
ygyno.2017.10.017. ian carcinoma after repeated stereotactic body
106. Jadon R, Higgins E, Hanna L, Evans M, Coles B, radiotherapy. Case Rep. 2020;106:6. https://doi.
Staffurth J. A systematic review of dose-volume pre- org/10.1177/0300891620952852.
dictors and constraints for late bowel toxicity follow- 114. Heike L, Kashani R, Robinson C, et al. Phase I trial
ing pelvic radiotherapy. Radiat Oncol. 2019;14(1):57. of stereotactic MR-guided online adaptive radiation
https://doi.org/10.1186/s13014-019-1262-8. therapy (SMART) for the treatment of oligometa-
107. Onderdonk BE, Chmura SJ. The Yin and Yang of static or unresectable primary malignancies of the
cytoreductive SBRT in oligometastases and beyond. abdomen. Radiother Oncol. 2018;126(3):519–26.
Front Oncol. 2019;9:1–6. https://doi.org/10.3389/ https://doi.org/10.1016/j.radonc.2017.11.032.
fonc.2019.00706. 115. Macchia G, Jereczek-Fossa BA, et al. Efficacy and
108. Park HJ, Chang AR, Seo Y, et al. Stereotactic safety of stereotactic body radiotherapy (SBRT) in
body radiotherapy for recurrent or oligometastatic oligometastatic/persistent/recurrent ovarian cancer: a
uterine cervix cancer: a cooperative study of the prospective, multicenter phase II study (MITO-RT3/
Korean Radiation Oncology Group (KROG 14-11). RAD). Int J Gynecol Cancer. 2022;32(7):939–43.
Anticancer Res. 2015;35(9):5103–10. https://doi.org/10.1136/ijgc-2021-002709.
109. Matsushita H, Jingu K, Umezawa R, et al. Stereotactic
radiotherapy for oligometastases in lymph nodes—
MR-Guided Radiotherapy
in the Pediatric and Adolescent
19
Patient
Margaret M. Kozak, Sean Mahase,

Bryan Traughber, Mitchell Machtay,
and John Buatti
Introduction cers underscores the importance of tailoring

cancer therapy with the goal of avoiding adverse
Disease Overview effects that may not become apparent until much
later in life [3, 4]. In this chapter, we will focus
In 2022 in the United States, an estimated 10,470 primarily on pediatric solid tumors, which com-
children aged 0–14 years and 5480 adolescents prise approximately 38% of cancer diagnoses in
aged 15–19 years were diagnosed with cancer [1]. children aged 0–14 years and 30% of diagnoses in
The most common types of cancer in children are adolescents aged 15–19 years; they are also the
leukemias, followed by brain and other central predominant type of pediatric malignancy seen in
nervous system (CNS) tumors and lymphomas. radiation oncology clinics around the world [5].
Based on Surveillance, Epidemiology, and End These include but are not limited to pediatric
Results (SEER) data, the 5-year survival for all brain and CNS tumors, soft-tissue and bone sar-
children diagnosed with brain and other CNS can- comas, neuroblastoma, and Wilms tumor.
cers is about 75%, compared to about 86% for
children diagnosed with leukemia [2]. This rela- odality Approach Based
M
tively high cure rate when compared to adult can- on Tumor Type
The key tenet of highly conformal radiotherapy
in children is to treat the tumor to a therapeutic
M. M. Kozak (*) dose while sparing as much surrounding normal
Department of Radiation Oncology, University of tissue from radiation exposure as possible. A
Iowa Hospitals and Clinics, Iowa City, IA, USA number of critical components need to be consid-
Department of Radiation Oncology, UT Southwestern ered in order for an appropriate radiation plan to
Medical Center, Dallas, TX, USA be created and safely delivered. These include
patient-, target-, and technique-specific factors
S. Mahase · B. Traughber · M. Machtay that allow for highly conformal RT delivery:
Department of Radiation Oncology, Penn State
Cancer Institute, Hershey, PA, USA
e-mail: [email protected]; –– Patient immobilization, localization, and
[email protected]; motion are critical to the design of a sophisti-
[email protected] cated RT plan. Basic patient positioning,
J. Buatti including location of the arms and legs, patient
Department of Radiation Oncology, University of comfort especially for treatments that are
Iowa Hospitals and Clinics, Iowa City, IA, USA
expected to take >20 min, and an understanding
356 M. M. Kozak et al.
of the beam directions are all taken into consid- tial benefit of protons rests in their physical
eration at the time of radiation simulation. properties, namely, the Bragg peak, which
–– Target and organ-at-risk (OAR) motion are yields a steep dose falloff with the added ability
both influenced by respiration, setup error, to decrease integral dose and exit dose. However,
and/or involuntary movement during RT deliv- there persist certain uncertainties with respect to
ery. Obtaining a 4D-CT during the simulation proton dose deposition and biologic effects that
process is critical for targets in the lungs and also must be considered when choosing the
abdomen. The use of a face or head and shoul- most appropriate treatment modality for a pedi-
ders mask is important to immobilize targets atric or adolescent cancer patient.
located in the brain and head and neck regions. The aforementioned considerations do not
Instructing the patient about appropriate blad- account for any positional or physiological
der and rectal filling can have an impact on changes that occur after the pre-treatment images
doses to critical structure located in the pelvis. are obtained. Additionally, intrafraction changes
–– The development of onboard imaging with due to respiratory motion, peristalsis, and bladder
cone beam CT allows for daily assessment of filling are also not corrected for, which may alter
patient and target positioning and verification the intended dose distribution. Potential conse-
of appropriate dose distribution. quences include increased dose to OARs and
–– The ability to modulate and sculpt the radia- decreased dose to the target. Methods to mitigate
tion beam was enabled by the development of these issues include treating to a more conserva-
intensity-modulated radiotherapy (IMRT), tive dose or using larger planning volumes to
which uses static gantry positions to deliver account for these uncertainties. Another solution
radiation from multiple angles. VMAT is an entails placing fiducials in the target, requiring
enhanced version of IMRT in which the gan- the patient to undergo an invasive procedure.
try continually rotates around the patient, Finally, major changes in anatomy or tumor vol-
which allows for both faster RT delivery and ume during a treatment course may require paus-
increased conformality. ing treatment while a new simulation and
–– Certain treatment techniques such as stereo- treatment plan are performed, thus lengthening
tactic body radiotherapy (SBRT) allow for the total treatment package time.
highly conformal dose distributions but are The advent of MRI linear accelerator
limited by adjacent OAR dose constraints. For (MR-linac)-based RT has revolutionized radiation
example, a lung nodule near the mainstem delivery by offering the ability for daily plan
bronchus may need to be underdosed to meet adaptation based on changes in both target and
the bronchial dose constraint, while SBRT of OAR anatomy. MRI-guided radiation therapy
a liver metastasis is determined by whether (MRgRT) utilizes a linear accelerator with MRI
sufficient normal liver can be spared from capabilities, enabling real-time MRI-based detec-
therapeutic radiation. Currently, SBRT for tion of the treatment target and organs at risk.
pediatric malignancies is primarily used to MR-based imaging provides superior visualiza-
treat metastatic sites. tion of soft-tissue structures over CT-based imag-
–– Minimizing the planning target volume ing [6, 7]. In contrast to conventional
(PTV), defined as the margin required for radiation-delivery machines reliant on static
setup error, decreases the overall treatment imaging, MRgRT platforms possess intrafraction
volume. For this to be feasible, patient immo- motion monitoring capabilities to account for
bilization and setup accuracy are of utmost tumor changes, internal anatomy movement, and
importance. Setup accuracy can be increased changes in bowel gas or bladder fill, providing a
by daily image-guided RT (IGRT) and can truer representation of the dose distribution. This
also be improved by daily adaptation of the is utilized for gating, in which radiation is deliv-
treatment plan. ered in parallel with a specific parameter, such as
–– Finally, proton therapy is commonly used in the the target volume being fully encompassed within
treatment of pediatric malignancies. The poten- a predefined boundary. Additionally, substantial
19 MR-Guided Radiotherapy in the Pediatric and Adolescent Patient 357
changes in target or normal structures can be The Elekta Unity combines a wide-bore 1.5-
accounted for in real time, recontoured, and Tesla MR imaging system and a next-generation
replanned using same-day imaging (adaptive Elekta 7-MV linear accelerator [14]. Like the
planning) without treatment breaks. Gating and MRIdian, the Unity has the ability for online
adaptive planning are distinct abilities, with gat-
adaptive planning and real-time automated gat-
ing accounting for intrafraction motion while ing in the near future. It contains a Monaco TPS
adaptive planning adjusts for interfraction capable of two types of adaptation—adapt-to-
changes. This allows for both reduced PTV vol- position (ATP) and adapt-to-shape (ATS) [15].
umes and optimal OAR sparing. It also allows for The ATP workflow adapts the plan according to
daily assessment of patient response to treatment,changes in daily patient positioning as compared
with the eventual goal of deriving biomarkers to the reference plan. For this workflow, the con-
from quantitative MRI techniques to better under- tours remain unchanged. The ATS workflow is
stand tumor biology [8–10]. Finally, an inherent employed when significant changes in patient
benefit of MRgRT for pediatric patients is elimi- anatomy require recontouring by the treating
nating radiation exposure from pre-treatment KV physician, and a new plan is generated in real
X-rays or cone beam CT scans. MRgRT platforms time with the goal of meeting dose constraints
have only been in routine use since 2014. Data on that were set at the time of initial RT planning.
this emerging technology is primarily retrospec- Many components of MRgRT planning are
tive, focusing on feasibility, safety, and dosimetry,
similar to conventional radiation planning,
with the first phase three prospective clinical trials
including simulation and patient immobilization,
reported recently [11]. While the ideal candidatesimaging (and adjunct imaging fusions) for target
for MRgRT are still under investigation, it likelyand organ-at-risk contouring, and treatment plan-
provides the greatest benefits in regions subject to
ning. MRgRT patients undergo an MR simula-
physiologic motion, such as the thorax, abdomen, tion for treatment planning as well as a CT
and pelvis. Thus far, the clinical experience in simulation to obtain electron density data. The
MRgRT has been primarily in the adult popula- CT scan can be deformably registered to the daily
tion. However, in certain scenarios, pediatric MR scan for adaptive dose calculations. Target
patient may also significantly benefit from this and organ-at-risk delineation are performed on
approach. The decision of whether to offer a pedi-the MRI dataset for plan generation, which obvi-
atric patient photon versus protons, VMAT versus ates uncertainties involved with using fused
MRgRT, is nuanced and integrates a variety of image sets which may not have been obtained in
factors. In this chapter, we will review some com-the treatment position [16]. However, adjunct
mon pediatric diagnoses, and propose ways in diagnostic scans can be fused to the reference
which MRgRT can be used to optimize radiation MRI planning scan to aid in target delineation.
delivery in this clinical setting. Certain structures with a predefined expansion
(e.g., using a 3-mm CTV expansion to create a
MR-Guided Linear Accelerator PTV) or Boolean operations (e.g., removing
Treatment Workflow overlap between two contoured structures) can be
The MRIdian linear accelerator (ViewRay set as “rules” automatically applied during adap-
Technologies, Inc.), approved in 2017, consists tive treatments to reduce table time. Treatment
of a 0.35-T MRI, a 70-cm bore diameter, and planning is via TPS software with Monte Carlo-
6-MV photons [12]. The MRIdian possesses based dose calculations accounting for the mag-
adaptive planning and real-time automated gat- netic field influence on dose distribution.
ing functions. A true fast imaging with steady- Treatment plans are limited to step-and-shoot
state free precession (TRUFI) sequence is IMRT with coplanar beam arrangements. In spite
obtained for both volumetric clinical imaging of these limitations, treatment plans are compa-
and 2D cine MRI for gating [13]. The MRIdian is rable to conventional linac plans [17–19].
currently pending FDA clearance to treat pediat- Additionally, the initial MRgRT plan represents a
ric malignancies. reference point that can be utilized for the entire
treatment course, or further improved through of pediatric CNS tumors arise in the supratento-
adaptive planning on interfraction changes. rial brain, the most common of which are low-
Each treatment entails screening the patient for grade astrocytomas. Other types of primary
MR compatibility. The coils must be placed ide- brain tumors include high-grade gliomas,
ally centered along imaging isocenter while cov- embryonal CNS tumors, germ cell tumors, and
ering the target area. It is important to ensure the craniopharyngiomas. Craniopharyngiomas have
coils aren’t in direct contact with the skin. Once the lowest 5-year overall survival rates among
the patient is appropriately positioned, image nonmalignant CNS tumors [23]. These tumors
acquisition and registration are performed. Upon require special attention as they frequently have
evaluating the original target volumes registered cystic components that can change in size and
to the new MRI, the patient can be treated using shape over the course of radiation delivery.
the original plan. The MRIdian also allows for Typically, repeat MRI every 1–2 weeks is indi-
couch shifts [16]. A predicted plan can also be cated for these patients with cyst aspiration
created, entailing recalculating the original plan either through indwelling Ommaya reservoirs or
with contours deformed with respect to same-day via a stereotactic approach for symptomatic
anatomy [13, 20]. After reviewing the treatment patients. The use of daily MRgRT for this patient
plan, the provider may elect to treat with the origi- population can allow for significantly improved
nal plan or create an adapted plan. The MRIdian visualization of the target, daily adaptation of the
offered two adaptive planning options. The first treatment volume as needed, and a reduction in
entails reoptimizing segment weighting while PTV volume. Typical PTV margins for cranio-
retaining their respective silhouettes, which can pharyngiomas range from 0.5 to 1 cm to account
be performed relatively quickly [21]. The alterna- for changes in tumor size over the course of
tive is complete reoptimization which includes treatment. In small children, the use of a 2- to
changing priorities for treatment targets and nor- 3-mm PTV would significantly decrease the
mal tissues in the planning algorithm. Given the dose to adjacent critical structures such as the
time constraints involved in replanning while the pituitary gland, optic chiasm, globes, and sur-
patient is in the treatment position, it is critical to rounding normal brain. This is of utmost impor-
perform adequate quality assurance, including tance when considering that the long-term
comparing recontoured structures to those created control of craniopharyngiomas following subto-
for the initial plan, and verifying electron density. tal resection and radiotherapy is excellent and
During treatment, the 2D cine MR acquisition long-term sequelae of treatment can be severe. A
functions in a single sagittal plane with images up study from St. Jude Children’s Hospital using
to 5–20 mm thick continuously obtained at 4–8 smaller PTV margins of 3–5 mm did not show
frames per second, enabling target volume posi- any decrement in tumor control [24].
tion verification [13]. The beam turns off when For other tumors in the supratentorial brain that
the target moves outside the defined boundary. A do not change in size and shape as significantly as
3-mm gating margin is commonly utilized for a craniopharyngioma, MRgRT may be useful if
many tumors treated on the MRIdian [22]. there are significant changes in the size of the
resection cavity or FLAIR signal s urrounding the
ationale for MRgRT Based on Tumor
R tumor. However, the primary research focus in this
Type and Location area in adult patients with glioblastoma has been
quantitative imaging to discover functional
Supratentorial Brain Tumors response-based adaptation during radiation treat-
Primary brain and other central nervous system ment. This has not yet been done in the pediatric
(CNS) tumors are the most common solid tumors population. At this time, we do not routinely treat
of childhood. There are a wide range of diagno- pediatric supratentorial brain tumors using
ses in this group, with heterogeneous tumor MRgRT; however, there may be select scenarios in
types and variation in tumor location. Over half which this would be beneficial.
Tumors of the Posterior Fossa sis. Though the posterior fossa is a rigid struc-
Tumors of the posterior fossa also form a hetero- ture, patients with residual disease may benefit
geneous group, the most common of which are from improved alignment using MRgRT,
low-grade astrocytic tumors, medulloblastoma, smaller treatment volumes, and adaptive replan-
and ependymoma. The primary treatment ning to help keep hot spots away from the brain-
approach for these tumors is maximal safe surgi- stem and minimize potential for radiation
cal resection. In patients with low-grade astro- necrosis [27–29]. However, a large multi-
cytic tumors, observation may be an appropriate institutional study of pediatric patients receiv-
next step especially in the setting of a gross total ing photon irradiation for posterior fossa tumors
resection. Medulloblastoma requires craniospi- demonstrated extremely low rates of brainstem
nal irradiation (CSI), which is not easily done toxicity with conventionally fractionated RT
using MRgRT due to multiple factors including [27]. In this series of 107 patients, the median
large field sizes and long daily treatment delivery. prescribed dose was 55.8 Gy with maximum
Many of these patients will benefit from proton doses to the brainstem ranging from 50.4 to
CSI, especially if they are of young age. The 93.7 Gy. Only 2 patients developed late brain-
tumor bed or posterior fossa boost in patients stem toxicity, and those patients received Dmax
with medulloblastoma is also very well defined, to the brainstem of 58.0 Gy and 57.6 Gy.
with CTV expansions of 1.0–1.5 cm. Providing Figure 19.1 depicts a case of a brainstem gli-
the patient does not require a posterior fossa oma with a posterior cystic component that fluc-
boost, the cochlea can be relatively well spared tuated in size during RT. The patient experienced
from high radiation doses with modern RT plan- worsening of bulbar symptoms at the time of cyst
ning techniques. Therefore, we do not recom- enlargement and therefore was transitioned from
mend that medulloblastoma patients be treated a regular Versa linear accelerator to the Unity
with MRgRT for reasons of target delineation or MR-linac for the remainder of his treatment.
OAR avoidance. Weekly adaptive replanning was performed on an
Intracranial ependymomas comprise approx- as-needed basis. The patient also received Avastin
imately 5–8% of brain tumors in children, most infusions to help with tumor swelling as well.
of which present in the posterior fossa, but can
be supratentorial as well. The most important Spinal Cord Tumors
determinant of outcome in ependymoma is Spinal tumors comprise a heterogeneous group,
extent of surgical resection. Complicating this is with intramedullary tumors frequently arising
the fact that these tumors arise from the floor of from the spinal cord, intradural/extramedullary
the fourth ventricle and can extend along cranial tumors arising from the dura, and extradural
nerves and into the foramina of Luschka and tumors originating from outside central nervous
Magendie making complete resection difficult system. Intramedullary tumors commonly
and at times leaving patients with permanent include ependymomas and low-grade astrocyto-
cranial nerve deficits. These tumors do not usu- mas. Meningiomas and nerve sheath tumors are
ally invade the brainstem, but due to their loca- the most prevalent intradural/extramedullary
tion, residual disease can be left at the brainstem tumors. Metastases account for the overwhelm-
interface. The most recent Children’s Oncology ing majority of extradural spinal tumors. Given
Group trial ACNS0121 included adjuvant RT the rarity of many of these tumors, multidisci-
for all patients with posterior fossa ependy- plinary management is tantamount to determin-
moma as well as all patients who underwent ing the optimal treatment paradigm which chiefly
subtotal resection [25]. The CTV is comprised entails resection.
of the tumor bed, any residual disease, and a 5- Malignant primary spinal tumors and spinal
to 10-mm margin [26]. The PTV is typically metastases can significantly impact quality of life
3–5 mm. CSI is not needed unless the patient due to pain, pathological fracture, and spinal cord
presents with neuraxis dissemination at diagno- compression [30, 31]. Postoperative radiotherapy
Fig. 19.1 MR-linac images of a patient with a brainstem ing treatment. A diagnostic MRI demonstrated interval
glioma with cystic component that fluctuated in size over enlargement of the cystic component of the tumor. The
the course of the patient’s treatment, causing intermit- patient was transitioned to the MR-linac for adaptive
tently worsening symptoms. The patient was initially replanning as needed. He eventually required Avastin
planned for standard linear accelerator treatment, however infusion due to worsening cyst enlargement and severe,
complained of worsening symptoms 1 week after initiat- unrelenting symptoms
is considered standard of care following resection coverage [37–39]. MRgRT platforms enable bet-
of malignant spinal tumors to minimize local ter soft-tissue delineation for planning, and real-
recurrence [32, 33]. Definitive radiotherapy is time visualization of the spinal cord to reduce
also considered for unresectable patients. In par- setup uncertainties and improve dose coverage
ticular, the frequency and complexity of spinal to the PTV [40]. Prospective studies in adults are
metastasis radiotherapy has increased in parallel under development, with initial experiences dem-
with systemic therapy advances conferring lon- onstrating feasibility and safety of this approach
ger overall survival to metastatic cancer patients. [41]. These studies will be imperative to deter-
Historically, low-dose palliative radiotherapy mining whether these putative benefits of MRgRT
was the sole option for symptom management, translate clinically in adults before applying this
but with disappointing local control rates [31, 34, technology in pediatric populations.
35]. Advances in image guidance, technology,
and planning software promoted use of SBRT for Head and Neck Cancers
spinal metastases to safely deliver a higher bio- Head and neck cancers comprise about 12% of
logically equivalent dose. While prospective tri- all childhood malignancies. The most common
als utilizing SBRT for spinal tumors are ongoing, histologies include lymphomas (~40%), rhabdo-
it’s utilization is increasing [36]. myosarcoma (~20%), and nasopharyngeal carci-
When treating a patient with definitive or post- nomas (~15%) [42–44]. The benefit of MRgRT
operative radiotherapy, radiation dose is chiefly in patients with head and neck malignancies has
limited by spinal cord or thecal sac constraints. primarily been studied in the adult population.
Visualization of the spinal cord and treatment vol- Due to the significantly improved soft-tissue con-
ume are essential to successful and safe radia- trast, MRgRT allows for more precise tumor
tion delivery. This often entails obtaining and delineation and margin reduction [45] and affords
fusing MRI or myelogram sequences to CT-based the added benefit of assessing tumor response
planning scans. Limitations to radiotherapy utiliz- over the course of treatment. Another problem
ing conventional linac-based platforms include that is often encountered in the head and neck
uncertainties regarding spinal cord location neces- adult population are alterations in patient and
sitating the use of PRVs, as well as setup uncer- tumor anatomy due to weight loss [46] caused by
tainties, which can decrease effective tumor oral toxicity secondary to treatment. Weekly
Fig. 19.2 Example of a palliative treatment plan for a treatment began to decrease in size. The purple line shows
neuroblastoma mandibular metastasis. The patient was the initial GTV, and, from left to right, the T2 images
initially planned for standard linear accelerator treatment, show decreasing size of the GTV (T2 bright mass) over
however complained of significant swelling of the mass a time. The treatment volume was adapated as needed with
few days into treatment. He was subsequently transferred the goals of decreasing acute oral toxicity during the
to the MR-linac and put under anesthesia to finish out his course of radiotherapy
course. The tumor initially enlarged, but after 5 days of
offline adaptation may be sufficient for address- ment and OAR sparing with the use of smaller
ing these issues; however, preliminary studies to PTV margins; however, this should be evaluated
date report mixed results [47–49]. Additionally, on a case-by-case basis.
some tumors may fluctuate in size more quickly
than others necessitating daily adaptive replan- Pediatric Leukemias
ning. This is not feasible with conventional treat- Acute lymphoblastic leukemia (ALL) is the most
ment approaches as re-simulating and replanning common malignancy in children accounting for
require routine imaging that places a strain on ~80% of pediatric leukemia cases. The next most
busy clinic schedules [47]. Especially for young common is acute myeloblastic leukemia (AML)
children who may require anesthesia for imaging comprising ~20% of cases. There are about
purposes, daily online adaptation can signifi- 2500–3000 children diagnosed with ALL each
cantly streamline the process (Fig. 19.2). year, with a peak incidence between ages 2 and 5.
Further improvements in outcomes using Treatment of ALL is heavily based on intensive
functional MRI technologies are being explored, chemotherapy regimens, as well as intrathecal
including the use of quantitative MRI biomarkers chemotherapy and potentially cranial irradiation
such as diffusion-weighted imaging (DWI), depending on the risk grouping of the patient
dynamic contrast-enhanced (DCE) MRI, MR [58–60]. In recent years, cranial irradiation has
relaxometry, and radiomics [50–54]. Adaptation fallen out of favor and been replaced by more
to shrinking tumor volumes continues to be an intensive chemotherapeutic regimens [61, 62].
active area of investigation in adult head and neck Similarly, AML is primarily treated with chemo-
malignancies with long-term outcomes results therapy with little role for RT unless certain clini-
not yet available [55–57]. Given the uncertainty cal scenarios arise.
in microscopic tumor extension, indistinct tumor One of such clinical scenarios includes the
borders, and tumor infiltration into adjacent tis- formation of solid tumors known as myeloid sar-
sues, the treating physician can consider the use comas, previously called chloromas; they are
of a CTV margin to account for microscopic dis- infrequent but have been seen to arise in patients
ease depending on the clinical scenario. There with relapsed or refractory AML [63, 64]. The
may be a significant advantage to tumor align- most commonly involved sites include the bone,
a b
Fig. 19.3 MR-guided radiation plan for a 6-month-old as much as possible. The GTV was defined on T1 post-
infant with relapsed refractory AML and a painful orbital contrast imaging. The PTV was a 2-mm expansion on the
chloroma. The patient was planned for MR-guided treat- GTV. Panel a shows the axial slice and panel b shows the
ment due to superior daily alignment and ability to use a coronal slice. Green = GTV; red = 6 Gy; yellow = 5.4 Gy;
small PTV expansion to limit dose to surrounding tissues teal = 4.8 Gy; blue = 4.2 Gy
skin, orbit, lymph nodes, GI tract, and CNS. In Hodgkin’s lymphoma (NHL) comprises about
children, the skin and orbit are most commonly 60% of lymphoma cases in children under the age
described [65]. Prior data has not shown a benefit of 15 years. In adolescents aged 15–18 years,
to the addition of RT for the treatment of chloro- Hodgkin’s lymphoma predominates. The three
mas; however, radiation is commonly used to pal- main types of NHL in children include Burkitt’s
liate chloroma-related vision loss or cord lymphoma, lymphoblastic lymphoma, and large
compression [66, 67]. Due to their inherent radio- cell lymphoma [77]. Treatment outcomes for NHL
sensitivity and a potentially rapid local response, are superior to those seen in adults. The role of
the use of adaptive treatment to treat myeloid sar- radiation is largely limited due to advances in che-
comas in children is particularly appealing motherapy regimens but can be considered on a
(Fig. 19.3). The proposed dose is 20–24 Gy in case-by-case basis in patients with relapsed or
10–12 fractions, which has been shown to result refractory disease [78–80]. At our institution, we
in excellent local control and palliation of symp- treated a case of relapsed refractory DLBCL of the
toms [68] and is also recommended by the jejunum using MR-guided radiotherapy. Daily
ILROG guidelines [69]. However, for palliation MRI allowed for improved visualization and align-
of symptoms, doses as low as 6 Gy have been ment to the tumor, as well as assessment of the
used [70]. These lesions can be well visualized tumor’s relationship to nearby critical structures,
using FDG-PET imaging and can also be seen such as the stomach. The patient required adaptive
well on MRI [71–76]. The use of daily MRgRT replanning over the course of treatment due to fluc-
may be especially useful for myeloid sarcomas of tuations in bowel and stomach position in order to
the orbit, where the use of small target volumes meet maximum dose constraints.
and adaptive replanning may help minimize Pediatric Hodgkin’s lymphoma accounts for
treatment toxicities. Especially in very young 5–6% of childhood cancers and is more common
children with poor prognosis, the use of lower in adolescents rather than young children.
doses for palliation should be considered. Patients most commonly present with painless
cervical adenopathy or large mediastinal masses
Pediatric Lymphoma in up to 66% of patients. Treatment paradigms
Lymphomas in children are the third most common have evolved from risk-adapted therapy to
type of cancer after ALL and CNS tumors. Non- response-adapted therapy based on PET scan
evaluations over the course of chemotherapy [81, tumor molecular findings and/or their response to
82]. Radiation volumes have also evolved from 6 weeks of chemotherapy [85–87]. Further defi-
involved field to involved site and now to involved nitions of staging for Wilms tumor and what con-
nodal RT. Radiation was historically adminis- stitutes anaplasia are beyond the scope of this
tered to all sites of initial disease involvement book.
and now is selectively administered to sites of With respect to radiation volumes, the stan-
bulky disease and sites of either slow early dard of care in the United States for flank and
response (SER) or partial response (PR). In the whole abdomen RT remains parallel-opposed
SWOG 1826 trial, only sites of residual disease (AP-PA) treatment fields using 4-MV or 6-MV
after completion of systemic therapy are irradi- photons [88]. In Europe, there has been increas-
ated (NCT03907488). To date, no reports of ing interest in the use of more streamlined radia-
MRgRT for Hodgkin’s lymphoma exist. tion fields for postoperative flank treatment. Mul
However, a survey of pediatric oncology practi- et al. [89] defined conformal postoperative RT
tioners (n = 12) indicated an expected benefit to volumes using VMAT. Using 20 consecutive
MR-guided treatment for patients with mediasti- cases from their institution, they evaluated dose
nal lymphomas [83]. Seventy-five percent of to OARs and total body volume (TBV) for both
respondents indicated a benefit of PTV margin conventional AP-PA and VMAT plans. Dose con-
reduction, and 67% indicated a benefit of daily straints included tail of the pancreas to <10 Gy,
adaptation to a shifting treatment volume, the use heart D50 <5 Gy, and mammary bud mean
of online replanning for OAR avoidance, and <10 Gy. Their results showed an estimated dose
compensation for intrafraction motion, and 50% reduction to OARs by 60%. VMAT increased the
felt there might be a benefit to the use of func- TBV receiving 10% prescription dose but
tional imaging. Furthermore, a large portion of decreased the irradiated TBV for all other doses.
respondents felt there would be a benefit to the If this technique were to be more widely adopted,
use of MR guidance to the treatment of lymph utilization of MR-guided technologies would
nodes located in the mediastinum, upper abdo- likely prove effective in further reducing doses to
men including peri-hilar and splenic lymph adjacent OARS such as the heart and stomach.
nodes, and those located in the pelvis. One unique clinical scenario in which we have
found a benefit to MR guidance for Wilms tumor
Abdominal Tumors is that of an unresectable inferior vena cava (IVC)
tumor thrombus [90]. IVC tumor thrombus is
Wilms Tumor found in 4–10% of Wilms tumor cases. Removal
Wilms tumor is the most common kidney tumor of the tumor is required at the time of surgery—in
in children and accounts for ~6% of childhood a large SIOP/GPOH study, thrombus was
malignancies. There are approximately 500–600 completely removed in 85%, incompletely in
new cases per year in the United States [84]. 12%, and not removed in 2% [91]. For patients
More than two-thirds of cases arise in children with incompletely removed thrombus, the recom-
under the age of 5 years, and 10–13% of cases are mendation is to deliver a boost dose of radiation
associated with genetic predisposition syndromes targeting residual disease to a total of 21.6 Gy.
including WAGR, Denys-Drash, and Beckwith- The thrombus region is included in the initial
Wiedemann. Patients are typically treated with flank treatment field, and then an additional
upfront surgical resection followed by chemo- 10.6 Gy in 1.8 Gy/fraction is delivered to a vol-
therapy with or without adjuvant RT. In general, ume encompassing the tumor thrombus. We have
indications for flank or whole abdomen RT are as been treating the tumor thrombus using MRgRT
follows: stage III favorable histology, stages I–III as the IVC is well visualized on T2 MRI and this
focal anaplasia, and stages I–III diffuse anapla- affords superior alignment to the region and
sia. Patients with lung metastases may or may not allows for the use of significantly smaller treat-
receive whole lung irradiation based on their ment volumes (Fig. 19.4). Our standard protocol
is to obtain a 4D CT simulation scan with con- of 5 and unable to lay still on their own for the
trast as well as an MRI with and without contrast duration of treatment.
for RT planning. The IVC and residual thrombus
are identified, and an internal target volume Neuroblastoma
(ITV) is created, if necessary. We then apply a Neuroblastoma is the most common malignancy
3-mm PTV. We have found that with this in infants under the age of 12 months, and 90%
approach, mean doses to OARs including the of cases are diagnosed prior to the age of 5 [94].
heart, bowel, liver, and stomach are all decreased Neuroblastoma cells derive from embryonic neu-
when compared to conventional linear accelera- ral crest cells of the peripheral nervous system,
tor treatment [90]. If the primary tumor was on and tumors are comprised of several different
the contralateral side, we also aggressively con- histologic subtypes including ganglioneuroma,
strain the remaining kidney to a composite mean ganglioneuroblastoma, and neuroblastoma.
of <10–11 Gy. We ensure that the dose gradient Neuroblastomas can arise anywhere along the
across the adjacent vertebral bodies is less than peripheral nervous system with most located in the
5 Gy to avoid future risk of scoliosis [92]. adrenal gland (35%), abdomen (30%), and poste-
Finally, we would be remiss without mention- rior mediastinum (20%) [95]. Treatment involves
ing the potential for whole abdomen RT using surgical resection and chemotherapy, and patients
IMRT or VMAT, which has been shown to better with high-risk neuroblastoma will also receive
spare OARs and provide superior PTV coverage RT. Other considerations for RT include Pepper
to standard AP-PA treatment fields [93]. The syndrome, in which infants experience respiratory
potential benefit of adaptive therapy in this set- failure due to liver enlargement from tumor.
ting would include superior visualization and Radiation treatment planning should be done
avoidance of the remaining kidney, as well as in accordance with standard guidelines. The use
adaptive replanning for patients whose abdomi- of 4D CT or 4D MRI will aid in the delineation of
nal circumference changes over the course of treatment volumes near the diaphragm. Target
treatment. Some potential limitations of this volumes in neuroblastoma are well defined, and
approach using MRgRT would be the larger field we instruct our US readers to review the most
size and longer time under anesthesia as most contemporary neuroblastoma protocols for fur-
patients with Wilms tumor will be under the age ther details. The GTV should encompass the
Fig. 19.4 MR-linac radiation plan for a child with Wilms dose to thrombus was 21.6 Gy) using MR-guided
tumor and unresectable IVC thrombus remaining after IMRT. The main objectives were to avoid a gradient
neoadjuvant chemotherapy. The patient received conven- across the vertebral body of >5 Gy, limit the dose to the
tional flank radiation using AP-PA fields on a standard remaining adjacent kidney to a mean of <3 Gy, and keep
linear accelerator to a dose of 10.8 Gy. The tumor throm- coverage of the PTV at >95%
bus was subsequently boosted to another 10.8 Gy (total
region of the primary tumor and any involved be surgically addressed, and others will require
lymph nodes at diagnosis but should not include radiotherapy [97–99].
fixed OARs including the kidney, bones, liver, Genitourinary rhabdomyosarcoma of the
etc. unless the tumor invaded into these regions. bladder and prostate is typically treated with che-
The CTV is an anatomically confined 1.0-cm motherapy and radiation to enable organ preser-
expansion, and its superior and inferior extent vation. Certain cases are amenable to
should encompass the initial extent of the presur- brachytherapy, which provides optimal sparing
gical tumor volume. Adjacent vertebral bodies of adjacent OARs such as the rectum and can sig-
are traditionally treated to 18 Gy to purposefully nificantly minimize late rectal toxicity and
cause symmetric growth stunting and prevent decrease rates of secondary malignancy [100].
scoliosis. The PTV is 0.3- to 0.8-cm expansion to However, for some patients, brachytherapy is
account for physiologic motion. The most recent either not accessible or feasible due to tumor
ANBL1531 protocol allows for the use of a 0.3- location, and external beam radiation is
cm PTV for most tumor sites provided the patient recommended.
undergoes daily cone beam CT. For bladder rhabdomyosarcoma treated with
Typical beam arrangements are AP-PA; how- external beam radiation, tumor targeting can
ever, over the past two decades, other approaches prove complex due to daily changes in bladder
including 3D-CRT, IMRT, and protons have been size and shape. In addition, the rectum frequently
investigated. Paulino et al. compared the use of changes in its filling, and this is more difficult to
traditional AP-PA plans to IMRT for the treat- control in small children than in adults. In these
ment of abdominal neuroblastoma and evaluated respects, daily adaptive MRgRT can be espe-
doses to adjacent OARs [96]. They found that cially useful in ensuring adequate target coverage
dose to kidneys was decreased with the use of while minimizing doses to adjacent critical struc-
IMRT at the expense of increased doses to the tures such as the rectum (Fig. 19.5). The bladder
liver, stomach, and spleen. IMRT was not found and rectum can be recontoured daily, and the
to be beneficial for well-lateralized tumors. The PTV margin can be kept to 3 mm to account for
authors concluded that treatment recommenda- any potential changes during the adaptation pro-
tions should be made on a case-by-case basis. cess. Even with the use of protons, margins of
At our institution, we have selectively 5–10 mm for aperture opening and additional
employed MR-guided radiotherapy for patients range to account for setup uncertainties and tis-
with neuroblastoma tumors located in the abdo- sue heterogeneity are required. A study by Cotter
men and for neuroblastoma metastases. We have et al. showed that protons were superior to IMRT
found that based on changes in kidney position, in sparing dose to bladder, testes, femoral heads,
daily plan adaptation is beneficial for abdominal and pelvic bones [101]. However, there was no
tumors near that structure. In addition, we have difference with respect to dose to the rectum,
found that neuroblastoma metastases can change prostate, penile bulb, or bowel. To date, no com-
rapidly in size over the course of treatment— parative study of the use of MRgRT and either
from tumor swelling to shrinkage—making daily proton radiation or IMRT has been done. We
adaptive therapy an excellent option in select have successfully treated bladder/prostate rhab-
clinical scenarios. domyosarcoma using MRgRT and felt that this
approach successfully minimized dose to adja-
Pelvic Tumors in Children cent critical structures through the ability to use
and Adolescents daily adapt-to-shape treatment planning.
Pelvic tumors in children are a rare, heteroge- Other instances of pelvic disease in the pediat-
neous group that includes urogenital tract rhab- ric and adolescent population include Ewing sar-
domyosarcoma (bladder, prostate, vagina), coma. The most common pelvic subsite for
neuroblastoma, Ewing sarcoma, and germ cell Ewing sarcoma is the sacrum (44%), followed by
tumors, among others. Some of these tumors can iliac/buttock region (40%) and the ischiopubic-
Fig. 19.5 Panel on the left shows the original treatment required daily adapt-to-shape (ATS) to account for
planning CT scan for a 3-year-old patient with bladder changes in bladder filling. The panels show the wide range
rhabdomyosarcoma (light orange = CTV). Due to the of changes that occurred over the course of the patient’s
patient’s age and need for anesthesia, it was not possible treatment. MR guidance allowed us to employ a small
to give bladder filling instructions. Because we expected PTV margin of 3 mm and therefore to significantly limit
bladder filling to vary significantly over the course of the dose to surrounding normal structures. The rectum
radiotherapy, we elected to treat the tumor using MR (not shown) was also recontoured as needed
guidance for adaptive planning as needed. The patient
acetabular region (16%). Historically, pelvic ment volume incorporated a 1-cm constrained
Ewing sarcoma has been associated with inferior margin around the gross tumor. The estimated
outcomes due to part to adjacent pelvic anatomy cumulative incidence of local failure was
such as bowel, nerves, etc. that limit the ability to 4.4 ± 3.1%. The most common acute toxicity was
achieve high radiation doses to the region [102]. radiation dermatitis. Late grade 3 toxicities were
However, the most recent data from AEWS1031 related to bone, joint, skin, or subcutaneous tis-
challenge that paradigm and show that RT is an sue injury in the region of the high-dose radiation
effective modality for local control. The cumula- field. A more recent analysis from the University
tive incidence of local failure by tumor site was of Florida comprised of large tumors with a
8.6% for pelvis, 12.5% for extremities, 6.7% for median size and volume of 10.5 cm and 472 cm3
spine/chest wall/skull, and 3.6% for extraosseous showed no local progression in five patients who
tumors (p = 0.61) [103]. That being said, local received doses of ≥59.4 Gy RBE [106]. Only two
failure was higher for tumors ≥200 mL (14.3%) patients treated with proton radiation alone devel-
versus those <200 mL (5.9%, p = 0.04), and oped local recurrence, and both those patients
Ewing sarcoma of the pelvis tends to present with received doses to tumor of <59.4 Gy. The overall
larger tumors overall compared to other subsites. local control was 91% for patients receiving radi-
Over the past several years, radiation oncolo- ation alone. There were five late grade 2+ toxici-
gists have pursued dose escalation in an effort to ties including premature ovarian failure after
improve local control outcomes for patients with 12-Gy RBE to the ovary on the affected side, hip
Ewing sarcoma. A retrospective study of 48 dislocation after treatment for a left ischium/
patients with pelvic Ewing sarcoma from Mayo pubic primary, bladder fibrosis in the setting of
Clinic demonstrated improved local control with bladder Dmax of 54-Gy RBE, grade 3 wound
definitive radiation doses of ≥56 Gy [104]. A infection after preoperative RT, grade 2 pelvic
phase II study from St. Jude documented zero pain of unclear etiology, and grade 2 limb-length
local failures with dose escalation to 64.8 Gy for discrepancy.
tumors >8 cm in size [105]. In addition, that For pelvic primaries, MR-guided treatment
study employed limited margin RT using confor- could be beneficial in decreasing treatment mar-
mal or intensity-modulated techniques. The treat- gins, allowing for accurate patient positioning,
and adapting to the changing shape of nearby tor patient body temperature during treatment.
OARs such as the bowel, bladder, and rectum. It Diapers should be checked prior to treatment to
could potentially allow for the daily identifica- minimize the risk of thermal injury secondary to
tion of ovarian position, depending on patient moisture. Hearing protection should be utilized
age. In general, we recommend approaching every treatment to minimize auditory toxicity.
these complex clinical scenarios on a case-by- Pediatric and young adults not requiring sedation
case basis. may not tolerate MRgRT due to claustrophobia.
They must also remain still for longer periods
compared to conventional linear accelerator
atient Selection and Technical
P treatments as positional intrafraction changes
Challenges will undermine adaptive planning. These consid-
erations may be addressed with a combination of
Potential pediatric MRgRT candidates are sub- anxiolytics, patient education, parental presence,
ject to many limitations as adults. They must pass and child life services [116]. Anesthesiology
the same screening criteria [107], entailing metal- should be informed of pending cases requiring
lic implants that may be ferromagnetic or cause sedation to ensure they have MRI zone
image distortion, and patient girth. The respec- 4-compatible equipment [117]. The MRgRT
tive MRgRT platform’s magnetic field strength’s suite must be equipped to monitor vitals during
impact on particle trajectory, inability to perform treatment. The distance from the vault to console
couch kicks, and utilizing of noncoplanar beams may require longer oxygen tubes. A dry run with
must be accounted for during treatment planning anesthesiology is recommended for the first seda-
[108–110]. MRgRT utilizing adaptive planning is tion case. This can determine the need for addi-
resource- and time-intensive: a radiation oncolo- tional time-outs throughout the adaptive process
gist, physicist, and dosimetrist (and possibly and points of no return. A volunteer adult patient
nursing and anesthesia for cases requiring seda- can wear patient-monitoring devices for treat-
tion) must be present for each treatment. Patient ment to test radiation interference with monitor-
setup, imaging, modification of volumes requiring equipment, as well as imaging artifacts.
ing familiarity with MRI-based anatomy, and Finally, the increased time under anesthesia due
recalculating a treatment plan may necessitate to adaptive planning must be weight against the
30–60 min per fraction. This also translates to benefits of MRgRT.
fewer patients treated daily compared to tradi-
tional linac-based plans averaging 15–30 min per
fraction [111–115]. These aforementioned fac- Future Directions
tors may prove to be financially limiting to a
practice, curbing widespread adoption of Improving the therapeutic ratio by reducing tar-
MRgRT, especially in community settings. get volumes, tracking intratumoral motion in real
Finally, prospective studies are necessitated to time, and daily adaption of OAR changes pro-
validate potential significant therapeutic benefits mote consideration of several concepts currently
conferred by dosimetric advantages as proven under investigation for adult malignancies.
improvements in toxicity and quality-of-life may Historically radioresistant malignancies, such as
ultimately prove to be cost-effective. This Ewing sarcoma and rhabdomyosarcoma, may
endeavor may require several years for trial benefit from dose escalation to the gross tumor.
development, implementation, accrual and data Similarly, inoperable gross disease in the pelvic
maturation. and abdominal regions can be safely dose-
There are several planning considerations escalated. The MR-linac adaptive planning fea-
unique to the pediatric population. The coil tures can be combined with functional imaging
weight must be safely tolerable; custom setups (e.g., FDG, MIBG) to selectively modify dose
may be necessitated. It is also important to moni- distribution based on therapeutic response.
Tumor regions demonstrating rapid regression c a n c e r . g o v / t y p e s / c h i l d h o o d -c a n c e r s /

child-adolescent-cancers-fact-sheet.
and/or decreased metabolic activity may be ame- 3. Fidler-Benaoudia MM, Oeffinger KC, Yasui Y,
nable to dose reduction to spare organs at risk Robison LL, Winter DL, Reulen RC, et al. A compar-
while resistant regions will benefit from dose ison of late mortality among survivors of childhood
escalation. Integrating artificial intelligence may cancer in the United States and United Kingdom. J
Natl Cancer Inst. 2021;113(5):562–71.
mitigate resource demands through facilitating 4. Phillips SM, Padgett LS, Leisenring WM, Stratton
adaptive planning and quality assurance [118]. KK, Bishop K, Krull KR, et al. Survivors of child-
Finally, development of a MR-guided particle hood cancer in the United States: prevalence and
therapy platform is under investigation [119]. burden of morbidity. Cancer Epidemiol Biomark
Prev. 2015;24(4):653–63.
Pediatric malignancies are heterogeneous and 5. Steliarova-Foucher E, Colombet M, Ries LAG,
relatively rare compared to typical adult malignan- Moreno F, Dolya A, Bray F, et al. International
cies. Prospective, multi-institutional clinical trials incidence of childhood cancer, 2001-10: a
are necessitated to sufficiently power studies population- based registry study. Lancet Oncol.
2017;18(6):719–31.
showing whether the putative dosimetric benefits 6. Fischer-Valuck BW, Henke L, Green O, Kashani R,
of MRgRT correlate with clinical outcomes. Acharya S, Bradley JD, et al. Two-and-a-half-year
Equally important is determining which tumor his- clinical experience with the world’s first magnetic
tologies or target sites derive benefits from MRgRT resonance image guided radiation therapy system.
Adv Radiat Oncol. 2017;2(3):485–93.
versus proton therapy or conventional linac-based 7. Metcalfe P, Liney GP, Holloway L, Walker A,
treatment. This requires selection of appropriate Barton M, Delaney GP, et al. The potential for
clinical endpoints for evaluation. Radiation oncol- an enhanced role for MRI in radiation-therapy
ogists must be adept in MRI-based imaging, as treatment planning. Technol Cancer Res Treat.
2013;12(5):429–46.
well as the benefits and limitations of MRgRT to 8. Dunlop A, Mitchell A, Tree A, Barnes H, Bower
appropriately select ideal patients who will benefit L, Chick J, et al. Daily adaptive radiotherapy for
from this emerging approach. patients with prostate cancer using a high field
MR-linac: initial clinical experiences and assess-
ment of delivered doses compared to a C-arm linac.
Clin Transl Radiat Oncol. 2020;23:35–42.
Conclusions 9. Henke L, Kashani R, Yang D, Zhao T, Green O,
Olsen L, et al. Simulated online adaptive magnetic
Photon- and proton-based radiation therapies are resonance-guided stereotactic body radiation ther-
apy for the treatment of oligometastatic disease of
planned and delivered through CT-based image the abdomen and central thorax: characterization of
guidance. MRgRT is a state-of-the-art technol- potential advantages. Int J Radiat Oncol Biol Phys.
ogy with the potential to revolutionize pediatric 2016;96(5):1078–86.
cancer treatment. Its real-time imaging and adap- 10. Winkel D, Bol GH, Werensteijn-Honingh AM,
Intven MPW, Eppinga WSC, Hes J, et al. Target
tive planning capabilities allow high precision coverage and dose criteria based evaluation of the
radiation delivery which may improve the thera- first clinical 1.5T MR-linac SBRT treatments of
peutic ratio compared to traditional radiotherapy lymph node oligometastases compared with con-
platforms. As technology continues to advance, it ventional CBCT-linac treatment. Radiother Oncol.
2020;146:118–25.
is expected that MRgRT will become an even 11. Kishan AU, Ma TM, Lamb JM, Casado M, Wilhalme
more crucial component of radiation therapy in H, Low DA, et al. Magnetic resonance imaging-
the treatment of pediatric cancers. guided vs computed tomography-guided stereo-
tactic body radiotherapy for prostate cancer: the
MIRAGE randomized clinical trial. JAMA Oncol.
2023;9(3):365–73.
References 12. Lagendijk JJ, Raaymakers BW, Raaijmakers AJ,
Overweg J, Brown KJ, Kerkhof EM, et al. MRI/linac
1. Siegel RL, Miller KD, Fuchs HE, Jemal A. Cancer integration. Radiother Oncol. 2008;86(1):25–9.
statistics, 2022. CA Cancer J Clin. 2022;72(1):7–33. 13. Klüter S. Technical design and concept of a
2. National Cancer Institute. Cancer in chil- 0.35 T MR-Linac. Clin Transl Radiat Oncol.
dren and adolescents. 2021. https://www. 2019;18:98–101.
14. Dunkerley DAP, et al. Clinical implementational and incompletely resected Ependymoma, and observa-
site-specific workflows for a 1.5 T MR-Linac. J Clin tion for completely resected, supratentorial ependy-
Med. 2022;11(6):1662. moma. J Clin Oncol. 2019;37(12):974–83.
15. Rammohan N, et al. History of technologi- 26. Peters S, Merta J, Schmidt L, Jazmati D, Kramer
cal advancements towards MR-Linac: the future PH, Blase C, et al. Evaluation of dose, volume, and
of image-guided radiotherapy. J Clin Med. outcome in children with localized, intracranial
2022;11(16):4730. ependymoma treated with proton therapy within
16. Christiansen RL, Dysager L, Bertelsen AS, Hansen the prospective KiProReg study. Neuro Oncol.
O, Brink C, Bernchou U. Accuracy of automatic 2022;24(7):1193–202.
deformable structure propagation for high-field 27. Devine CA, Liu KX, Ioakeim-Ioannidou M, Susko
MRI guided prostate radiotherapy. Radiat Oncol. M, Poussaint TY, Huisman T, et al. Brainstem
2020;15(1):32. injury in pediatric patients receiving posterior fossa
17. Yadav P, Musunuru HB, Witt JS, Bassetti M, photon radiation. Int J Radiat Oncol Biol Phys.
Bayouth J, Baschnagel AM. Dosimetric study for 2019;105(5):1034–42.
spine stereotactic body radiation therapy: mag- 28. Orukari I, Perkins S, Zhao T, Huang J, Caruthers DF,
netic resonance guided linear accelerator versus Duriseti S. Brainstem toxicity in pediatric patients
volumetric modulated arc therapy. Radiol Oncol. treated with protons using a single-vault synchrocy-
2019;53(3):362–8. clotron system. Int J Part Ther. 2022;9(1):12–7.
18. Chamberlain M, Krayenbuehl J, van Timmeren 29. Yock TI, Constine LS, Mahajan A. Protons, the
JE, Wilke L, Andratschke N, Garcia Schüler H, brainstem, and toxicity: ingredients for an emerging
et al. Head and neck radiotherapy on the MR dialectic. Acta Oncol. 2014;53(10):1279–82.
linac: a multicenter planning challenge amongst 30. Rades D, Veninga T, Stalpers LJ, Basic H, Rudat
MRIdian platform users. Strahlenther Onkol. V, Karstens JH, et al. Outcome after radiother-
2021;197(12):1093–103. apy alone for metastatic spinal cord compression
19. Musunuru HB, Yadav P, Olson SJ, Anderson in patients with oligometastases. J Clin Oncol.
BM. Improved ipsilateral breast and Chest Wall 2007;25(1):50–6.
sparing with MR-guided 3-fraction accelerated par- 31. Rades D, Schild SE, Abrahm JL. Treatment of
tial breast irradiation: a dosimetric study comparing painful bone metastases. Nat Rev Clin Oncol.
MR-Linac and CT-Linac plans. Adv Radiat Oncol. 2010;7(4):220–9.
2021;6(3):100654. 32. Pica A, Miller R, Villà S, Kadish SP, Anacak Y,
20. Acharya S, Fischer-Valuck BW, Kashani R, Parikh Abusaris H, et al. The results of surgery, with or
P, Yang D, Zhao T, et al. Online magnetic reso- without radiotherapy, for primary spinal myxopap-
nance image guided adaptive radiation therapy: first illary ependymoma: a retrospective study from the
clinical applications. Int J Radiat Oncol Biol Phys. rare cancer network. Int J Radiat Oncol Biol Phys.
2016;94(2):394–403. 2009;74(4):1114–20.
21. van Timmeren JE, Chamberlain M, Krayenbuehl J, 33. Kotecha R, Tom MC, Naik M, Angelov L, Benzel
Wilke L, Ehrbar S, Bogowicz M, et al. Comparison EC, Reddy CA, et al. Analyzing the role of adjuvant
of beam segment versus full plan re-optimization in or salvage radiotherapy for spinal myxopapillary
daily magnetic resonance imaging-guided online- ependymomas. J Neurosurg Spine. 2020:1–6.
adaptive radiotherapy. Phys Imaging Radiat Oncol. 34. Zeng L, Chow E, Bedard G, Zhang L, Fairchild
2021;17:43–6. A, Vassiliou V, et al. Quality of life after palliative
22. van Sörnsen de Koste JR, Palacios MA, Bruynzeel radiation therapy for patients with painful bone

AME, Slotman BJ, Senan S, Lagerwaard metastases: results of an international study validat-
FJ. MR-guided gated stereotactic radiation therapy ing the EORTC QLQ-BM22. Int J Radiat Oncol Biol
delivery for lung, adrenal, and pancreatic tumors: Phys. 2012;84(3):e337–42.
a geometric analysis. Int J Radiat Oncol Biol Phys. 35. Mizumoto M, Harada H, Asakura H, Hashimoto T,
2018;102(4):858–66. Furutani K, Hashii H, et al. Radiotherapy for patients
23. Ostrom QT, Cioffi G, Waite K, Kruchko C, with metastases to the spinal column: a review of
Barnholtz-Sloan JS. CBTRUS statistical report: pri- 603 patients at Shizuoka cancer center hospital. Int J
mary brain and other central nervous system tumors Radiat Oncol Biol Phys. 2011;79(1):208–13.
diagnosed in the United States in 2014-2018. Neuro 36. Pan H, Simpson DR, Mell LK, Mundt AJ,
Oncol. 2021;23(12 Suppl 2):iii1–iii105. Lawson JD. A survey of stereotactic body
24. Merchant TE, Kun LE, Hua CH, Wu S, Xiong X, radiotherapy use in the United States. Cancer.
Sanford RA, et al. Disease control after reduced 2011;117(19):4566–72.
volume conformal and intensity modulated radia- 37. Redmond KJ, Robertson S, Lo SS, Soltys SG, Ryu
tion therapy for childhood craniopharyngioma. Int J S, McNutt T, et al. Consensus contouring guidelines
Radiat Oncol Biol Phys. 2013;85(4):e187–92. for postoperative stereotactic body radiation therapy
25. Merchant TE, Bendel AE, Sabin ND, Burger PC, for metastatic solid tumor malignancies to the spine.
Shaw DW, Chang E, et al. Conformal radiation ther- Int J Radiat Oncol Biol Phys. 2017;97(1):64–74.
apy for pediatric ependymoma, chemotherapy for
38. Ryu SI, Chang SD, Kim DH, Murphy MJ, Le QT, radiomics during radiotherapy to predict recurrence
Martin DP, et al. Image-guided hypo-fractionated in patients with head and neck squamous cell car-
stereotactic radiosurgery to spinal lesions. cinoma. Clin Transl Radiat Oncol. 2021;28:62–70.
Neurosurgery. 2001;49(4):838–46. 52. McGee KP, Hwang KP, Sullivan DC, Kurhanewicz
39. Gerszten PC, Ozhasoglu C, Burton SA, Vogel WJ, J, Hu Y, Wang J, et al. Magnetic resonance biomark-
Atkins BA, Kalnicki S, et al. CyberKnife frame- ers in radiation oncology: the report of AAPM task
less stereotactic radiosurgery for spinal lesions: group 294. Med Phys. 2021;48(7):e697–732.
clinical experience in 125 cases. Neurosurgery. 53. Shukla-Dave A, Obuchowski NA, Chenevert TL,
2004;55(1):89–98; discussion 98–9. Jambawalikar S, Schwartz LH, Malyarenko D,
40. Han EY, Aima M, Hughes N, Briere TM, Yeboa et al. Quantitative imaging biomarkers alliance
DN, Castillo P, et al. Feasibility of spinal stereotac- (QIBA) recommendations for improved precision
tic body radiotherapy in Elekta Unity. J Radiosurg of DWI and DCE-MRI derived biomarkers in mul-
SBRT. 2020;7(2):127–34. ticenter oncology trials. J Magn Reson Imaging.
41. Han EY, Yeboa DN, Briere TM, Yang J, Wang 2019;49(7):e101–e21.
H. Dosimetric analysis of MR-LINAC treatment 54. Tran WT, Suraweera H, Quaioit K, Cardenas D,
plans for salvage spine SBRT re-irradiation. J Appl Leong KX, Karam I, et al. Predictive quantitative
Clin Med Phys. 2022;23(10):e13752. ultrasound radiomic markers associated with treat-
42. Albright JT, Topham AK, Reilly JS. Pediatric head ment response in head and neck cancer. Future Sci
and neck malignancies: US incidence and trends OA. 2019;6(1):FSO433.
over 2 decades. Arch Otolaryngol Head Neck Surg. 55. Bahig H, Yuan Y, Mohamed ASR, Brock KK, Ng SP,
2002;128(6):655–9. Wang J, et al. Magnetic resonance-based response
43. Arboleda LP, Perez-de-Oliveira ME, Hoffmann IL, assessment and dose adaptation in human papilloma
Cardinalli IA, Gallagher KP, Santos-Silva AR, et al. virus positive tumors of the oropharynx treated with
Clinical manifestations of head and neck cancer in radiotherapy (MR-ADAPTOR): an R-IDEAL stage
pediatric patients, an analysis of 253 cases in a sin- 2a-2b/Bayesian phase II trial. Clin Transl Radiat
gle Brazilian center. Med Oral Patol Oral Cir Bucal. Oncol. 2018;13:19–23.
2022;27(3):e285–e93. 56. McDonald BA, Vedam S, Yang J, Wang J, Castillo P,
44. Sengupta S, Pal R, Saha S, Bera SP, Pal I, Tuli Lee B, et al. Initial feasibility and clinical implemen-
IP. Spectrum of head and neck cancer in children. J tation of daily MR-guided adaptive head and neck
Indian Assoc Pediatr Surg. 2009;14(4):200–3. cancer radiation therapy on a 1.5T MR-Linac sys-
45. Boeke S, Monnich D, van Timmeren JE, Balermpas tem: prospective R-IDEAL 2a/2b systematic clini-
P. MR-guided radiotherapy for head and neck can- cal evaluation of technical innovation. Int J Radiat
cer: current developments, perspectives, and chal- Oncol Biol Phys. 2021;109(5):1606–18.
lenges. Front Oncol. 2021;11:616156. 57. Subesinghe M, Scarsbrook AF, Sourbron S, Wilson
46. Gurney-Champion OJ, McQuaid D, Dunlop A, DJ, McDermott G, Speight R, et al. Alterations in
Wong KH, Welsh LC, Riddell AM, et al. MRI-based anatomic and functional imaging parameters with
assessment of 3D intrafractional motion of head and repeated FDG PET-CT and MRI during radiother-
neck cancer for radiation therapy. Int J Radiat Oncol apy for head and neck cancer: a pilot study. BMC
Biol Phys. 2018;100(2):306–16. Cancer. 2015;15:137.
47. Chatterjee S, Maulik S, Prasath S, Arun B, Das 58. Laningham FH, Kun LE, Reddick WE, Ogg
A, Chakrabarty S, et al. Xerostomia quality of life RJ, Morris EB, Pui CH. Childhood central ner-
and resource requirements following parotid spar- vous system leukemia: historical perspectives,
ing adaptive radiotherapy in head and neck can- current therapy, and acute neurological sequelae.

cers: results of a prospective cohort study (study Neuroradiology. 2007;49(11):873–88.
ID CTRI/2017/11/010683). Radiother Oncol. 59. Nachman J, Sather HN, Cherlow JM, Sensel MG,
2022;168:250–5. Gaynon PS, Lukens JN, et al. Response of chil-
48. Castelli J, Simon A, Lafond C, Perichon N, Rigaud dren with high-risk acute lymphoblastic leukemia
B, Chajon E, et al. Adaptive radiotherapy for head treated with and without cranial irradiation: a report
and neck cancer. Acta Oncol. 2018;57(10):1284–92. from the Children’s cancer group. J Clin Oncol.
49. Zhang P, Simon A, Rigaud B, Castelli J, Ospina 1998;16(3):920–30.
Arango JD, Nassef M, et al. Optimal adaptive IMRT 60. Waber DP, Turek J, Catania L, Stevenson K, Robaey P,
strategy to spare the parotid glands in oropharyngeal Romero I, et al. Neuropsychological outcomes from
cancer. Radiother Oncol. 2016;120(1):41–7. a randomized trial of triple intrathecal chemother-
50. Bernstein JM, Homer JJ, West CM. Dynamic apy compared with 18 Gy cranial radiation as CNS
contrast-enhanced magnetic resonance imaging treatment in acute lymphoblastic leukemia: findings
biomarkers in head and neck cancer: potential to from Dana-Farber Cancer Institute ALL consortium
guide treatment? A systematic review. Oral Oncol. protocol 95-01. J Clin Oncol. 2007;25(31):4914–21.
2014;50(10):963–70. 61. McNeer JL, Schmiegelow K. Management of CNS
51. Fatima K, Dasgupta A, DiCenzo D, Kolios C, disease in pediatric acute lymphoblastic leukemia.
Quiaoit K, Saifuddin M, et al. Ultrasound delta- Curr Hematol Malig Rep. 2022;17(1):1–14.
62. Thastrup M, Duguid A, Mirian C, Schmiegelow K, 77. Sandlund JT, Martin MG. Non-Hodgkin lymphoma
Halsey C. Central nervous system involvement in across the pediatric and adolescent and young adult
childhood acute lymphoblastic leukemia: challenges age spectrum. Hematology. 2016;2016(1):589–97.
and solutions. Leukemia. 2022;36(12):2751–68. 78. Burkhardt B, Woessmann W, Zimmermann M,
63. Brooks HW, Evans AE, Glass RM, Pang Kontny U, Vormoor J, Doerffel W, et al. Impact of
EM. Chloromas of the head and neck in childhood: cranial radiotherapy on central nervous system pro-
the initial manifestation of myeloid leukemia in three phylaxis in children and adolescents with central
patients. Arch Otolaryngol. 1974;100(4):306–8. nervous system-negative stage III or IV lymphoblas-
64. Lusher JM. Chloroma as a presenting feature of tic lymphoma. J Clin Oncol. 2006;24(3):491–9.
acute leukemia: a report of two cases in children. Am 79. Link MP, Shuster JJ, Donaldson SS, Berard CW,
J Dis Child. 1964;108(1):62–6. Murphy SB. Treatment of children and young adults
65. Samborska M, Derwich K, Skalska-Sadowska J, with early-stage non-Hodgkin’s lymphoma. N Engl
Kurzawa P, Wachowiak J. Myeloid sarcoma in J Med. 1997;337(18):1259–66.
children—diagnostic and therapeutic difficulties. 80. Seidemann K, Tiemann M, Schrappe M, Yakisan
Contemp Oncol (Pozn). 2016;20(6):444–8. E, Simonitsch I, Janka-Schaub G, et al. Short-pulse
66. Dusenbery KE, Howells WB, Arthur DC, Alonzo T, B-non-Hodgkin lymphoma-type chemotherapy is
Lee JW, Kobrinsky N, et al. Extramedullary leuke- efficacious treatment for pediatric anaplastic large
mia in children with newly diagnosed acute myeloid cell lymphoma: a report of the Berlin-Frankfurt-
leukemia: a report from the Children’s cancer group. Munster group trial NHL-BFM 90. Blood.
J Pediatr Hematol Oncol. 2003;25(10):760–8. 2001;97(12):3699–706.
67. Majdoul S, Colson-Durand L, To NH, Belkacemi 81. Bair SM, Svoboda J. Response-adapted treatment
Y. Adaptive radiotherapy for an uncommon chlo- strategies in Hodgkin lymphoma using PET imag-
roma. Case Rep Oncol. 2016;9(3):593–8. ing. PET Clin. 2019;14(3):353–68.
68. Bakst R, Wolden S, Yahalom J. Radiation therapy for 82. Hall MD, Terezakis SA, Lucas JT, Gallop-Evans E,
chloroma (granulocytic sarcoma). Int J Radiat Oncol Dieckmann K, Constine LS, et al. Radiation therapy
Biol Phys. 2012;82(5):1816–22. across pediatric Hodgkin lymphoma research group
69. Bakst RL, Dabaja BS, Specht LK, Yahalom J. Use protocols: a report from the staging, evaluation,
of radiation in extramedullary leukemia/chlo- and response criteria harmonization (SEARCH) for
roma: guidelines from the International Lymphoma childhood, adolescent, and young adult Hodgkin
Radiation Oncology Group. Int J Radiat Oncol Biol lymphoma (CAYAHL) group. Int J Radiat Oncol
Phys. 2018;102(2):314–9. Biol Phys. 2022;112(2):317–34.
70. Lee DY, Baron J, Wright CM, Plastaras JP, Perl 83. Seravalli E, Kroon PS, Buatti JM, Hall MD,
AE, Paydar I. Radiation therapy for chemotherapy Mandeville HC, Marcus KJ, et al. The potential role
refractory gingival myeloid sarcoma. Front Oncol. of MR-guided adaptive radiotherapy in pediatric
2021;11:671514. oncology: results from a SIOPE-COG survey. Clin
71. AlSemari MA, Perrotta M, Russo C, Alkatan Transl Radiat Oncol. 2021;29:71–8.
HM, Maktabi A, Elkhamary S, et al. Orbital 84. National Cancer Institute. Childhood Cancer Data
myeloid sarcoma (chloroma): report of 2 cases Initiative, National Childhood Cancer Registry
and literature review. Am J Ophthalmol Case Rep. Explorer. 2023. https://nccrexplorer.ccdi.cancer.
2020;19:100806. gov/.
72. Matsui M, Yamanaka J, Shichino H, Sato N, Kubota 85. Gadd S, Huff V, Skol AD, Renfro LA, Fernandez
K, Matsushita T. FDG-PET/CT for detection of extra- CV, Mullen EA, et al. Genetic changes associated
medullary disease in 2 pediatric patients with AML. J with relapse in favorable histology Wilms tumor: a
Pediatr Hematol Oncol. 2016;38(5):398–401. Children’s oncology group AREN03B2 study. Cell
73. Shinagare AB, Krajewski KM, Hornick JL, Rep Med. 2022;3(6):100644.
Zukotynski K, Kurra V, Jagannathan JP, et al. MRI 86. Gratias EJ, Dome JS, Jennings LJ, Chi YY, Tian J,
for evaluation of myeloid sarcoma in adults: a Anderson J, et al. Association of chromosome 1q
single-institution 10-year experience. AJR Am J gain with inferior survival in favorable-histology
Roentgenol. 2012;199(6):1193–8. Wilms tumor: a report from the Children’s oncology
74. Singh A, Kumar P, Chandrashekhara SH, Kumar group. J Clin Oncol. 2016;34(26):3189–94.
A. Unravelling chloroma: review of imaging find- 87. Grundy PE, Breslow NE, Li S, Perlman E, Beckwith
ings. Br J Radiol. 2017;90(1075):20160710. JB, Ritchey ML, et al. Loss of heterozygosity for
75. Wu J, Jiang G, Wu J, Ou L, Zhang C. 18F-FDG PET/ chromosomes 1p and 16q is an adverse prognostic
CT imaging of testicular myeloid sarcoma in a pedi- factor in favorable-histology Wilms tumor: a report
atric patient. Clin Nucl Med. 2021;46(1):84–5. from the National Wilms Tumor Study Group. J Clin
76. Zhang M, Yang X, Si Y, Kan Y, Yang J. 18 F-FDG Oncol. 2005;23(29):7312–21.
PET/CT findings of myeloid sarcoma involving the 88. Wilm’s tumor. In: Gunderson LL, editor. Gunderson
left ventricle and kidney in a pediatric patient. Clin and Tepper’s clinical radiation oncology. 4th ed.
Nucl Med. 2023;48(1):103–5. Philadelphia: Elsevier; 2015.
89. Mul J, Seravalli E, Bosman ME, van de Ven CP, sarcoma: clinical outcomes and dosimetry compared
Littooij AS, van Grotel M, et al. Estimated clini- to intensity-modulated radiation therapy. Int J Radiat
cal benefit of combining highly conformal target Oncol Biol Phys. 2011;81(5):1367–73.
volumes with volumetric-modulated arc therapy 102. Ahmed SK, Witten BG, Harmsen WS, Rose PS,
(VMAT) versus conventional flank irradiation in Krailo M, Marcus KJ, et al. Analysis of local control
pediatric renal tumors. Clin Transl Radiat Oncol. outcomes and clinical prognostic factors in localized
2021;29:20–6. pelvic Ewing sarcoma patients treated with radiation
90. Kozak MM, Crompton D, Gross BA, Harshman L, therapy: a report from the Children’s oncology group.
Dickens D, Snyder J, et al. Initial clinical applica- Int J Radiat Oncol Biol Phys. 2023;115(2):337–46.
tions treating pediatric and adolescent patients 103. Ahmed SK, Indelicato DJ, Chuba PJ, Krailo M,
using MR-guided radiotherapy. Front Oncol. Buxton A, Randall RL, et al. Local failure in non-
2022;12:962926. metastatic Ewing sarcoma patients treated with
91. Meier CM, Furtwangler R, von Schweinitz D, Stein definitive radiation therapy on AEWS1031: a report
R, Welter N, Wagenpfeil S, et al. Vena cava throm- from the Children’s oncology group. Int J Radiat
bus in patients with Wilms tumor. Cancers (Basel). Oncol Biol Phys. 2022;114(3 Suppl):S69–70.
2022;14(16):3924. 104. Ahmed SK, Robinson SI, Arndt CAS, Petersen IA,
92. Hoeben BA, Carrie C, Timmermann B, Mandeville Haddock MG, Rose PS, et al. Pelvis Ewing sarcoma:
HC, Gandola L, Dieckmann K, et al. Management local control and survival in the modern era. Pediatr
of vertebral radiotherapy dose in paediatric patients Blood Cancer. 2017;64(9).
with cancer: consensus recommendations from the 105. Talleur AC, Navid F, Spunt SL, McCarville MB, Wu
SIOPE radiotherapy working group. Lancet Oncol. J, Mao S, et al. Limited margin radiation therapy
2019;20(3):e155–e66. for children and young adults with Ewing sarcoma
93. Chen MJ, Leao CR, Simoes RCP, Belletti FS, achieves high rates of local tumor control. Int J
Figueiredo MLS, Cypriano MS. Kidney-sparing Radiat Oncol Biol Phys. 2016;96(1):119–26.
whole abdominal irradiation in Wilms tumor: poten- 106. Uezono H, Indelicato DJ, Rotondo RL, Mailhot
tial advantages of VMAT technique. Pediatr Blood Vega RB, Bradfield SM, Morris CG, et al. Treatment
Cancer. 2020;67(5):e28223. outcomes after proton therapy for Ewing sar-
94. Ward E, DeSantis C, Robbins A, Kohler B, Jemal coma of the pelvis. Int J Radiat Oncol Biol Phys.
A. Childhood and adolescent cancer statistics, 2014. 2020;107(5):974–81.
CA Cancer J Clin. 2014;64(2):83–103. 107. Adaptive radiation therapy. Med Phys.
95. Vo KT, Matthay KK, Neuhaus J, London WB, Hero 2011;38(7):4467–4468.
B, Ambros PF, et al. Clinical, biologic, and prog- 108. Prior P, Chen X, Botros M, Paulson ES, Lawton C,
nostic differences on the basis of primary tumor site Erickson B, et al. MRI-based IMRT planning for
in neuroblastoma: a report from the international MR-linac: comparison between CT- and MRI-based
neuroblastoma risk group project. J Clin Oncol. plans for pancreatic and prostate cancers. Phys Med
2014;32(28):3169–76. Biol. 2016;61(10):3819–42.
96. Paulino AC, Ferenci MS, Chiang KY, Nowlan 109. Chen X, Prior P, Chen GP, Schultz CJ, Li
AW, Marcus RB Jr. Comparison of conventional XA. Technical note: dose effects of 1.5 T transverse
to intensity modulated radiation therapy for magnetic field on tissue interfaces in MRI-guided
abdominal neuroblastoma. Pediatr Blood Cancer. radiotherapy. Med Phys. 2016;43(8):4797.
2006;46(7):739–44. 110. Raaijmakers AJ, Raaymakers BW, Lagendijk
97. Fleming ID, Etcubanas E, Patterson R, Rao B, Pratt JJ. Magnetic-field-induced dose effects in
C, Hustu O, et al. The role of surgical resection MR-guided radiotherapy systems: dependence
when combined with chemotherapy and radiation in on the magnetic field strength. Phys Med Biol.
the management of pelvic rhabdomyosarcoma. Ann 2008;53(4):909–23.
Surg. 1984;199(5):509–14. 111. Li XA, Liu F, Tai A, Ahunbay E, Chen G, Kelly T,
98. Groff DB. Pelvic neoplasms in children. J Surg et al. Development of an online adaptive solution
Oncol. 2001;77(1):65–71. to account for inter- and intra-fractional variations.
99. Zobel M, Zamora A, Sura A, Wang L, Stein J, Radiother Oncol. 2011;100(3):370–4.
Marachelian A, et al. The clinical management and 112. Ahunbay EE, Li XA. Gradient maintenance: a new
outcomes of pelvic neuroblastic tumors. J Surg Res. algorithm for fast online replanning. Med Phys.
2020;249:8–12. 2015;42(6):2863–76.
100. Romano E, Simon R, Minard-Colin V, Martin V, 113. Zhang Y, Paulson E, Lim S, Hall WA, Ahunbay E,
Bockel S, Espenel S, et al. Analysis of radiation Mickevicius NJ, et al. A patient-specific autoseg-
dose/volume effect relationship for anorectal mor- mentation strategy using multi-input deformable
bidity in children treated for pelvic malignancies. Int image registration for magnetic resonance imaging-
J Radiat Oncol Biol Phys. 2021;109(1):231–41. guided online adaptive radiation therapy: a feasibil-
101. Cotter SE, Herrup DA, Friedmann A, Macdonald ity study. Adv Radiat Oncol. 2020;5(6):1350–8.
SM, Pieretti RV, Robinson G, et al. Proton radio- 114. Ates O, Ahunbay EE, Moreau M, Li XA. Technical
therapy for pediatric bladder/prostate rhabdomyo- note: a fast online adaptive replanning method for
VMAT using flattening filter free beams. Med Phys. ing in pediatric age totally safe? A critical reprisal.
2016;43(6):2756–64. Radiol Med. 2018;123(9):695–702.
115. Zhang J, Ahunbay E, Li XA. Technical note: 118. Huynh E, Hosny A, Guthier C, Bitterman DS,
acceleration of online adaptive replanning with Petit SF, Haas-Kogan DA, et al. Artificial intelli-
automation and parallel operations. Med Phys. gence in radiation oncology. Nat Rev Clin Oncol.
2018;45(10):4370–6. 2020;17(12):771–81.
116. Boik N, Hall MD. Psychosocial support for pediat- 119. Fuchs H, Padilla-Cabal F, Oborn BM, Georg
ric patients at proton therapy institutions. Int J Part D. Commissioning a beam line for MR-guided parti-
Ther. 2020;7(1):28–33. cle therapy assisted by in silico methods. Med Phys.
117. Salerno S, Granata C, Trapenese M, Cannata V, 2023;50(2):1019–28.
Curione D, Rossi Espagnet MC, et al. Is MRI imag-
MR-Linac-Guided Adaptive
Radiotherapy for Brain Tumors
20
Tugce Kutuk, Jay Detsky, Arjun Sahgal,
and Rupesh Kotecha
Disease Overview different primary tumor histologies and biologic

behaviors. Metastases significantly outpace the
The term “brain tumors” refers to malignancies incidence of primary brain tumors with a 10:1
that originate in the brain (primary brain tumors) ratio and an annual estimated incidence exceed-
as well as those that metastasize to the central ing 150,000 [3]. Primary and metastatic brain
nervous system (CNS) with a primary site of dis- tumors can cause a wide range of symptoms
ease origin arising from an extracranial site. depending on their size, location, and rate of
Primary brain tumors consist of various histo- growth, as well as the functional characteristics
logic types with different tumor growth rates, and eloquence of the substructures of the brain
patterns of disease progression, and treatment affected. Comprehensive history, neurologic
paradigms [1]. Approximately 30% of primary examination, and diagnostic neuroimaging stud-
brain tumors in the USA are malignant, while the ies are necessary in the initial evaluation of the
remainder are benign. Meningiomas and glial patient suspected to have a brain tumor.
tumors (such as glioblastoma multiforme [GBM], GBM is a highly malignant tumor that grows
astrocytoma, oligodendroglioma, etc.) account quickly and infiltrates into the surrounding brain
for approximately two-thirds of all primary brain tissue [4]. The most common symptoms of GBM
tumors in adults, with GBM as the most common include headaches, focal neurologic deficits, sei-
primary malignant brain tumor [2]. On the other zures, nausea, and vomiting. As measured from
hand, metastases to the CNS most commonly initial diagnosis, the median survival is less than
arise from lung, melanoma, and breast cancers, 2 years, and the overall survival remains poor [5].
and are treated in a similar manner despite the Despite advancements in radiotherapy technol-
ogy, developments of novel systemic therapies,
T. Kutuk · R. Kotecha (*) and improvements in surgical techniques, only
Department of Radiation Oncology, Miami Cancer 5–10% of patients survive up to 5 years [6]. For
Institute, Baptist Health South Florida, patients with newly diagnosed GBM, the current
Miami, FL, USA standard of care includes maximal safe resection,
e-mail: [email protected]; RupeshK@
baptisthealth.net followed by concurrent chemoradiotherapy and
adjuvant temozolomide. Tumor-treating fields
J. Detsky · A. Sahgal
Department of Radiation Oncology, Odette Cancer (TTFields) can also be applied adjuvantly [7].
Centre, Sunnybrook Health Sciences Centre, Conventionally fractionated radiotherapy courses
University of Toronto, Toronto, ON, Canada (i.e., 60 Gy in 30 fractions) are typically used for
e-mail: [email protected]; younger patients with good performance status or
[email protected]
376 T. Kutuk et al.
those treated in clinical trials, whereas hypofrac- Overcoming Challenges

tionated regimens (i.e., 40 Gy in 15 fractions) are with Traditional Approaches
used for elderly patients [8–10]. Extreme hypo-
fractionation (i.e., 25 Gy in 5 fractions) can also Computed tomography (CT) is a widely used
be used as a palliative measure to stabilize imaging technique for simulation, target volume
symptomatology over medical management
delineation, treatment planning, and image guid-
alone [11]. ance during treatment. However, CT alone may
Brain metastases occur in approximately not be sufficient to achieve the ultimate goal of
15–30% of all cancer patients at some point dur- delivering radiation dose efficiently and precisely
ing their disease course [12]. The incidence of to the target volume while sparing surrounding
brain metastases is increasing due to a combina- normal tissues and organs at risk (OARs). In the
tion of factors such as patients living longer with treatment of brain tumors such as glioblastoma
their systemic disease, increased screening and (GBM) or brain metastases, MRI provides supe-
surveillance for asymptomatic patients, and the rior resolution of the tumor extent and OARs and
improved sensitivity of imaging techniques such is the standard of care for treatment planning.
as magnetic resonance imaging (MRI) [13]. Despite these benefits, MRI has a limited role in
Treatment for brain metastases typically involves image guidance during treatment. To address
a combination of approaches including surgery, this, cone-beam CT is routinely applied for posi-
radiotherapy, chemotherapy, targeted agents, and tional accuracy. Nonetheless, the technology’s
immunotherapies [14]. The specific treatment capabilities with respect to image guidance are
plan depends on factors such as the number and not yet sufficient for the brain (Fig. 20.1).
location of the brain metastases, the patient’s The integration of MRI technologies into the
overall performance status, and the type of pri- radiotherapy workflow has been growing rapidly
mary cancer. over the past decade, as there has been greater
Whole-brain radiation therapy (WBRT) has availability and access to diagnostic and dedi-
been the historical standard of care for treating cated radiotherapy simulation MRI scanners. In
brain metastases. However, the development addition to the use of MRI for anatomic visual-
of stereotactic radiosurgery (SRS) has revolu- ization at the time of treatment planning, specific
tionized the treatment of brain metastases, due MRI sequences can be applied to investigate the
to improved disease control, reduced detri- physiology and metabolism of CNS tumors (i.e.,
ment to neurocognitive function, and improved diffusion, perfusion, chemical exchange satura-
quality of life compared to WBRT [15–19]. tion transfer, magnetic transferase, and spectros-
For patients with small intact brain metastasis, copy sequences) [24]. Diffusion-weighted
SRS is typically delivered in a single fraction imaging (DWI), for example, can detect changes
[20]. Given the inferior local control rates in water mobility within the tumor, which repre-
with large intact brain metastasis (>2 cm), sents changes in the tumor’s cellularity and
fractionated SRS (FSRS), delivered in 3–5 necrosis, and this information can be used to pre-
fractions, is frequently recommended [21]. dict the tumor’s response to chemotherapy and
SRS or FSRS can also be delivered preopera- radiotherapy, as well as to predict overall survival
tively or postoperatively to improve local con- rates [25–28]. Alternatively, perfusion imaging
trol over surgery alone and as an alternative to can measure blood flow within the tumor, which
WBRT [22, 23]. can provide information about the tumor’s blood
As radiotherapy represents a primary treat- supply and metabolism. This can also be used to
ment modality used in the management of GBM predict the tumor’s response to therapy and over-
and brain metastases, the focus of this chapter is all survival [29]. The metabolic profile of a tumor
to review the potential applications of magnetic can also be determined by MRI spectroscopy,
resonance-guided radiation therapy (MRgRT) for chemical exchange saturation transfer, and mag-
these two categories of CNS malignancies. netic transferase imaging as potential imaging
20 MR-Linac-Guided Adaptive Radiotherapy for Brain Tumors 377
a b
Fig. 20.1 Imaging comparison of a cerebellopontine images, whereas both were clearly visualized on the
angle meningioma using (a) cone-beam computed tomog- MRIdian image. This demonstrates the superior resolu-
raphy (CBCT) and (b) ViewRay MRIdian™ TrueFISP tion and sensitivity of MRIdian MRI for detecting and
magnetic resonance imaging (MRI). It was difficult to localizing such tumors
delineate the tumor and adjacent brainstem on the CBCT
surrogates of tumor aggressiveness [30–36]. The unlocking the potential to use this real-time infor-
time course of MRI acquisition has been limited mation during the treatment course to potentially
to the time of treatment and then approximately affect patient outcomes [42, 43].
1–3 months after the completion of chemoradio- One of the key benefits of MR-Linac devices
therapy. However, recent developments in these is their ability to overcome the challenge of invol-
sequences are significant, and parameters derived untary movements, such as respiration and diges-
from these sequences are being integrated into tion, which can cause changes in the position of
the radiotherapy process to guide boost volumes tumors during the course of radiotherapy. This is
in clinical trials [37]. not an issue in the brain, making functional inter-
MRgRT also provides the unique advantage of fraction sequences more robust and clinically
visualizing tumors on a daily basis, both during applicable. Therefore, the focus has been on
treatment (intra-treatment) and in between treat- understanding tumor dynamics during radiother-
ments (inter-treatment), using advanced imaging apy; a recent example of this is a deeper under-
sequences that can provide additional informa- standing of the migration of glioma during a
tion about the tumor’s physiology during therapy course of radiotherapy and the potential to adapt
which opens up the possibility of personalizing [44].
the response of the tumor to therapy [38–41]. There are other potential benefits from
MRI-guided linear accelerator (MR-Linac) MRgRT as certain targets are sensitive to
devices have unique advantages for CNS tumors changes that can occur with little external
to frequently image tumors during a course of motion. Although not common for brain metas-
radiotherapy using both anatomically guided and tasis or GBMs, optic pathway meningiomas or
physiologically focused sequences [32], hence metastases to the orbit or retina are known to be
378 T. Kutuk et al.
affected by optic nerve motion [45, 46]. In fact, cession (SSFP) sequence provides the highest
the displacement of the optic nerves in certain signal-to-noise ratio per unit of time to yield rap-
positions can even exceed 1 cm [46]. Even if idly acquired, high-quality images. With
intrafraction motion is not a concern, interfrac- TrueFISP, the resulting signal is related to the
tion changes in the tumor size and shape can ratio of T2 to T1, but ultimately yields a predomi-
still occur between treatment sessions, which nantly T2-weighted sequence [48–50]. This is
can be problematic. These changes can be because the SSFP sequence produces a bright
caused by factors such as edema, inflammation, signal from tissues with short T2 relaxation
or changes in the tumor’s blood supply [44]. To times, such as water while suppressing signals
address this issue, it is common practice to per- from tissues with longer T2 relaxation times.
form diagnostic MRI scans weekly or biweekly Therefore, TrueFISP is particularly useful for
to ensure adequate target dose coverage and to imaging tissues with high water content, such as
adapt the radiotherapy plan accordingly, if nec- the brain [51, 52].
essary, based on anatomical changes. However,
with MRgRT, the diagnostic and treatment plan-
ning can be done simultaneously in the same Glioblastoma
session, thus avoiding the need for additional
diagnostic MRI scans, which can save time and Although limited in long-term follow-up, recent
resources. studies have demonstrated evolving tumor
Two commercial systems are currently avail- dynamics, geographic shifts, and changes in
able for MRgRT: the Elekta Unity™ (Elekta AB, enhancement and edema throughout a fraction-
Stockholm, Sweden) and the ViewRay ated radiotherapy course for GBM. In a pro-
MRIdian™ A3i (ViewRay Inc., Cleveland, OH, spective study that evaluated sequential MRI
USA) systems. The Elekta Unity™ utilizes a 1.5- scans conducted during chemoradiotherapy for
Tesla (T) magnet, the most powerful magnet patients with GBM, target dynamics were evalu-
commercially available for MRgRT. In the ated [44]. In that study, the gross tumor volume
Unity™ system, the photon beam is placed on a (GTV) and clinical target volume (CTV), along
rotating gantry, which passes through a supercon- with tumor migration, were evaluated at fraction
ducting cryostat with a configuration that avoids 0, fraction 10, fraction 20, and 1 month after
angles that may interfere with imaging quality, in treatment. Among the majority of patients who
order to overcome interference from the magnet. were enrolled, the GTV decreased from a
Also, the electron gun is mounted in a zone with median of 18.4 cm3 at fraction 0, 14.7 cm3 at
zero field, and specific settings are used for shims fraction 10, 13.7 cm3 at fraction 20, and 13 cm3
to minimize image degradation [47]. The at 1 month after treatment. Moreover, in their
ViewRay MRIdian™ system uses similar design report, tumor shrinkage was associated with
characteristics to the Elekta Unity™. However, it tumor migration. The median GTV migrations
uses a less powerful magnet (0.35 T) which leads were greater than 5 mm in 54% of patients at
to a weak magnetic field that places less stress on fraction 10, 58% of patients at fraction 20, and
the linear accelerator yet allows for sufficient 57% of patients 1 month after treatment. This
image quality for image guidance and treatment amount of tumor migration during chemoradio-
adaptation [47]. The image quality is not the therapy exposes patients to the possibility of
same as a 1.5 T MRI and that has to be consid- geographic tumor miss, in particular, if applying
ered as functional sequences cannot be obtained margin reduction strategies. Prior studies have
at low-field-strength MRI. Additionally, the also demonstrated the potential geographic miss
ViewRay MRIdian™ system uses a balanced due to tumor migration, although to a smaller
steady-state free precession sequence, also degree [53, 54]. Consistent findings of tumor
known as true fast imaging with steady-state pre- migration during treatment further reinforce the
cession (TrueFISP). This steady-state free pre- need to evaluate the role of MRgRT with on-
table adaptation for patients undergoing chemo- eated on MRI before each of the 30 fractions
radiotherapy for GBM. [57]. In their study, 4 of 14 patients demonstrated
In addition to tumor migration, changes in the increases of at least 25% in T2 hyperintense vol-
primary tumor and cavity are also observed and ume through treatment. This volume expansion
reported in MRgRT series. In a case series, Mehta correlated with both T2/FLAIR and contrast-
et al. evaluated the daily MRgRT images of three enhanced volume expansion on post-radiotherapy
GBM patients treated on the MRIdian™ system diagnostic MRIs. The observed growth in early
to determine whether changes in resection cavity edema was suggestive of early pseudoprogres-
volume and cerebral edema could be observed on sion, and these patients exhibited improved out-
daily MRIs [55]. There was a general trend of comes. In three of these four cases, a significant
daily decrease in the resection cavity volume in volume increase started at week 3, with the most
these three patients analyzed. The edema volume prominent changes occurring during weeks 4 and
of 1 patient initially increased during the first 13 5. In addition to visualization of these structural
fractions, but then began to decrease until the end changes, advanced imaging techniques can add
of the treatment. In a recent case report of a potential information during a radiotherapy
patient treated with the new MRIdian® A3i course. For example, in one series, it was detected
BrainTx™, imaging changes were also observed that patients who developed progressive disease
during a course of conventionally fractionated had significantly lower relative cerebral blood
chemoradiotherapy [56]. While starting radio- volume (rCBV) during their third week of treat-
therapy, the surgical cavity volume on T1 was ment [58]. Apparent diffusion coefficient (ADC)
6.4 cc (representing 100%); on weekly repeat full values were also shown to increase during treat-
MRI, the postoperative cavity relative decrease ment of tumors responding to radiotherapy [59].
was 21.9%, 29.7%, 50.0%, 58.6%, and 67.5% Therefore, the potential for these advanced imag-
from baseline. Additionally, the initial ing features to be integrated into treatment plan
T2-weighted-fluid-attenuated inversion recovery delivery review is key to unlocking the full power
(T2/FLAIR) signal abnormality volume at base- of MRgRT systems.
line was 60.3 cc (representing 100%) and showed
a relative decrease in volume of 79.6%, 80.0%,
81.5%, 81.8%, and 86.5% on weekly MRIs Brain Metastasis
(Fig. 20.2a and b). It is likely that regular interval
imaging with MRgRT, based on the gradual Significant volume changes can also occur with
decrease in cavity size and edema volume, can brain metastases during treatment or from the
assist in adaptive planning which results in time of simulation to delivery of radiotherapy;
reduced irradiated volume and a better sparing of adaptation may therefore be useful for improving
normal brain tissue. In an alternative scenario, a treatment accuracy, especially if limited addi-
brain tumor may grow during treatment in some tional planning margins are employed. Salkeld
cases, and it can be necessary to enlarge the tar- and colleagues demonstrated that a change in
get volume. Additionally, daily imaging can also management was required in 46% of lesions with
allow for the monitoring of cerebral edema, 7-day interscan intervals, increasing to 62% of
which may facilitate the correlation of clinical lesions at intervals longer than 7 days in their
symptoms and the follow-up of corticosteroid study. In their analysis, the overall mean change
response. Future studies between both MRgRT was 0.15 cm3 for brain metastases in approxi-
platforms are needed to understand which tim- mately 1 week [60]. In a study with 531 brain
ings are critical for tumor and edema evaluation. metastases in 101 consecutive patients at our cen-
In another study from Jones et al., 14 GBM ter, 20.5% of the lesions would require a change
patients underwent tri-Cobalt-60 MRgRT in 30 in management within 7 days, and 33.3% of
fractions over 6 weeks delivered with concurrent lesions would require replanning if the interval
temozolomide, and the tumor volume was delin- exceeded 7 days [61]. Additionally, the
380 T. Kutuk et al.
Fig. 20.2 Axial T1 post-contrast (a) and T2/FLAIR (b) ing enhancement and the peri-tumoral FLAIR volumes
MRI sequences obtained weekly during radiotherapy. On decreased during the chemoradiotherapy course. (a) Axial
sequential imaging, both the resection cavity and surround- T1 post-contrast images. (b) Axial T2/FLAIR images
d isplacement of the tumor centroid was observed metastasis treatment. Even though these short
in 66.8% of lesions, and in 12.4% of cases, the courses only allow for a limited amount of time
magnitude exceeded 1 mm. Ultimately, each day for adaptation, studies have shown that changes
between treatment planning MRI acquisition and in MRI from 1 day to 1 week are correlated with
treatment was associated with an increased tumor treatment responses [63, 64]. It may be possible
margin expansion beyond 1 mm by 5%. Similarly, to adjust the treatment plan or to predict early
tumor dynamic changes have also been reported response in cases of SRS by daily monitoring
for cavity volumes after resection of brain metas- with MRgRT. An example brain metastasis case
tases [62]. This suggests that in some situations, treated with FSRS is demonstrated in Fig. 20.3 to
anatomic adaptation might be useful for brain show the tumor volume change during the treat-
Fig. 20.3 An example brain metastasis resection cavity fractions. The MRIs were performed at the time of each
case treated with fractionated stereotactic radiosurgery fraction to assess the target volume change. The interval
(FSRS). The patient was a 54-year-old male with a history tumor shrinkage observed between the first (volume,
of lung cancer who presented with a single brain metasta- 6.5 cc) and third (volume, 5.0 cc) fractions is
sis and underwent gross total resection. He underwent demonstrated
FSRS to the resection cavity to a dose of 27 Gy in three
ment course. A recent study conducted by Tan tissue and reducing the risk of radiation necrosis.
et al. evaluated the changes in the CTV during However, the implementation of SRS on
five-fraction FSRS for resected brain metastases MR-Linac devices is a significant challenge. SRS
to quantify the interfraction cavity dynamic requires highly conformal planning and delivery,
changes [65]. Their findings showed that there accurate patient setup, and position confirmation,
were significant cavity dynamic changes, with and management of intrafraction motion and sev-
80% of patients experiencing a reduction in T1 eral different SRS platforms handle these issues
post-contrast CTV volume at third fraction com- in different ways. Previous feasibility studies
pared to simulation MRI. Additionally, the have demonstrated that MR-Linac is capable of
observed variability in CTV contouring between generating SRS plans that meet clinical require-
T1 post-contrast and T2/FLAIR sequences in the ments for brain metastases [65–67]. However,
study indicated that T1 post-contrast sequence these studies still demonstrated that plan quality
should be the preferred choice for precise target is still limited in comparison with other dedicated
delineation. SRS systems.
Based on the findings of the aforementioned In conclusion, MRgRT is a new and evolving
studies, it can be inferred that personalized treat- technology, and it is still being researched to
ment adaptation through MRgRT facilitates the determine its optimal use in the treatment of
detection of significant changes in the tumor GBM and brain metastases. The advantages of
dynamics and utilizing MR-Linacs for treatment MRgRT include improved targeting accuracy
adaptation would enhance CTV coverage while which will ultimately translate into improved dis-
minimizing radiation exposure to healthy brain ease control and reduced risk of side effects.
382 T. Kutuk et al.
Contouring and Planning study, when the target volumes delineated using
the diagnostic MRI sequences (MPRAGE and
To deliver radiation to targets in the brain, a non- T2/FLAIR) were compared to the MRIdian™
contrast CT (1- to 3-mm slice thickness) simula- sequences (T1 post-contrast and T2/FLAIR) in
tion scan is performed initially. Patients are placed one example GBM case, a good concordance
in an optimal position utilizing a customized ther- between the volumes was observed: 68.4 cc vs
moplastic mask with head support which is used 68.3 cc for GTV 46 Gy and 6.2 cc versus 6.4 cc
for daily positioning. An MRI simulation scan is for GTV 14 Gy, respectively. CTVs are defined
also acquired to fuse to the CT simulation scan for as expansions of each of the GTVs by 1.5–2 cm,
target volume delineation; a TrueFISP sequence respecting natural anatomic barriers to tumor
for ViewRay MRIdian™ treatments; and a stan- spread. Finally, PTVs are defined as CTVs plus
dard T1-weighted gradient echo series with con- uniform expansions of 3 mm. Comparisons of
trast (T1 + c) and a volumetric T2/FLAIR series image qualities between diagnostic (3 T) and
for Unity treatments. Additional sequences, such simulation (0.5 T) MRIs are illustrated in
as T1 post-contrast, T2, and FLAIR sequences, Fig. 20.4.
can also be obtained and coregistered to the treat- For brain metastasis treatments, the GTV is
ment planning TrueFISP sequence. With the defined as the lesion itself for intact brain metas-
ViewRay MRIdian™, the simulation study is tasis and the resection cavity in the postoperative
exported to the planning software for coregistra- setting. No additional CTV margin is added for
tion with the previous diagnostic MRI sequences intact brain metastasis, and a 2-mm margin is
(i.e., T1-magnetization-prepared rapid gradient added for resection cavities [68, 69]. These tar-
echo [T1-MPRAGE] and T2/FLAIR), using a gets are delineated using the T1 post-contrast
rigid registration to the area of interest. MR-Linac- imaging study. Finally, PTVs are defined as CTV
acquired brain imaging sequences for MRIdian™ plus uniform expansions of 1 mm.
and Unity™ are shown in Table 20.1. In the MRIdian™ treatment planning system
The standard Radiation Therapy Oncology (TPS), the density from the CT is rigidly regis-
Group (RTOG) planning approach for GBM tered to the planning MRI for dose calculation
involves a two-phase approach, with the GTVs purposes. To simplify the treatment planning pro-
being defined using diagnostic postoperative cess, bilateral structures such as lens left + right
MRIs. The GTV46 Gy is defined as the T2/ (L + R) and optic nerve L + R, which have a max-
FLAIR contrast-enhanced volume, and the imum point dose constraint, are merged into a
GTV14 Gy (boost) is defined as the MRI T1 single structure. Furthermore, to ensure that the
contrast-enhanced resection cavity. In our recent dose to the OARs is within acceptable limits, a
Table 20.1 MRgRT-acquired brain imaging sequences

ViewRay MRIdian Elekta Unity
Scan Details Scan Details
Daily imaging
TrueFISP 1-mm slice thickness; FOV 25 × 25 × 25 cm3 T1-weighted 1.1 × 1.1 × 2.2 mm voxels; FOV
GRE 27 × 20 × 20 cm3
T2/FLAIR 1.1 × 1.1 × 2.2 mm voxels; FOV
27 × 20 × 20 cm3
Weekly imaging—IV contrast
T1 1-mm slice thickness; FOV 20 × 20 × 20 cm3
T2 1.5-mm slice thickness; FOV 20 × 20 × 20 cm3
FLAIR 2.5-mm slice thickness; FOV 15 × 15 × 25 cm3
(in-plane 1.5 mm)
TrueFISP True fast imaging with steady state precession, FOV Field of view, GRE Gradient recalled echo, FLAIR Fluid-
attenuated inversion recovery
Fig. 20.4 Comparison

between diagnostic (3 T)
and simulation (0.5 T)
MRIs for a gross totally
resected glioblastoma.
Represented in the left
side are two axial slices
(T1-MPRAGE and T2/
FLAIR sequences). On
the right side are the
corresponding
simulation MRIs
(T1-post Gd and T2/
FLAIR sequences)
planning organ-at-risk volume (OAR_PRV) is MRI. The selection of the radiation beam posi-
created, which encompasses any OARs that have tions in the treatment plan is meticulously deter-
a maximum point dose constraint and overlap mined based on the size and location of the target
with the target volumes. In order to optimize the volume. For GBM cases, for example, a pseudo-
radiation dose, the planning target volume opti- half-
arc beam arrangement is typically used,
mized structure (PTV_opt) is generated by sub- which involves delivering radiation from multi-
tracting the OAR_PRV from the PTV. To control ple angles around the involved hemisphere of the
low- and high-dose spills, two rings are created. brain. In order to ensure that the total radiation
A high-dose ring is used to control prescription dose between the initial and boost plans achieves
dose conformality, which is created as a function the desired constraints for each OAR, the cumu-
of rings PTV + 4 mm minus PTV + 2 mm. Low- lative constraints of OARs are adjusted based on
dose spilling is controlled by creating a low-dose the constraint limitations in each plan. For
ring which is a function of the skin minus the instance, if the optic chiasm constraint is
PTV + 4 mm ring. To help with daily localization D0.03cc < 54 Gy and the prescribed dose to the
matching, a GTV_TrueFISP is delineated as the target volume is 60 Gy, each plan takes into

enhanced region of the GTV on the TrueFISP account the overlapping structures with a lower
384 T. Kutuk et al.
tolerance. To meet the maximum dose constraint, plan planned for ViewRay MRIdian is demon-
the chiasm is limited to 44 Gy in the primary plan strated in Fig. 20.5. For the comparison purpose,
and 10 Gy in the boost plan. Thus, this guarantees the isodose distributions for a right frontal brain
that the D0.03cc < 54 Gy requirement is satisfied in metastasis planned with Gamma Knife® (Elekta,
the composite of the primary and boost plans. A Stockholm, Sweden), CyberKnife® (Accuray,
composite dose distribution is utilized at the end Wisconsin, USA), and TrueBeam® Linac (Varian,
of planning to assess the summation of the initial Palo Alto, California, USA) SRS are illustrated
and boost plans. An example GBM radiotherapy in Fig. 20.6.
Fig. 20.5 An example glioblastoma radiotherapy plan and corresponding isodose distribution with ViewRay MRIdian.
A 2-phase technique was employed in this case, 46 Gy in 23 fractions with a boost to 60 Gy in 30 fractions
Fig. 20.6 Comparison of MRgRT to other radiation are illustrated to demonstrate the differences in isodose
modalities. Isodose distributions for a left frontal brain distribution among the radiotherapy modalities. All plans
metastasis planned with Gamma Knife radiosurgery, were planned 27 Gy in 3 fractions
CyberKnife, MRIdian, and TrueBeam Linac radiosurgery
Monaco (Elekta CMS, Maryland Heights, ance” constraints for both CTV and PTV cover-
MO, USA) is the TPS used for the Unity™ sys- age and OAR limits based on institutional
tem. Similar to the MRIdian™, the CT simula- guidelines. A green light demonstrates passing
tion is rigidly registered to the MRI simulation the ideal constraint; yellow is for beyond ideal
to use electron densities to calculate delivered but still within tolerance; and red light indicates
radiation dose. PRVs for OARs in close proxim- an unacceptable level of target coverage or over-
ity to (or overlapping with) the PTV are used dosing of an OAR. This same system is used for
specifically for the brainstem, optic chiasm, and rapid plan evaluation during daily onboard
optic nerves. Reference OARs using a 2-mm replanning. Further technical details on treat-
margin on the globes and skull (split into left/ ment planning for the Unity™ system are
right) are created for verification during onboard described by Ruschin et al. [70].
treatment to ensure proper patient positioning
since rotations cannot be corrected on the
Unity™ treatment bed. An IMRT plan with a Dose Constraints
default 9 beam (which can be adjusted to
between 7 and 11) is created with a 2-mm dose Institutional dosimetric constraints for the nearby
grid. Dose calculation is performed with Monaco CNS OARs and treatment planning directives for
TPS Monte Carlo algorithm coined GPUMCD. A glioblastoma and brain metastases are presented
stoplight system is used with “ideal” and “toler- in Tables 20.2 and 20.3.
Table 20.2 Institutional glioblastoma dose constraints for organ-at-risk (OAR) and target volume coverage directives
Glioblastoma dose constraints
OAR Target planning dose
Uninvolved brain Dmean ≤30 Gy ≥99% of the GTV receives ≥100% of
PD
Total brain D0.03cc <66 Gy ≥99% of the CTV receives ≥100% of
PD
Brainstem D0.03cc ≤55 Gy ≥95% of the PTV receives ≥100% of PD
Cochlea D5% 55 Gy
Dmean ≤35 Gy
Spinal cord D0.03cc ≤45 Gy
Ocular globe D0.03cc ≤45 Gy
D15% ≤10 Gy
Lens D0.03cc ≤7 Gy
Optic nerves D0.03cc ≤55 Gy
Chiasm D0.03cc ≤56 Gy
Retina D0.03cc ≤45 Gy
Hippocampus Dmean ≤20 Gy
D40% ≤7.3 Gy
Pituitary gland Dmean ≤35 Gy
Lacrimal gland Dmean ≤25 Gy
Temporal lobe Dmean ≤20 Gy
OARs organs at risk
386 T. Kutuk et al.
Table 20.3 Institutional brain metastasis dose constraints for organ-at-risk (OAR) and target volume coverage
directives
Brain metastasis dose constraints for SRS and FSRS
One fraction Three fractions Five fractions
OAR OAR OAR
Uninvolved brain Dmean ≤8 Gy Uninvolved brain Dmean ≤18 Gy Uninvolved brain Dmean ≤18 Gy
V12Gy ≤10 cc V20Gy ≤20 cc V20Gy ≤20 cc
Brainstem D0.03cc ≤15 Gy Brainstem D0.03cc ≤23.1 Gy Brainstem D0.03cc ≤31 Gy
D0.5cc ≤10 Gy D0.5cc ≤18 Gy D0.5cc ≤23 Gy
Optic pathway D0.03cc ≤10 Gy Optic pathway D0.03cc ≤17.4 Gy Optic pathway D0.03cc ≤25 Gy
D0.2cc ≤8 Gy D0.2cc ≤15.3 Gy D0.2cc ≤23 Gy
Cochlea D0.03cc ≤9 Gy Cochlea D0.03cc ≤17.1 Gy Cochlea D0.03cc ≤25 Gy
Spinal cord and D0.03cc ≤14 Gy Spinal cord and D0.03cc ≤21.9 Gy Spinal cord and D0.03cc ≤30 Gy
medulla medulla medulla
D0.35cc ≤10 Gy D0.35cc ≤18 Gy D0.35cc ≤23 Gy
Target planning dose
≥99.5% of the PTV receives ≥100% of PD
OARs organs at risk, PD prescribed dose, FSRS fractionated stereotactic radiosurgery
Dose Escalation [76, 77]. The most significant aspect of these tri-
als is that they were conducted either without
The median survival for GBM remains poor at chemotherapy or with nitrosourea-based chemo-
approximately 15–16 months in contemporary therapy instead of the use of temozolomide, as a
series, despite surgical intervention, conven- radiosensitizer. In a historical joint study of the
tional radiotherapy, chemotherapy, and in some RTOG and Eastern Cooperative Oncology Group
cases with TTFields [6–8, 71, 72]. Radiotherapy (ECOG), no survival benefit was observed with a
is typically aimed at treating the surgical bed, 10-Gy focal boost to 60-Gy WBRT [77].
enhancing residual tumor on MRI, and a margin However, a recent systematic review and meta-
of 1–2 cm around the surgical site [73]. In addi- analysis of 22 prospective trials demonstrated
tion to prolonging the time to tumor progression that dose-escalated radiotherapy (defined as the
and increasing survival, the addition of temo- equivalent dose in 2-Gy fractions) >60 Gy with-
zolomide to radiotherapy enhances its therapeu- out chemotherapy was more effective in terms of
tic efficacy with a synergistic effect [6]. Despite progression-free and overall survival compared
radiotherapy and temozolomide improving sur- to 60-Gy standard-dose radiotherapy without
vival, most of the recurrences are local, and chemotherapy [78]. In the contemporary era of
locoregional recurrences occur where the high- concurrent and adjuvant temozolomide, a pro-
est radiotherapy dose is delivered (e.g., >57 Gy spective phase I/II study was conducted to evalu-
or the 95% isodose line of prescribed 60 Gy) ate the efficacy of dose escalation using an IMRT
[74, 75]. technique [79]. They reported that temozolomide
Increasing the dose of radiotherapy, either was tolerated to a maximum of 75 Gy in 30 frac-
through concomitant escalation of the total dose tions. Interestingly, the probability of local fail-
or dose per fraction, represents approaches that ure decreased as dose escalation increased, and
have been tested to overcome hypoxia-related the median survival was 20.1 months, suggesting
resistance of GBM. Standard radiotherapy in a higher degree of efficacy compared to previous
postoperative setting currently employs a total studies. These findings paralleled results from
dose of 60 Gy in 30 fractions [6]. Dose-escalated prior studies of radiotherapy dose escalation
radiotherapy has not proven to be beneficial for without temozolomide, where local control
GBM patients in previous prospective studies improved with radiotherapy dose escalation [80,
81]. Most recently, the NRG Oncology BN001 membrane. The amount of N-acetyl aspartate
study was designed to test the hypothesis that (NAA), a protein found in neurons, is diminished
radiotherapy intensification with temozolomide when the neuronal environment is destroyed in
can prolong survival in patients with newly diag- tissue samples. The ratio of Cho to NAA (Cho/
nosed GBM based on the promising results of NAA) has been previously shown to be elevated
radiotherapy dose escalation combined with con- in patients with GBM and can serve as a bio-
current and adjuvant temozolomide. In this trial, marker to differentiate regions of tumor from
75 Gy was administered to the surgical bed, along normal healthy brain tissue [84, 85]. Based on
with residual enhancing tumor plus a volumetric these results, studies have assessed the potential
margin of 5 mm, with a differential dose painting of MRSI for radiotherapy planning in GBM
of 50 Gy in 30 fractions to the non-enhancing patients [85–87]. However, very few clinical tri-
tumor and a volumetric margin of 2 cm, along als have managed to integrate MRSI for treat-
with concurrent and adjuvant temozolomide [82]. ment of brain tumors prospectively.
Although no benefit to dose escalation was SPECTRO-GLIO is one such trial where regions
observed in the photon group (and proton of the brain with a ratio of Cho/NAA >2 received
therapy-specific results are pending at this time), an escalated radiation dose of 72 Gy; however,
a detailed pattern of failure analysis is needed to the results have not been reported yet [88]. In a
fully understand the potential and limitations of pilot trial at 3 institutions with 30 patients, the
that dose-escalation approach. authors assessed the feasibility, safety, and effi-
Instead of dose escalation to the postoperative cacy of MRI spectroscopy-guided RT dose esca-
tumor bed, MRgRT offers the ability to deliver a lation to 75 Gy for newly diagnosed GBMs [89].
more personalized form of dose escalation. If, as For radiotherapy, standard tumor volumes based
an example, MRI indicates that a tumor becomes on T2/FLAIR and T1 post-contrast MRIs were
increasingly cellular, metabolic, or angiogenic as treated in 30 fractions to 50.1 and 60 Gy, respec-
it goes through treatment, it may be appropriate tively. An additional high-risk volume based on
to consider RT dose escalation as a method of residual tumor and Cho/NAA ≥2 was treated to
treatment intensification. One way to answer this 75 Gy. At a median follow-up of 21.4 months, the
question is to use MRgRT to obtain median overall survival and progression-free sur-
multidimensional MRIs with high temporal fre- vival were 23.0 and 16.6 months, respectively.
quency with MR-Linac devices throughout treat- Eleven of 30 patients experienced grade 3 or
ment to reveal changes in tumor physiology that greater toxicities that were at least possibly due
may be useful in identifying treatment decisions to their treatment. Of the seven patients who
that are adaptive to tumor physiology. MRI-based experienced these by 9 months post-radiotherapy,
physiologic metrics that are used to predict most were attributable to temozolomide, and
patient outcomes should be examined in future only one case could potentially be ascribed to
studies to determine which changes and at what radiotherapy. The integration of MRgRT imaging
time points are most predictive of treatment out- into the treatment workflow may help to facilitate
come. Obtaining and evaluating these images this process for radiotherapy planning.
during radiotherapy will allow for early adapta- SRS for brain metastases can be delivered as
tion of treatment in an effort to improve survival either a single fraction (generally 18–24 Gy) or
outcomes. By utilizing the nuclear magnetic res- as FSRS ranging from 24 to 30 Gy in 3 frac-
onance phenomenon in conjunction with mag- tions, or 25–35 Gy in 5 fractions [90]. The
netic resonance spectroscopic imaging (MRSI), RTOG 90-05 dose-escalation trial was the initial
it is possible to identify the chemical composi- prospective study to understand the maximum
tion of tissues noninvasively [83]. Rapidly prolif- tolerated single- fraction doses for SRS for
erating cells are found to be enriched with recurrent primary brain tumors and brain metas-
choline-containing compounds (Cho), which is a tases, although the relationship between dose
structure in the phospholipid bilayer of the cell and local control was not reported [20]. Large
388 T. Kutuk et al.
retrospective series have supported the RTOG On-Table Adaptation

90-05 dose prescriptions, for example, in one
series of 3034 brain metastases ≤2 cm treated ViewRay MRIdian™
with SRS, a dose of 24 was associated with
reduced local failure compared to 19–23 Gy During daily setup for treatment, a volumetric
(dose de-escalation) [91]. Further dose-escala- MRI is obtained to visualize the target volume
tion efforts with single-fraction SRS, however, and OARs. Tumor target volumes are registered
are limited due to the perceived increased risk of rigidly from the simulation MRI to the daily
radiation necrosis. A recent systematic review localization MRI scan, and OAR contours are
including 10 studies with 720 brain metastases propagated to the MRI scan of the day using
treated with FSRS showed that the tumor con- deformable image registration. GTV and OAR
trol probability increases with the higher BED10. contours within 3 cm radially of the PTV and
Local control at 1 and 2 years was 84% and 2 cm cranial or caudal to the PTV are edited to
73%, respectively, for lesions treated with a reflect the anatomy of the day. Online plans are
BED10 of at least 57.6 Gy compared to 72% and reoptimized with the TPS, using the same beam
60% for those treated with BED10 less than parameters and optimization objectives.
57.6 Gy [92]. In one of the few randomized tri- Following this, patient-specific quality assur-
als, Kim et al. compared the local control and ance is performed with the patient in treatment
radiation necrosis rate with 24, 27, or 30 Gy in 3 position, using an independent Monte Carlo
fractions (8, 9, or 10 Gy per fraction, respec- dose calculation followed by a gamma analysis
tively) for brain metastases larger than 3 cm comparison with dose distribution from the
treated on the Gamma Knife® [93]. In their MRIdian™ planning system [38, 98]. After plan
dose-escalation study, 27 Gy in 3 fractions approval, treatment delivery occurs using visual
appeared to be the best regimen of FSRS feedback provided to the patient on a monitor
because 30 Gy in 3 fractions was associated projecting the real-time MRI.
with an increased risk of radiation necrosis
(hazard ratio, 7.2, p = 0.047), and 24 Gy in 3
fractions was associated with an increased risk Elekta Unity™
of local failure (hazard ratio, 2.5, p = 0.037).
Despite the early postoperative SRS clinical There are two primary modes for daily adapta-
trials utilizing single-fraction SRS, which yielded tion on the Unity™ system: adapt-to-position
modest disease recurrence rates (15–39%), vari- (ATP) and adapt-to-shape (ATS). ATP uses a
ous subsequent studies have explored the poten- single MRI sequence (typically a T1-weighted
tial of postoperative FSRS to enhance local 3D non-contrast scan, but a volumetric FLAIR
control rates and reduce the risk of radiation scan can also be used) that is fused to the MRI
necrosis [23, 94–96]. The consensus guidelines simulation, providing rigid deformations for the
of the International Society of Stereotactic target and OARs. A virtual couch shift is applied
Radiosurgery recommend postoperative dose and to adjust for differences in patient positioning
fractionation schedules ranging from approxi- [70]. The same IMRT beam angles are used,
mately 30 to 50 Gy EQD210, 50 to 70 EQD25, and reoptimization is performed, and the plan is
70 to 90 EQD22 [97]. The A071801 trial evaluated using the green/yellow/red stoplight
(NCT04114981) recently completed accrual and system described in section “Contouring and
involved randomizing patients with resected Planning”. An example online adaptive interface
brain metastasis larger than 2 cm to either single- for a GBM plan on Unity system is shown in
fraction SRS or FSRS. The trial’s primary end- Fig. 20.7. Target and OAR contours are not man-
point was the time to local recurrence, and it is ually adjusted during an ATP fraction. An inde-
expected to generate further prospective compar- pendent dose check by medical physics is
ative data to aid in guiding clinical practice. performed to ensure the dose difference between
20 MR-Linac-Guided Adaptive Radiotherapy for Brain Tumors
Fig. 20.7 The figure illustrates the online adaptive interface for a glioblastoma imaging (MRI) reconstructed from the day’s MRI scan. The bottom-middle panel
multiforme (GBM) treatment plan for Unity™ system. The top panel shows the displays the optimization window, while the bottom-right panel shows the target
beam arrangement over the CTsim, along with the collimator arrangement for one and organ-at-risk (OAR) constraints. Once the plan is ready, the stoplight system
beam. The top-right panel displays the dose-volume histogram (DVH), which is (green/yellow/red) will indicate the plan’s quality based on ideal, tolerance, and
currently empty. The bottom-left panel shows a sagittal magnetic resonance unacceptable numbers
389
390 T. Kutuk et al.
the treatment planning software and the physics- (NCT03137888)—This trial is investigating
independent verification software is within 4% the use of MR spectroscopy to guide radia-
[99]. During the contouring, plan optimization, tion dose escalation in patients with newly
and beam-on phases, additional MRI sequences diagnosed GBM.
such as diffusion, perfusion, magnetization (b) UNIty-Based MR-Linac Guided AdapTive
transfer, quantitative relaxometry, and functional RadiothErapy for High-GraDe Glioma: A
MRI can be acquired. Cine MRI during beam-on Phase 2 Trial (UNITED) (NCT04726397)—
can also be acquired to monitor for motion but This trial is studying the use of a small-
has been not found to be helpful due to the lack margin (5 mm from GTV to CTV), daily
of motion with proper immobilization. The ini- adaptive MRgRT approach for GBM using
tial experience using ATP in ten glioma patients the Unity™ system.
on the Unity™ reported a mean fraction time of (c) UNITy-BasED MR-Linac Adaptive
37.3 min, and all fractions met the independent Simultaneous Integrated Hypofractionationed
physics tolerance check [99]. Boost Trial for High-Grade Glioma in the
The ATS mode for intracranial radiotherapy Elderly (UNITED2) (NCT05565521)—This
was developed for the UNITED (NCT04726397) study is using a small-margin approach as
clinical trial. Once a week during a course of well when treating to 40 Gy/15 fractions for
fractionated radiotherapy for GBM, gadolinium elderly GBM, but adding a simultaneously
contrast is injected (outside the Unity™), the integrated boost to 52.5 in 15 fractions to the
patient is positioned, and a 3D T1 + c sequence, GTV only (with a PTV expansion).
as well as a volumetric FLAIR sequence, is
acquired. The MRI images are fused to the MRI
simulation images, and either rigid or deformable References
registrations can be used for the targets and
OARs. The treating radiation oncologist is pres- 1. Lah TT, Novak M, Breznik B. Brain malignancies:
glioblastoma and brain metastases. Semin Cancer
ent and edits the GTV (on the T1 + c images) and Biol. 2020;60:262–73.
OARs; a 5-mm volumetric expansion is applied 2. Ostrom QT, Price M, Neff C, Cioffi G, Waite KA,
respecting anatomical boundaries, and the CTV Kruchko C, et al. CBTRUS statistical report: pri-
can then be manually expanded beyond 5 mm to mary brain and other central nervous system tumors
diagnosed in the United States in 2015-2019. Neuro
incorporate involved areas of FLAIR hyperin- Oncol. 2022;24(Suppl 5):v1–v95.
tense signal. A new radiation plan is optimized 3. Sacks P, Rahman M. Epidemiology of brain metasta-
(using the same IMRT beam angles as used dur- ses. Neurosurg Clin N Am. 2020;31(4):481–8.
ing the reference plan optimization) and reviewed 4. Seker-Polat F, Pinarbasi Degirmenci N, Solaroglu I,
Bagci-Onder T. Tumor cell infiltration into the brain
by the radiation oncologist. This new ATS plan in glioblastoma: from mechanisms to clinical per-
then becomes the reference plan for the subse- spectives. Cancers. 2022;14(2):443.
quent four ATP fractions until the next ATS frac- 5. Bell EH, Pugh SL, McElroy JP, Gilbert MR, Mehta
tion is developed a week later. M, Klimowicz AC, et al. Molecular-based recur-
sive partitioning analysis model for glioblastoma in
the Temozolomide era: a correlative analysis based
on NRG oncology RTOG 0525. JAMA Oncol.
Other 2017;3(6):784–92.
6. Stupp R, Hegi ME, Mason WP, van den Bent MJ,
Taphoorn MJ, Janzer RC, et al. Effects of radio-
There are currently several clinical trials investi- therapy with concomitant and adjuvant temozolo-
gating the use of MRgRT for the treatment of mide versus radiotherapy alone on survival in
GBM. Some examples of ongoing trials include: glioblastoma in a randomised phase III study: 5-year
analysis of the EORTC-NCIC trial. Lancet Oncol.
2009;10(5):459–66.
(a) A multi-institutional pilot clinical trial of
7. Stupp R, Taillibert S, Kanner A, Read W, Steinberg
spectroscopic MRI-guided radiation dose D, Lhermitte B, et al. Effect of tumor-treating fields
escalation for newly diagnosed GBM plus maintenance Temozolomide vs maintenance
Temozolomide alone on survival in patients with 20. Shaw E, Scott C, Souhami L, Dinapoli R, Kline R,
glioblastoma: a randomized clinical trial. JAMA. Loeffler J, et al. Single dose radiosurgical treat-
2017;318(23):2306–16. ment of recurrent previously irradiated primary
8. Stupp R, Mason WP, van den Bent MJ, Weller M, brain tumors and brain metastases: final report of
Fisher B, Taphoorn MJ, et al. Radiotherapy plus con- RTOG protocol 90-05. Int J Radiat Oncol Biol Phys.
comitant and adjuvant temozolomide for glioblas- 2000;47(2):291–8.
toma. N Engl J Med. 2005;352(10):987–96. 21. Masucci GL. Hypofractionated radiation therapy for
9. Roa W, Brasher PM, Bauman G, Anthes M, Bruera E, large brain metastases. Front Oncol. 2018;8:379.
Chan A, et al. Abbreviated course of radiation ther- 22. Prabhu RS, Dhakal R, Vaslow ZK, Dan T, Mishra
apy in older patients with glioblastoma multiforme: MV, Murphy ES, et al. Preoperative radiosurgery
a prospective randomized clinical trial. J Clin Oncol. for resected brain metastases: the PROPS-BM mul-
2004;22(9):1583–8. ticenter cohort study. Int J Radiat Oncol Biol Phys.
10. Malmström A, Grønberg BH, Marosi C, Stupp R, 2021;111(3):764–72.
Frappaz D, Schultz H, et al. Temozolomide versus 23. Mahajan A, Ahmed S, McAleer MF, Weinberg JS, Li
standard 6-week radiotherapy versus hypofraction- J, Brown P, et al. Post-operative stereotactic radiosur-
ated radiotherapy in patients older than 60 years with gery versus observation for completely resected brain
glioblastoma: the Nordic randomised, phase 3 trial. metastases: a single-centre, randomised, controlled,
Lancet Oncol. 2012;13(9):916–26. phase 3 trial. Lancet Oncol. 2017;18(8):1040–8.
11. Roa W, Kepka L, Kumar N, Sinaika V, Matiello 24. Padhani AR, Liu G, Koh DM, Chenevert TL, Thoeny
J, Lomidze D, et al. International Atomic Energy HC, Takahara T, et al. Diffusion-weighted magnetic
Agency randomized phase III study of radiation resonance imaging as a cancer biomarker: consensus
therapy in elderly and/or frail patients with newly and recommendations. Neoplasia. 2009;11(2):102–25.
diagnosed glioblastoma multiforme. J Clin Oncol. 25. Ellingson BM, Malkin MG, Rand SD, Connelly JM,
2015;33(35):4145–50. Quinsey C, LaViolette PS, et al. Validation of func-
12. Kotecha R, Gondi V, Ahluwalia MS, Brastianos PK, tional diffusion maps (fDMs) as a biomarker for
Mehta MP. Recent advances in managing brain metas- human glioma cellularity. J Magn Reson Imaging.
tasis. F1000Res. 2018;7:F1000 Faculty Rev-1772. 2010;31(3):538–48.
13. Suh JH, Kotecha R, Chao ST, Ahluwalia MS, Sahgal 26. Hein PA, Eskey CJ, Dunn JF, Hug EB. Diffusion-
A, Chang EL. Current approaches to the manage- weighted imaging in the follow-up of treated high-
ment of brain metastases. Nat Rev Clin Oncol. grade gliomas: tumor recurrence versus radiation
2020;17(5):279–99. injury. AJNR Am J Neuroradiol. 2004;25(2):201–9.
14. Vogelbaum MA, Brown PD, Messersmith H, 27. Hamstra DA, Rehemtulla A, Ross BD. Diffusion
Brastianos PK, Burri S, Cahill D, et al. Treatment for magnetic resonance imaging: a biomarker for
brain metastases: ASCO-SNO-ASTRO guideline. J treatment response in oncology. J Clin Oncol.
Clin Oncol. 2022;40(5):492–516. 2007;25(26):4104–9.
15. Gondi V, Bauman G, Bradfield L, Burri SH, Cabrera 28. Jabehdar Maralani P, Myrehaug S, Mehrabian H,
AR, Cunningham DA, et al. Radiation therapy for Chan AKM, Wintermark M, Heyn C, et al. Intravoxel
brain metastases: an ASTRO clinical practice guide- incoherent motion (IVIM) modeling of diffusion MRI
line. Pract Radiat Oncol. 2022;12(4):265–82. during chemoradiation predicts therapeutic response
16. Chang EL, Wefel JS, Hess KR, Allen PK, Lang FF, in IDH wildtype glioblastoma. Radiother Oncol.
Kornguth DG, et al. Neurocognition in patients with 2021;156:258–65.
brain metastases treated with radiosurgery or radio- 29. Larsson C, Groote I, Vardal J, Kleppesto M, Odland
surgery plus whole-brain irradiation: a randomised A, Brandal P, et al. Prediction of survival and progres-
controlled trial. Lancet Oncol. 2009;10(11):1037–44. sion in glioblastoma patients using temporal perfu-
17. Aoyama H, Shirato H, Tago M, Nakagawa K, Toyoda sion changes during radiochemotherapy. Magn Reson
T, Hatano K, et al. Stereotactic radiosurgery plus Imaging. 2020;68:106–12.
whole-brain radiation therapy vs stereotactic radiosur- 30. Pirzkall A, McKnight TR, Graves EE, Carol MP,
gery alone for treatment of brain metastases: a random- Sneed PK, Wara WW, et al. MR-spectroscopy guided
ized controlled trial. JAMA. 2006;295(21):2483–91. target delineation for high-grade gliomas. Int J Radiat
18. Shinde A, Akhavan D, Sedrak M, Glaser S, Amini Oncol Biol Phys. 2001;50(4):915–28.
A. Shifting paradigms: whole brain radiation therapy 31. Mehrabian H, Chan RW, Sahgal A, Chen H, Theriault
versus stereotactic radiosurgery for brain metastases. A, Lam WW, et al. Chemical exchange saturation
CNS. CNS Oncol. 2019;8(1):CNS27. transfer MRI for differentiating radiation necro-
19. Kocher M, Soffietti R, Abacioglu U, Villa S, Fauchon sis from tumor progression in brain metastasis-
F, Baumert BG, et al. Adjuvant whole-brain radiother- application in a clinical setting. J Magn Reson
apy versus observation after radiosurgery or surgical Imaging. 2022;57(6):1713–25.
resection of one to three cerebral metastases: results 32. Chan RW, Lawrence LSP, Oglesby RT, Chen H,
of the EORTC 22952-26001 study. J Clin Oncol. Stewart J, Theriault A, et al. Chemical exchange
2011;29(2):134–41. saturation transfer MRI in central nervous system
392 T. Kutuk et al.
tumours on a 1.5 T MR-Linac. Radiother Oncol. dynamics during concurrent chemoradiation for glio-
2021;162:140–9. blastoma: a prospective serial imaging study. Int J
33. Maralani PJ, Myrehaug S, Mehrabian H, Chan AK, Radiat Oncol Biol Phys. 2021;109(3):736–46.
Wintermark M, Heyn C, et al. ADC, D, f dataset 45. Mesny E, Jacob J, Culot F, Calugaru V, Jenny C, Fonti
calculated through the simplified IVIM model, with B, et al. Optic nerve motion and gaze direction: their
MGMT promoter methylation, age, and ECOG, in 38 impact on intraorbital tumor radiotherapy. Cancer
patients with wildtype IDH glioblastoma. Data Brief. Radiother. 2022;26(5):678–83.
2021;35:106950. 46. Qing K, Nie K, Liu B, Feng X, Stone JR, Cui T, et al.
34. Chan RW, Chen H, Myrehaug S, Atenafu EG, Stanisz The impact of optic nerve movement on intracranial
GJ, Stewart J, et al. Quantitative CEST and MT at radiation treatment. Front Oncol. 2022;12:803329.
1.5T for monitoring treatment response in glioblas- 47. Liney GP, Whelan B, Oborn B, Barton M, Keall
toma: early and late tumor progression during chemo- P. MRI-linear accelerator radiotherapy systems. Clin
radiation. J Neurooncol. 2021;151(2):267–78. Oncol (R Coll Radiol). 2018;30(11):686–91.
35. Karami E, Soliman H, Ruschin M, Sahgal A, 48. Bonert M, Schneider M, Solyanik O, Hellbach K,
Myrehaug S, Tseng CL, et al. Quantitative MRI bio- Bondesson D, Gaass T, et al. Diagnostic accuracy
markers of stereotactic radiotherapy outcome in brain of magnetic resonance imaging for the detection of
metastasis. Sci Rep. 2019;9(1):19830. pulmonary nodules simulated in a dedicated porcine
36. Mehrabian H, Detsky J, Soliman H, Sahgal A, Stanisz chest phantom. PLoS One. 2020;15(12):e0244382.
GJ. Advanced magnetic resonance imaging tech- 49. Kluter S. Technical design and concept of a 0.35 T
niques in management of brain metastases. Front MR-Linac. Clin Transl Radiat Oncol. 2019;18:98–101.
Oncol. 2019;9:440. 50. Dajani S, Hill VB, Kalapurakal JA, Horbinski CM,
37. Kim MM, Sun Y, Aryal MP, Parmar HA, Piert M, Nesbit EG, Sachdev S, et al. Imaging of GBM in the
Rosen B, et al. A phase 2 study of dose-intensified age of molecular markers and MRI guided adaptive
chemoradiation using biologically based target vol- radiation therapy. J Clin Med. 2022;11(19):5961.
ume definition in patients with newly diagnosed 51. Huang TY, Huang IJ, Chen CY, Scheffler K,
glioblastoma. Int J Radiat Oncol Biol Phys. Chung HW, Cheng HC. Are TrueFISP images T2/
2021;110(3):792–803. T1-weighted? Magn Reson Med. 2002;48(4):684–8.
38. Chuong MD, Bryant J, Mittauer KE, Hall M, Kotecha 52. Schmitt P, Griswold MA, Jakob PM, Kotas M, Gulani
R, Alvarez D, et al. Ablative 5-fraction stereotactic V, Flentje M, et al. Inversion recovery TrueFISP:
magnetic resonance-guided radiation therapy with quantification of T(1), T(2), and spin density. Magn
on-table adaptive replanning and elective nodal irra- Reson Med. 2004;51(4):661–7.
diation for inoperable pancreas cancer. Pract Radiat 53. Manon R, Hui S, Chinnaiyan P, Suh J, Chang E,
Oncol. 2021;11(2):134–47. Timmerman R, et al. The impact of mid-treatment
39. Kutuk T, Herrera R, Mustafayev TZ, Gungor G, MRI on defining boost volumes in the radiation treat-
Ugurluer G, Atalar B, et al. Multi-institutional out- ment of glioblastoma multiforme. Technol Cancer
comes of stereotactic magnetic resonance image Res Treat. 2004;3(3):303–7.
guided adaptive radiation therapy with a median 54. Tsien C, Gomez-Hassan D, Ten Haken RK, Tatro D,
biologically effective dose of 100 Gy(10) for Junck L, Chenevert TL, et al. Evaluating changes in
non-bone oligometastases. Adv Radiat Oncol. tumor volume using magnetic resonance imaging
2022;7(6):100978. during the course of radiotherapy treatment of high-
40. Kutuk T, McCulloch J, Mittauer KE, Romaguera T, grade gliomas: implications for conformal dose-
Alvarez D, Gutierrez AN, et al. Daily online adaptive escalation studies. Int J Radiat Oncol Biol Phys.
magnetic resonance image (MRI) guided stereotactic 2005;62(2):328–32.
body radiation therapy for primary renal cell cancer. 55. Mehta S, Gajjar SR, Padgett KR, Asher D, Stoyanova
Med Dosim. 2021;46(3):289–94. R, Ford JC, et al. Daily tracking of glioblastoma
41. Ugurluer G, Mustafayev TZ, Gungor G, Atalar B, resection cavity, cerebral edema, and tumor vol-
Abacioglu U, Sengoz M, et al. Stereotactic MR-guided ume with MRI-guided radiation therapy. Cureus.
online adaptive radiation therapy (SMART) for 2018;10(3):e2346.
the treatment of liver metastases in oligometastatic 56. La Rosa A, Mittauer KE, Rzepczynski AE, Chuong
patients: initial clinical experience. Radiat Oncol J. MD, Kutuk T, Bassiri N, et al. Treatment of glio-
2021;39(1):33–40. blastoma using MRIdian(R) A3i BrainTx: imaging
42. Cao Y, Tseng CL, Balter JM, Teng F, Parmar HA, and treatment workflow demonstration. Med Dosim.
Sahgal A. MR-guided radiation therapy: transforma- 2023;48(3):127–33.
tive technology and its role in the central nervous sys- 57. Kolton K Jones SD, Maziero D, et al. MRI-guided
tem. Neuro Oncol. 2017;19(suppl_2):ii16–29. radiotherapy identifies early pseudoprogression
43. Maziero D, Straza MW, Ford JC, Bovi JA, Diwanji T, of glioblastoma. Preprint (Version 1) available at
Stoyanova R, et al. MR-guided radiotherapy for brain Research Square 2020.
and spine tumors. Front Oncol. 2021;11:626100. 58. Tsien C, Galbán CJ, Chenevert TL, Johnson TD,
44. Stewart J, Sahgal A, Lee Y, Soliman H, Tseng CL, Hamstra DA, Sundgren PC, et al. Parametric response
Detsky J, et al. Quantitating interfraction target map as an imaging biomarker to distinguish progres-
sion from pseudoprogression in high-grade glioma. J patients on an integrated high-field magnetic reso-
Clin Oncol. 2010;28(13):2293–9. nance imaging linear accelerator. Int J Radiat Oncol
59. Yang Y, Cao M, Sheng K, Gao Y, Chen A, Kamrava Biol Phys. 2017;98(3):699–708.
M, et al. Longitudinal diffusion MRI for treatment 71. Fabian D, Guillermo Prieto Eibl MDP, Alnahhas I,
response assessment: preliminary experience using an Sebastian N, Giglio P, Puduvalli V, et al. Treatment
MRI-guided tri-cobalt 60 radiotherapy system. Med of glioblastoma (GBM) with the addition of tumor-
Phys. 2016;43(3):1369–73. treating fields (TTF): a review. Cancers (Basel).
60. Salkeld AL, Hau EKC, Nahar N, Sykes JR, Wang 2019;11(2):174.
W, Thwaites DI. Changes in brain metastasis during 72. Kotecha R, Odia Y, Khosla AA, Ahluwalia MS. Key
radiosurgical planning. Int J Radiat Oncol Biol Phys. clinical principles in the management of glioblas-
2018;102(4):727–33. toma. JCO Oncol Pract. 2023;19:180–9.
61. Kutuk T, Tolakanahalli R, Williams A, Tom MC, 73. Tseng CL, Stewart J, Whitfield G, Verhoeff JJC, Bovi
Vadhan JD, Appel H, et al. Impact of MRI timing on J, Soliman H, et al. Glioma consensus contouring
tumor volume and anatomic displacement for brain recommendations from a MR-Linac International
metastases undergoing stereotactic radiosurgery. Consortium Research Group and evaluation of a
Neurooncol Pract. 2021;8(6):674–83. CT-MRI and MRI-only workflow. J Neurooncol.
62. Atalar B, Choi CY, Harsh GR 4th, Chang SD, Gibbs 2020;149(2):305–14.
IC, Adler JR, et al. Cavity volume dynamics after 74. Chan JL, Lee SW, Fraass BA, Normolle DP,
resection of brain metastases and timing of postresec- Greenberg HS, Junck LR, et al. Survival and fail-
tion cavity stereotactic radiosurgery. Neurosurgery. ure patterns of high-grade gliomas after three-
2013;72(2):180–5; discussion 5. dimensional conformal radiotherapy. J Clin Oncol.
63. Taunk NK, Oh JH, Shukla-Dave A, Beal K, Vachha 2002;20(6):1635–42.
B, Holodny A, et al. Early posttreatment assessment 75. Milano MT, Okunieff P, Donatello RS, Mohile
of MRI perfusion biomarkers can predict long-term NA, Sul J, Walter KA, et al. Patterns and timing of
response of lung cancer brain metastases to stereotac- recurrence after temozolomide-based chemoradia-
tic radiosurgery. Neuro Oncol. 2018;20(4):567–75. tion for glioblastoma. Int J Radiat Oncol Biol Phys.
64. Jakubovic R, Sahgal A, Soliman H, Milwid R, Zhang 2010;78(4):1147–55.
L, Eilaghi A, et al. Magnetic resonance imaging-based 76. Mallick S, Kunhiparambath H, Gupta S, Benson R,
tumour perfusion parameters are biomarkers predict- Sharma S, Laviraj MA, et al. Hypofractionated accel-
ing response after radiation to brain metastases. Clin erated radiotherapy (HART) with concurrent and
Oncol (R Coll Radiol). 2014;26(11):704–12. adjuvant temozolomide in newly diagnosed glioblas-
65. Tan H, Stewart J, Ruschin M, Wang MH, Myrehaug toma: a phase II randomized trial (HART-GBM trial).
S, Tseng CL, et al. Inter-fraction dynamics dur- J Neurooncol. 2018;140(1):75–82.
ing post-operative 5 fraction cavity hypofraction- 77. Nelson DF, Diener-West M, Horton J, Chang CH,
ated stereotactic radiotherapy with a MR LINAC: Schoenfeld D, Nelson JS. Combined modality
a prospective serial imaging study. J Neurooncol. approach to treatment of malignant gliomas—re-
2022;156(3):569–77. evaluation of RTOG 7401/ECOG 1374 with long-
66. Tseng CL, Eppinga W, Seravalli E, Hackett S, Brand term follow-up: a joint study of the Radiation Therapy
E, Ruschin M, et al. Dosimetric feasibility of the Oncology Group and the Eastern Cooperative
hybrid magnetic resonance imaging (MRI)-linac sys- Oncology Group. NCI Monogr. 1988;6:279–84.
tem (MRL) for brain metastases: the impact of the 78. Singh R, Lehrer EJ, Wang M, Perlow HK, Zaorsky
magnetic field. Radiother Oncol. 2017;125(2):273–9. NG, Trifiletti DM, et al. Dose escalated radiation
67. Wen N, Kim J, Doemer A, Glide-Hurst C, Chetty therapy for glioblastoma multiforme: an interna-
IJ, Liu C, et al. Evaluation of a magnetic resonance tional systematic review and meta-analysis of 22
guided linear accelerator for stereotactic radiosurgery prospective trials. Int J Radiat Oncol Biol Phys.
treatment. Radiother Oncol. 2018;127(3):460–6. 2021;111(2):371–84.
68. Choi CY, Chang SD, Gibbs IC, Adler JR, Harsh 79. Tsien CI, Brown D, Normolle D, Schipper M, Piert
GR 4th, Lieberson RE, et al. Stereotactic radiosur- M, Junck L, et al. Concurrent temozolomide and
gery of the postoperative resection cavity for brain dose-escalated intensity-modulated radiation therapy
metastases: prospective evaluation of target mar- in newly diagnosed glioblastoma. Clin Cancer Res.
gin on tumor control. Int J Radiat Oncol Biol Phys. 2012;18(1):273–9.
2012;84(2):336–42. 80. Tanaka M, Ino Y, Nakagawa K, Tago M, Todo
69. Kutuk T, Kotecha R, Tolakanahalli R, Wieczorek T. High-dose conformal radiotherapy for supratento-
DJJ, Lee YC, Ahluwalia MS, et al. Zero setup mar- rial malignant glioma: a historical comparison. Lancet
gin mask versus frame immobilization during gamma Oncol. 2005;6(12):953–60.
knife((R)) icon stereotactic radiosurgery for brain 81. Nakagawa K, Aoki Y, Fujimaki T, Tago M, Terahara
metastases. Cancers (Basel). 2022;14(14):3392. A, Karasawa K, et al. High-dose conformal radio-
70. Ruschin M, Sahgal A, Tseng CL, Sonier M, Keller therapy influenced the pattern of failure but did not
B, Lee Y. Dosimetric impact of using a virtual couch improve survival in glioblastoma multiforme. Int J
shift for online correction of setup errors for brain Radiat Oncol Biol Phys. 1998;40(5):1141–9.
394 T. Kutuk et al.
82. Gondi V, Pugh S, Tsien C, Chenevert T, Gilbert 91. Mohammadi AM, Schroeder JL, Angelov L, Chao
M, Omuro A, et al. Radiotherapy (RT) dose- ST, Murphy ES, Yu JS, et al. Impact of the radiosur-
intensification (DI) using intensity-modulated gery prescription dose on the local control of small
RT (IMRT) versus standard-dose (SD) RT with (2 cm or smaller) brain metastases. J Neurosurg.
Temozolomide (TMZ) in newly diagnosed glio- 2017;126(3):735–43.
blastoma (GBM): preliminary results of NRG 92. Baliga S, Garg MK, Fox J, Kalnicki S, Lasala PA,
oncology BN001. Int J Radiat Oncol Biol Phys. Welch MR, et al. Fractionated stereotactic radiation
2020;108(3):S22–3. therapy for brain metastases: a systematic review with
83. Nelson SJ, Vigneron DB, Dillon WP. Serial evaluation tumour control probability modelling. Br J Radiol.
of patients with brain tumors using volume MRI and 2017;90(1070):20160666.
3D 1H MRSI. NMR Biomed. 1999;12(3):123–38. 93. Kim KH, Kong DS, Cho KR, Lee MH, Choi JW,
84. Narayana A, Chang J, Thakur S, Huang W, Karimi S, Seol HJ, et al. Outcome evaluation of patients treated
Hou B, et al. Use of MR spectroscopy and functional with fractionated gamma knife radiosurgery for large
imaging in the treatment planning of gliomas. Br J (>3 cm) brain metastases: a dose-escalation study. J
Radiol. 2007;80(953):347–54. Neurosurg. 2019:1–10.
85. Pirzkall A, Li X, Oh J, Chang S, Berger MS, Larson 94. Brown PD, Ballman KV, Cerhan JH, Anderson SK,
DA, et al. 3D MRSI for resected high-grade gliomas Carrero XW, Whitton AC, et al. Postoperative ste-
before RT: tumor extent according to metabolic activ- reotactic radiosurgery compared with whole brain
ity in relation to MRI. Int J Radiat Oncol Biol Phys. radiotherapy for resected metastatic brain dis-
2004;59(1):126–37. ease (NCCTG N107C/CEC.3): a multicentre, ran-
86. Chang J, Thakur SB, Huang W, Narayana A. Magnetic domised, controlled, phase 3 trial. Lancet Oncol.
resonance spectroscopy imaging (MRSI) and brain 2017;18(8):1049–60.
functional magnetic resonance imaging (fMRI) for 95. Brennan C, Yang TJ, Hilden P, Zhang Z, Chan K,
radiotherapy treatment planning of glioma. Technol Yamada Y, et al. A phase 2 trial of stereotactic radio-
Cancer Res Treat. 2008;7(5):349–62. surgery boost after surgical resection for brain metas-
87. Parra NA, Maudsley AA, Gupta RK, Ishkanian F, tases. Int J Radiat Oncol Biol Phys. 2014;88(1):130–6.
Huang K, Walker GR, et al. Volumetric spectro- 96. Minniti G, Niyazi M, Andratschke N, Guckenberger
scopic imaging of glioblastoma multiforme radiation M, Palmer JD, Shih HA, et al. Current status and
treatment volumes. Int J Radiat Oncol Biol Phys. recent advances in resection cavity irradiation of brain
2014;90(2):376–84. metastases. Radiat Oncol. 2021;16(1):73.
88. Laprie A, Ken S, Filleron T, Lubrano V, Vieillevigne 97. Redmond KJ, De Salles AAF, Fariselli L, Levivier
L, Tensaouti F, et al. Dose-painting multicenter M, Ma L, Paddick I, et al. Stereotactic radiosurgery
phase III trial in newly diagnosed glioblastoma: the for postoperative metastatic surgical cavities: a criti-
SPECTRO-GLIO trial comparing arm A standard cal review and international stereotactic radiosurgery
radiochemotherapy to arm B radiochemotherapy with society (ISRS) practice guidelines. Int J Radiat Oncol
simultaneous integrated boost guided by MR spectro- Biol Phys. 2021;111(1):68–80.
scopic imaging. BMC Cancer. 2019;19(1):167. 98. Li HH, Rodriguez VL, Green OL, Hu Y, Kashani R,
89. Ramesh K, Mellon EA, Gurbani SS, Weinberg BD, Wooten HO, et al. Patient-specific quality assurance
Schreibmann E, Sheriff SA, et al. A multi-institutional for the delivery of (60)Co intensity modulated radia-
pilot clinical trial of spectroscopic MRI-guided radiation therapy subject to a 0.35-T lateral magnetic field.
tion dose escalation for newly diagnosed glioblas- Int J Radiat Oncol Biol Phys. 2015;91(1):65–72.
toma. Neurooncol Adv. 2022;4(1):vdac006. 99. Tseng CL, Chen H, Stewart J, Lau AZ, Chan RW,
90. Singh K, Saxena S, Khosla AA, McDermott MW, Lawrence LSP, et al. High grade glioma radiation
Kotecha RR, Ahluwalia MS. Update on the man- therapy on a high field 1.5 Tesla MR-Linac—work-
agement of brain metastasis. Neurotherapeutics. flow and initial experience with daily adapt-to-
2022;19(6):1772–81. position (ATP) MR guidance: a first report. Front
Oncol. 2022;12:1060098.
Treatment of Head and Neck
Cancers with MR-Linac
21
Julia Khriguian, Laila Gharzai, Jolien Heukelom,
Brigid McDonald, and Clifton D. Fuller
Introduction treatment toxicity. This chapter will explore the

current state of MRL use for the treatment of head
Advances in magnetic resonance imaging (MRI), and neck cancers and review the advantages, chal-
image-guided radiotherapy (IGRT), and radiation lenges, and potential impact on patient outcomes of
delivery, such as with MR-Linac (MRL) technol- this technology as well as future directions.
ogy, have revolutionized the field of radiation
oncology by providing superior soft-tissue contrast
imaging for better tumor and surrounding normal Disease Overview
structure delineation and real-time imaging during
radiotherapy for monitoring and treatment adapta- Epidemiology and Risk Factors
tion. Treatment of head and neck cancers poses a
unique challenge to clinicians as tumours are in Head and neck squamous cell carcinomas
close proximity to multiple critical organs. Precise (HNSCC) account for 90% of all head and neck
radiation delivery to the tumor while sparing cancers. HNSCC includes cancers that affect the
healthy tissues is thus paramount to avoid excess oral cavity, pharynx, larynx, and sinuses. In this
chapter, we will primarily discuss pharyngeal
and laryngeal HNSCC as these comprise the
J. Khriguian · B. McDonald · C. D. Fuller (*)
majority of patients treated with definitive radio-
Department of Radiation Oncology, The University
of Texas MD Anderson Cancer Center, therapy (RT) with or without chemotherapy. The
Houston, TX, USA estimated incidence of HNSCC varies between
e-mail: [email protected]; regions of the world and populations, with higher
[email protected];
rates seen in developing countries [1]. In the
[email protected]
United States, the estimated incidence of oral
L. Gharzai
cavity and pharyngeal HNSCC is 54,000 new
Memorial Hospital, Northwestern University cases per year with a mortality rate of 11,000
Feinberg School of Medicine, Chicago, IL, USA deaths per year [2]. The incidence of oral cavity
e-mail: [email protected] and pharyngeal HNSCC has been increasing by
J. Heukelom 1% annually, primarily driven by the increased
Department of Radiation Oncology (Maastro), incidence of oral human papillomavirus (HPV)
GROW School for Oncology and Reproduction,
infection [2]. HNSCC is also a male-predominant
Maastricht University Medical Centre,
Maastricht, Netherlands disease with a 2:1 to 4:1 ratio [2] with the largest
e-mail: [email protected] difference in incidence being for oropharyngeal
396 J. Khriguian et al.
SCC (OPSCC). The average age at diagnosis of treatment modalities may carry different toxic-
patients with HPV-related OPSCC is also lower ity profiles. Surgery is more commonly used for
than that for non-HPV-related (mean of 56 versus the treatment of early-stage HNSCC and may
63 years old) [3]. In the United States, a 2% per involve the removal of the tumor and surround-
year increase in HPV-related HNSCC incidence ing tissue including elective nodal dissection of
was seen in men and 1% in women from 2010 to nodal basins at high risk of harboring micromet-
2019 [2]. Conversely, there was a 0.8% per year astatic nodal disease. In advanced disease, sur-
decline in smoking-associated HNSCC [2]. gery may be more extensive and combined with
In the United States, it is estimated that 70% RT and/or chemotherapy. Due to the morbidity
of OPSCC are caused by HPV infection [4] with that may be associated with extensive surgeries
the majority being from HPV16, which often needed to treat advanced disease, definitive
leads to an overexpression of the p16 protein. treatment with RT and chemotherapy alone may
HPV-related OPSCC is distinct from the rest of be preferred.
HNSCC with better treatment response and prog- The most commonly used definitive RT dose
nosis, and this has led to changes in OPSCC stag- regimen for HNSCC is 69.96–70 Gy in 33–35
ing accounting for HPV status in the new fractions delivered daily 5 days a week. In
American Joint Committee on Cancer eighth edi- early-stage cancers, patients are typically
tion. Recent evidence also shows that nearly 20% treated with RT alone; in more advanced dis-
of hypopharyngeal HNSCC are HPV-associated eases, patients typically receive concomitant
and may carry a better prognosis [5], although platinum-based chemotherapy (usually cispla-
further investigation is warranted. In fact, non- tin) [10, 11]. Induction chemotherapy may also
OPSCC p16-positive tumors in the head and neck be used for stage III/IV nasopharyngeal cancer
may also, in general, carry a better prognosis [6– [12], hypopharyngeal cancer [13], or when
8], although further investigation is warranted. In tumor shrinkage may be needed prior to local
the setting of nasopharyngeal HNSCC, Epstein- therapy (i.e., decreasing treatment volumes if
Barr virus (EBV) infection is a major risk factor, very extensive disease or near critical organs at
and is seen more commonly in endemic coun- risk (OAR)). In cisplatin-ineligible patients,
tries. The most important risk factors for non- carboplatin [14], docetaxel [15], cetuximab
HPV-related HNSCC are tobacco smoking and [16], or accelerating RT to six treatments per
alcohol consumption. This also has led to the week instead of five may be considered [17]. In
theory of field cancerization in HNSCC [9] due the postoperative setting, 60–66 Gy in 30–33
to these common etiologic factors, and high rates fractions is given depending on pathologic risk
of development of secondary HNSCC. Many factors and margin status. In cases with posi-
other environmental and lifestyle risk factors tive margins or extracapsular extension (ECE),
exist as well, depending on the site of HNSCC, cisplatin-based chemotherapy is added con-
but are beyond the scope of this chapter. comitantly to adjuvant RT. The mainstay ther-
apy for oral cavity and sinus SCC is surgery
with or without postoperative RT or chemoRT
Current Treatment Paradigms depending on high-risk pathological features
for HNSCC such as advanced T stage, N stage, presence of
ECE, close or positive margins, perineural
The treatment of HNSCC typically involves a invasion (PNI), and high-grade disease. In
multidisciplinary approach that includes sur- cases of laryngeal SCC with extensive thyroid
gery, RT, and/or chemotherapy. The choice of cartilage invasion, up-front total laryngectomy
treatment depends on the stage and location of and neck dissection may be preferred prior to
the tumor, the patient’s performance status and chemoRT due to high risk of local failures with
overall health, and their preferences as different chemoRT.
21 Treatment of Head and Neck Cancers with MR-Linac 397
hanging Landscape of HNSCC

C spread as well as understanding of its unique RT
Treatments planning complexities. HNSCC subtypes are
highly diverse with different molecular profiles
The landscape of HNSCC treatments is changing and response to treatment. Although HNSCC RT
as a result of a better understanding of the molec- doses and volumes have been relatively stable
ular and genetic mechanisms underlying this het- and very similar throughout most head and neck
erogeneous disease. One of the key discoveries in sites for the past few decades, there is an increas-
HNSCC has been the recognition that a significant ing body of evidence and potentially an upcom-
proportion of these tumors, particularly arising ing paradigm shift in favor of dose and/or volume
from the oropharynx, are caused by HPV infec- adaptation, particularly in the setting of HPV-
tion. This has led to a better risk stratification of related oropharyngeal cancers. There is also an
the disease as these tumors respond better to ther- increasing awareness of the need for highly con-
apy and carry a better prognosis [18]. Another formal and precise RT techniques in order to
important discovery is the recognition of the role decrease dose to OARs and limit RT-related tox-
of EBV in the development of nasopharyngeal icities and their consequent impact on quality of
cancer and the use of plasma EBV DNA titers to life.
follow treatment response and potentially drive The use of MRL in the setting of HNSCC has
treatment decisions. Both HPV- and EBV-derived not yet been well-established and data is still
biomarkers are currently being investigated in maturing. However, two major areas of improve-
treatment de-escalation studies as these tumors ment in HNSCC treatment may be satisfied by
may be more responsive to standard therapy. this technique—the ability to provide high-
Additionally, the identification of genetic muta- quality anatomic and functional imaging during
tions and other molecular changes as well as the treatment allowing for a better understanding of
discovery of the tumor microenvironment (TME) this heterogeneous group of tumors and the
has led to the investigation of targeted therapies opportunity for volume and/or dose adaptation
and immunotherapy in the setting of based on tumor response allowing for an increase
HNSCC. These discoveries, along with the tech- in therapeutic ratio.
nologic advances with less invasive surgeries
such as transoral robotic surgery and more pre-
cise RT with 3D planning with CT and MRI, Heterogeneity of HNSCC
intensity-modulated RT (IMRT), and volumetric
modulated arc therapy (VMAT), as well as IGRT Many discoveries in the past two decades have
with daily kV imaging, cone beam CT (CBCT), led us to understand that HNSCC constitute a
and now MRL, have led to the development of heterogeneous group of tumors with varying
more effective and personalized treatments for genetic and molecular features, locations, and
HNSCC. radiosensitivity. Although current standards of
treatment have not yet changed to account for
these features, many trials are currently under-
Overcoming Challenges way to investigate treatment personalization by
either revisiting the role of different systemic
The head and neck region is a small anatomic therapies or adapting RT doses and/or volumes.
area containing many structures responsible for One way to better understand the different behav-
vital functions such as breathing, swallowing, iors of HNSCC is through pre-, intra-, and post-
and phonation. Limiting dose to OARs is particu- therapy imaging—an endeavor that may be
larly challenging due to the close proximity of achieved through MR-guided RT (MRgRT).
targets and high-dose areas to these structures. Treatment personalization may also be facilitated
Consequently, RT for HNSCC requires a strong with MRgRT as tumor shrinkage may lead to vol-
knowledge of anatomy and pattern of disease ume adaptation and functional sequences such as
diffusion-weighted (DWI), and perfusion imag- Another particularity of HNSCC RT plan-

ing may provide additional information on tumor ning is the use of different prescription doses
kinetics allowing for dose de-escalation. depending on risk of disease involvement. In
general, the gross primary and nodal disease
receive 70 Gy in 35 fractions, and elective nodal
Complex Anatomy basins receive 56 Gy in 35 fractions (or equiva-
lent dose) using either a simultaneous boost
The head and neck region has a complex anatomy (SIB) or sequential cone-down method to deliver
and contains many radiosensitive OARs, such as higher tumoricidal doses to gross disease or
the salivary glands, mandible, pharyngeal con- areas at increased risk of involvement. Some
strictor muscles, brachial plexus, spinal cord, may also use an intermediate dose level of 59.5–
brainstem, and brain, in close proximity to high- 63 Gy in 35 fractions (or equivalent dose) to
dose RT target volumes. High contrast imaging cover areas at intermediate risk of harboring
with crisp anatomic resolution is needed for ade- disease or containing suspicious nodes. When
quate soft-tissue and tumor delineation and pri- using the SIB technique, different areas receive
mary tumor staging. MRI in particular may be different doses in the same amount of fractions
superior to CT in primary tumor delineation in (e.g., 2 Gy/fraction for high-risk volumes,
certain instances such as laryngeal cartilage inva- 1.8 Gy/fraction for intermediate-risk volumes,
sion [19] and perineural spread [20]. and 1.6 Gy/fraction for low-risk/elective vol-
Retropharyngeal node detection may also be supe- umes). This was possible with the advent of
rior with MRI and fludeoxyglucose (FDG)- IMRT and, particularly, VMAT. Conformality
positron emission tomography (PET/CT) [21]. may also be better with SIB technique [26]. As
Any nodal disease may also be better assessed by VMAT is not possible with MRL, however, a
PET/CT [22, 23]. Ideally, all three imaging modal- sequential cone-down method is more often
ities should be used for accurate target volume and employed with MRL treatments. Where there
OAR delineation [24]. may be an advantage for MRL is for dose-
painting based on functional multiparametric
MRI, but further investigation is needed to
Complex Planning determine feasibility and time points during the
course of RT to obtain this imaging.
As the head and neck region has a small surface
area, changes in weight and/or tumor size may
lead to changes in RT dose distribution warrant- Patterns of Failure of HNSCC
ing RT replanning and volume adaptation. With
IGRT, these changes can be monitored more Unfortunately, the most common type of failure
readily. Adaptive RT (ART) is also used when in HNSCC is local and/or regional with 65% of
tumors are close to or abutting critical OARs recurrences arising in the primary/local location
such as the base of skull where the brain, brain- and 36% in the nodal/regional location [27].
stem, and optic chiasm may not tolerate the pre- Most recurrences also occur within the first
scription RT dose. Volume adaptation as the 3 years post-therapy [27]. Another common
tumor shrinks may thus be particularly helpful in occurrence in patients with HNSCC is the devel-
limiting dose to these structures. The optimal opment of second primary tumors, with a 20-year
time for RT replanning is not well-established, cumulative risk of as high as 36% [28]. This is
but is frequently done at around fraction 15 [25] attributed to the relative high incidence of alcohol
as tumors start to have apparent shrinkage or consumption and smokers among patients with
weight loss becomes significant as toxicities such HNSCC in the pre-HPV era. These tumors may
as mucositis, dysphagia, xerostomia, and dysgeu- also initially be mistaken for recurrences until
sia develop and limit food intake. further investigation is undertaken (e.g., specific
pathological assessment of the clonal relation of positive OPSCC. Lastly, SBRT for early-stage
different tumors). glottic SCC has also shown promise in well-
There is increasing evidence for and use of selected patients, with early results showing fea-
reirradiation (re-RT) in HNSCC due to the high sibility [36–38]. However, potential toxicity
numbers of local/locoregional recurrences and concerns caused a phase I study to close early
second primaries in the head and neck region. due to severe toxicity with vocal cord ulceration
This is important as it stresses the up-front need and arytenoid cartilage necrosis in 2 patients out
to minimize dose to OARs as much as possible of 13 enrolled [38]. The University of Texas
and have conformal RT plans as some patients Southwestern Medical Center, however, has pub-
may require additional RT in the future. Altered lished encouraging preliminary results [36, 37]
fractionated RT [29, 30], stereotactic body RT and continues to recruit patients to the LT-SABR
(SBRT) [31], proton therapy [32], and brachy- trial (NCT03548285).
therapy [33] are all options that may be used in
cases of re-RT as well in order to limit dose to
OARs. urrent Literature on MRL Use
C
in HNSCC
he Role of Hypofractionation
T The current literature on MRL use in HNSCC is
and SBRT in HNSCC limited. Prior to widespread implementation, it
should be determined whether local and locore-
Hypofractionation refers to the use of fraction gional control outcomes can be maintained or
sizes of >2 Gy/fraction, and SBRT refers to a improved with this new technology and treatment
highly precise technique delivering high doses of adaptation. Additionally, confirmation is needed
RT in typically five or fewer fractions. Moderate to show that dose to OARs can be decreased and
hypofractionation has the strongest data in the that this ultimately results in less RT-related tox-
setting of nasopharyngeal and early glottic icities. Establishing these benefits and the non-
HNSCC where fraction sizes of 2.12–2.75 Gy/ inferiority of MRL will justify the use of this
fraction have been commonly used [34, 35]. The technique that is currently resource- and time-
use of SBRT has mostly been reserved for cases intensive. The standardization of IGRT schedules
of re-RT. The SOAR-HN (NCT03164460) phase and best MRI sequences for HNSCC would also
II trial is currently accruing patients to compare be beneficial.
SBRT to IMRT/proton therapy for recurrent Evidence is accumulating for the feasibility of
HNSCC in patients having received previous MRgRT for HNSCC, and a few ongoing trials are
RT. There are also ongoing trials evaluating dose- accruing patients globally, with the majority
escalated hypofractionation and incorporation of being for OPSCC. The MD Anderson MRL
SBRT boost in the definitive RT setting. The Development Working Group conducted a pro-
DEHART (NCT04477759) phase I trial is cur- spective in silico study assessing the feasibility
rently investigating the feasibility of hypofrac- and potential dosimetric advantages of
tionated dose escalation to target areas with MR-guided dose adaptation in HPV-associated
residual tumor and dose de-escalation to areas of OPSCC [39]. Two plans were generated—a stan-
tumor regression using MRgRT and concurrent dard one and an adaptive one with serial imaging
atezolizumab. The MRL-02 (NCT05160714) done every 2 weeks. The prescription dose to the
dose-finding phase I trial is assessing SBRT initial planning target volume (PTV) was
boost to high-risk subvolumes based on apparent 69.96 Gy in 33 fractions. For the adaptive plan, a
diffusion coefficient (ADC) changes, and the PTVadaptive was created (if detectable tumor
SHORT-OPC (NCT04178174) phase II trial is shrinkage) encompassing the residual tumor, and
comparing de-escalated chemoRT with SBRT the previously involved volume received a mini-
boost to GTV to standard chemoRT in HPV- mum of 50.16 Gy, which was the same dose used
for the elective nodal volumes. This adaptation mal response to therapy. In the latter arm, treat-
resulted in a decrease of 12.2 Gy to 95% of the ment effect will be evaluated after fraction 10
PTVinitial volume (70.7 Gy for the standard plan based on ADC response with dose escalation
and 58.5 Gy for the adaptive plan) [39]. Dose to being done to the non-responding clinical target
OARs was also reduced, and this translated into a volume (CTV).
decreased risk of RT-related toxicity, particularly
dysphagia, use of feeding tube 6 months post-RT,
and hypothyroidism when using the normal tis- Contouring and PTV
sue complication probability (NTCP) models
[39]. The next step is to confirm these findings Knowledge in anatomy and radiology are crucial
in vivo in the setting of a trial. McDonald et al. for accurate target and OAR contouring in
did an initial feasibility study with ten patients HNSCC. CT, MRI, and positron emission tech-
and described a clinical workflow for HNSCC nology CT (PET/CT) are often used to help guide
MRL [40] with the Unity system, which will be treatment and target delineation. CT is useful to
described later in this chapter. Chen et al. also evaluate bone involvement, and contrast helps
reported their experience with HNSCC MRL clearly visualize the tumor and any surrounding
with the ViewRay system in both primary and tissue changes. PET/CT may be used to confirm
recurrent/second primary disease re-RT setting or locate gross tumor volumes (GTV), including
[41, 42]. suspicious nodes [44–46]. MRI is particularly
The MR-ADAPTOR (NCT03224000), ART- useful for better visualization of soft tissues,
OPC (NCT04901234), MARTHA (NCT03972072), bone marrow involvement, retropharyngeal
and INSIGHT-2 (NCT04242459) trials are currently lymph node involvement, and presence and
accruing and evaluating MRL adapted RT for extent of perineural spread [47, 48], all of which
HNSCC. MR-ADAPTOR is a phase II non- affect staging and/or RT planning.
inferiority trial comparing standard IMRT with
69.96 Gy in 33 fractions versus weekly MR-guided
dose adaptation and elective volume de-escalation in MRI Sequences
low-risk HPV-associated OPSCC being treated with
RT alone, with the primary endpoint being locore- The most commonly used MRI sequences for tar-
gional control at 6 months [43]. The ART-OPC trial get delineation are T1-weighted (T1w) and T1
is a phase II trial comparing standard IMRT to contrast-enhanced (T1wC+), T2-weighted
MR-guided RT with a single mid-treatment volume (T2w), DWI, and dynamic contrast-enhanced
adaptation at week 3 in patients being treated with (DCE). T1w MRI sequences provide anatomical
chemoRT with AJCC eighth edition T3-4N0-three information and are typically used to evaluate the
tumors. The primary endpoint of this study is patient- tumor’s size and location, particularly when com-
reported dysphagia at 6 months post-RT. The bined with contrast administration and with fat
MARTHA trial is a single-arm study evaluating suppression [49], while T2w sequences can be
RT-induced xerostomia in the setting of daily used to evaluate the extent of tumor infiltration
MR-IGRT with once-weekly offline volume and associated edema (i.e., ECE in nearby mus-
adaptation in patients with stage II–IVB HNSCC cles such as the sternocleidomastoid). DWI
requiring bilateral neck RT either definitively or sequences provide information on tumor aggres-
postoperatively. Lastly, the INSIGHT-2 trial is a siveness and potential for lymph node involve-
volume and dose adaptation phase I/II study with ment by assessing hypercellularity via diffusion
three arms (HPV-associated OPSCC, HPV- restriction with low ADC values, and DCE is
negative OPSCC, and base of skull HNSCC) used to evaluate the tumor blood flow and vascu-
evaluating MR-guided ART with volume adapta- larity which may correlate with histological
tion at weeks 2 and 4 in all three arms, and dose grade [47]. Additionally, steady-state free preces-
escalation to HPV-negative disease with subopti- sion (SSFP) sequences provide better assessment
of cranial nerves, particularly at the base of skull contrast during RT, it is also crucial to under-
[48]. Lastly, the true fast imaging with steady- stand how the presence of gadolinium impacts
state free precession (TRUFI) sequence, which the radiation dose. In a study investigating
will be further explained later in this chapter and dosimetric differences between HNSCC treat-
is a mixture of T1w/T2w images, is often used ment plans generated on CT images without
with the ViewRay system and provides excellent contrast and with simulated gadolinium con-
soft-tissue contrast by suppressing background trast, the presence of gadolinium resulted in
tissue signal while enhancing the signal from median dose deviations of less than 1.2% for
tumor. all reported dosimetric parameters [57]. These
For MRL treatments, anatomical non-fat- results are encouraging for the safe and feasi-
suppressed T2w or T1w/T2w TRUFI scans are ble use of gadolinium contrast for daily MRL
used for setup imaging and plan adaptation in the treatments for HNSCC, but further in vivo
online MRL workflows for HNSCC [40–42, 50]. studies are needed.
However, depending on clinical needs, other In addition to anatomical sequences, func-
sequences with improved tumor contrast may be tional MRI techniques such as DWI are of inter-
of interest. Salzillo et al. optimized a fat- est for both treatment planning and response
suppressed T2w MRI sequence for HNSCC assessment using MR-Linacs. Changes in the
imaging on the 1.5T Unity MRL and found that associated quantitative imaging biomarker, the
the optimized sequence improved conspicuity of ADC, during RT have been shown to predict
the GTV, lymph nodes, and parotid glands com- ultimate oncological outcomes in HNSCC [58],
pared to the standard non-fat-suppressed T2w so DWI acquired on MR-Linacs may potentially
sequence used on the Unity system [51]. be used for treatment plan adaptation based on
Currently, the ViewRay system only offers the mid-therapy response. DWI has been acquired
TRUFI sequence for anatomical imaging, but in HNSCC for both the Unity 1.5T MR-Linac
other MRI sequences may be useful for disease [59, 60] and the ViewRay 0.35T MR-Linac
site-specific applications and should be explored [61], although the low-field strength of the
with this system. ViewRay system limits the signal-to-noise ratio
T1wC + MRI is another technique that can (SNR) and may pose difficulties for accurate
improve lesion contrast in HNSCC [52], but and robust ADC measurements. While the 1.5T
the safety and feasibility of gadolinium con- field strength of the Unity system matches that
trast agents used during the daily online adap- of many diagnostic MRI scanners, differences
tive workflow is still being explored. The in the gradient and receive coil hardware also
gadolinium in commonly used contrast agents limit SNR and have major implications for
has been shown in two separate studies to be acquisition of DWI on this system [62].
stable when irradiated to doses up to 100 Gy However, McDonald et al. showed that DWI
with 7 MV photons on the Unity MRL [53, sequences on the MR-Linac had similar in vivo
54]. In both of these studies, no un-chelated repeatability/reproducibility and phantom SNR
aqueous gadolinium(III) was found in the irra- values to a 1.5T MR sim [60]. Furthermore,
diated samples, which is toxic and accumu- geometric distortion and susceptibility artifacts
lates in the kidneys [55]. Petronek et al. are major concerns with echo planar imaging,
investigated the effects of gadolinium irradia- the most common type of DWI sequence. Other
tion in mouse glioblastoma xenograft models types of DWI sequences such as turbo spin
by injecting gadolinium contrast and irradiat- echo-based acquisitions may improve spatial
ing the tumors up to 18 Gy with a 220 kVp accuracy [63] and, when properly optimized,
small animal irradiator [56]. Based on mea- have been shown to have comparable or
sured blood urea nitrogen and creatinine lev- improved repeatability/reproducibility and SNR
els, no evidence of acute kidney toxicity was compared to echo planar imaging on the 1.5T
found. In addition to the safety of gadolinium MR-Linac [60].
Contouring coverage of nodal areas at risk of harboring

microscopic disease in the neck. Weissmann
In order to be able to safely adapt RT volumes et al. have proposed an elective nodal volume
during therapy without compromising tumor reduction contouring approach, but validation via
control, one must be able to confidently delineate incorporation into a prospective randomized trial
residual tumors. Strong knowledge for patterns is still pending [64]. For this chapter, the focus
of recurrence of HNSCC with nodal drainage lies on standard and widely used contouring
based on disease site involvement is also crucial. approaches based on published guidelines [65–
As opposed to many cancer sites treated with RT, 71]. Table 21.1 describes target volume contours
HNSCC treatments often encompass elective based on disease site. As the main benefits of
Table 21.1 HNSCC Target Volumes

Disease site Target volumes Notes
Oropharynx GTV = GTVp + GTVn • Contralateral LN RT may be
[66, 67] CTVhigh = GTV + 5 mm omitted in well-lateralized
Optional: CTVint = CTVhigh + at least involved lymph node tonsillar SCC with ipsilateral
levels nodes only and limited to the
CTVlow = CTVhigh + 5 mm and bilateral levels II–IV (include tonsillar fossa or anterior
level Ib if base of the tongue or level II node involved, include pillar without base of tongue
level V ± retropharyngeal nodes if N1 disease or level IV involved, invasion or soft palate
include retrostyloid nodes is level II node involved) invasion and >1 cm from
CTV = CTVhigh + CTVint + CTVlow midline [72–75]
PTV = CTV + 3–5 mm depending on institutional standards
Nasopharynx GTV = GTVp + GTVn • Unlike in the other disease
[71] CTVhigh = GTV + 5 mm sites, the use of CTVint is
CTVint = CTVhigh + 5 mm and high-risk regions including the commonly used for
entire nasopharynx, posterior 1/3 of the nasal cavity, nasopharyngeal SCC
pterygopalatine fossa, parapharyngeal space, anterior 1/2 of the • In the post-induction
clivus (or entirety if involved), foramen ovale and rotundum, 1/2 of chemotherapy setting, the
sphenoid sinus (or entirety if T3–T4 disease), posterior 1/2 of pre-induction volume is
ethmoid sinus, 1/3 of maxillary sinus and cavernous sinus if T3–T4 included in CTVint [76]
disease ± at least involved lymph node levels
CTVlow = CTVint + bilateral retropharyngeal nodes and levels
Ib–V (include retrostyloid nodes if nasal cavity/hard palate/
maxillary sinus and/or level II involved—may omit level Ib in
node-negative disease)
CTV = CTVhigh + CTVint + CTVlow
PTV = CTV + 3–5 mm depending on institutional standards
Larynx and GTV = GTVp + GTVn • Although guidelines have
hypopharynx CTVhigh = GTV + 5 mm and including the entire larynx from the been published including
[66, 67] top of the thyroid cartilage to the bottom of the cricoid cartilage smaller volumes for early
Optional: CTVint = CTVhigh + at least involved lymph node glottic SCC not including the
levels entire larynx, this has yet to
CTVlow = CTVhigh + 5 mm and bilateral levels II–IV (include be validated in a prospective
retropharyngeal nodes, levels Ib, V, and VI if hypopharyngeal study. The VOCAL-RT is
SCC, ipsilateral level Ib if base of the tongue or level II involved, currently open and addressing
level VI if subglottic extension and/or anterior soft-tissue extension this question of vocal cord
and/or if non-oncologic emergency tracheostomy was performed) versus complete larynx RT
CTV = CTVhigh + CTVint + CTVlow (NCT03759431)
* CTVint and CTVlow are not applicable for early glottic SCC • In the post-induction
where elective nodal RT is not performed chemotherapy setting, the
PTV = CTV + 3–5 mm for nodal volumes and up to 10 mm for pre-induction volume may be
primary tumor volume depending on institutional standards included in CTVint
(may also consider non-isotropic expansion for laryngeal SCC
primary, accounting for glottic movement with deglutition and
breathing [77, 78]
GTVp primary gross tumor volume, GTVn nodal gross tumor volume, CTVhigh high-risk clinical target volume, CTVint
intermediate-risk clinical target volume, CTVlow low-risk clinical target volume, PTV planning target volume
MR-guided ART are primarily suspected in the OAR Contouring

definitive setting, postoperative and oral cavity
SCC volumes will not be discussed in this chap- OAR contouring follows standard computed
ter. In certain centers, CTV intermediate risk may tomography (CT)-based guidelines [69], and all
be omitted in non-nasopharyngeal OAR structures seen on CT are readily identifi-
HNSCC. Further, as VMAT is not currently pos- able with MRI (either 0.35 T or 1.5 T). CT-based
sible with either of the commercially available consensus guidelines for OAR contouring exist
MRLs, it may be simpler to use two prescription [69] (Figs. 21.1 and 21.2) with concise descrip-
dose levels only. tions of the various OARs within the head/neck
region. Critical OARs in the head/neck region
include the spinal cord, brainstem, brain, optic
PTV Margins nerves, and chiasm. Important noncritical OARs
include the parotid glands, submandibular glands,
PTV margins are often specific to the treatment oral cavity, lips, mandible, pharyngeal constric-
machine, IGRT schedule/modality, and immobi- tors, larynx, esophagus, thyroid gland, and swal-
lization devices and consequently can vary lowing muscles. No MRI-based contouring
widely from one institution to another. In guidelines exist, and the nuances of identifying
HNSCC, tightening PTV margins is particularly structures on MRI as compared to CT have not
important in order to minimize dose to OARs in yet been defined.
close proximity to high-dose regions and decrease
treatment-related toxicity [79]. This is achieved
by limiting inter- and intra-fraction motion with Constraints
the use of rigid immobilization and daily onboard
image verification. MRL machines and daily The Quantitative Analysis of Normal Tissue
MR-guided IGRT provide an opportunity for Effects in the Clinic (QUANTEC) [69, 80]
PTV margin reduction as daily images provide review updated the classic paper by Emami et al.
superior soft-tissue contrast and inter-fraction [81] and provided more modern dose constraints
uncertainty is decreased. However, prior to estab- in the three-dimensional era. While this provided
lishing an optimal PTV margin expansion, MRL- important dose constraints relevant for treatment
compatible devices and appropriate IGRT planning the head/neck region, the QUANTEC
protocols/schedules must be standardized, which review did not provide a comprehensive over-
is an area of ongoing investigation. view of OAR constraints specific to the head/
neck region, including lacking guidance on
structures such as the mandible, brachial plexus,
AR/Dose-Volume Histogram (DVH)
O oral cavity, and submandibular glands. Ongoing
Constraints head/neck clinical trials by NRG/RTOG provide
additional input on modern dose constraints. The
OARs are a unique challenge in the head/neck three-arm NRG HN005 trial (NCT03952585)
region due to the number of structures important examining IMRT dose reduction and altering
for vital function such as breathing, phonation, systemic therapy offers additional guidance on
and swallowing. HNSCC treatment planning has critical structures (Table 21.2).
typically consisted of defining critical structures More recent studies in the IMRT era have sug-
such as the brainstem and optic nerves, which are gested that as compared to traditional
prioritized over tumor coverage, as well as lower 3D-conformal treatment planning, tighter dose
priority structures such as swallowing muscles constraints may be met with acceptable target
and/or salivary glands for which tumor coverage coverage and may minimize long-term toxicities.
is prioritized. A single-institution retrospective review from the
Fig. 21.1 Axial (left) and sagittal (right) view of the con- Jean Bourhis, Wilfried Budach, Cai Grau, Vincent
sensus delineations of the parotid glands (1), pharyngeal Grégoire, Marcel van Herk, Anne Lee, Philippe Maingon,
constrictor muscles (2), carotid arteries (3), spinal cord Chris Nutting, Brian O’Sullivan, Sandro V. Porceddu,
(4), mandible (5), extended oral cavity (6), buccal mucosa David I. Rosenthal, Nanna M. Sijtsema et al., CT-based
(7), lips (8), brain (9), chiasm (10), pituitary gland (11), delineation of organs at risk in the head and neck region:
brainstem (12), supraglottic larynx (13), glottic area (14), DAHANCA, EORTC, GORTEC, HKNPCSG, NCIC
cricopharyngeal inlet (15), cervical esophagus (16), and CTG, NCRI, NRG Oncology and TROG consensus
thyroid (17) (Reprinted from Radiotherapy and Oncology, guidelines, p. 8, Copyright (2015), with permission from
Vol. 117, Charlotte L. Brouwer, Roel J.H.M. Steenbakkers, Elsevier [69])
University of Pennsylvania [82] examining treat- of longer-term toxicities for which dose con-
ment plans in postoperative OPSCC found that straints are lacking, such as minimizing the effect
dose to various organs could be consistently of radiation-induced lymphedema or trismus
reduced, offering additional tighter dose con- through contouring structures such as the anterior
straints. For example, mean dose to the larynx neck or pterygoid muscles, respectively. The
can usually be reduced to a mean of 20 Gy, down improved soft-tissue delineation of the MRI may
from 40 Gy as suggested by NRG and 50 Gy as allow for improvements in understanding of dose
suggested by QUANTEC (Table 21.2). constraints to prevent these acute and late
Despite these constraints, there remain a num- toxicities.
ber of acute toxicities which are challenging to
manage or prevent during treatment, such as dys-
geusia [87]. With improvements in outcomes for Planning Challenges
patients with head/neck cancer in the modern era,
particularly in the setting of increased incidence The novel techniques utilized to integrate an MRI
of HPV-related OPSCC, the effect of longer-term into a linear accelerator present challenges for
toxicities on patients’ quality of life is becoming the creation of optimal radiation plans during
increasingly important. There remain a number treatment planning, which is already difficult due
a b
c d
Fig. 21.2 Axial CT slices showing the delineation of the Grau, Vincent Grégoire, Marcel van Herk, Anne Lee,
supraglottic larynx (A) (a), glottic area (B) (b), cricopha- Philippe Maingon, Chris Nutting, Brian O’Sullivan,
ryngeal inlet muscle (C) (c), and cervical esophagus (D) Sandro V. Porceddu, David I. Rosenthal, Nanna
(d). Other organs at risks visible are the submandibular M. Sijtsema et al., CT-based delineation of organs at risk
glands (1), pharyngeal constrictor muscles (2), carotid in the head and neck region: DAHANCA, EORTC,
arteries (3), brachial plexus (4), spinal cord (5), arytenoids GORTEC, HKNPCSG, NCIC CTG, NCRI, NRG
(6), and thyroid (7). (Reprinted from Radiotherapy and Oncology and TROG consensus guidelines, p. 8,
Oncology, Vol. 117, Charlotte L. Brouwer, Roel Copyright (2015), with permission from Elsevier [69])
J.H.M. Steenbakkers, Jean Bourhis, Wilfried Budach, Cai
to the number of critical structures in the head VMAT to balance these competing priorities,
and neck region. Treatment planning typically which offers an optimal method for sparing criti-
requires balancing competing priorities of tumor cal organs with faster treatment times [88] as
coverage and sparing nearby OAR. Many stan- compared to IMRT. However, VMAT cannot be
dard head and neck treatment plans now utilize delivered with an MRL due to the machine’s
Table 21.2 OAR/DVH constraints

Structure QUANTEC [80] NRG HN005 (NCT03952585) Single institution
Brainstem Max ≤54 Gy Max ≤50 Gy N/A
Spinal cord Max ≤50 Gy Max ≤45 Gy N/A
Brachial plexus N/A Max ≤66 Gy N/A
Parotid Mean ≤ 25 Gy Mean ≤ 26 Gy Mean ≤ 20 Gy [82]
Larynx Mean ≤ 50 Gy Mean ≤ 40 Gy Mean ≤ 20 Gy [82]
Pharynx/constrictors Mean ≤ 50 Gy Mean ≤ 45 Gy Mean ≤ 40 Gy [82]
Oral cavity N/A Mean ≤ 35 Gy Mean ≤ 20 Gy [82]
Esophagus Mean ≤ 34 Gy Mean ≤ 30 Gy Mean ≤ 20 Gy [82]
Submandibular gland N/A Mean ≤ 39 Gy Mean ≤ 30 Gy [82]
Cochlea Mean ≤ 45 Gy Mean ≤ 35 Gy Mean ≤ 35 Gy [83]
Mandible N/A Max ≤73.5 Gy V35 Gy < 30% [84]
Thyroid N/A Mean ≤ 50 Gy Mean ≤ 42 Gy [85]
V50 Gy ≤ 24% [86]
inability to rotate from the weight of the shield- ple optimization structures for each dose level.
ing present on the machine. Despite this limita- Typically at least four additional structures are
tion, early work on treatment plan comparison often drawn during treatment planning: two
with IMRT on an MRL suggests that high-quality structures of air and water to allow for density
treatment plans are feasible [89]. overrides and an optional tracking structure and
The number of critical organs present in the accompanying boundary structure. Thus, a rela-
head and neck region presents a significant chal- tively simple head and neck plan often rapidly
lenge. Most treatment plans require multiple tar- approaches the structure limits available for the
get volumes, nearby OARs, and numerous ViewRay system. Newer software updates on the
optimization structures, and the balancing of so ViewRay system may allow for modest increases
many structure constraints has implications for in these structure limits. Additionally, treatment
plan quality. For typical VMAT planning, patients plans utilizing the ViewRay system are typically
typically have 2–3 target volumes, approximately heterogeneous, with higher hot spots than seen in
10 OARs, 5–6 optimization structures (such as VMAT plans. Thus, numerous planning chal-
rings), and approximately 10 substructures for lenges exist that may degrade the quality of treat-
overlapping areas (e.g., left parotid minus a ment plans available utilizing the MRL system.
nearby target area). Finally, additional structures
are often drawn to account for areas of hot spots
and cold spots during planning, for a total of over Dose Escalation
30 structures for a typical plan.
For the ViewRay MRL system, in particular, Locoregional recurrences are the most common
treatment planning is currently limited to 32 type of treatment failure for HNSCC, with distant
structures. Additionally, optimization structures metastases seen less often as the first site of
within the ViewRay planning system must be relapse [90]. The locoregional failures predomi-
rule-based, and the rule-based operations do not nantly occur inside the high-dose volumes [91].
allow for complex operations. For example, to It has been stipulated that high-risk subvolumes
create a 1 cm ring around a target volume, three within the high-dose/GTV volume might be the
structures must be created: a planning volume cause of the recurrences, as primary tumors are
plus 0.2 cm, a planning volume plus 1.2 cm, and known to be heterogeneous. Identification and
then a third structure created by the subtraction of subsequent targeting of these subvolumes might
the previous two. For multiple dose levels, this lead to increased tumor control, although this is
rule-based requirement creates a need for multi- challenging and remains to be elucidated. The
determination of the exact location of these sub- uptake (i.e., determining if the increased dose is
volumes is critical but difficult. To achieve this, more important than the location of the boost).
identification of several possible high-risk sub- Several years after the start of accrual to the
volumes has been proposed using PET/CT scans. aforementioned trials, the MRL-02 study opened
Examples of these are hypoxic and/or FDG-avid (NCT05160714). This was a novel trial because
regions in which a tumor may be more likely to it is the first HNSCC dose escalation trial done on
recur [91, 92]. In the last decade, the increased the MRL. It aims to define the maximal tolerated
use of MRI has led to improved knowledge of dose in an ADC-based subvolume using an adap-
tumor characteristics and behavior. Therefore, tive SBRT boost. A real-time adaptive boost to
MR-guided RT trials may lead to the discovery of the ADC-high-risk volume is delivered once a
other high-risk subvolumes. Dose escalation to week in this study. This boost is gradually
such subvolumes (or to the whole primary tumor) increased using four dose levels of 5 × 5 Gy,
is the topic of several ongoing clinical trials. Both 5 × 5.5 Gy, 5 × 6 Gy, and 5 × 6.5 Gy in a classical
the boost subvolume and the optimal boost dose 3 + 3 dose-finding design where three patients
are currently being investigated. per radiation dose level are enrolled. The primary
endpoint is the dose-limiting toxicity, and
regional tumor control is a secondary objective.
Optimal Dose Another MRI-based phase I dose escalation
study is PIRATES study which investigates dose
To date, there is no consensus on the optimal escalation with particle therapy [96]. The study is
escalation dose and is a field of active investiga- a safety and feasibility trial of MRI-guided adap-
tion. A pioneering trial in the field of dose escala- tive particle RT [96]. Interestingly, this trial
tion in search of an optimal dose was conducted includes a boost regime (MRI guided at week
by Madani et al. [93]. In this phase I study, FDG- 5–7) as well as a hyperfractionation regime with
PET dose painting by number approach was twice-daily fractions. To what extent results from
used, and the maximum median tolerated dose MRI-directed dose escalation with proton ther-
was determined to be 80.9 Gy [93]. Higher dose apy are applicable to photon therapy on a MRL
was tolerated in the acute phase, but led to ulcers device is unclear. It is to be noted that this trial
4–10 months after treatment for several patients did not use MRL but instead used high-frequency
receiving the higher doses. Long-term follow-up offline MRI simulation studies for adaptation
did not show improved survival after treatment guidance.
with higher doses [94]. Lastly, the University of Michigan published a
Another FDG-PET-based dose escalation multicenter randomized phase II study assessing
study is the ARTFORCE trial (NCT01504815). physiologic MRI-based dose escalation in
In this trial, dose to a high FDG uptake subvol- patients with poor prognosis HNSCC and showed
ume is escalated to a maximum total dose of decreased locoregional failure with this tech-
84 Gy to 2% of the subvolume, with a minimum nique (40% versus 18% at 2 years) [97]. However,
dose of 70 Gy and a mean dose of approximately no significant difference was seen in disease-free
77 Gy [95]. Interestingly, this study does not survival, overall survival, or toxicity. The boost
boost the entire GTV but a subvolume defined by subvolume was based on low blood volume/ADC
high FDG uptake and does not randomize maps at fraction 10 of RT and was escalated to
between a boost directed at the area of high FDG 80 Gy [97].
uptake and an alternate subvolume determined The implicit assumption of many dose escala-
otherwise. However, this study design does not tion trials is that optimizing tumor control by
allow us to differentiate whether the effect of a dose escalation does not require increasing the
random boost might achieve the same results as dose uniformly across the primary tumor. Partly
specifically boosting the area of higher FDG driven by the attempt to spare surrounding OARs,
these trials will oftentimes only escalate dose to a fractions to primary and nodal tumor volume.
predefined subvolume assumed to harbor an Dose to HPV-associated OPSCC and base of
increased risk for tumor recurrence. Finding the skull HNSCC is to be adapted based on anatomi-
optimal subvolume to boost is an important topic cal and/or volumetric changes during therapy. Of
of investigation, as well as which imaging modal- note, this is not quite dose escalation for these
ity/sequences to use in order to determine such tumors. The dose to HPV-negative tumors is,
subvolumes. Dose escalation trials, particularly however, escalated based on a classification as
those in search of an optimal boost subvolume, responders vs. nonresponders, which is based on
have thus developed alongside the adoption of ADC response at week 2 of chemoRT. Responders
increasingly advanced imaging techniques. get target volumes adapted at weeks 2 and 4
For a long time, a PET/CT scan was the most based on anatomical changes, and nonresponders
informative scan regarding biological behavior of get an increase in dose/fraction to the CTV for
HNSCC. FDG has been and remains the most fractions 11–30. The INSIGHT-2 trial is also a
commonly used tracer for this disease. Studies good example of how dose escalation to volumes
have shown that tumor recurrences occur in the based on pre-treatment images can be substituted
FDG uptake areas [91]. Ensuingly, the investiga- by ART to strategically increase dose to the pri-
tion of a boost to this area seems logical. Both the mary tumor. This strategy can be used to direct
trial by Madani et al. [93] and the ARTFORCE dose with online (real-time) adaptive RT treat-
trial direct the high dose to the high FDG uptake ments, as done with MRL devices [102].
region. Moreover, in the original protocol, the Interestingly, these same devices have also
ARTFORCE study included a PET/CT scan with opened up possibilities for ART with the purpose
zirconium-labeled cetuximab (89Zr-Cetuximab) of dose de-escalation for tumors with a good
to investigate whether high uptake on that scan prognosis.
was predictive of tumor response to cetuximab,
and to assess whether low uptake might indicate
patients with tumors requiring higher RT dose for Dose De-escalation
disease control. This intervention, however, was
taken out of the protocol after it was amended. Dose de-escalation for HNSCC may be done on
Other PET/CT tracers that have been coined as two levels: to the nodal volumes and to the pri-
possibly indicative of high-risk areas for tumor mary tumor based on imaging and pathologic
recurrence and include areas of low Zr89- risk factors. Dose de-escalation to elective nodal
cetuximab uptake [98], FAZA [99], and F-miso fields is an area of active investigation. The gen-
[100, 101]. eral idea is that increased accuracy of diagnostic
Recent advancements and the increased avail- imaging and RT techniques may allow safe dose
ability and adoption of MRI scanners have cre- de-escalation to the elective nodal volumes with-
ated new opportunities for MRI-guided dose out compromising tumor control, potentially
escalation. Examples of MRI-guided dose esca- reducing treatment-related toxicity. Many of the
lation trials include the MRL-02 and INSIGHT elective doses that are being used currently are
trials for photon therapy and the PIRATES study based either on historical experience or on trials
for proton therapy. The MRL-02 and PIRATES that relied on physical examination to determine
studies have been described above. The the nodal status in the head and neck region.
INSIGHT-2 study (NCT04242459) includes Because clinical practice and available imaging
patients with OPSCC and base of skull tumors. It are more accurate nowadays, a node-negative
is a phase I/II trial consisting of three cohorts neck in the current era is thought to harbor a
(HPV-associated OPSCC, HPV-negative OPSCC, lower risk of involvement than node-negative
and base of skull HNSCC). Prescribed doses are necks from the era of trials done decades ago,
54 Gy in 30 fractions to nodal areas at risk of and, therefore, may be treated effectively with
harboring microscopic disease and 65 Gy in 30 less dose than was used at the time. Examples of
trials that investigated this hypothesis include the anatomical changes throughout RT. The exact
UPGRADE-RT trial [103] and a study by workflow differs between the Unity and ViewRay
Deschuymer et al. [104]. Additionally, for MRIdian systems and may vary among different
patients with HNSCC with a relatively good users of the same device, but many consider-
prognosis (i.e., HPV-associated OPSCC [18]), ations remain the same. Site-specific workflows
dose de-escalation to the primary tumor might be for HNSCC have been described by McDonald
feasible. This can be achieved by an up-front et al. [40] and Gupta et al. [50] for the Elekta
dose reduction to the target volumes or by imple- Unity and in two separate studies by Chen et al.
menting ART regimens where the treatment vol- [41, 42] for the ViewRay MRIdian cobalt-60 unit.
umes are cropped and adjusted according to For both systems, patients undergo CT and/or
anatomic and volumetric changes during RT with MR simulation for treatment planning purposes
resulting dose reduction to surrounding areas in prior to the start of treatment. The treatment plan
consequence (e.g., ARTreduco trial [102]). created on the pre-treatment image serves as the
Based on retrospective studies, dose de- reference plan for future adaptive plans created
escalation to the primary tumor seems feasible during the online workflow. After the first treat-
and safe for specific low-risk patient cohorts ment, any prior treatment plan may be selected as
[105]. Clinical trials are currently ongoing. An the reference plan, including the initial (pre-
example is the pilot DELPHI study treatment) plan.
(NCT05544136), in which dose to the high-risk
volume is lowered stepwise from 66 to 55 Gy.
Another example of a dose de-escalation trial is Elekta Unity Online Workflow
from Memorial Sloan Kettering (NCT05544136)
where positive nodes receive an experimental, First, the patient is positioned on the treatment
lower-than-standard dose of 50 Gy, 44 Gy, or table with a custom thermoplastic head and neck
40 Gy if 18F-FMISO PET is also hypoxia- immobilization device. The treatment area is
negative. Hypoxic patients are taken off protocol imaged using a T2w MRI sequence. Next, the
and receive the standard 70 Gy. Depending on the daily MRI is rigidly registered to the reference
results, these studies may prelude online real- image in Monaco, the associated treatment plan-
time adaptive dose de-escalation strategies ning system, to determine the isocenter shift to
guided by functional imaging. account for daily setup variations and any posi-
In conclusion, the employment of dose escala- tional changes. Because the Unity table can only
tion as well as dose de-escalation for HNSCC is move in a single direction (in and out of the bore),
an area of active investigation and may poten- the isocenter shift must be incorporated into the
tially offer important disease control and toxicity adaptive plan (i.e., “virtual isocenter shift”).
benefits. We are currently awaiting more data on Next, the user decides between the Adapt-
dose (de-)escalation with multiple trials open for to-
Position (ATP) or Adapt-to-Shape (ATS)
enrollment. These trials will establish which workflows [106]. ATP applies the virtual iso-
HNSCC (sub)volumes need to be boosted or center shift to the reference plan and either
treated less aggressively, the sequences to be recalculates the dose from the reference plan
used for their identification as well as the optimal or reoptimizes the reference plan based on the
dose to which they should be treated. new target position. However, ATP does not
allow segmentation on the daily setup image
and uses the reference image for dose calcula-
On-Table Adaptation tion. In ATS, the reference image and daily
image are deformably registered and the con-
Both currently available commercial MRL plat- tours propagated onto the daily image. The
forms support online treatment plan adaptation target volume and OAR contours may be
for HNSCC patients based on both positional and edited to account for anatomical changes
throughout RT. Electron density information more experience. For the first two patients treated
for dose calculation is transferred to the con- with the ATS-Lite approach reported by Gupta
tours on the MRI using bulk density assign- et al. [50], the median (range) treatment times
ment [106]. The treatment plan is then were 40 (28–47) and 32 (25–44) min.
reoptimized on the current anatomy. The seg-
mentation editing step tends to be a major
bottleneck of the online ATS workflow due to ViewRay MRIdian Online Workflow
the large number of structures in the head and
neck region and the poor performance of the Similar to the Unity workflow, patients are first
current deformable image registration algo- set up on the table with a custom immobilization
rithm within Monaco [40, 50]. For this reason, mask and imaged using a true fast imaging with
some sites have adopted an offline ATS/online steady-state free precession (TRUFI) sequence,
ATP strategy in which a prior online image is which has a combined T1 and T2 contrast weight-
used to create an ATS plan offline that is used ing [41, 42]. The daily image is fused with the
as the reference plan for ATP during the next reference image. If any isocenter shifts need to be
treatment [40]. An alternative “ATS-Lite” performed, the table can be physically moved in
approach was proposed by Gupta et al. [50] all three dimensions. Next, contours are propa-
and involves daily online ATS using deform- gated to the daily image with either rigid or
able image registration for the external patient deformable registration, and the contours may be
contour but rigid registration for all other edited as necessary. Electron densities are trans-
structures with weekly contour review by a ferred using bulk density assignment. The dose
physician. distribution from the reference plan is then recal-
After the treatment plan has been reoptimized culated on the current anatomy, and the dose dis-
and/or recalculated, the dose distribution and tribution and DVHs are reviewed and compared
DVHs are reviewed by the clinician. The dosi- to the reference plan. The clinician chooses to
metric criteria and DVHs for both the reference treat with the reference plan or to proceed with
plan and the adaptive plan can be viewed side by treatment plan adaptation. If they decide to adapt
side in Monaco for easy comparison. Factors the plan, it undergoes reoptimization and is sub-
considered for plan evaluation are discussed in sequently reviewed by the clinicians and deliv-
the following section (“Evaluation”). Once the ered after final approval [109]. During beam
plan has been approved by the clinician, the beam delivery, motion is tracked with a two-dimensional
is delivered. A two-dimensional balanced fast- TRUFI sequence for beam gating. The beam will
field echo MRI sequence in three orthogonal automatically shut off if the target falls outside of
directions is used for motion monitoring of the a pre-specified margin for a given percentage of
target volume or a reference OAR during beam time [110].
delivery [107]. The Unity system is currently not In the reports by Chen et al. [41, 42], the
capable of motion gating, but the beam can be majority of patients did not undergo any adaptive
stopped manually if gross motion is observed. replanning and were treated with the reference
However, that is less of a concern for HNSCC plan during every fraction. Although adaptation
patients in an immobilization mask unless severe was generally not performed, treatment with
weight loss during RT causes the mask not to fit online MR-guided RT was particularly advanta-
properly [108]. geous for patients undergoing re-RT of recurrent
The median (range) total treatment time was and new primary HNSCC [41]. The enhanced
46 (31–85) min in the online ATP workflow soft-tissue visualization during treatment setup
reported by McDonald et al. [40]. However, these allowed clinicians to reduce PTV margins, limit-
treatment times were for the first ten HNSCC ing dose to previously irradiated normal tissues,
patients treated on the MRL at their institution, and improving confidence in the precision of
and treatments may now be streamlined with each delivered treatment. However, Chen et al.
stated that online ART for HNSCC is an area of many clinics have implemented autosegmenta-
active investigation within their institution [42]. tion tools such as deep learning models or atlas-
Furthermore, in these studies, motion was tracked based autosegmentation for contouring OARs on
for observational purposes during treatments, but CT, which speed up the treatment planning pro-
no margin was set for automatic beam termina- cess considerably [111, 112]. A number of
tion based on target motion. research-grade autosegmentation tools have
shown promise for use on MRI, but a large vol-
ume of high-quality training data/atlases and
Additional Considerations thorough validation are needed for these tools to
for On-Table Adaptation be clinically implemented [113, 114].
MRI can be used for the reference plan instead
reatment Plan Adaptation Strategies
T of CT as done currently and may potentially
Beyond the general workflows for MR-guided streamline the online adaptive process for
ART for HNSCC described above, the actual HNSCC. MRI offers better soft-tissue visualiza-
treatment plan adaptation paradigm can vary con- tion compared to CT, which can potentially
siderably depending on clinical needs. Many improve segmentation accuracy. Also, one of the
treatment factors are not currently standardized, major challenges in online plan adaptation for
such as the frequency of adaptive replanning and HNSCC is that extensive manual contour correc-
whether target volumes and/or prescription doses tion is often necessary after propagating contours
should be modified throughout treatment. from the reference image to the daily image.
Heukelom and Fuller have introduced a nomen- McDonald et al. [40] compared the accuracy of
clature for describing various clinical ART inten- OAR contours propagated via deformable image
tions including serial plan verification, dose registration in Monaco between the pre-treatment
escalation to tumors or tumor subvolumes, dose CT or MRI and daily MRI images throughout
de-escalation for OAR sparing, and dose modifi- RT. The results showed that MR-to-MR outper-
cation based on shrinking tumor volumes [102]. formed CT-to-MR on average for all structures,
They also describe implementation techniques but that some individual outlier cases demon-
such as fixed-interval adaptation and adaptation strated very poor registration accuracy with
triggered by plan deviation from a pre-specified MR-to-MR. Nonetheless, using MRI for the ini-
threshold such as tumor volume, target dose cov- tial reference plan may help improve contour
erage, or accumulated dose. As MR-guided ART propagation accuracy for many patients, thus
becomes increasingly utilized for HNSCC, it is reducing the time spent manually correcting con-
important to report specific details about plan tours while the patient is on the table.
adaptation in any studies and to prospectively One practical consideration of MRI-based
evaluate outcomes of each adaptation strategy treatment planning is that it requires electron
with clinical trials. A number of recently com- density mapping for dose calculation. In both
pleted and current clinical trials for MR-guided MRL treatment planning systems, bulk density
ART have been described previously in this assignment is used, which involves assigning a
chapter. single electron density value to all voxels within
a structure [106, 109]. The density value can be
se of CT Vs. MRI for Reference Plan
U the mean value from the patient’s CT simulation
On both MRL platforms, either CT or MRI can if it is available, or it can be provided from a
be used as the planning image for the initial refer- population-based look-up table. While no studies
ence plan. The primary advantage of CT is that have specifically evaluated bulk density assign-
the pixel values contain quantitative information ment in MRI using structure-specific density val-
that can be converted to electron densities for ues, studies that assigned bulk density values to
dose calculation. Also, since CT simulation is water (soft tissue), bone, and air resulted in dose
universally used for standard RT workflows, differences within 2.5% of CT-based dose calcu-
lation across all target structures [115, 116]. Chin PTV expansions may not be suitable for every
et al. noted that dental implants were not patient. Bradley et al. assessed the impact of
adequately visualized on MRI and would require swallowing motion and duration on delivered
additional density corrections for accurate dose dose in HNSCC RT [123]. For most patients,
calculation [115]. Still, if standardized electron swallowing was infrequent enough to have mini-
density values are used for each structure type, mal impact on delivered dose, but for patients
the MRI simulation can be used for the reference with long and frequent swallows, they estimated
plan without the need to acquire a CT simulation, that delivered dose to the PTV could be reduced
which reduces the time burden on both patients by up to 10% in the worst-case scenario. Based
and clinical staff. on these results, real-time motion tracking and
adaptation could be beneficial for head and neck
Intrafraction Motion cancer patients with large tumor displacements
The Elekta Unity system is currently capable of and long swallowing durations to ensure that the
real-time motion monitoring during treatment full dose is delivered during each treatment.
delivery [107], while the ViewRay MRIdian sys-
tem has both motion monitoring and beam gating MRI Coils
capabilities (although only in one plane cur- The MRI coils used for imaging have important
rently) [110]. Unlike Unity, in which the user implications for both radiation treatment planning
would have to manually terminate the beam if and image quality. The Unity system uses a
gross motion is observed, the ViewRay beam can 2 × 4-channel radiotranslucent coil system, with a
be set to automatically turn off if the target falls posterior coil segment beneath the table and an
outside of a specific margin for a certain percent- anterior coil segment that is mounted on a ring
age of time. The current implementation of the and suspended above the patient [124]. Any radio-
ViewRay allows tracking/gating in a single plane, sensitive and highly attenuating coil components
but the next software release will enable two- are positioned on the sides of the coil, producing
plane tracking/gating. For both systems, the abil- a rectangular radiotranslucent window that allows
ity for the beam to move [42] in real time with the a maximum radiation field size of 22 cm in the
moving target are is an area of active research and superior-inferior direction. In the ViewRay sys-
development [117–121]. While these features tem, users can choose from two types of radio-
may be especially beneficial for lung and abdom- translucent flexible foam coils, a 2 × 6 channel
inal tumors subject to large displacements from torso coil or a 2 × 5 channel head and neck coil,
respiratory motion, their utility for HNSCC may which can both be positioned posteriorly and
be more limited because patients are well- anteriorly [125]. The maximum super-inferior
immobilized in thermoplastic head and neck field size is 24 cm. The limited field sizes of the
masks. systems restrict the selection of HNSCC patients
However, tumors are not completely static in who can be treated on the devices, potentially
the head and neck; they may be displaced by excluding patients with multi-level lymph node
respiratory, swallowing, and/or tongue motion involvement or patients with cancers of the naso-
[42, 108, 122]. Bruijnen et al. quantified the pat- pharynx/paranasal sinuses who would require
terns and extent of motion in nasopharyngeal, larger field sizes [126]. In studies at two different
oropharyngeal, and laryngeal tumors and found institutions, the treatment plans of HNSCC
that laryngeal tumors had the largest overall max- patients who were previously treated with con-
imum displacements and nasopharyngeal tumors ventional RT were analyzed, and the studies found
the smallest [108]. They used the displacements that 66% and 95% of previous patients had field
quantified among a large cohort of patients to cal- sizes small enough to be treated on the Unity
culate modified PTV margins, but they noted that MRL [127, 128].
tumor motion varies considerably among indi- In addition to field size considerations, the
vidual patients and that the population-based design of the coils also impacts image quality.
The flexible coils of the ViewRay system can be using statistical methods to determine whether
positioned very close to and conformed around QA actually needs to be performed for every
the patient’s surface to maximize signal-to-noise adaptive plan [129–131]. Kim et al.’s statistical
ratio. However, reproducibility of the coil setup analysis of brain cancer treatments indicated that
can be challenging, which is important during the the gamma pass rates of the reference plan and
online adaptive workflow so that the dose attenu- first fraction were sufficient to calculate a 99.9%
ation of the coils can be properly accounted for confidence interval that accurately encompassed
during treatment planning. The Unity system the mean gamma pass rate of all remaining frac-
uses a rigid anterior coil, which is helpful for tions [131]. Strand et al. found that ATS and
setup reproducibility but limits the signal-to- SBRT plans had lower overall gamma pass rates
noise ratio for HNSCC imaging because the coil but still passed QA [130]. Their results also indi-
cannot be conformed to the patient’s contour. cated that if reference plans pass QA, then adap-
Although the image quality of T1w and T2w tive plans may also be expected to pass. While
images using this coil is adequate for MR simula- these results are promising, similar analyses
tion and online plan adaptation [40], signal-to- should be performed specifically for HNSCC
noise ratio is slightly hindered compared to treatment plans due to the complexity of these
conventional MRI scanners and may have impli- plans.
cations for advanced quantitative MRI sequences In addition to IMRT QA, adaptive plans are
such as DWI. Thus, the development of dedicated evaluated prior to treatment delivery based on
head and neck coils for the Unity system is of whether pre-specified dosimetric criteria are met.
great interest to the community. Head and neck treatment plans can be very com-
plex and often involve a large number of target
volumes and OARs. Dosimetric constraint viola-
Evaluation tions occur frequently with adaptive plans gener-
ated with Monaco for the Unity MRL, although
Treating HNSCC on MRL introduces several the frequency of violation depends on the optimi-
new considerations for evaluating treatment plan zation strategy used [40, 50]. No data is currently
quality. Treatment plans must not only be evalu- available in the literature for ViewRay. In
ated before the start of treatment as with conven- Monaco, users are able to specify both optimal
tional RT, but adaptive plans must also be and mandatory dosimetric constraints, but there
reviewed during each treatment prior to delivery. is currently no easy way to track whether each
Patient-specific IMRT quality assurance (QA) is constraint was met in previous plans during the
typically performed for each reference plan online adaptation workflow. Furthermore, there
before treatment, but it cannot be performed on is no way to accumulate the delivered dose from
adaptive plans before each treatment because the previous fractions during online plan adaptation
patient is on the treatment table. Therefore, it is or to see if each constraint is being met in the sum
typically done after each treatment as a sanity of all delivered plans [60]. If we could see this
check, but if QA were to fail, it may indicate that information in the treatment planning system
a new reference plan should be created to ensure when adapting plans, it would help clinicians
that future adaptive plans meet quality control decide whether to accept a constraint failure for a
standards. In a report of the first 10 head and neck single adaptive plan or whether replanning needs
patients treated on the Unity MRL at the MD to be done. Some clinics export treatment data for
Anderson Cancer Center, all 253 adaptive plans dose accumulation in third-party software, but
passed QA, and the median gamma pass rate was this process is time-consuming and introduces
99.9% [40]. High pass rates have been reported uncertainties and variability between centers
for other disease sites as well, leading to studies using different platforms.
Outcome showed that, for xerostomia, one of the most fre-

quent late toxicities after HNSCC treatment,
To assess the true benefit of the introduction of patient-, and physician-rated toxicity showed
MRL devices, standardization of outcome mea- weak correlation. Moreover, patients tended to
sures across the field of head and neck radiation report more severe toxicity compared to physi-
oncology is paramount. This is important for cians [142].
both pillars of successful treatment: tumor con- It is hypothesized that the use of MRL devices
trol and toxicity reporting. may lead to a reduction in the toxicity profile for
HNSCC due to increased accuracy of contouring
combined with innumerable ART options. Yet,
Toxicity Reporting this can only be measured if standardization of
toxicity reporting as well as tumor control out-
Currently, toxicity reporting is done by physicians come reporting has been achieved.
and/or patients, typically using different question-
naires or scales. Well-known toxicity reporting
scales include the Common Terminology Criteria Tumor Control Outcome Reporting
for Adverse Events (CTCAE) and Radiation
Therapy Oncology Group (RTOG)/European Traditionally, outcomes of clinical studies include
Organisation for Research and Treatment of overall survival, distant metastases, local control,
Cancer (EORTC) scoring systems that allow scor- and locoregional control. However, with
ing of general complaints such as weight loss and increased options for treatment modulation, the
fatigue, as well as site-specific toxicities such as definition of subvolumes of tumor to (de-)esca-
xerostomia and dysphagia. The Late Effects late dose to outcome reporting has become
Normal Tissue Task Force-Subjective, Objective, increasingly complex. Moreover, the traditional
Management, Analytic (LENT-SOMA) scale is reporting simply does not provide sufficient
similar but not used as regularly. In addition to detail for assessment of treatment strategy effi-
these general scoring systems for toxicity report- cacy. A standardized means of reporting outcome
ing, there are more focused scoring systems for measures in the era of IMRT was coined in 2016
specific toxicities available, such as the Edinburgh by Mohamed et al. [143], but this strategy has not
Dysphagia Score (EDS) [132] for dysphagia or yet been widely implemented. In (real-time,
Functional Oral Intake Scale (FOIS) [133] for online) ART with changing boost volumes, what
oral intake. The FOIS scale has been validated is the dose to report? Accurate dose reporting
using data from several different countries such as might only be possible if an accurate approxima-
Iran [134], Italy [135], and Germany [136]. The tion of the actual dose delivered to the tumor/
Danish Head and Neck Cancer (DAHANCA) studied subvolume/voxel can be documented.
group has developed its own scoring system, the However, as mentioned above, dose accumula-
DAHANCA toxicity scale. Equally important tion for MRL devices does not currently allow
compared to physician-rated toxicity are patient- differentiation of dose to that detail.
reported outcome measures. For these, similar to
physician-rated toxicity, scoring can be quite gen-
eral or HNSCC-specific. Examples of these uture Perspectives on Outcome
F
include MDASI-HN [70], EORTC QLQ-HN35 Measurement
[137]/QLQ-C30 [138]/EQ-D5-L5 [137],
FACT-HN [139], MDADI for dysphagia [140], MRL devices allow for the collection of immense
and VHI-10 for voice quality [141]. amounts of data, either pre-treatment, post-
Unfortunately, physician-rated and patient- treatment, or even with on-beam imaging using
rated outcome measures regarding toxicity can several different sequences. Big data in HNSCC
differ substantially. For example, Kaae et al. may further drive technological advancements
such as the identification of cellular mechanisms MRI is less frequently used than CT and FDG-
driving HNSCC pathogenesis, new prognostic PET in the diagnosis and surveillance of
and predictive factors, or prediction of the effec- HNSCC. When MRI is used in treatment plan-
tiveness and tolerance of RT [144]. These predic- ning, interobserver variability in tumor delinea-
tion models, in turn, can aid physicians in tion exists [149], which raises concern for tumor
personalized treatment choices, patient education coverage and appropriate dosing to microscopic
(e.g., CITOR calculator for comprehensive indi- margins. With the better soft-tissue delineation
vidual toxicity risk assessment [145]), and/or available on the MRL, as well as potential real-
selection of treatment modality (e.g., proton vs. time gating as in the ViewRay system, opportu-
photon treatment [146]). nity exists to create smaller margins on tumor,
which further requires confidence in treatment
volume delineation and minimization of interob-
Challenges and Future Directions server variability. Furthermore, MRI-specific
contouring guidelines to minimize interobserver
Technological Challenges variability in contouring OARs are needed.
MRI acquisition requires patient coils for
General MRL challenges are particularly acute in optimal images. Given the lower magnet strength,
the head and neck region. There remain a number coils must be placed close to the area of interest
of technological challenges that need to be solved on the patient in order to optimize image quality.
for wider use and implementation of MRL in The immobilization required to treat HNSCC
HNSCC. As discussed in previous sections, the patients, which typically includes a mask to mini-
inability to deliver VMAT with the MRL leads to mize motion, adds to additional devices that must
challenges in creating treatment plans that offer be placed upon the patient in addition to the
optimal balance of tumor coverage and organ required coils; given the small size of the MRI
sparing. Given machine constraints, patients bore, this can present patient setup challenges.
must be treated with step-and-shoot IMRT. As Finally, current approaches to treatment plan-
discussed previously, this may lead to less opti- ning require obtaining a standard CT simulation
mal organ sparing and heterogeneity or hotter image to obtain electron density information to
plans [88]. This leads to important patient- and allow dose calculation. Ongoing efforts to
plan-selection considerations, where the benefits develop synthetic CTs from MR images [150] to
of a treatment plan on an MRL with better soft- eliminate this additional CT-based step would
tissue imaging and ability to adapt treatment may allow for streamlining of treatment planning
be outweighed by better planning techniques pathways. Different development of algorithms
available on a conventional linear accelerator. for both the Unity and the ViewRay MRL are
The lower magnet strength of the two cur- needed, as well as evaluation of ability to clini-
rently commercially available MRL systems cally implement these synthetic CTs.
(0.35T ViewRay and 1.5T Unity) as compared to
standard diagnostic quality magnet strength (3T)
may lead to lower-quality images utilizing the Staffing and Time-Related Challenges
MRL system [147]. Identification of optimal
MRI sequences for these lower-strength magnets Standard HNSCC RT typically consists of con-
that offer the best delineation of head and neck ventionally fractionated RT given over a period
structures is needed. Ongoing development of of 6–7 weeks. The potential incorporation of
additional sequences that can be deployed on the MRL into these long treatment courses raises sig-
MRL, such as DWI sequences [148], could pro- nificant time-based and staffing-based
vide additional tumor or treatment response data challenges.
that allows for more personalized treatment The average time for a single head and neck
adaptation. treatment on a CT-based linear accelerator is less
than 10 min. CBCTs for IGRT can be obtained in Physician input remains necessary for volume
a fraction of a minute, and patients can be aligned adaptation, which can variably take 10–20 min
quickly for treatment delivery. For MRL treat- depending on the treatment volumes, but the
ments, due to the requirement of utilizing step- reoptimization process is significantly shorter
and-shoot IMRT, patients typically require a utilizing the MRL platform. This replanning pro-
minimum of 25 min of table time. This includes a cess can occur online or offline, after the patient
need to obtain MR images, which is a minimum has completed treatment and with plans to initi-
of 2 min. For patients undergoing adaptation, ate a new plan on the next fraction. This faster
additional editing of contours adds a minimum of reoptimization can allow for more frequent plan
10–15 additional minutes. adaptations allowing for rapid response to patient
For long treatment courses, such as standard changes, although with unclear dosimetric bene-
HNSCC plans, this difference in treatment deliv- fits of frequent adaptation [152] and unknown
ery time equates to a large difference in table optimal timing or frequency of replanning. Of
time needed for each patient. For example, a note, for patients who experience very significant
patient undergoing a definitive course of 35 frac- anatomic shifts such as significant weight loss,
tions of radiation would require less than 6 h of reoptimization on the MRL does require repeat
total treatment time for a CT-based linear accel- simulation in order to ensure accurate electron
erator, and 14.5 h for an MRL treatment (assum- density calculations, potentially decreasing the
ing no adaptation). This total treatment time and time-saving advantage of rapid replanning.
required table time represent a significant burden Physician manual input on contouring struc-
on staffing and limitations for treating additional tures remains critical and, given the number of
patients. In contrast to the typical several-week structures that need to be contoured in the head
timeframe of treatment of HNSCC, RT for other and neck region, can require a significant amount
areas such as early-stage lung cancers can be of physician time. Autocontouring options exist
treated in as few as 3–5 treatments; thus, 7–10 for CT-based images [153], but there are cur-
early-stage lung cancer patients could be treated rently no commercially available autocontouring
in the same timeframe as a single conventionally software options available for the MRL. To
treated HNSCC patient utilizing the MRL plat- increase the feasibility of widespread use of the
form. The use of hypofractionation in HNSCC MRL, in addition to creating MRI-based contour-
remains limited, as mentioned above, with typi- ing guidelines, autocontouring software would
cal uses including SBRT for small re-RT cases. increase accessibility and reduce time burden for
Replanning or treatment adaptation during a patient treatment on the MRL.
course of head and neck RT is common; patients
frequently experience anatomical changes from
weight loss or tumor response [151]. In standard Patient-Specific Challenges
treatment planning workflows, patients who are
identified as requiring replanning are scheduled General MRI safety precautions must be fol-
for repeat CT simulation. Treatment volumes lowed when treating patients on an MRL, includ-
from the prior plan are transferred and potentially ing monitoring for implantable metal devices
deformed from the original simulation scan, and such as pacemakers. However, some challenges
then updated by the physician. Once this has may be more likely to occur for HNSCC patients
been completed, the plan is reoptimized, typi- as compared to others.
cally with some prior knowledge of achievable Patient setup for head and neck treatment
objectives for this particular plan. The whole pro- planning often requires custom devices for opti-
cess takes approximately 24–48 h. mal patient anatomy. Reproducible shoulder
In contrast, reoptimization on the MRL plat- positioning is important for patients getting neck
form once contours have been edited can take treatment [154]; shoulder pulls can be used as an
less than 5 min for a complex treatment plan. additional treatment device to monitor this. Bite
blocks can be used to displace adjacent mucosa nology, including further refinement of the
and increase dose falloff. A bolus may also be machine to allow for faster gantry rotation and
needed for tumors near the skin, including potential VMAT treatment. Imaging-based
patients with a thin neck receiving laryngeal RT refinements will continue to add to the ability to
or having undergone emergency tracheostomy define target volumes and potentially allow for
with tumor cut-through. Current commercially treatment adaptation based on novel sequences
available devices are intended for use with CT such as DWI. This may allow for a future of func-
simulations and CT-based linear accelerators. tional adaptation, allowing for treatment changes
Use of such additional patient-specific devices that prevent acute toxicities or minimize the risk
needs to be MRI compatible for use with the of late toxicities.
MRL. Further clinical trial development for the role
Screening for claustrophobia and ensuring and safety of hypofractionation in HNSCC may
patient comfort during treatment are critical for further increase the appeal of the use of the
HNSCC treatment on the MRL, given both the MRL in HNSCC treatment, and may help over-
longer treatment times and the patient devices come the staffing-/time-related challenges of
required for setup. Patients require a treatment conventionally fractionated treatment. Novel
mask for immobilization and placement of coils work on hypofractionated larynx treatment [37]
for optimal image quality, and the MRL treat- lends itself particularly well to the MRL plat-
ment bore is small (70 cm for both Unity and form, and the ViewRay platform’s ability to
ViewRay systems). Some patients may be more allow for real-time tracking may allow for addi-
at risk of experiencing claustrophobia. tional developments in this space. MRL may
Anecdotally, our experience has been that also be interesting in the setting of re-RT where
patients tolerate the mask/coil placement in the SBRT and hypofractionation may be used more
smaller bore without issues when appropriately commonly.
counseled beforehand.
Older patients are at risk of hearing loss, and a
proportion of patients may require hearing aids Summary
as assistive devices during treatment. Within
treatment of HNSCC, concurrent chemotherapy The use of MRL in HNSCC has potential for
agents also carry the risk of inducing and further treatment personalization and may offer an
exacerbating ototoxicity. Patients with pre- opportunity to study treatment changes and
existing hearing loss may not be candidates for response through the collection of serial MRIs
treatment on the MRL given the need to be able with functional sequences throughout therapy.
to communicate with patients without these Future works assessing treatment and toxicity
metal-containing assistive devices during the outcomes as well as ways to optimize workflow
course of MRL treatment (setup, image acquisi- and address current challenges with MRL for
tion, potential reoptimization, and treatment). HNSCC are warranted and under way with mul-
tiple trials open internationally.
Future Directions
References
MRL is a promising new technology for the treat-
ment of HNSCC, but will require additional work 1. Warnakulasuriya S. Global epidemiology of
to optimally identify the best patients, best dis- oral and oropharyngeal cancer. Oral Oncol.
2009;45(4–5):309–16.
eases, and best anatomy that will lend itself
favorably to the benefits of this technology. statistics, 2022. CA Cancer J Clin. 2022;72(1):7–33.
As noted above, ongoing research is needed to 3. Rettig EM, Fakhry C, Khararjian A, Westra WH. Age
overcome challenges associated with this tech- profile of patients with oropharyngeal squamous cell
carcinoma. JAMA Otolaryngol Head Neck Surg. 5-year survival data from a phase 3 randomised trial,
2018;144(6):538–9. and relation between cetuximab-induced rash and
4. Chaturvedi AK, Engels EA, Pfeiffer RM, Hernandez survival. Lancet Oncol. 2010;11(1):21–8.
BY, Xiao W, Kim E, et al. Human papillomavirus 17. Overgaard J, Hansen HS, Specht L, Overgaard M,
and rising oropharyngeal cancer incidence in the Grau C, Andersen E, et al. Five compared with six
United States. J Clin Oncol. 2011;29(32):4294–301. fractions per week of conventional radiotherapy
5. Patel EJ, Oliver JR, Jacobson AS, Li Z, Hu KS, of squamous-cell carcinoma of head and neck:
Tam M, et al. Human papillomavirus in patients DAHANCA 6 and 7 randomised controlled trial.
with hypopharyngeal squamous cell carcinoma. Lancet. 2003;362(9388):933–40.
Otolaryngol Head Neck Surg. 2022;166(1):109–17. 18. Ang KK, Harris J, Wheeler R, Weber R, Rosenthal
6. Ohno S, Arnheiter H, Dubois-Dalcq M, Lazzarini DI, Nguyen-Tan PF, et al. Human papillomavirus
RA. Immunocytochemical localization of vesicular and survival of patients with oropharyngeal cancer.
stomatitis virus proteins N and NS with monoclonal N Engl J Med. 2010;363(1):24–35.
antibodies. Histochemistry. 1985;82(2):185–96. 19. Becker M, Zbaren P, Casselman JW, Kohler R,
7. Chung CH, Zhang Q, Kong CS, Harris J, Fertig EJ, Dulguerov P, Becker CD. Neoplastic invasion of
Harari PM, et al. p16 protein expression and human laryngeal cartilage: reassessment of criteria for diag-
papillomavirus status as prognostic biomarkers of nosis at MR imaging. Radiology. 2008;249(2):551–9.
nonoropharyngeal head and neck squamous cell car- 20. Ong CK, Chong VF. Imaging of perineural spread
cinoma. J Clin Oncol. 2014;32(35):3930–8. in head and neck tumours. Cancer Imaging. 2010;10
8. Stephen JK, Divine G, Chen KM, Chitale D, Havard Spec no A(1A):S92–8.
S, Worsham MJ. Significance of p16 in site-specific 21. Kim JH, Choi KY, Lee SH, Lee DJ, Park BJ, Yoon
HPV positive and HPV negative head and neck DY, et al. The value of CT, MRI, and PET-CT in
squamous cell carcinoma. Cancer Clin Oncol. detecting retropharyngeal lymph node metastasis of
2013;2(1):51–61. head and neck squamous cell carcinoma. BMC Med
9. Slaughter DP. Multicentric origin of intraoral carci- Imaging. 2020;20(1):88.
noma. Surgery. 1946;20(1):133–46. 22. Adams S, Baum RP, Stuckensen T, Bitter K, Hor
10. Adelstein DJ, Li Y, Adams GL, Wagner H Jr, Kish G. Prospective comparison of 18F-FDG PET with
JA, Ensley JF, et al. An intergroup phase III compari- conventional imaging modalities (CT, MRI, US) in
son of standard radiation therapy and two schedules lymph node staging of head and neck cancer. Eur J
of concurrent chemoradiotherapy in patients with Nucl Med. 1998;25(9):1255–60.
unresectable squamous cell head and neck cancer. J 23. Schmid DT, Stoeckli SJ, Bandhauer F, Huguenin
Clin Oncol. 2003;21(1):92–8. P, Schmid S, von Schulthess GK, et al. Impact of
11. Pignon JP, le Maitre A, Maillard E, Bourhis J, positron emission tomography on the initial stag-
MACH-NC Collaborative Group. Meta-analysis ing and therapy in locoregional advanced squamous
of chemotherapy in head and neck cancer cell carcinoma of the head and neck. Laryngoscope.
(MACH-NC): an update on 93 randomised trials and 2003;113(5):888–91.
17,346 patients. Radiother Oncol. 2009;92(1):4–14. 24. Chauhan D, Rawat S, Sharma MK, Ahlawat P, Pal
12. Zhang Y, Chen L, Hu GQ, Zhang N, Zhu XD, Yang M, Gupta G, et al. Improving the accuracy of target
KY, et al. Gemcitabine and cisplatin induction che- volume delineation by combined use of computed
motherapy in nasopharyngeal carcinoma. N Engl J tomography, magnetic resonance imaging and posi-
Med. 2019;381(12):1124–35. tron emission tomography in head and neck carcino-
13. Lefebvre JL, Andry G, Chevalier D, Luboinski B, mas. J Cancer Res Ther. 2015;11(4):746–51.
Collette L, Traissac L, et al. Laryngeal preservation 25. Figen M, Colpan Oksuz D, Duman E, Prestwich
with induction chemotherapy for hypopharyngeal R, Dyker K, Cardale K, et al. Radiotherapy for
squamous cell carcinoma: 10-year results of EORTC head and neck cancer: evaluation of triggered adap-
trial 24891. Ann Oncol. 2012;23(10):2708–14. tive replanning in routine practice. Front Oncol.
14. Sun L, Candelieri-Surette D, Anglin-Foote T, Lynch 2020;10:579917.
JA, Maxwell KN, D’Avella C, et al. Cetuximab-based 26. Mohan R, Wu Q, Manning M, Schmidt-Ullrich
vs carboplatin-based chemoradiotherapy for patients R. Radiobiological considerations in the design of
with head and neck cancer. JAMA Otolaryngol Head fractionation strategies for intensity-modulated radi-
Neck Surg. 2022;148(11):1022–8. ation therapy of head and neck cancers. Int J Radiat
15. Patil VM, Noronha V, Menon N, Singh A, Ghosh- Oncol Biol Phys. 2000;46(3):619–30.
Laskar S, Budrukkar A, et al. Results of phase III 27. Pagh A, Grau C, Overgaard J. Failure pattern and
randomized trial for use of docetaxel as a radiosen- salvage treatment after radical treatment of head and
sitizer in patients with head and neck cancer, unsuit- neck cancer. Acta Oncol. 2016;55(5):625–32.
able for cisplatin-based chemoradiation. J Clin 28. Chuang SC, Scelo G, Tonita JM, Tamaro S, Jonasson
Oncol. 2023;41(13):2350–61. JG, Kliewer EV, et al. Risk of second primary can-
16. Bonner JA, Harari PM, Giralt J, Cohen RB, Jones cer among patients with head and neck cancers: a
CU, Sur RK, et al. Radiotherapy plus cetuximab pooled analysis of 13 cancer registries. Int J Cancer.
for locoregionally advanced head and neck cancer: 2008;123(10):2390–6.
29. Kim YS. Reirradiation of head and neck cancer in 41. Chen AM, Cao M, Hsu S, Lamb J, Mikaeilian
the era of intensity-modulated radiotherapy: patient A, Yang Y, et al. Magnetic resonance imaging
selection, practical aspects, and current evidence. guided reirradiation of recurrent and second pri-
Radiat Oncol J. 2017;35(1):1–15. mary head and neck cancer. Adv Radiat Oncol.
30. You R, Liu YP, Xie YL, Lin C, Duan CY, Chen DP, 2017;2(2):167–75.
et al. Hyperfractionation compared with standard 42. Chen AM, Hsu S, Lamb J, Yang Y, Agazaryan N,
fractionation in intensity-modulated radiotherapy Steinberg ML, et al. MRI-guided radiotherapy for
for patients with locally advanced recurrent naso- head and neck cancer: initial clinical experience.
pharyngeal carcinoma: a multicentre, randomised, Clin Transl Oncol. 2018;20(2):160–8.
open-label, phase 3 trial. Lancet. 2023;401:917. 43. Bahig H, Yuan Y, Mohamed ASR, Brock KK, Ng SP,
31. Yamazaki H, Kodani N, Ogita M, Sato K, Himei Wang J, et al. Magnetic resonance-based response
K. Reirradiation of head and neck cancer focusing assessment and dose adaptation in human papilloma
on hypofractionated stereotactic body radiation ther- virus positive tumors of the oropharynx treated with
apy. Radiat Oncol. 2011;6:98. radiotherapy (MR-ADAPTOR): an R-IDEAL stage
32. Phan J, Sio TT, Nguyen TP, Takiar V, Gunn GB, 2a-2b/Bayesian phase II trial. Clin Transl Radiat
Garden AS, et al. Reirradiation of head and neck Oncol. 2018;13:19–23.
cancers with proton therapy: outcomes and analyses. 44. Kyzas PA, Evangelou E, Denaxa-Kyza D, Ioannidis
Int J Radiat Oncol Biol Phys. 2016;96(1):30–41. JP. 18F-fluorodeoxyglucose positron emission
33. Rodin J, Bar-Ad V, Cognetti D, Curry J, Johnson tomography to evaluate cervical node metasta-
J, Zender C, et al. A systematic review of treating ses in patients with head and neck squamous cell
recurrent head and neck cancer: a reintroduction of carcinoma: a meta-analysis. J Natl Cancer Inst.
brachytherapy with or without surgery. J Contemp 2008;100(10):712–20.
Brachytherapy. 2018;10(5):454–62. 45. Kim SY, Kim JS, Doo H, Lee H, Lee JH, Cho KJ,
34. Le QT, Fu KK, Kroll S, Ryu JK, Quivey JM, et al. Combined [18F]fluorodeoxyglucose positron
Meyler TS, et al. Influence of fraction size, total emission tomography and computed tomography for
dose, and overall time on local control of T1-T2 detecting contralateral neck metastases in patients
glottic carcinoma. Int J Radiat Oncol Biol Phys. with head and neck squamous cell carcinoma. Oral
1997;39(1):115–26. Oncol. 2011;47(5):376–80.
35. Kachhwaha A, Jakhar SL, Syiem T, Sharma N, 46. Nguyen A, Luginbuhl A, Cognetti D, Van Abel K,
Kumar HS, Sharma A. Hypofractionated radiother- Bar-Ad V, Intenzo C, et al. Effectiveness of PET/
apy versus conventional radiotherapy in early glot- CT in the preoperative evaluation of neck disease.
tic cancer T1-2N0M0: a randomized study. J Cancer Laryngoscope. 2014;124(1):159–64.
Res Ther. 2021;17(6):1499–502. 47. Vishwanath V, Jafarieh S, Rembielak A. The role of
36. Schwartz DL, Sosa A, Chun SG, Ding C, Xie XJ, imaging in head and neck cancer: an overview of dif-
Nedzi LA, et al. SBRT for early-stage glottic larynx ferent imaging modalities in primary diagnosis and
cancer-initial clinical outcomes from a phase I clini- staging of the disease. J Contemp Brachytherapy.
cal trial. PLoS One. 2017;12(3):e0172055. 2020;12(5):512–8.
37. Sher DJ, Timmerman RD, Nedzi L, Ding C, Pham 48. Widmann G, Henninger B, Kremser C, Jaschke
NL, Zhao B, et al. Phase 1 fractional dose-escalation W. MRI sequences in head & neck radiology—state
study of equipotent stereotactic radiation therapy of the art. Rofo. 2017;189(5):413–22.
regimens for early-stage Glottic larynx cancer. Int J 49. Ross MR, Schomer DF, Chappell P, Enzmann
Radiat Oncol Biol Phys. 2019;105(1):110–8. DR. MR imaging of head and neck tumors: com-
38. Kang BH, Yu T, Kim JH, Park JM, Kim JI, Chung parison of T1-weighted contrast-enhanced fat-
EJ, et al. Early closure of a phase 1 clinical trial suppressed images with conventional T2-weighted
for SABR in early-stage glottic cancer. Int J Radiat and fast spin-echo T2-weighted images. AJR Am J
Oncol Biol Phys. 2019;105(1):104–9. Roentgenol. 1994;163(1):173–8.
39. Mohamed ASR, Bahig H, Aristophanous M, 50. Gupta A, Dunlop A, Mitchell A, McQuaid D,
Blanchard P, Kamal M, Ding Y, et al. Prospective in Nill S, Barnes H, et al. Online adaptive radiother-
silico study of the feasibility and dosimetric advan- apy for head and neck cancers on the MR linear
tages of MRI-guided dose adaptation for human pap- accelerator: introducing a novel modified adapt-
illomavirus positive oropharyngeal cancer patients to-shape approach. Clin Transl Radiat Oncol.
compared with standard IMRT. Clin Transl Radiat 2022;32:48–51.
Oncol. 2018;11:11–8. 51. Salzillo TC, Dresner AM, Way A, Wahid KA,
40. McDonald BA, Vedam S, Yang J, Wang J, Castillo P, McDonald BA, Mulder S, et al. Development and
Lee B, et al. Initial feasibility and clinical implemen- implementation of optimized endogenous contrast
tation of daily MR-guided adaptive head and neck sequences for delineation in adaptive radiotherapy
cancer radiation therapy on a 1.5T MR-Linac sys- on a 1.5T MR-Linear-accelerator (MR-Linac):
tem: prospective R-IDEAL 2a/2b systematic clini- a prospective R-IDEAL Stage 0-2a quantita-
cal evaluation of technical innovation. Int J Radiat tive/qualitative evaluation of in vivo site-specific
Oncol Biol Phys. 2021;109(5):1606–18. quality-assurance using a 3D T2 fat-suppressed
platform for head and neck cancer. medRxiv. advanced head and neck cancer-comparison of a
2022:2022.06.24.22276839. prospective trial with a revised simulated contouring
52. Phillips CD, Gay SB, Newton RL, Levine approach. J Clin Med. 2021;10(20):4653.
PA. Gadolinium-enhanced MRI of tumors of the 65. Gregoire V, Ang K, Budach W, Grau C, Hamoir
head and neck. Head Neck. 1990;12(4):308–15. M, Langendijk JA, et al. Delineation of the neck
53. Wang J, Salzillo T, Jiang Y, Mackeyev Y, David node levels for head and neck tumors: a 2013
Fuller C, Chung C, et al. Stability of MRI contrast update. DAHANCA, EORTC, HKNPCSG, NCIC
agents in high-energy radiation of a 1.5T MR-Linac. CTG, NCRI, RTOG, TROG consensus guidelines.
Radiother Oncol. 2021;161:55–64. Radiother Oncol. 2014;110(1):172–81.
54. Mahmood F, Nielsen UG, Jorgensen CB, Brink 66. Gregoire V, Evans M, Le QT, Bourhis J, Budach V,
C, Thomsen HS, Hansen RH. Safety of gadolin- Chen A, et al. Delineation of the primary tumour
ium based contrast agents in magnetic resonance clinical target volumes (CTV-P) in laryngeal, hypo-
imaging- guided radiotherapy—an investigation of pharyngeal, oropharyngeal and oral cavity squa-
chelate stability using relaxometry. Phys Imaging mous cell carcinoma: AIRO, CACA, DAHANCA,
Radiat Oncol. 2022;21:96–100. EORTC, GEORCC, GORTEC, HKNPCSG,
55. Ersoy H, Rybicki FJ. Biochemical safety profiles HNCIG, IAG-KHT, LPRHHT, NCIC CTG, NCRI,
of gadolinium-based extracellular contrast agents NRG Oncology, PHNS, SBRT, SOMERA, SRO,
and nephrogenic systemic fibrosis. J Magn Reson SSHNO, TROG consensus guidelines. Radiother
Imaging. 2007;26(5):1190–7. Oncol. 2018;126(1):3–24.
56. Petronek MS, Steinbach EJ, Kalen AL, Builta 67. Gregoire V, Grau C, Lapeyre M, Maingon P. Target
ZJ, Callaghan CM, Hyer DE, et al. Assessment of volume selection and delineation (T and N) for pri-
Gadobutrol safety in combination with ionizing radi- mary radiation treatment of oral cavity, oropharyn-
ation using a preclinical MRI-guided radiotherapy geal, hypopharyngeal and laryngeal squamous cell
model. Radiat Res. 2021;195(3):230–4. carcinoma. Oral Oncol. 2018;87:131–7.
57. Hales RB, Chuter R, McWilliam A, Salah A, Dubec 68. Machiels JP, Rene Leemans C, Golusinski W, Grau
M, Freear L, et al. The impact of gadolinium-based C, Licitra L, Gregoire V, et al. Squamous cell car-
MR contrast on radiotherapy planning for oropha- cinoma of the oral cavity, larynx, oropharynx and
ryngeal treatment on the MR Linac. Med Phys. hypopharynx: EHNS-ESMO-ESTRO clinical prac-
2022;49(1):510–20. tice guidelines for diagnosis, treatment and follow-
58. Payabvash S. Quantitative diffusion magnetic reso- up. Ann Oncol. 2020;31(11):1462–75.
nance imaging in head and neck tumors. Quant 69. Brouwer CL, Steenbakkers RJ, Bourhis J, Budach
Imaging Med Surg. 2018;8(10):1052–65. W, Grau C, Gregoire V, et al. CT-based delinea-
59. Habrich J, Boeke S, Nachbar M, Nikolaou K, tion of organs at risk in the head and neck region:
Schick F, Gani C, et al. Repeatability of diffusion- DAHANCA, EORTC, GORTEC, HKNPCSG, NCIC
weighted magnetic resonance imaging in head and CTG, NCRI, NRG oncology and TROG consensus
neck cancer at a 1.5 T MR-Linac. Radiother Oncol. guidelines. Radiother Oncol. 2015;117(1):83–90.
2022;174:141–8. 70. Cleeland CS, Mendoza TR, Wang XS, Chou C, Harle
60. McDonald BA, Zachiu C, Christodouleas J, Naser MT, Morrissey M, et al. Assessing symptom distress
MA, Ruschin M, Sonke JJ, et al. Dose accumulation in cancer patients: the M.D. Anderson symptom
for MR-guided adaptive radiotherapy: from practical inventory. Cancer. 2000;89(7):1634–46.
considerations to state-of-the-art clinical implemen- 71. Lee AW, Ng WT, Pan JJ, Poh SS, Ahn YC, AlHussain
tation. Front Oncol. 2022;12:1086258. H, et al. International guideline for the delineation of
61. Yang Y, Cao M, Sheng K, Gao Y, Chen A, Kamrava the clinical target volumes (CTV) for nasopharyngeal
M, et al. Longitudinal diffusion MRI for treatment carcinoma. Radiother Oncol. 2018;126(1):25–36.
response assessment: preliminary experience using 72. O’Sullivan B, Warde P, Grice B, Goh C, Payne D,
an MRI-guided tri-cobalt 60 radiotherapy system. Liu FF, et al. The benefits and pitfalls of ipsilateral
Med Phys. 2016;43(3):1369–73. radiotherapy in carcinoma of the tonsillar region. Int
62. Kooreman ES, van Houdt PJ, Keesman R, Pos FJ, J Radiat Oncol Biol Phys. 2001;51(2):332–43.
van Pelt VWJ, Nowee ME, et al. ADC measure- 73. Chronowski GM, Garden AS, Morrison WH, Frank
ments on the Unity MR-linac - a recommendation SJ, Schwartz DL, Shah SJ, et al. Unilateral radio-
on behalf of the Elekta Unity MR-linac consortium. therapy for the treatment of tonsil cancer. Int J
Radiother Oncol. 2020;153:106–13. Radiat Oncol Biol Phys. 2012;83(1):204–9.
63. Schakel T, Hoogduin JM, Terhaard CHJ, Philippens 74. Dan TD, Raben D, Schneider CJ, Hockstein NG,
MEP. Technical note: diffusion-weighted MRI with Witt RL, Dzeda M, et al. Freedom from local and
minimal distortion in head-and-neck radiotherapy regional failure of contralateral neck with ipsilateral
using a turbo spin echo acquisition method. Med neck radiotherapy for node-positive tonsil cancer:
Phys. 2017;44(8):4188–93. updated results of an institutional clinical manage-
64. Weissmann T, Speer S, Putz F, Lettmaier S, Schubert ment approach. Oral Oncol. 2015;51(6):616–21.
P, Shariff M, et al. Reduction of elective radiotherapy 75. Kennedy WR, Herman MP, Deraniyagala RL,
treatment volume in definitive treatment of locally Amdur RJ, Werning JW, Dziegielewski PT, et al.
Ipsilateral radiotherapy for squamous cell carci- receiving intensity-modulated radiotherapy. JAMA
noma of the tonsil. Eur Arch Otorhinolaryngol. Otolaryngol Head Neck Surg. 2022;148(7):604–11.
2016;273(8):2151–6. 88. Ouyang Z, Liu Shen Z, Murray E, Kolar M, LaHurd
76. Xue F, Hu C, He X. Induction chemotherapy fol- D, Yu N, et al. Evaluation of auto-planning in IMRT
lowed by intensity-modulated radiotherapy with and VMAT for head and neck cancer. J Appl Clin
reduced gross tumor volume delineation for stage Med Phys. 2019;20(7):39–47.
T3-4 nasopharyngeal carcinoma. Onco Targets Ther. 89. Chamberlain M, Krayenbuehl J, van Timmeren
2017;10:3329–36. JE, Wilke L, Andratschke N, Garcia Schüler H,
77. Perillo A, Landoni V, Farneti A, Sanguineti G. Organ et al. Head and neck radiotherapy on the MR
motion in linac-based SBRT for glottic cancer. linac: a multicenter planning challenge amongst
Radiat Oncol. 2021;16(1):106. MRIdian platform users. Strahlenther Onkol.
78. Kwa SL, Al-Mamgani A, Osman SO, Gangsaas A, 2021;197(12):1093–103.
Levendag PC, Heijmen BJ. Inter- and Intrafraction 90. Lacas B, Carmel A, Landais C, Wong SJ, Licitra
target motion in highly focused single vocal cord L, Tobias JS, et al. Meta-analysis of chemotherapy
irradiation of T1a larynx cancer patients. Int J Radiat in head and neck cancer (MACH-NC): an update
Oncol Biol Phys. 2015;93(1):190–5. on 107 randomized trials and 19,805 patients, on
79. Navran A, Heemsbergen W, Janssen T, Hamming- behalf of MACH-NC group. Radiother Oncol.
Vrieze O, Jonker M, Zuur C, et al. The impact 2021;156:281–93.
of margin reduction on outcome and toxicity in 91. Soto DE, Kessler ML, Piert M, Eisbruch
head and neck cancer patients treated with image- A. Correlation between pretreatment FDG-PET
guided volumetric modulated arc therapy (VMAT). biological target volume and anatomical location of
Radiother Oncol. 2019;130:25–31. failure after radiation therapy for head and neck can-
80. Marks LB, Yorke ED, Jackson A, Ten Haken RK, cers. Radiother Oncol. 2008;89(1):13–8.
Constine LS, Eisbruch A, et al. Use of normal tissue 92. Mohamed ASR, Cardenas CE, Garden AS, Awan
complication probability models in the clinic. Int J MJ, Rock CD, Westergaard SA, et al. Patterns-of-
Radiat Oncol Biol Phys. 2010;76(3 Suppl):S10–9. failure guided biological target volume definition for
81. Emami B, Lyman J, Brown A, Coia L, Goitein M, head and neck cancer patients: FDG-PET and dosi-
Munzenrider JE, et al. Tolerance of normal tissue to metric analysis of dose escalation candidate subre-
therapeutic irradiation. Int J Radiat Oncol Biol Phys. gions. Radiother Oncol. 2017;124(2):248–55.
1991;21(1):109–22. 93. Madani I, Duprez F, Boterberg T, Van de Wiele C,
82. Su W, Wright CM, Lee DY, Kim M, Anstadt EJ, Teo Bonte K, Deron P, et al. Maximum tolerated dose
BK, et al. Stricter postoperative oropharyngeal can- in a phase I trial on adaptive dose painting by num-
cer radiation therapy normal tissue dose constraints bers for head and neck cancer. Radiother Oncol.
are feasible. Pract Radiat Oncol. 2022;12(4):e282–5. 2011;101(3):351–5.
83. Yock TI, Murphy B, Bass JK, Ronckers CM, Kremer 94. Berwouts D, Madani I, Duprez F, Olteanu AL,
L, Baliga S, et al. Modeling the risk of hearing loss Vercauteren T, Boterberg T, et al. Long-term out-
from radiotherapy in childhood cancer survivors: come of (18) F-fluorodeoxyglucose-positron emis-
initial results from the pediatric normal tissue effects sion tomography-guided dose painting for head
in the clinic (PENTEC) hearing loss task force. Int J and neck cancer: matched case-control study. Head
Radiat Oncol Biol Phys. 2019;105(1):S191. Neck. 2017;39(11):2264–75.
84. van Dijk LV, Abusaif AA, Rigert J, Naser MA, 95. Heukelom J, Hamming O, Bartelink H, Hoebers F,
Hutcheson KA, Lai SY, et al. Normal tissue com- Giralt J, Herlestam T, et al. Adaptive and innovative
plication probability (NTCP) prediction model for radiation treatment FOR improving cancer treatment
osteoradionecrosis of the mandible in patients with outcomE (ARTFORCE); a randomized controlled
head and neck cancer after radiation therapy: large- phase II trial for individualized treatment of head
scale observational cohort. Int J Radiat Oncol Biol and neck cancer. BMC Cancer. 2013;13:84.
Phys. 2021;111(2):549–58. 96. van Dijk LV, Frank SJ, Yuan Y, Gunn B, Moreno AC,
85. Kamal M, Peeler CR, Yepes P, Mohamed ASR, Mohamed ASR, et al. Proton image-guided radiation
Blanchard P, Frank S, et al. Radiation-induced assignment for therapeutic escalation via selection
hypothyroidism after radical intensity modulated of locally advanced head and neck cancer patients
radiation therapy for oropharyngeal carcinoma. Adv [PIRATES]: a phase I safety and feasibility trial of
Radiat Oncol. 2020;5(1):111–9. MRI-guided adaptive particle radiotherapy. Clin
86. Zhou L, Chen J, Shen W, Chen ZL, Huang S, Tao Transl Radiat Oncol. 2022;32:35–40.
CJ, et al. Thyroid V(50) is a risk factor for hypothy- 97. Mierzwa ML, Aryal M, Lee C, Schipper M, VanTil
roidism in patients with nasopharyngeal carcinoma M, Morales K, et al. Randomized phase II study of
treated with intensity-modulated radiation therapy: physiologic MRI-directed adaptive radiation boost
a retrospective study. Radiat Oncol. 2020;15(1):68. in poor prognosis head and neck cancer. Clin Cancer
87. Chen WC, Hsu CM, Tsai YT, Lin MH, Tsai MS, Res. 2022;28(23):5049–57.
Chang GH, et al. Prospective evaluation of taste 98. Zhang Y, Hong H, Cai W. PET tracers based on
function in patients with head and neck cancer Zirconium-89. Curr Radiopharm. 2011;4(2):131–9.
99. Grosu AL, Souvatzoglou M, Roper B, Dobritz M, 110. Green OL, Rankine LJ, Cai B, Curcuru A, Kashani
Wiedenmann N, Jacob V, et al. Hypoxia imaging R, Rodriguez V, et al. First clinical implementation
with FAZA-PET and theoretical considerations of real-time, real anatomy tracking and radiation
with regard to dose painting for individualization of beam control. Med Phys. 2018;45:3728.
radiotherapy in patients with head and neck cancer. 111. Zhong Y, Yang Y, Fang Y, Wang J, Hu W. A pre-
Int J Radiat Oncol Biol Phys. 2007;69(2):541–51. liminary experience of implementing deep-learning
100. Bittner MI, Wiedenmann N, Bucher S, Hentschel M, based auto-segmentation in head and neck cancer:
Mix M, Weber WA, et al. Exploratory geographical a study on real-world clinical cases. Front Oncol.
analysis of hypoxic subvolumes using 18F-MISO- 2021;11:638197.
PET imaging in patients with head and neck can- 112. Nikolov S, Blackwell S, Zverovitch A, Mendes
cer in the course of primary chemoradiotherapy. R, Livne M, De Fauw J, et al. Clinically appli-
Radiother Oncol. 2013;108(3):511–6. cable segmentation of head and neck anatomy for
101. Zschaeck S, Haase R, Abolmaali N, Perrin R, Stutzer radiotherapy: deep learning algorithm develop-
K, Appold S, et al. Spatial distribution of FMISO in ment and validation study. J Med Internet Res.
head and neck squamous cell carcinomas during 2021;23(7):e26151.
radio-chemotherapy and its correlation to pattern of 113. Kieselmann JP, Kamerling CP, Burgos N, Menten
failure. Acta Oncol. 2015;54(9):1355–63. MJ, Fuller CD, Nill S, et al. Geometric and dosimet-
102. Heukelom J, Fuller CD. Head and neck cancer adap- ric evaluations of atlas-based segmentation methods
tive radiation therapy (ART): conceptual consid- of MR images in the head and neck region. Phys
erations for the informed clinician. Semin Radiat Med Biol. 2018;63(14):145007.
Oncol. 2019;29(3):258–73. 114. Korte JC, Hardcastle N, Ng SP, Clark B, Kron T,
103. van den Bosch S, Dijkema T, Kunze-Busch MC, Jackson P. Cascaded deep learning-based auto-
Terhaard CH, Raaijmakers CP, Doornaert PA, segmentation for head and neck cancer patients:
et al. Uniform FDG-PET guided GRAdient Dose organs at risk on T2-weighted magnetic resonance
prEscription to reduce late Radiation Toxicity imaging. Med Phys. 2021;48(12):7757–72.
(UPGRADE-RT): study protocol for a randomized 115. Chin AL, Lin A, Anamalayil S, Teo BK. Feasibility
clinical trial with dose reduction to the elective neck and limitations of bulk density assignment in MRI
in head and neck squamous cell carcinoma. BMC for head and neck IMRT treatment planning. J Appl
Cancer. 2017;17(1):208. Clin Med Phys. 2014;15(5):4851.
104. Deschuymer S, Nevens D, Duprez F, Daisne JF, 116. Karotki A, Mah K, Meijer G, Meltsner
Dok R, Laenen A, et al. Randomized clinical trial on M. Comparison of bulk electron density and voxel-
reduction of radiotherapy dose to the elective neck based electron density treatment planning. J Appl
in head and neck squamous cell carcinoma; update Clin Med Phys. 2011;12(4):3522.
of the long-term tumor outcome. Radiother Oncol. 117. Crijns SP, Kok JG, Lagendijk JJ, Raaymakers
2020;143:24–9. BW. Towards MRI-guided linear accelerator con-
105. Gabani P, Lin AJ, Barnes J, Oppelt P, Adkins DR, trol: gating on an MRI accelerator. Phys Med Biol.
Rich JT, et al. Radiation therapy dose de-escalation 2011;56(15):4815–25.
compared to standard dose radiation therapy in 118. Kontaxis C, Bol GH, Lagendijk JJ, Raaymakers
definitive treatment of HPV-positive oropharyn- BW. A new methodology for inter- and intrafraction
geal squamous cell carcinoma. Radiother Oncol. plan adaptation for the MR-linac. Phys Med Biol.
2019;134:81–8. 2015;60(19):7485–97.
106. Winkel D, Bol GH, Kroon PS, van Asselen B, 119. Zachiu C, Papadakis N, Ries M, Moonen C, Denis
Hackett SS, Werensteijn-Honingh AM, et al. de Senneville B. An improved optical flow track-
Adaptive radiotherapy: the Elekta Unity MR-linac ing technique for real-time MR-guided beam
concept. Clin Transl Radiat Oncol. 2019;18:54–9. therapies in moving organs. Phys Med Biol.
107. de Muinck Keizer DM, Kerkmeijer LGW, 2015;60(23):9003–29.
Willigenburg T, van Lier A, Hartogh MDD, van 120. Willigenburg T, Zachiu C, Lagendijk JJW, van der
der Voort van Zyp JRN, et al. Prostate intrafrac- Voort van Zyp JRN, de Boer HCJ, Raaymakers
tion motion during the preparation and delivery BW. Fast and accurate deformable contour propaga-
of MR-guided radiotherapy sessions on a 1.5T tion for intra-fraction adaptive magnetic resonance-
MR-Linac. Radiother Oncol. 2020;151:88–94. guided prostate radiotherapy. Phys Imaging Radiat
108. Bruijnen T, Stemkens B, Terhaard CHJ, Lagendijk Oncol. 2022;21:62–5.
JJW, Raaijmakers CPJ, Tijssen RHN. Intrafraction 121. Uijtewaal P, Borman PTS, Woodhead PL, Kontaxis
motion quantification and planning target volume C, Hackett SL, Verhoeff J, et al. First experimen-
margin determination of head-and-neck tumors tal demonstration of VMAT combined with MLC
using cine magnetic resonance imaging. Radiother tracking for single and multi fraction lung SBRT on
Oncol. 2019;130:82–8. an MR-linac. Radiother Oncol. 2022;174:149–57.
109. Kluter S. Technical design and concept of a 122. Yang J, Mohamed ASR, Bahig H, Ding Y, Wang J,
0.35 T MR-Linac. Clin Transl Radiat Oncol. Ng SP, et al. Automatic registration of 2D MR cine
2019;18:98–101.
images for swallowing motion estimation. PLoS optic endoscopic evaluation of swallowing and nutri-
One. 2020;15(2):e0228652. tional status. Dysphagia. 2022;37(1):137–47.
123. Bradley JA, Paulson ES, Ahunbay E, Schultz C, Li 136. Hamzic S, Braun T, Juenemann M, Butz M,
XA, Wang D. Dynamic MRI analysis of tumor and Voswinckel R, Belly M, et al. Validation of the
organ motion during rest and deglutition and margin German version of Functional Oral Intake Scale
assessment for radiotherapy of head-and-neck cancer. (FOIS-G) for flexible endoscopic evaluation of swal-
Int J Radiat Oncol Biol Phys. 2011;81(5):e803–12. lowing (FEES). Dysphagia. 2021;36(1):130–9.
124. Hoogcarspel SJ, Zijlema SE, Tijssen RHN, 137. Bjordal K, Hammerlid E, Ahlner-Elmqvist M, de
Kerkmeijer LGW, Jurgenliemk-Schulz IM, Graeff A, Boysen M, Evensen JF, et al. Quality of
Lagendijk JJW, et al. Characterization of the first life in head and neck cancer patients: validation
RF coil dedicated to 1.5 T MR guided radiotherapy. of the European Organization for Research and
Phys Med Biol. 2018;63(2):025014. Treatment of cancer quality of life questionnaire-
125. Cuccia F, Alongi F, Belka C, Boldrini L, Hörner- H&N35. J Clin Oncol. 1999;17(3):1008–19.
Rieber J, McNair H, et al. Patient positioning and 138. Kaasa S, Bjordal K, Aaronson N, Moum T, Wist E,
immobilization procedures for hybrid MR-Linac Hagen S, et al. The EORTC core quality of life ques-
systems. Radiat Oncol. 2021;16(1):183. tionnaire (QLQ-C30): validity and reliability when
126. Boeke S, Monnich D, van Timmeren JE, Balermpas analysed with patients treated with palliative radio-
P. MR-guided radiotherapy for head and neck can- therapy. Eur J Cancer. 1995;31A(13–14):2260–3.
cer: current developments, perspectives, and chal- 139. List MA, D’Antonio LL, Cella DF, Siston A,
lenges. Front Oncol. 2021;11:616156. Mumby P, Haraf D, et al. The performance status
127. Chuter RW, Whitehurst P, Choudhury A, van Herk scale for head and neck cancer patients and the
M, McWilliam A. Technical note: investigating the functional assessment of cancer therapy-head and
impact of field size on patient selection for the 1.5T neck scale. A study of utility and validity. Cancer.
MR-Linac. Med Phys. 2017;44(11):5667–71. 1996;77(11):2294–301.
128. Ng-Cheng-Hin B, Nutting C, Newbold K, Bhide 140. Chen AY, Frankowski R, Bishop-Leone J, Hebert
S, McQuaid D, Dunlop A, et al. The impact of T, Leyk S, Lewin J, et al. The development and
restricted length of treatment field and anthropo- validation of a dysphagia-specific quality-of-life
metric factors on selection of head and neck cancer questionnaire for patients with head and neck can-
patients for treatment on the MR-Linac. Br J Radiol. cer: the M. D. Anderson dysphagia inventory. Arch
2020;93(1111):20200023. Otolaryngol Head Neck Surg. 2001;127(7):870–6.
129. Monnich D, Winter J, Nachbar M, Kunzel L, Boeke 141. Rosen CA, Lee AS, Osborne J, Zullo T, Murry
S, Gani C, et al. Quality assurance of IMRT treat- T. Development and validation of the voice handicap
ment plans for a 1.5 T MR-linac using a 2D ioniza- index-10. Laryngoscope. 2004;114(9):1549–56.
tion chamber array and a static solid phantom. Phys 142. Kaae JK, Johnsen L, Hansen CR, Kristensen MH,
Med Biol. 2020;65(16):16NT01. Brink C, Eriksen JG. Relationship between patient
130. Strand S, Boczkowski A, Smith B, Snyder JE, and physician-rated xerostomia and dose distribu-
Hyer DE, Yaddanapudi S, et al. Analysis of patient- tion to the oral cavity and salivary glands for head
specific quality assurance for Elekta Unity adaptive and neck cancer patients after radiotherapy. Acta
plans using statistical process control methodology. Oncol. 2019;58(10):1366–72.
J Appl Clin Med Phys. 2021;22(4):99–107. 143. Mohamed AS, Rosenthal DI, Awan MJ, Garden AS,
131. Kim A, Ruschin ME, McCann C, Au P, Singh A, Kocak-Uzel E, Belal AM, et al. Methodology for
Sahgal A, et al. Streamlining patient-specific qual- analysis and reporting patterns of failure in the era
ity assurance for the 1.5T MR-Linac using a clas- of IMRT: head and neck cancer applications. Radiat
sic statistics tool. Int J Radiat Oncol Biol Phys. Oncol. 2016;11(1):95.
2020;108(3):e350–1. 144. Resteghini C, Trama A, Borgonovi E, Hosni H,
132. Rhatigan E, Tyrmpas I, Murray G, Plevris Corrao G, Orlandi E, et al. Big data in head and neck
JN. Scoring system to identify patients at high risk of cancer. Curr Treat Options Oncol. 2018;19(12):62.
oesophageal cancer. Br J Surg. 2010;97(12):1831–7. 145. Van den Bosch L, van der Schaaf A, van der Laan
133. Crary MA, Mann GD, Groher ME. Initial psycho- HP, Hoebers FJP, Wijers OB, van den Hoek JGM,
metric assessment of a functional oral intake scale et al. Comprehensive toxicity risk profiling in radia-
for dysphagia in stroke patients. Arch Phys Med tion therapy for head and neck cancer: a new con-
Rehabil. 2005;86(8):1516–20. cept for individually optimised treatment. Radiother
134. Bakhtiyari J, Tohidast SA, Mansuri B, Azimi H, Oncol. 2021;157:147–54.
Ebadi A. The Persian version of the Functional Oral 146. Langendijk JA, Hoebers FJP, de Jong MA,
Intake Scale (FOIS-P): a validation study on stroke Doornaert P, Terhaard CHJ, Steenbakkers R, et al.
patients with dysphagia. Logoped Phoniatr Vocol. National protocol for model-based selection for pro-
2022;47(2):133–8. ton therapy in head and neck cancer. Int J Part Ther.
135. Ninfa A, Pizzorni N, Eplite A, Moltisanti C, 2021;8(1):354–65.
Schindler A. Validation of the Italian version of the 147. Lewis BC, Shin J, Maraghechi B, Quinn B, Cole M,
Functional Oral Intake Scale (FOIS-It) against fiber- Barberi E, et al. Assessment of a novel commercial
large field of view phantom for comprehensive MR 151. Morgan HE, Sher DJ. Adaptive radiotherapy for
imaging quality assurance of a 0.35T MRgRT sys- head and neck cancer. Cancers Head Neck. 2020;5:1.
tem. J Appl Clin Med Phys. 2022;23(4):e13535. 152. Ahn PH, Chen CC, Ahn AI, Hong L, Scripes PG, Shen
148. Ng SP, Cardenas CE, Bahig H, Elgohari B, Wang J, et al. Adaptive planning in intensity-modulated
J, Johnson JM, et al. Changes in apparent diffu- radiation therapy for head and neck cancers: single-
sion coefficient (ADC) in serial weekly MRI during institution experience and clinical implications. Int J
radiotherapy in patients with head and neck cancer: Radiat Oncol Biol Phys. 2011;80(3):677–85.
results from the PREDICT-HN study. Curr Oncol. 153. Oktay O, Nanavati J, Schwaighofer A, Carter D,
2022;29(9):6303–13. Bristow M, Tanno R, et al. Evaluation of deep learn-
149. Anderson CM, Sun W, Buatti JM, Maley JE, Policeni ing to augment image-guided radiotherapy for head
B, Mott SL, et al. Interobserver and intermodality and neck and prostate cancers. JAMA Netw Open.
variability in GTV delineation on simulation CT, 2020;3(11):e2027426.
FDG-PET, and MR images of head and neck cancer. 154. Neubauer E, Dong L, Followill DS, Garden AS,
Jacobs J Radiat Oncol. 2014;1(1):006. Court LE, White RA, et al. Assessment of shoul-
150. Li X, Yadav P, McMillan AB. Synthetic computed der position variation and its impact on IMRT and
tomography generation from 0.35T magnetic reso- VMAT doses for head and neck cancer. Radiat
nance images for magnetic resonance-only radiation Oncol. 2012;7:19.
therapy planning using perceptual loss models. Pract
Radiat Oncol. 2022;12(1):e40–8.
The Australian MRI-Linac:
A 10-Year Journey
22
Gary Liney
Introduction My involvement in the project began in mid-

2012 when I applied to the position of MRI phys-
Having been asked many times to contribute to icist at Ingham Institute in southwest Sydney.
papers and book chapters on MRI-Linac technol- This was to lead the scientific input into two key
ogy over the years, it becomes increasingly difficult programs of research, namely, the dedicated MRI
to provide the reader with new material—even in simulator at Liverpool cancer therapy center and
such an evolving field. In this case, serendipity the prototype MRI-Linear accelerator being
means this latest invitation comes at the end of my developed in a research bunker within that same
contract and association with the Australian MRI- department. I was leaving a job-for-life in the
Linac (AMRL) project. This has given me the UK’s national health service for a 3-year con-
opportunity to look back on my last 10 years’ work tract, which on the face of it seemed rather
on this system. What follows is a personal reflection strange. However, I was also swapping a position
of the progress that was made as well as some of the of a lead imaging physicist in a UK radiotherapy
issues that were not always possible to overcome. I department that had a perpetual “5-year plan” to
will talk about some of the distinguishing features acquire an MRI scanner—that still has not mate-
and technical specifications only briefly, and for a rialized—to one in Australia that would have
more detailed discussion, the reader is directed else- access to dedicated MRI equipment on both a
where [1, 2]. The points made are my own and not clinical and research basis. Or so I thought.
necessarily the opinion of others on the project. Imagine my horror when I arrived to see that nei-
However, as someone that led the scientific devel- ther of these pieces of MRI equipment actually
opment from an empty bunker way back in 2012 to even existed: The magnet for the MRI-Linac had
a fully functioning system, I feel I am uniquely encountered serious delays, and the business case
qualified to do this. Furthermore, while acknowl- for the simulator had not started. And so began
edging there were (and still are) countless people what turned into my decade-long involvement.
associated with the project, I will instead be focus-
ing on the work undertaken by the small team I led
as part of the development of the AMRL, which he Australian MRL: A 10-Year
T
ultimately delivered the working system. Journey
The Australian MRI-Linac (AMRL) was a col-

G. Liney (*)
Sydney, NSW, Australia laboration across multiple institutions including
e-mail: [email protected] the Universities of NSW, Sydney, Queensland,
426 G. Liney
and Wollongong. It was operated by the Ingham decision to finally go ahead was granted at New
Institute and housed in a dedicated research bun- South Wales state level, and this ultimately led to
ker which was completed in August 2012 within the installation of a 3-T Siemens Skyra midway
Liverpool hospital in southwest Sydney. The through 2013. The implementation of this scan-
project was the brain child of the “three amigos” ner into clinical practice was a principal focus of
Paul Keall, Michael Barton, and Stuart Crozier the first couple of years—especially with the
who were the lead CIs that had procured a signifi- ongoing delays on the AMRL—but in so doing
cant amount of NHMRC funding [3]. there were clearly many indirect benefits to the
The original intention of the Australian MRL AMRL program. A good example of this was the
had been to provide a system that allowed for a testing of a manually driven rotating couch proto-
degree of flexibility in its operation with the type that fitted inside the 70-cm-wide bore sys-
potential for both inline and perpendicular orien- tem (built by the late Rob Wilkins) and allowed
tations. At the time there was still much debate as us to replicate and study the effects and imaging
to which beam-magnetic field configuration performance of patient rotation [4]. Beyond the
would win the day, with both having advantages protocol development and training on this scan-
and disadvantages in terms of interactions and ner, I was asked to negotiate a research agree-
engineering complexities. The AMRL system ment with Siemens. This also led to us to access
was designed around a large open bore that could many early preclinical sequences such as ultra-
be accessed by the patient or linear accelerator in short TE and 4D with these being relevant to
either orientation. In order to achieve this degree MR-only planning and motion management.
of flexibility, the Linac gantry was to be fixed Around the end of 2013, we recruited our
during treatment, and instead the patient would chief engineer Bin Dong—initially having little
be moved and rotated in front of the beam. experience of radiotherapy—he went on to
Similar-looking split-bore designs had been become a vital cog in the operation of the AMRL
used in the 1990s for surgical interventions. and was a fantastic support over the years. By
However, at 1 Tesla, the field strength selected for mid-2014, with the projected 12-month delay to
the AMRL was double that previously achieved; the magnet already slipping, it was decided that I
this ambitious figure was a best guess compromise would return to England on a fact-finding mis-
between optimal image quality and practicality. sion. In a combined road trip of 1000 miles, I vis-
Furthermore, a 30-cm working imaging volume ited the factory producing the gradient coils on
and a near-zero field region 1.4 m from the isocen- the south coast; the Agilent factory providing the
ter had also been stipulated in the blueprint. The magnet, near Oxford; and the manufacturer of the
homogeneity and active shielding requirements of remaining fourth magnet former in the northeast.
such a design required extreme precession engi- At this point only three of the four formers had
neering of four superconducting formers, and been successfully made with cracks having
these took nearly 2 years to successfully manufac- appeared in the last of the original set. The added
ture. I entered my first meeting of the project in time pressure was that Agilent were getting out of
November 2012 and left an hour or so later (armed the bespoke magnet business and instead concen-
with a photograph of the “high level” marker trating on small-bore systems going forward,
board plan) having being told I was expected to meaning that ours was to be the last magnet to
lead the entire project through to its completion! leave their factory. I returned to Australia with
Looking back, I could not have imagined just how many reassurances and some trepidation con-
much work and responsibility was falling on my cerning the completion of the magnet.
shoulders, but it would certainly be challenging. In the meantime, there was plenty of activity
The first period of my tenure was however proceeding in the bunker and the adjacent control
spent on producing a business case for the afore- and equipment rooms. The portable linear accel-
mentioned MRI simulator and its ensuing tender erator (a Linatron from Varian and now Varex)
process. After a very nervous few months, the was delivered and over the next few months was
22 The Australian MRI-Linac: A 10-Year Journey 427
installed into a fixed position in the middle of the driven 2000 miles cross country to be installed in
bunker. The gradient cabinets were also installed our empty bunker. The agreement would see us
at this stage, composed from a tailor-made list of use this MRI for the next 5 or 6 months allowing
Siemens Avanto components. I became con- us to do anything other than human imaging.
cerned that no thought was being given to the Both myself and Bin worked over the Christmas
proximity of the Linatron and committing to a period to gain as much experimental data of the
fixed position seemed too risky. At this point in beam-field interactions as we could [5]. This
the MRL story, interactions between components included our first ever beam-on image which,
was still very much trial and error. Out of an with a nod to the Utrecht group’s first pork chop
abundance of caution, I began exploring a rail image, was a kangaroo steak! As it turned out, we
system that could position the Linac at varying only got about half of the expected time with the
distances. Initially a design was considered that magnet owing to a water leak and an unintended
moved it on an arc between the two proposed ori- quench. The prohibitive cost of refilling the
entations, but this was quickly abandoned for a helium put a sudden end to “Phase I” of the proj-
design that allowed it to more simply translate ect. Nevertheless, in the intervening period, we
backwards and forwards at the end of the bore. had managed to achieve a fair amount of prog-
We had taken the first stage of committing the ress. This included a firsthand practical demon-
MRI-Linac to an inline orientation—but it also stration of the electron focusing effect and its
meant we now had the ability to vary the interact- potential mitigation by shooting the beam off
ing field strength by virtue of its proximity to the axis. We set up radiochromic films at various
magnet. A couple of photographs from the early points through and out the other side of the bore
stages of the project are shown in Fig. 22.1. and could see the primary photon divergence
The research relationship with Siemens would while the electrons were first pulled and then
pay dividends from the MR-Linac perspective in repelled from opposite sides of the isocenter.
one other major way. As the delays to magnet These films generated a lot of excitement,
production dragged on, I approached Siemens to although we struggled to convince some of our
see whether we could make use of an ex-clinical colleagues as to the explanation! We had also
scanner that was being removed from Royal demonstrated a dosimetric application for the
Perth Hospital and potentially scrapped. And so radiation induction we were observing in some
in November 2015, this 1.5-Tesla Sonata was receiver coils, which would lead to a US patent
Fig. 22.1 Early photographs from the Australian MRI- eventually be mounted onto a sliding rail system. (Right)
Linac project. (Left) The Varian Linatron being installed The RF cage nearing completion; the quench pipe awaits
on fixed footplates toward the end of 2014. This would connection to the magnet in this photo taken in mid-2015
428 G. Liney
being granted [6]. Sadly this was never exploited

further and in fact Ingham’s first and only tech-
nology patent has since been allowed to lapse.
The Agilent magnet was finally installed in
April 2017 to a flurry of media interest. The
12-ton magnet looked immense as it was being
craned overhead. I spoke to a local reporter and
jokingly said that the only thing that could go
wrong was if it was too big to fit in the room. And
of course, as if to tempt fate, that is exactly what
happened! Errors on the factory blueprint had
shown a shorter scaffold overhang above the
magnet meaning it would not fit through the bun-
Fig. 22.2 Photograph of the bore of the system showing
ker door in the way intended. Some last-minute the patient opening circa 2017. A glass RF window was
robust adjustments were made to the entrance of installed at this time to further increase the light and the
the bunker, and it eventually went through feeling of space. The patient opening is 78 × 50 cm, and
“lengthways.” Once inside the room, it then took the table can be removed to permit imaging in the stand-
ing position
many hours of iterations to move the magnet with
specialist lifting equipment into its final position.
This had to be as close as possible to the cage Bin, as he grappled with third party and homebuilt
wall—the thickness of a cable—and exactly apparatus to link up the manufactured compo-
square on, to allow the MLC to benefit from the nents. Having completed this, attention turned to
zero field just 1.4 m from the middle of the software integration led by Kevin Zhang with
magnet. some assistance from Siemens. All things relevant
To complete the hardware, a dedicated split to a 1.5-Tesla system had to be reconfigured to the
gradient module and RF transmit receive coil new (actually just shy of 1.0 Tesla) field strength.
were designed to fit inside the opening with as In addition, the reconstruction engine was modi-
little intrusion into the gap as possible as can be fied so images could be sent to control the
seen in Fig. 22.2. The gradient coils were manu- MLC. Our first ever test image of a very distorted
factured as a single modular unit for each magnet QA phantom was acquired via remote help from
half by Tesla, UK. Each half of the module, Erlangen in Germany. A few tweaks to the gradi-
which contains a series of passive shim trays, had ent parameters later and we were ready to image
been test fitted in the UK, shipped separately, and our first subject. I was more than happy to volun-
was then reinserted by a fork lift once the magnet teer (admittedly the queue to do so was very
was inside the bunker. Also shown in Fig. 22.2 short!) and climbed in between the magnet pillars
are two of the four rungs that make up the bird- keeping 40 tons of attractive force apart. The first
cage design of the integrated transceiver coil. images are shown in Fig. 22.3, and seeing these at
This was designed (Magnetica, Brisbane) and the end of that test session will rank as a most
tested using a full-scale mock-up of the magnet memorable moment of my career. In the May of
gap and was the last of the MRI hardware to be that year, we had got the MLC tracking to work
installed [7]. The four legs of the coil run along just in time for our planned bunker tour on the
the support pillars in the magnet gap to maximize final day of the MRI in RT symposium being held
the patient space. In maintaining this opening, in Sydney. A demonstration involved Bin moving
the primary beam is unobstructed creating no a bottle of water backward and forward manually
radiation induction effects, and image quality is in front of the MLC whose leaf positions dutifully
unaffected [8]. followed the sub-second images. This managed to
Much of the next few months were spent on impress all of our visitors even the more seasoned
the hardware integration chiefly by our engineer MRI-Linac experts among the contingent.
Fig. 22.3 (Left) The official first image on the Australian transceiver coil. (Right) The actual first human image
MRI-Linac. This is a 5-min T2-weighted TSE image of (also of the author) which was acquired indirectly as part
the author taken in the “beams’ eye view” using the open of initial RF heating measurements
Attention turned to completing the commis- By the end of 2018, the commercial systems
sioning of the beam and dosimetry. Urszula Jelen, were up and running in other parts of the world,
a medical physicist with experience working on and a first installation in Australia was scheduled
particle accelerators, was brought on board to for Townsville, northern Queensland, in 2019. At
complete this work [9]. At this moment in 2017, this point I was loudly advocating the need for us
no one had yet been treated on an MRL anywhere to adopt a more general MRI guidance approach
in the world, and although we were clearly many in response to these developments. Clearly there
years behind the commercial developments, it were a lot of things we could achieve with the
really did feel like we could make some impact. system from a patient perspective alone. It was
We started to examine various planning and treat- extremely open and well tolerated, and at one
ment strategies including how we could deter- point we inserted an RF window to maximize this
mine patient position from rapid imaging as well feeling of space even further. We’d already started
as demonstrating MR-MR registration possibili- using this space to great effect to perform vertical
ties between MRL and high-resolution data from imaging which even attracted the attention of the
the 3 T. Australian defense force who were interested in
Once the AMRL was up and running a num- upright imaging as a way of designing tailored
ber of PhD projects from the various collaborat- body armor. The magnet had other possibilities
ing institutions were able to begin in earnest. One too, in terms of adjunct therapeutic equipment
of the more fruitful of these was a study investi- that could fit in and around the bore enabling
gating the effects of gadolinium dose enhance- MRI guidance of various complementary
ment in rodents [10, 11]. This actually led to the treatments.
first treatment use of an inline MRL anywhere in The final component of the AMRL was the
the world. It also enabled skin dose mitigation in rotating patient couch which was developed over
prospective patient trials via a simple bolus to be three prototypes and over several years working
studied. Ongoing solutions to this issue are still closely with Craig Andrews and his colleagues at
being examined including off-axis irradiation to Design Momentum as shown in Fig. 22.4. After a
keep electrons down the isocenter and out of the number of attempts at trying to accommodate
beam, work which builds directly on earlier tests motors inside the scan room, I got them to com-
performed on the Sonata magnet during Phase I pletely rethink the approach and instead we came
[12]. up with a solution that worked from outside the
430 G. Liney
Fig. 22.4 (Left) Early days of imaging using a wooden about to pass earplugs to Gary Liney prior to some imag-
bench top that had to be lifted into place and climbed onto ing tests. (Right) The final version of the rotating couch
by subjects. By mid-2017 this had been replaced with a shown in situ taken in 2022...after all that, I hear this is
moveable table. In this photo chief engineer Bin Dong is now broken!
Fig. 22.5 Illustration of

the extended FOV
possible by a moving
table acquisition. In total
an 80-cm length is
possible by
concatenating smaller
sections to provide a
final distortion-free
image
room. The final design is now driven from exter- also allows extended FOV imaging which effec-
nal motors that connect to the bed via shafts that tively increases the DSV in the x-direction by a
pass through a waveguide; this reduces artifacts factor of three. By concatenating smaller image
and makes for a more accurate and quiet perfor- sections with simultaneous incremental table
mance. The patient rotation provides a unique movements, we could provide distortion-free
solution to the fixed gantry design and is images of up to 80 cm in length with no need for
potentially very attractive in situations where the gradient correction, B0 shimming, or additional
beam line may be much harder to move as in the RF coils as shown in Fig. 22.5. A very similar
case of proton therapy. The couch is easily implementation now exists on some clinical
docked and retracted and once removed still scanners. I had described this method in our first
allows clear access to the bore for vertical imag- group meeting in 2012, and we had actually tried
ing providing an alternative treatment position. this with the manual table previously, but it
The large couch translation from the isocenter wasn’t until the accuracy of the new couch could
we achieve it properly. Example images shown in develop a more general MRI guidance approach.
Fig. 22.5 show a longer FOV than what is even It remains to be seen whether there are any plans
possible on clinical scanners, with greater unifor- to change direction in the future.
mity and no need for surface coils on the subject. A much more tangible legacy has been the
This may have benefits in the spine and as the impetus that the project provided generally to the
couch can also move in two other directions MRI in RT field within Australia. It was a focal
could be applied to other anatomy. And with that point of numerous meetings and conferences
last project, and the couch installation now com- over the years, including the international MRI in
plete, marking just over 10 years in post, my RT symposium, which I was very excited to chair
involvement was no longer required. in Sydney in 2017. This remains the only time the
event has taken pace outside of Europe or the
USA. This conference led to a national spin-off
Summary which is still very active. The landscape of MRI
in RT in Australia has certainly changed in the
The Australian MRI-Linac was certainly a sig- last decade. When I arrived in 2012, there was no
nificant part of my life. With neither the financial dedicated MRI use in radiotherapy and of course
or personnel resources of the other big MRL no MRI-Linacs. As of today, there are four MRI
developments, the scale of the task cannot be simulators and four installations of the Elekta
overstated. Engagement of wider clinical staff in Unity MRI-Linac system. These figures stand up
a department unused to this type of work was very well against other countries when you con-
always tentative. The development ultimately sider they serve a combined population of only
depended on a very small number of enthusiastic 26 million. Interestingly only one of these depart-
and dedicated people. Many careers have benefit- ments has both an MRL and MRI scanner in situ.
ted enormously from their association with the Three of the MRI-simulator sites operate in isola-
AMRL, although sadly those of us directly tion and have no access to MRI-Linac. These are
involved have had to move on. The failure to quite different working models, and it will be
retain our engineer—so key to its continued oper- interesting to see how this plays out in the next
ation—was a particularly significant blow. few years. The dedicated MRI simulator at
So what of the future of the AMRL? Well as Liverpool hospital—after a 10-year lifespan—
part of the proposed clinical trial, the Ingham has still never exceeded 40% patient throughput;
Institute handed over temporary ownership of its Access to a nearby MRI-guided treatment system
once flagship research program to the cancer ther- would clearly be an advantage, and those centers
apy team in 2020. This team was charged with lucky enough to have both would seem to be in
delivering the first patient treatment. The trial is a the ideal position to exploit MRI use to its maxi-
basic parallel pair radiation treatment for pallia- mum. In any case, there is clearly a strong future
tive brain cases and is now the sole focus of the for MRI in RT in Australia, and although I am no
AMRL. The acquisition of the MRI plays no part longer included in it, I am proud to have contrib-
in the therapy, a move that looks increasingly out uted to this in some way.
of step with the adaptive MR-guided treatments
being performed at other hospitals. As of time of
writing, the animal study remains the only treat- References
ment yet performed on the system, and while this
could be seen as a failure with respect to the com- 1. Whelan B, Liney GP, Oborn B, Keall P. MRI linac
peting technologies, it should be remembered that systems. In: Liney GP, van der Heide U, editors. MRI
the original intention had never been to treat (or for radiotherapy. Springer. 2019. pp. 155–168.
2. Liney GP, Whelan B, Oborn B, Barton M, Keall
even image) a patient. In my opinion it would P. MRI-linear accelerator radiotherapy systems. Clin
have been much better to concentrate on the Oncol (R Coll Radiol). 2018;30(11):686–91. https://
unique aspects of the system—the magnet—and doi.org/10.1016/j.clon.2018.08.003.
432 G. Liney
3. Keall PJ, Barton M, Crozier S. The Australian mag- on a 1.0 Tesla inline MRI-linac. Phys Med Biol.
netic resonance imaging linac program. Semin Radiat 2018;63:135005.
Oncol. 2014;24:203–6. 9. Jelen U, Dong B, Begg J, et al. Dosimetric optimi-
4. Whelan BM, Liney GP, Dowling J, et al. An MRI- zation and commissioning of a high-field inline
compatible patient rotation system design, construc- MRI-Linac. Front Oncol. 2020;10:136. https://doi.
tion, and first organ deformation results. Med Phys. org/10.3389/fonc.2020.00136.
2017;44(2):581–8. https://doi.org/10.1002/mp.12065. 10. Liney GP, Jelen U, Byrne H, et al. Technical note:
5. Liney GP, Dong B, Begg J, et al. Experimental the first live treatment on a 1.0 Tesla inline MRI-
results from a prototype high-field inline linac. Med Phys. 2019;46(7):3254–8. https://doi.
MRI-
linac. Med Phys. 2016;43:5188. https://doi. org/10.1002/mp.13556.
org/10.1118/1.4961395. 11. Byrne HL, et al. Enhanced MRI-guided radiotherapy
6. Liney G, Dong B. Radio-frequency coil for simul- with gadolinium-based nanoparticles: preclinical eval-
taneous imaging and dosimetry on an MRI-Linac. uation with an MRI-linac. Cancer Nano. 2020;11:9.
Patent no. 10391335. https://doi.org/10.1186/s12645-020-00065-5.
7. Destruel A, Weber E, Hughes I, Li Y, Liu F, Crozier 12. Roberts NF, Patterson E, Jelen U, et al. Experimental
S. Design of a whole-body radiofrequency coil for characterization of magnetically focused electron
image-guided radiotherapy treatment in a MRI-Linac contamination at the surface of a high-field inline
system 3076 Proc. of ISMRM. 2015. MRI-linac. Med Phys. 2019;46(12):5780–9. https://
8. Liney GP, Dong B, Begg J, et al. Imaging perfor- doi.org/10.1002/mp.13847. Epub 2019 Oct 21.
mance of a dedicated radiation transparent RF coil
Low-Field MR Imaging
23
Yingli Yang, Shaolei Li, Daniel A. Low, Zhimin Li,
and Peng Hu
Introduction 1980s and early 1990s, superconducting magnets

were developed which allowed higher field
Magnetic resonance imaging (MRI) is a mainstay strengths and better field homogeneity. Since
diagnosis tool in modern medicine, providing then, MRI has been on a continuous trajectory of
significant value in both scientific and clinical higher field strengths in pursuit of greater signal
applications. MRI offers excellent soft-tissue to noise ratio (SNR), with 3-T scanners becom-
contrast and the capability of probing a wealth of ing a clinical mainstay and 7-T and greater sys-
physiological and physical parameters. The tems being more widely adopted.
development of MRI systems in the early days However, in recent years, there has been a
started with low-field systems and permanent renaissance in low-field MRI systems because of
magnets. These permanent magnet designs had economic, technological, and scientific factors.
an upper limit in achievable field strength and From an economic perspective, MRI is widely
relatively poor B0 field uniformity. In the late used in high-income countries (HICs) with 0.19
scans/persons annually in the USA [1, 2] but has
limited adoption in low-income and middle-
Y. Yang (*) · S. Li income countries (LMICs) due to high costs and
Institute for Medical Imaging Technology, Ruijin many technical barriers. Even inside HICs, MRI
Hospital, Shanghai Jiaotong University School of
is usually unavailable to rural areas or for patients
Medicine, Shanghai, China
with disability or device constraints [3].
Department of Radiology, Ruijin Hospital, Shanghai
Worldwide, only 10 percent of the population has
Jiaotong University, School of Medicine,
Shanghai, China access to MRI, and even within LMICs an uneven
e-mail: [email protected] distribution of this important diagnostic tool per-
D. A. Low sists [4, 5]. There are two major factors respon-
Department of Radiation Oncology, University of sible for this situation: (1) the cost of installation
California, Los Angeles, CA, USA and post-installation maintenance and (2) the
complexity of operating MR systems and (3)
Z. Li technical know-how. Given the extraordinary
Department of Radiology, Northwestern University,
imbalance in healthcare availability in LMICs,
Chicago, IL, USA
e-mail: [email protected] there has been increasing interest in revisiting
low-field MRI, which, while not producing the
P. Hu
School of Biomedical Engineering, ShanghaiTech highest-quality images, should be able to provide
University, Shanghai, China diagnostically useful information. From a tech-
434 Y. Yang et al.
nological perspective, the past 30 years have seen it was generally accepted that higher static field
tremendous advances in relevant technology, strength would inevitably lead to improved per-
from magnet design to gradient and radiofre- formance. This was based on the notion that the
quency system performance. Many technologies MR signal amplitude is a quadratic function of the
developed after the era of lower-performance static magnetic field strength while the noise is
low-field systems, such as active shielding, paral- only linearly proportional to the field strength.
lel imaging [6], and compressed sensing [7], can According to this principle, a 3-T MRI system
now be adopted by modern low-field systems, would theoretically have 15 times the SNR com-
greatly boosting their performance. More impor- pared to a 0.2-T MRI system [11, 12].
tantly, such a setup could potentially allow for The concept of linear SNR dependence on
interesting applications that would be challeng- magnetic field strength has since been deter-
ing for conventional higher-field strength sys- mined to be not entirely true [13, 14]. T1 and T2
tems. For example, MRI-guided catheter changes at higher field strengths could partially
interventions, where the patient undergoes a negate those SNR gains due to higher magnetic
catheterization procedure under continuous or polarization at higher field strengths. The pro-
intermittent MRI guidance and real-time feed- gression in both software, including pulse
back, are potentially more feasible at lower field sequences and hardware, including gradient and
strengths due to reduced heating of various imaging coils, has enabled relatively high-
devices [8, 9] in the MRI environment. Lung performance imaging techniques even at lower
imaging is another example where low-field sys- field strengths. In the past, imaging sequences for
tems could provide superior imaging due to the low-field MRI systems were relatively limited
reduction of magnetic susceptibility effects [10]. and basic, but now they have become quite
Low-field MRI is usually defined as systems sophisticated and are comparable to the imaging
with <1.0-T main magnetic fields [2]. In Fig. 23.1, sequences of 1.5-T MRI systems that are widely
low-field MRI systems are shown with a variety used clinically [8]. In addition, the low SNR
of main magnetic fields. Originally, the perfor- inherent in low-field MRI systems has been
mance of low-field MRI systems was rated lower improved by image reconstruction methods using
than their actual clinical usefulness, and they were deep learning and de-noising techniques. This
viewed as low-cost but poorly performing sys- has further improved image quality, and studies
tems. The older type of MRI systems had limited have shown sufficient imaging performance for
spatial resolution associated with poor image various clinical applications [15, 16]. Low field
quality, limited available receiver coils, limited also provides benefits from a patient safety and
types of imaging sequences and parameters, and comfort perspective, in terms of reduced risks of
insufficient temporal resolution associated with projectiles, decreased implanted device heating,
low SNR. In the early phase of MRI development, and ability to image tissues closer to implants due
Hyperfine Swoop Promaxo Seimens Magnetom Synaptive Evry Siemens Magnetom Aspect Embrace
0.064T 0.066T 0.2T 0.5T Free.Max 0.55T 1.0T
Fig. 23.1 Defining “low-field” MRI. This figure defines a field strength of 1 Tesla or lower are illustrated in this
how scanners at difference field strengths will be catego- figure. Scanners are not to scale. Scanner images are
rized in the article, with the following boundaries illus- copyright of the respective manufacturers. (Images used
trated on the bottom: low-field (LF) ≤ 1.0 T < high-field with permission or in accordance with manufacturer pol
(HF) ≤ 3 T < very-high-field (VHF) < 7 T ≤ ultra-high- [2])
field (UHF). Select commercially available scanners with
23 Low-Field MR Imaging 435
to smaller magnetic susceptibility, reduced spe- 3–5 years. The homogeneity of the magnetic field
cific absorption rate (SAR) for high-SAR pulse should be on the order of parts per million (ppm)
sequences, and reduced acoustic noise because of over an ellipsoidal imaging volume, with fluctua-
lowered forces on the gradient coil windings with tions during the scan period of less than 100 nT
a given current amplitude [12]. The other attrac- [12]. For a low-field system operating at between
tive points of low-field systems are mobility due 0.25 and 0.5 T, there are two basic choices of the
to the lower weight, which could take MRI to the magnet, one using a permanent neodymium-iron-
point-of-care, similar to an ultrasound unit. boron magnet or one that combines a permanent
Lower-field MRI still faces significant chal- magnet with an additional electromagnet. A more
lenges, and it remains to be seen how newer recent system, the Siemens Free.Max, uses a
devices will be deployed clinically. In this chap- 0.55-T superconducting magnet.
ter, we discuss the advantages of and challenges
for low-field MRI systems. In addition to diag-
nostic imaging, we will also further illustrate the Designing Coils for Low-Field MRI
potential application of low-field MRI to radio-
therapy in the setting of MRI-guided linear An MRI system as shown in Fig. 23.2 is a mix of
accelerators. several subsystems which provide the necessary
functions for imaging.
The general consideration for designing coils
hysical Properties and Engineering
P (especially RF coils) includes the direction of the
Challenges of Low-Field MRI main magnetic field, bore diameter, size of the
homogeneous volume, and the relation between
Since the invention of MRI, researchers have electric current and magnetic field. The coils are
made efforts to pursue higher magnetic fields. designed as resonant structures to maximize their
Worldwide, there have been almost the same efficiency and sensitivity. The resonant frequency
number of 1.5-T and 3-T systems sold [17]. The is given by the formula:
efforts of increasing static field already focused 1
on 5-T or even 7-T whole-body systems, some of f = (23.1)
2π LC
which have already advanced to the stage of tar-
geted clinical and clinical research studies. where L is the inductance in Henries and C is the
Although they were not developed for low-field capacitance in Faradays. The orientation of the
MRI, many advances in hardware and post- static (B0) field in most conventional low-field sys-
processing algorithms can benefit low-field imag- tems is vertical, which means that a solenoid coil
ing. In this section, we review the current trends can be used, which generally provides severalfold
and developments in the various hardware com- greater efficiency than a transverse resonator.
ponents and the engineering challenges of low- The main factor for designing a coil is to max-
field MRI scanners. imize SNR, which is proportional to the static
magnetic field. The major source of noise in the
MRI signal is different between low-field and
Physical Properties of Low-Field MRI high-field systems. At a lower field of <0.5 T, the
MRI signal noise in general is dominated by
One of the key properties of MRI systems is the noise from the RF signal processing chain, such
static magnetic field. MRI needs a very strong as A/D converters, amplifiers, etc. However, at
magnetic field strength, and conventional MRI higher field strengths, noise from the magnetiza-
uses a superconducting coil that requires a cryo- tions in the subject becomes more significant.
gen system to remain superconducting. Permanent One of the major advances in the past two decades
and/or resistive electromagnets had been the in clinical systems has been the incorporation of
mainstay for low-field systems until the past multiple receiver elements (receiver arrays) both
436 Y. Yang et al.
Fig. 23.2 The basic MRI Scanner Components

components of any MRI
system: the main magnet Gradient Coils
produces the B0 field, RF Coils
necessary to align the
spins and achieve
equilibrium. Gradient Main Magnet
coils enable image
encoding in the x, y, and
z directions (i.e., the
frequency, phase, and
x y z
slice-encoding
directions). The RF coil
Control
is the part of the MRI Console
Electronics
system that excites the RF Gradient
aligned spins and Electronics Amplifiers
receives an RF signal
back from the sample.
All the components are
Pulse Sequence
controlled and interfaced Computer
with the user via a
console (Gruber et al.
JMRI 2018) [18]
Image Reconstruction Computer
Fig. 23.3 Examples of RF coils used on low-field MRI research-phased array designed for 0.25 T with a loop/
systems. (a) Quadrature transmit/receive coil on the verti- butterfly coil arrangement. (a–c courtesy of FONAR
cal 0.6-T Fonar system. (b) Four-element head array on Corporation, Esaote and Siemens Healthineers, respec-
the 0.25-T Esaote. (c) Shoulder-phased array for the tively) (J.P Marques JMRI 2019) [12]
Siemens 0.35-T Magnetom. (d) One example of a
for higher SNR and reduced imaging time when body, they tend to couple to each other. The inter-
using parallel imaging techniques (examples are coil coupling is calculated as
shown in Fig. 23.3). In this way, there is an SNR X L = ωL (23.2)
gain mainly in the periphery of the image, and the
use of multiple receive elements also enables where L is the inductance and ω is the angular
faster scanning times through sparse sampling of frequency. Two decoupling strategies, capacitive
k-space [19, 20]; see details of k-space in Chap. 5 decoupling and geometric decoupling, are usu-
of this book . ally considered solutions, but they both have
With a very large inductance at low field, the drawbacks. Preamp decoupling is another solu-
decoupling of receive coil elements is challeng- tion. Based on the factors from the previous
ing because instead of being well coupled to the paragraphs, some designs, such as optimized
wire patterns and a separate transmitter/receiver allel to the magnetization, if being considered as
including quadrature coil and array coils, are dipole moment as shown in Fig. 23.4, can be as
engendering interest. In this case, many low- high as the order of 104 A. Even though this is a
field systems start to use receive arrays with their simplified example, it highlights how powerful
elements being of relatively large size so that these magnets are compared to electromagnets.
coil noise does not dominate. Additional research One category of permanent magnet systems
in this area is being done that might provide that is popular in both research and commercial
a flexible coil [21]. applications are yoked magnets. An example [22]
of a yoked magnet that is shown in Fig. 23.5 is
able to generate a strong static magnetic field
Permanent Magnet System with good homogeneity.
Despite the achievements of yoked magnets,
The earliest low-field MRI systems were con- they are still limited by their heavy weight,
structed using various forms of electromagnets requiring a supporting structure usually made of
and/or permanent magnets. Nowadays, a static steel and iron. Conversely, the category termed
field of 1.5 T has become the clinical standard in yokeless magnets is more diverse and can broadly
HICs, almost completely replacing the older be categorized into two subtypes: homogeneous
lower-field strength (0.2–1 T) systems that had an magnets and single-sided magnets. The homoge-
important role during the 1980s [17]. The early neous magnets are typically cylindrical and bipo-
designs of low-field MRI used permanent magnet lar shaped. Efforts have been made to reduce
arrays (PMAs), with and without iron yokes. A field inhomogeneity by increasing the diameter
PMA has no power consumption and is low-cost. of the cylinder or decreasing the space between
Compared to a superconducting magnet, a PMA the rings; it is also possible to use inhomogeneity
does not require sophisticated cooling systems to for spatial encoding as an example shown in
operate, although the field strength is relatively Fig. 23.6 from Vogel’s work [23].
low. Also, compared to a resistive electromagnet, The same principle can be used for single-
it has no heat dissipation resulting from provid- sided magnets as well even though it is harder to
ing the main static field. The materials used in control the field. Single-sided magnets usually
PMAs vary, as does the maximum achievable have small sizes and are also easier to position.
field strength. For example, a magnet based on The cylindrical Halbach array is used for generat-
neodymium can generate a magnetic field of ing a homogeneous encoding field as shown in
around 1.4 T. The equivalent surface current par- Fig. 23.7.
Fig. 23.4 Permanent

magnets generate a
magnetic field similar to
that of a coil with an
equivalent current.
(From Wikipedia)
438 Y. Yang et al.
b c
Fig. 23.5 Prototype of a low-cost low-power and shielding-free brain ultra-low-field (ULF) magnetic resonance imag-
ing (MRI) scanner with homogenous 0.055-Tesla magnetic field and small 5-Gauss fringe field (Liu et al. [22])
A series of applications for using Halbach safety manual in which four zones with increas-
magnets with external shimming for portable ing levels of danger are defined.
low-field scanners has shown optimal results, and As shown in Fig. 23.9, only zone I is consid-
researchers are continuously optimizing the field ered completely safe to the public where the
pattern of cylindrical Halbach arrays in a way magnetic field is 5 Gauss (0.5 mT) or less [26].
that retains construction simplicity. However, a locked-off safety is feasible for por-
table devices. One solution is to use low-field
magnets with a small magnetic footprint and
ardware Requirement for Point-of-
H fringe field as mentioned in the previous section.
Care and Portable MRI A physical barrier or at least an indicator has also
been used in some commercial products.
A major advantage of low-field MRI systems is Hyperfine [27] uses a barrier for the 5-Gauss
their accessibility, which makes them ideal as zone. Another tactic is to use an on/off magnet.
portable MRI scanners. We can even consider a Synaptive [28] 0.5-T product adopts this solution
single-handed device in the future. A group of allowing a relatively high magnetic field since the
representative MRI systems with different field magnetic field can be shut down.
strengths are shown in Fig. 23.8. The infrastructure and logistics are other
Regardless of the field strength, safety is important aspects of designing portable MRI
always of critical importance in clinical work- scanners. To increase accessibility, infrastructure
flow. The major hazards during MRI scanning are requirements, specifically cooling and power,
projectile accidents and injury to patients with need to be minimized. Portable scanners
implanted devices. Hence, a safety zone is neces- also need to be able to operate from standard wall
sary for conventional MRI. The American outlets. Several groups and companies have dem-
College of Radiology has published an MRI onstrated the achievability of this requirement
a b
Fig. 23.6 (Top) Spatial encoding with magnet arrays [23]. (Bottom) Single-sided magnets from McDaniel et al. [24]
(e.g., Fig. 23.5). Pushing even further, it is con- design of a portable MRI needs to be lightweight
ceivable to power low-field MRI systems using and compact in order to be practically
batteries, a feature that exists for some portable transportable.
CT scanners. A portable MRI scanner operates in locations
The portable MRI also needs a different cool- that typically lack the RF shielding that conven-
ing system than conventional scanners. Since a tional high-field systems use to prevent external
cryogenic environment is not practical, a water- artifact-generating RF interference. As a result,
cooling system is becoming a popular choice. portable scanners operate without shielding or
Most of the portable MRI permanent magnets retrospectively removing interference from the
and the water-cooling systems are used in con- images. Also, compact magnet coils can generate
junction with compact unshielded gradient coils. nonlinear gradient fields and field inhomogene-
For portable systems, a critical logistical con- ities that cause image distortion. Nonlinearity
sideration is the integration of the patient table corrections and shimming are necessary to reduce
with the scanner. For example, Hyperfine uses a these effects, and retrospective corrections dur-
fold-down patient table and integrates it with a ing the image reconstruction process are likely
standard patient bed for scanning. In general, the necessary. One approach is shown in Fig. 23.10.
440 Y. Yang et al.
Fig. 23.7 Field

direction and simulation
of higher-order circular
Halbach array (from
Wikipedia)
Fig. 23.8 The MRI systems with different magnitude of static magnetic fields. The hardware requirements for these
systems are different
Fig. 23.9 The MRI zone sign set from the American College of Radiology MRI safety manual [25]
442 Y. Yang et al.
a FFT CPR MB MB CS
A.u.
Image
0
1500
Hz
1000
500
DBo map 0
-500
-1000
-1500
150
Hz
75
DBo map
0
difference
-75
-150
b FFT CPR MB MB CS
A.u.
Image
0
1500
Hz
1000
500
DBo map
0
-500
-1000
-1500
150
Hz
75
DBo map
0
difference
-75
-150
Fig. 23.10 Comparison of FFT, iterative MB, and itera- of the field map estimation when using the iterative CPR
tive CPR approach for an in vivo brain scan. (a) Mild B0 and MB approaches. The MB CS reconstruction for an
inhomogeneities (~ 600 Hz). (Top row) The FFT recon- undersampling factor of 2 shows the flexibility of the MB
struction results in a stretch of the brain along the readout approach toward undersampling. (b) For strong B0 inho-
direction (left to right). Both iterative CPR and iterative mogeneities (~ 1500 Hz), here obtained by attaching an
MB image reconstruction result in low errors in the esti- external magnet to the Faraday cage, the iterative MB
mated field map, and correct for the corresponding image approach still results in a uniform signal intensity, while
deformations. The estimated ΔB0 maps (middle row) and the iterative CPR approach shows more signal shading in
the differences with respect to that of the FFT approach the lower-left corner of the brain [29]
(bottom row) show a similar improvement in the accuracy
Advantages of Low-Field MRI imaged well due to off-resonance artifacts.

Numerous metal implants cause artifacts on MRI
Conventional MRI, although widely used in most images. Metal causes spatial variations in the
clinical areas, still has limitations. For some static magnetic field (B0) and T2* dephasing as
patients with metal implants, conventional MRI well as the frequency shift ∆f, resulting in slice
can cause safety risks, and the metal implants can shifts by the magnitude (∆f/BWRF * slice width)
cause artifacts as well. Low-field MRI has advan- and readout displacements of (∆f/BWpix * pixel
tages in resolving these issues. Low-field MRI is width). It can also cause background gradients
also good at imaging near air-tissue interfaces and a B1 shading effect. Examples of B0-
(e.g., lung, airway). In this section, a review of inhomogeneity metal artifacts are shown in
the advantages and limitations of low-field MRI Fig. 23.11. Figure 23.12 shows the dependence
in these cases is presented. of signal void artifacts on the field strength. The
device-dependent signal void tends to be larger at
higher field strengths. Although the artifacts can-
Imaging of Patients with Implants not be completely removed, low-field scanners
inherently have smaller artifacts than higher-field
When preparing an MRI scan, the MRI system scanners due to their reduced device-dependent
operator needs to follow their institutional MRI off-resonance.
safety protocols, which typically involve a safety Many methods have been proposed to correct
questionnaire about whether the patient has distortion artifacts caused by metal implants.
metallic implants. In cases where patients do For example, to remove in-plane distortion arti-
have implants, the operator may need to consult facts, a sequence named view-angle tilting
resources, such as calling the device manufac- (VAT) can be used. VAT corrects for in-plane
turer, reading books, or online resources by distortions by adding a gradient on the slice-
Shellock et al. [26] to determine whether the select axis during readout. The sequence does
patient can be scanned. not require additional scanning time than the
There are several issues with metallic implants normal spin-echo sequence. Figure 23.13 illus-
and devices. The most apparent one is the mag- trates the use of VAT.
netic field interaction in form of both force and VAT does not fix through-slice distortion, and
torque exerted on the devices, which can cause to fix this problem, another sequence called mul-
them to dislodge. Another significant concern is tispectral imaging (MSI) is considered. This
device heating caused by the RF field. At a higher method uses a combination of spin echo and VAT
field with correspondingly higher Larmor fre- to generate 3D images of each slice. After adding
quencies, the fast-changing electromagnetic field up the slices, the cumulative distortion-free
of the RF pulse can substantially raise the device images are reconstructed.
surface temperature, which can cause tissue Table 23.1 shows a comparison of 0.55 T vs.
damage. This heating issue is generally mitigated 3 T in terms of several parameters related to MRI
at lower field strengths, resulting in overall better safety and image quality.
safety profiles. Another issue is the artifact caused At the lower field, the SNR is lower, but all the
by metal implants. Metal implants disturb the other parameters from Table 23.1 are more desir-
main field because the implant’s magnetic sus- able. The lower SAR means better implant safety,
ceptibility differs dramatically from human tis- and high-SAR sequences are easier to perform at
sue. These susceptibility differences cause a a lower field; the lower off-resonance means the
large-scale main field variation that extends well smaller range of frequency to correct for; the lon-
beyond the physical boundary of the devices. As ger T2* enables longer readout; a higher peak B1
a result, tissues at a short distance from these means a higher bandwidth and a more efficient
implants (often within centimeters) are not encoding.
444 Y. Yang et al.
Fig. 23.11 Metal artifact from 50 mT to 7 T. Here we show Fig. 7 from Van Speybroek et al. [44] where the “worst-
case” scenario images for four types of metal implant are demonstrated. (Images reused via open-access license) [30]
I maging of Organs with Air-Tissue The lung has long been neglected by conven-
Interfaces tional proton MR imaging methods. Numerous
studies have described lung MRI methods using
MRI signal strength depends on water proton den- hyperpolarized gases [33]. This is due to unique
sity. This causes an issue for imaging organs with intrinsic difficulties such as the lower proton den-
air-tissue interfaces, such as the lungs. The lungs sity and susceptibility artifacts caused by multiple
have much lower proton density and diminished air-tissue interfaces. Clinically speaking, CT is
intrinsic MRI signal. Usually, there is a relatively commonly used for lung imaging. However, CT
low SNR and long acquisition speed when imag- provides excellent spatial morphological informa-
ing these organs. Low-field imaging can improve tion but limited functional information. Pulmonary
morphological imaging because, at low field, function tests (PETs) are global measures of lung
there is a relatively longer T2* value and suscep- function but lack highly resolved spatial informa-
tibility also decreases. In this section, we will tion and are inadequate in assessing disease het-
review the role of low-field imaging in the lungs. erogeneity. Then, the question becomes whether
a b
c d
Fig. 23.12 An orange containing six ferromagnetic tion at (a) 0.2 T than at (c) 1.5 T for comparable screws
metallic surgical screws (5 mm in diameter) is imaged at (those parallel to the direction of B0; long arrows), the
0.2 T in the axial plane using (a) spin-echo T1-weighted frequency-encoding gradient being perpendicular to the
and (b) gradient-echo T1-weighted sequences. The same direction of B0. Note that when the phase and frequency-
phantom is imaged at 1.5 T in the coronal plane (c, d), encoding axes are switched (c, d), the image distortion in
using spin-echo T1-weighted sequences, the frequency- the screws parallel to the direction of B0 (long arrows) is
encoding axis being switched between the two series. less marked when the frequency encoding axis is parallel
Note that the screws parallel to the direction of B0 (long to the direction of B0 (d) and the distortion increases in
arrows) produce less artifacts than those perpendicular to the screws perpendicular to the direction of B0 (short
the direction of B0 (short arrows). The artifacts are more arrows). Interchange of the frequency and phase-encoding
important on (b) gradient-echo than on (a) spin-echo axes also changes the shape and orientation of the artifacts
sequences because of loss of signal intensity due to intra- in the direction of the frequency-encoding axis [31]
voxel dephasing (T2* effect). There is less image distor-
there exists a one-stop shop for morphologic and Campbell-Washburn comparing T2-weighed
spatially resolved functional imaging. imaging with CT using a 0.55-T MRI shows a
As we have mentioned in the beginning, con- good correlation between them.
ventional MRI suffers from low SNR, and it can- Another example is shown in Fig. 23.14 from
not provide enough details of the structure of the a COVID-19 pneumonia patient.
lung. It can be boosted by techniques using ultra- Other than morphologic imaging, functional
short echo (UTE) [34]. A study [35] by imaging is also an important aspect to consider.
446 Y. Yang et al.
Original Prosthesis Coronal STIR Coronal T1
VAT & SEMAC
VAT & SEMAC

Standard
Standard
VAT
VAT
TiAl6V4 with silver coating
Proximal Femur
Proximal Femur
TiAl6V4
Distal Femur
CoCrMo
CoCrMo with TiN coating
Distal Femur
Fig. 23.13 Experimental model. Tumor endoprostheses technique (standard; VAT, view-angle tilting;
were embedded in water. Specifications and materials are VAT&SEMAC, view-angle tilting and slice-encoding
indicated on the left. Central slices of coronal STIR and metal artifact correction) [32]
coronal T1-w MR pulse sequences are presented for each
Table 23.1 Comparing low-field and high-field parame- Fortunately, low-field MRI is still useful to provide
ters from Siemens. Bold, advantages; italics,
information about lung function such as lung com-
disadvantages
pliance, ventilation, and perfusion. One example is
Relation 0.55 vs 3
Parameter to B0 (T) Comment
a 4D imaging technique called the free-breathing
SNR ∝B0 Overall A proportional acquisition scheme that records respiration. By
efficiency ~4× lower loss of SNR recording spirometry measures of forced expira-
T1
∝ B00.3
TR 1.7× Small SNR tory volume (FEV) and forced vital capacity (FVC)
shorter efficiency gain normal breathing and deep breathing, we notice a
SAR 30× lower Reduced SAR
∝ B02 good correlation between MR and traditional mea-
Off- ∝B0 5.5× Reduced Δf surements. The MRI information can tell a better
resonance lower story than spirometry measures. For instance, MR
(ΔB0)
can help to calculate the flow volume loop for the
T2* ≈1/B0 Much Longer readout
longer left/right lung separately, which is impossible to
Peak B1 – 3× higher High BW, short get from spirometry measures. Pushing even fur-
RF ther, even the information about subregions of the
Fig. 23.14 Images in a 61-year-old man with coronavi- ties on CT image and peripheral distribution of
rus disease 2019 (COVID-19) infection. Shown are two consolidative opacities characteristic of COVID-2019 are
axial sections from, A, MRI scan obtained with prototype observed with both imaging modalities. In some regions,
0.55-T Magnetom Aera scanner (Siemens Healthineers) severity of abnormality appears worse on MRI scan, as
with spatial resolution of 1.2 mm × 1.2 mm × 6 mm depicted by red rectangles. (Campbell-Washburn et al.
(respiratory-gated fast-BLADE turbo spin-echo acquisi- [36])
tion. Multifocal lesions equivalent to ground-glass opaci-
448 Y. Yang et al.
lung is available. What’s more, with the help of agent and T1 relaxivity of oxygen increases at
Fourier decomposition, we can even obtain micro- low field. One could measure the percentage of
scopic information such as ventilation and perfu- signal enhancement from oxygen inhalation, an
sion maps as shown in Fig. 23.15. interesting and innovative method for probing
Oxygen-enhanced MRI is another option of lung ventilation. A low oxygen enhancement can
promising functional MRI in low field as 100% be used to identify areas of lung function deficit.
oxygen can be a T1-shortening MR contrast Figure 23.16 gives an example.
Slice Vw Qw
300 9
10
250 8
log(Qw-Signal)
Intensity [a.u.]
log(Vw-Signal)
8
200 7
NB 2
6
150 6
4
100 5
2
50 4
300 9
10
250 8
log(Vw-Signal)
Intensity [a.u.]
log(Qw-Signal)
8
200 7
IB
6
150 6
100 4
5
50 2
4
Fig. 23.15 Exemplary ventilation-(Vw) and perfusion- slice of Volunteer 4. The heart and large vessels were
weighted (Qw) maps. The unnormalized V- and Q-maps excluded from the segmentation of the Vw-maps. The
for the normal (NB 2) and the first irregular breathing scan maps were Gaussian filtered and logarithmically plotted
(IB, deep breathing) exemplarily presented for the aorta [37]
Fig. 23.16 Percent signal-enhancement maps generated slices are presented for two different patients classified
by oxygen-enhanced functional lung MRI. The color moderate (a) and severe (b) by FEV1, DLCO, and CT
scale represents the percent signal enhancement within images. Regional variation in enhancement is indicative
the lung tissue from inhalation of 100% oxygen. Three of local functional deficits (Bhattacharya et al. [38])
There are also techniques that can offer tissue In radiotherapy, the MR-guided radiotherapy
characterization such as MR fingerprinting [39] (MRgRT) is playing a more important role.
which can provide parameter maps such as T1/T2 Conventional image-guided radiotherapy which
maps. MR fingerprinting obtains quantitative is more commonly based on CT images is not able
images by template matching of the MRI signal to fully address the challenges that lead to errors
evolution over time, rather than conventional in treatment such as differentiating tumor from
Fourier encoding methods. Therefore, it is con- organ or patient position changes and motions.
ceivable that MR fingerprinting may be less sen- The soft-tissue contrast is often low, and it lacks
sitive to the intrinsic low SNR at a lower field. the ability to monitor the treatment responses
In conclusion, there exists a need for one-stop during radiation delivery. Last few years,
shop spatially resolved morphologic and func- MR-guided linear accelerators (MR-Linacs) such
tional imaging of air-tissue interface organs like as ViewRay MRIdian (0.35 T) and Elekta Unity
lungs. Low-field MRI coupled with advances in (1.5 T) have helped researchers expand the appli-
image acquisition and reconstruction can be an cation boundary of MR-Linac. In this section, we
enabling technology to deliver on this promise. review current clinical applications of MRI with
great potential in radiotherapy, especially low-
field scanners.
linical Applications of Low-Field
C
MRI in Radiotherapy
R-Linac Design with Low-Field
M
In the past decade, low-field MRI products have Magnet
proliferated and are quickly gaining popularity in
a variety of clinical fields including radiotherapy. Hybrid magnetic resonance (MR)-guided linear
Reviewing the history of MRI, the differences accelerators represent a new development in the
between high-field and low-field MRI scanners field of radiation oncology. By harnessing the
appeared during the early phase of MRI history favorable combination of onboard MR imaging
in the 1980s [40], and the gap expanded after with the possibility of daily recalculating the
1985 when the first 1.5-T scanner was introduced treatment plan based on real-time anatomy, the
[41]. The high-field scanners became more popu- accuracy in target and organs-at-risk identifica-
lar and gained a dominant market share because tion is expected to improve, with the aim of pro-
of their higher SNR per unit time, which permits viding the best-tailored treatment. To date, two
faster imaging, higher resolution, greater contrast main MR-Linac hybrid machines are available,
sensitivity, and more advanced sequences. With Elekta Unity (1.5 T) and ViewRay MRIdian
the technical advances mentioned earlier, some (0.35 T). Given the relative novelty of this tech-
commercial systems have received FDA clear- nology, these devices still raise practical issues
ance and been put on the market since 2018. The and require groups of researchers to address the
commercial products include 0.064-T Hyperfine issues.
Swoop head scanner, 0.066-T Promaxo prostate It is not a simple job to combine MR and
scanner, 0.5-T Synaptive Evry intraoperative RT. MR receives very weak RF signals from
scanner, and 0.55-T Siemens Magnetom patients, and even small noises can contaminate
FreeMax. While high-field devices still dominate the image. The strong magnetic field of MR is
the market because of higher image quality, two also able to influence the function of the linear
main drivers support this recent lower-field revi- accelerator. The linear accelerator is also a sig-
talization: (1) lower costs and (2) recent techno- nificant source of RF interference. Given the cur-
logical developments leading to improved image rent situation, the lower field MR-Linac offers
quality. These advantages also result in MRI the benefits of reduced engineering and shielding
scanners’ accessibility in all the other areas challenges in designing these systems.
beyond the diagnostic imaging field [42]. Theoretically, it may also offer favorable charac-
450 Y. Yang et al.
Fig. 23.17 ViewRay

MRIdian MRL model
that replaced the initial
tri-cobalt design in
2017. (Rammohan et al. Gantry and RT
Components - Redesigned
[43])
MRIS (Unchanged)
Patient Handling
System (Unchanged)
teristics regarding the impact of the electron erogeneities. Two sets of receive coils, the torso
return effect on the radiation dose distribution. coil and the head-and neck coil, consist of six-
Currently, the ViewRay MRIdian, as shown in and five-element phase arrays, respectively, in
Fig. 23.17, is the only FDA-approved low-field each half. Both coils are made with flexible
MR-Linac, and in this section, we will discuss foam. The system shields from leaking RF using
more details about the design of this system. copper lined with carbon fiber to absorb inter-
The details of MRIdian systems are described nally reflecting RF. The MRI system is able to
as follows [43]. The system consists of two apply intra-fraction soft-tissue-based tracking
hardware components: an MRI system and an by acquiring ciné planar images simultaneously
RT delivery system [43]. The MRI scanner is with radiation delivery.
double donut shaped with a 28-cm gap in The first generation of MRIdian contained a
between, generating a 0.35-T static magnetic ring gantry housing three cobalt-60 sources, pro-
field and a 50-cm-diameter spherical field of viding a nominal dose rate of 550 cGy/min with
view. The isocenter of the MRI matches that of fresh sources [43]. The system was later upgraded
the RT system. The gradient coils have an inner by replacing the cobalt with a linear accelerator
diameter of 80 cm. The coils are self-shielded to which is able to generate a 6-MV flattening filter-
help against thrust and torque. The maximum free photon beam, as shown in Fig. 23.17. The
magnetic field gradient and slew rate can reach beam has a nominal dose rate of 600 MU per
10 mT/m and 200 T/m/s on each axis, respec- minute at the 90-cm isocenter. The linear accel-
tively. The transmit coil can cover the whole erator and its subcomponents are shielded from
body. There is also an integrated RT shield to the magnetic field by five ferromagnetic steel
span the magnet gap while maintaining uniform concentric cylinders as well as three additional
radiation beam attenuation to prevent beam het- metal shields on the electron gun.
Thanks to great soft-tissue contrast, The approaches of fast acquisition need to

MR-Linacs have shown their potential in adap- harness low-field physics changes. For example,
tive RT and intra-factional motion management. at low field, T1 and T2 decrease differently, with
MR-Linacs have advantages in treating tumors T1 approaching T2. In this case, when people use
that grow in areas prone to deformation and balanced steady-state free precession (bSSFP)
motion such as the pancreas and liver. This sys- approaches, the steady-state magnetization, Mss,
tem is able to minimize the target area by track- depends on the ratio T2/T1 as [44]:
ing in real-time the tumor motion, saving as much 1
normal tissue as possible. M ss = M 0 T2 / T1 . (23.3)
2
Unlike CT, MRI scanning is quite versatile,
given multiple choices of functional MRI where M0 is the initial magnetic moment. Hence,
sequences such as diffusion weight imaging or at a low-field limit, this steady-state magnetiza-
contrast-enhanced imaging. MRIdian has tion can reach up to 50% of the total magnetiza-
installed these functional MRI sequences which tion, making it one of the most efficient sequences.
are able to provide additional physiological and The bSSFP sequence also needs good mag-
biological information during the whole treat- netic field homogeneity which is incompatible
ment. Combining all this information, it is prom- with the small footprint scanners. Nevertheless,
ising to imagine a personalized RT treatment some most recent initiatives to resolve the field
plan that maximizes the treatment effectiveness. homogeneity with passive shimming and first-
order gradient DC shimming show very promis-
ing results, enabling bSSFP imaging without
otentials of Low-Field MRI
P artifacts with reasonable TR on the order of
in Radiotherapy 7–10 ms [22].
In the MR fingerprinting approach [39], the
High-Speed MRI signal’s pattern gives a unique set of magnetic
There are already low-field MRI scanners being properties. The images of a series of bSSFP with
used in point-of-care testing. Both commercial varied flip angles (FA) and TR are generated, and
and experimental products have reached a rea- the signals are matched to the precomputed dic-
sonable image resolution using conventional tionary, as shown in Fig. 23.18. From this
sequences such as FSE, GRE, etc. The scanner approach, several quantitative maps can be
from Hyperfine is able to operate T1-/ extracted at once. For instance, an initiative of
T2-weighted, FLAIR, and DWI sequences, while low-field MRF like OPTIMUM is able to gener-
the other experimental groups also have the abil- ate six quantified parametric maps in a small
ity to offer comparable options. We are even footprint scanner of 0.1 T (Fig. 23.19) [45].
impressed by the image quality with noise can- Furthermore, with bSSFP sequences, OPTIMUM
cellation techniques which are often combined generates 270 2D images in around 8.5 min.
with deep learning reconstruction. The whole We can even go faster by using non-Cartesian
scan usually takes around 30 min. It is not diffi- sampling such as spiral or radial samplings.
cult to conclude that since the image quality of There are already proof-of-concept works from
the low-field scanner is already comparable with AMT center at the University of Basel, which
the state-of-the-art techniques, a fast acquisition used radial k-space sampling of 100 spokes with
scheme or high-speed MRI with low field is an image matrix in size of 129*129*15. The total
going to be helpful to show the clinical value of acquisition time is around 2.5 min. In conclusion,
the low-field MRI. For MRgRT, increasing the it is optimal to improve treatment quality by
images per unit time is considered very valuable, applying high-speed low-field MRI techniques,
especially for quantitative MRI which is already and there still exist many more to be explored in
quite time-consuming.
452 Y. Yang et al.
a 80
70
b Signal
Flip Angle
(degrees)
60
50 MRF
40
30 1
20 0.9
Signal Intensity
10
0
0.8
0 50 100 150 200 250 300 350 400 450 500 0.7
TR Index 0.6
14
0.5
13 0.4
(ms)
0.3
TR
12
0.2
11
0.1
10 0
0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450 500
TR Index
TR Index
c 3000
2500
2000
1500
1000
500
Fig. 23.18 The paradigm of MRF from acquisition (a) to its match to the dictionary (b) and the quantitative maps (c)
(Ma et al. 2013—with editing [39])
the future. In radiotherapy, such a technique can cuss the potential of 4D MRI applications in
be used to improve the efficiency of workflow. radiotherapy.
Many efforts have already been made on this
4DMRI topic. These include the 2D-CINE imaging and
4D MRI is a quantified assessment of the breath- 2D navigators which both have disadvantages
ing motion of tumors and organs. In radiotherapy like “stitching” artifacts or low scan efficiency.
of abdominal tumors, this motion needs to be More recently, 3D k-space binning-based tech-
considered when calculating the delivered dose niques such as SG-KS-4D-MRI and ROCK-4D-
accurately. Currently, 4D CT is conventionally MRI have been used to replace 2D image sorting
used in radiotherapy for motion evaluation with methods. The ROCK-4D-MRI uses a 4D Rotating
high spatial resolution. However, it has several Cartesian K-space (ROCK) sampling trajectory
limitations, such as low temporal resolution, poor which improves scan efficiency so that the high-
soft-tissue contrast, and motion-induced artifacts. resolution 4D images of the entire torso may be
These problems undermine tumor visualization acquired within 5 min using a 1.5-T diagnostic
and accurate motion quantification. To overcome MRI scanner [46].
these disadvantages, 4D MRI is a promising The same techniques were recently tested on
alternative due to its good soft-tissue contrast and the 0.35-T MRI-guided radiotherapy (MRgRT)
lack of ionizing radiation. In this section, we dis- system [47]. The motion quantification based on
a Mo (A.U.) ×104 b B1 Map (% nominal)

6
1.2
4
1
2
0.8
0
T1 Map (ms) Off Resonance Map (Hz)
300 100
50
200
5
100 -50
0 -100
T2 Map (ms) Dot product Map (%)

50 1
40 0.9
30
0.8
20
0.7
10
0 0.6
T2* Map (ms) 4 bone marrow

50
bone
40 tendon
3
muscle
30
2
20
10 1
0 0
2 4 6 8 10 12 14 16 18
c
Fig. 23.19 LF OPTIMUM results in vivo at 0.1 T. (a) voxel size is [1.5 × 1.5 × 6.5] mm3. Total imaging time is
Quantified maps in the human hand/wrist that includes 8.5 min with number of averages = 2. (b) Signal evolution
proton density M0, T1, T2, T2*, B0, FA homogeneity (i.e., MR fingerprints) reported for different tissue types in
B1+, and matching efficiency (dot product map) for six a central slice. (c) Corresponding MR image series
central slices. Matrix size is 90 × 81 × 15; corresponding (Sarracanie et al. [45])
4D MRI was compared with those based on was originally developed for the diagnosis of
2D-CINE, and the motion measurements based stroke and later for cancer diagnosis. In radio-
on 4D MRI matched well with 2D-CINE. Then, therapy, DWI is considered a noninvasive method
the image quality of 4D MRI was evaluated to assess the tissue microenvironment before,
against 4D CT, and the image quality scores of during, and after cancer therapy. The apparent
4D MRI were significantly higher than 4D CT, diffusion coefficient (ADC) calculated from DWI
with better tumor contrast and fewer motion arti- is closely related to tissue diffusivity, with low
facts, as shown in Figs. 23.20 and 23.21. ADC values corresponding to more restricted dif-
These researchers have shown the clinical fea- fusion and vice versa. An increase in ADC values
sibility of using 4D MRI to acquire high-quality, after treatment reflects a decrease in tumor cel-
respiratory motion-resolved 4D MRI in a low- lularity because of tumor cell death, indicating a
field MRgRT system. good response. This parameter offers a useful and
early tool to predict treatment response with a
Diffusion-Weighted MRI diagnosis MRI [48].
The diffusion-weighted MRI (DWI) is a com- Researchers have shown the feasibility of per-
monly used functional MRI technique for prob- forming DWI on the 0.35-T MRIdian MR-Linac
ing tissue cellularity and tumor response to system [49] and continued studies about DWI
radiotherapy and chemotherapy. This technique technical development as well as clinical applica-
454 Y. Yang et al.
Fig. 23.20 The 4D-MRI images (top row) and 4D-CT are free of the stitching artifacts. The boundary of the
images (bottom row) on a patient with liver metastasis. tumor is much better defined in 4D MRI than in 4D-CT
The 4D-MRI images have better soft-tissue contrast and (Han et al. [47])
Fig. 23.21 End-expiration phase of 4D-MRI and 4D-CT images acquired on a patient with liver cancer, reformatted to
coronal, sagittal, and transversal slice orientations (Han et al. [47])
tions at 0.35 T. Some groups tried to compensate Researchers from UCLA [50] have developed a
for image distortion by using improved sequences specially designed DWI sequence for the low-
with special designs or better positioning. field 0.35-T scanner called diffusion-prepared
turbo spin-echo (DP-TSE) readout to generate such as to visualize small critical structures for
distortion-free DWI images. Another group from better protection with improved patient position-
Washington University [51] discovered that ing and online adaptive RT. One example is the
results showed that variation of the ADC quanti- genitourinary (GU) toxicity of stereotactic body
fication accuracy and geometric fidelity on the radiotherapy (SBRT). SBRT is now a widely
0.35-T MR-Linac system was dependent on the accepted standard-of-care option for localized
gantry angle. Following this direction, another prostate cancer, but its late grade GU toxicity
group from Fondazione Policlinico Universitario remains a significant challenge. Typically, blad-
in Italy focused on optimizing the SE-EPI-based der sparring is a common method to reduce toxic-
DWI protocol on the low-field MR-Linac aiming ity, but it was discovered that urethral injury is
to maximize imaging quality and minimize also a contributor. In the past, delineating the ure-
acquisition time [52]. The value of using longitu- thra on CT images was not easy due to the lack of
dinal DWI acquired on the low-field MR-Linac contrast between the prostatic urethra and pros-
for treatment response prediction was also dem- tate tissue. On MRIdian, researchers optimized a
onstrated for both rectal cancer and sarcoma [53– T2-weighted MRI sequence for delineating the
55]. DWI has already been confirmed as a quite prostatic urethra. The sequence they used was a
practical tool on low-field MRI with broad 3D half-Fourier acquisition single-shot turbo
application prospects and can potentially bring spin-echo (3D HASTE). The UCLA group first
more clinical value when combined with other demonstrated the clinical value of acquiring MR
advanced techniques such as advanced diffusion urethra imaging on an MR-Linac [60]. Onboard
models, intravoxel incoherent motion (IVIM) urethra MRI was also acquired at two time points
[56, 57], and continuous-time random walk during the entire radiation course: after MR sim-
(CTRW) [58, 59]. ulation and after the final treatment fraction.
Their results show that good visualization of the
ptimized Protocols for Specific
O prostatic urethra (Fig. 23.22) can be a valuable
Clinical Purpose: Urethra MRI tool to help better protect the urethra through
With the low-field MR-guided RT introduced in patient setup or online adaptive RT, given the
RT, there are new possibilities and needs to opti- large variation visualized using the proposed MR
mize MR protocols for specific clinical purposes, technique [61].
Patient A Patient B Patient C Patient D

a b c d
Simulation
V42Gy = 0.00 cc V42Gy = 0.01 cc V42Gy = 0.00 cc V42Gy = 0.00 cc
e f g h
Final Fraction
V42Gy = 0.01 cc V42Gy = 2.55 cc V42Gy = 0.75 cc V42Gy = 0.02 cc
Fig. 23.22 A radiation oncologist’s calculated and esti- final HASTE urethra contour, purple/blue. Orange con-
mated dose and urethral V42Gy for patient A–D’s simula- tour, PTV. Red, 105% (42 Gy) isodose region; yellow,
tion (a–d) and final fraction (e–h) bSSFP. RO1 simulation/ 95% (38 Gy) isodose region (Pham et al. [61])
456 Y. Yang et al.
Conclusion tomography and magnetic resonance imaging equip-

ment in U.S. emergency departments. Acad Emerg
Med. 2008;15(8):780–3.
Great advances in hardware and imaging tech- 4. Health equipment—magnetic resonance imaging
niques of low-field MRI systems beyond com- (MRI) units—OECD data. [Internet]. OECD; 2018.
mon perception have happened. Since the 1990s, https://data.oecd.org/healtheqt/magnetic-resonance-
imaging-mri-units.htm.
it has been possible to apply advanced techniques 5. WHO|Atlas: multiple sclerosis resources in
such as DWI and contrast-enhanced imaging. the world 2008. [Internet]. WHO; 2018. http://
Compared with conventional MRI, low-field w w w. w h o . i n t / m e n t a l _ h e a l t h / n e u r o l o g y /
MRI is more economic and more convenient. atlas_multiple_sclerosis_resources_2008/en/.
6. Niendorf T, Sodickson DK. Parallel imaging in car-
Moreover, researchers have a better understand- diovascular MRI: methods and applications. NMR
ing of the physical characteristics of low-field Biomed. 2006;19(3):325–41.
systems which can be superior to high-field MRI 7. Lustig M, Donoho D, Pauly JM. Sparse MRI: the
in many ways. For instance, T1-shortening effects application of compressed sensing for rapid MR
imaging. Magn Reson Med. 2007;58(6):1182–95.
and low magnetic susceptibility artifacts are 8. Campbell-Washburn AE, Ramasawmy R, Restivo
associated with static field strength. Recent MC, Bhattacharya I, Basar B, Herzka DA, et al.
developments in hardware such as the use of per- Opportunities in interventional and diagnostic imag-
manent magnet arrays, sophisticated radiofre- ing by using high-performance low-field-strength
MRI. Radiology. 2019;293(2):384–93.
quency coils, and optimized sequences have 9. Basar B, Sonmez M, Yildirim DK, Paul R, Herzka
shown that the static field strength itself does not DA, Kocaturk O, et al. Susceptibility artifacts from
necessarily need to be high. Numerous products metallic markers and cardiac catheterization devices
in both academia and industry have given option a high-performance 0.55 T MRI system. Magn
Reson Imaging. 2021;77:14–20.
mistic results, some of which are better than 10. Hori M, Hagiwara A, Goto M, Wada A, Aoki S. Low-
those from conventional MRI for certain tasks. In field magnetic resonance imaging: its history and
addition, with the recent remarkable develop- renaissance. Invest Radiol. 2021;56(11):669–79.
ments in low-field MRgRT, the MRIdian system, 11. Hayashi N, Watanabe Y, Masumoto T, Mori H, Aoki
S, Ohtomo K, et al. Utilization of low-field MR scan-
the onboard low-field MRI as one of the solutions ners. Magn Reson Med Sci. 2004;3(1):27–38.
for MR and RT integration, shows great poten- 12. Marques JP, Simonis FFJ, Webb AG. Low-field MRI:
tial. Clinically, the fundamental value of MRI is an MR physics perspective. J Magn Reson Imaging.
superior tissue contrast information. For this rea- 2019;49(6):1528–42.
13. Ejbjerg BJ, Narvestad E, Jacobsen S, Thomsen HS,
son, the low-field MRI can surely help in opti- Østergaard M. Optimised, low cost, low field dedi-
mizing the treatment workflow, combined with cated extremity MRI is highly specific and sensitive
advanced imaging techniques including those we for synovitis and bone erosions in rheumatoid arthritis
have mentioned in the previous section. Overall, wrist and finger joints: comparison with conventional
high field MRI and radiography. Ann Rheum Dis.
we are enthusiastic about low-field MRI systems 2005;64(9):1280–7.
and believe that they will continue to develop and 14. Yamada K, Miyahara K, Sato M, Hirose T, Yasugi Y,
gain more implementations in the future. Matsuda Y, et al. Optimizing technical conditions for
magnetic resonance imaging of the rat brain and abdo-
men in a low magnetic field. Vet Radiol Ultrasound.
1995;36(6):523–7.
References 15. Iturri-Clavero F, Galbarriatu-Gutierrez L, Gonzalez-
Uriarte A, Tamayo-Medel G, de Orte K, Martinez-
1. Hong AS, Levin D, Parker L, Rao VM, Ross- Ruiz A, et al. “Low-field” intraoperative MRI:
Degnan D, Wharam JF. Trends in diagnostic imag- a new scenario, a new adaptation. Clin Radiol.
ing utilization among Medicare and commercially 2016;71(11):1193–8.
insured adults from 2003 through 2016. Radiology. 16. Klein HM. Clinical low-field strength magnetic
2020;294(2):342–50. resonance imaging: a practical guide to accessible
2. Arnold TC, Freeman CW, Litt B, Stein JM. Low-field MRI. Springer; 2016.
MRI: clinical promise and challenges. J Magn Reson 17. Magnetic Resonance Imaging Systems Market by
Imaging. 2023;57(1):25–44. Architecture (Open MRI Systems and Closed MRI
3. Ginde AA, Foianini A, Renner DM, Valley M, Systems (Standard Bore and Wide Bore)), Field
Camargo CA. Availability and quality of computed Strength (Low-to-mid Field, High-field (1.5T and
3T), and Very-high Field)—Global Forecast to 2023 33. Mugler JP, Altes TA. Hyperpolarized 129ssMRI
[Internet]. Markets and Markets. https://www.market- of the human lung. J Magn Reson Imaging.
sandmarkets.com/requestsampleNew.asp?id=99. 2013;37(2):313–31.
18. Gruber B, Froeling M, Leiner T, Klomp DWJ. RF 34. Togao O, Tsuji R, Ohno Y, Dimitrov I, Takahashi
coils: a practical guide for nonphysicists. J Magn M. Ultrashort echo time (UTE) MRI of the lung:
Reson Imaging. 2018;48(3):590–604. assessment of tissue density in the lung parenchyma.
19. Pruessmann KP, Weiger M, Scheidegger MB, Magn Reson Med. 2010;64(5):1491–8.
Boesiger P. SENSE: sensitivity encoding for fast 35. Campbell-Washburn AE, Malayeri AA, Jones EC,
MRI. Magn Reson Med. 1999;42(5):952–62. Moss J, Fennelly KP, Olivier KN, et al. T2-weighted
20. Griswold MA, Jakob PM, Heidemann RM, Nittka lung imaging using a 0.55-T MRI system. Radiol
M, Jellus V, Wang J, et al. Generalized autocalibrat- Cardiothorac Imaging. 2021;3(3):e200611.
ing partially parallel acquisitions (GRAPPA). Magn 36. Campbell-Washburn AE, Suffredini AF, Chen
Reson Med. 2002;47(6):1202–10. MY. High-performance 0.55-T lung MRI in
21. McGee KP, Campeau NG, Witte RJ, Rossman PJ, patient with COVID-19 infection. Radiology.
Christopherson JA, Tryggestad EJ, et al. Evaluation 2021;299(2):E246–7.
of a new, highly flexible radiofrequency coil for MR 37. Klaar R, Rabe M, Gaass T, Schneider MJ, Benlala I,
simulation of patients undergoing external beam radi- Eze C, et al. Ventilation and perfusion MRI at a 0.35
ation therapy. J Clin Med. 2022;11(20):5984. T MR-Linac: feasibility and reproducibility study.
22. Liu Y, Leong ATL, Zhao Y, Xiao L, Mak HKF, Radiat Oncol. 2023;18(1):58.
Tsang ACO, et al. A low-cost and shielding-free 38. Bhattacharya I, Ramasawmy R, Javed A, Lowery
ultra-low-field brain MRI scanner. Nat Commun. M, Henry J, Mancini C, et al. Assessment of lung
2021;12(1):7238. structure and regional function using 0.55 T MRI
23. Vogel MW, Guridi RP, Su J, Vegh V, Reutens DC. in patients with lymphangioleiomyomatosis. Invest
3D-spatial encoding with permanent magnets for Radiol. 2022;57(3):178–86.
ultra-low field magnetic resonance imaging. Sci Rep. 39. Ma D, Gulani V, Seiberlich N, Liu K, Sunshine JL,
2019;9(1):1522. Duerk JL, et al. Magnetic resonance fingerprinting.
24. McDaniel PC, Cooley CZ, Stockmann JP, Wald Nature. 2013;495(7440):187–92.
LL. The MR cap: a single-sided MRI system designed 40. PubMed by Year—The Finest Worklog [Internet].
for potential point-of-care limited field-of-view brain https://esperr.github.io/pubmed-by-year/.
imaging. Magn Reson Med. 2019;82(5):1946–60. 41. Edelman RR. The history of MR imaging as seen
25. Kodzwa R. ACR manual on contrast media: 2018 through the pages of radiology. Radiology. 2014;273(2
updates. Radiol Technol. 2019;91(1):97–100. Suppl):S181–200.
26. Sammet S. Magnetic resonance safety. Abdom Radiol 42. Sarracanie M, Salameh N. Low-field MRI: how low
(NY). 2016;41(3):444–51. can we go? A fresh view on an old debate. Front Phys.
27. Swoop® Portable MR Imaging System™ Details and 2020;8:172.
specifications [Internet]. Hyperfine Inc. 2023. https:// 43. Rammohan N, Randall JW, Yadav P. History of
hyperfine.io/swoop/details-and-specifications. technological advancements towards MR-Linac: the
28. SYNAPTIVE MRI Everywhere for everyone future of image-guided radiotherapy. J Clin Med.
[Internet]. [42] Synaptive Medical Inc. 2023. https:// 2022;11(16):4730.
www.synaptivemedical.com/products/mri/. 44. Scheffler K, Lehnhardt S. Principles and applica-
29. Koolstra K, O’Reilly T, Börnert P, Webb A. Image tions of balanced SSFP techniques. Eur Radiol.
distortion correction for MRI in low field permanent 2003;13(11):2409–18.
magnet systems with strong B0 inhomogeneity and 45. Sarracanie M. Fast quantitative low-field magnetic res-
gradient field nonlinearities. Magn Reson Mater Phys onance imaging with OPTIMUM-optimized magnetic
Biol Med. 2021;34(4):631–42. resonance fingerprinting using a stationary steady-
30. Haskell MW, Nielsen JF, Noll DC. Off-resonance state Cartesian approach and accelerated acquisition
artifact correction for MRI: a review. NMR Biomed. schedules. Invest Radiol. 2022;57(4):263–71.
2023;36(5):e4867. 46. Han F, Zhou Z, Han E, Gao Y, Nguyen KL, Finn JP,
31. Guermazi A, Miaux Y, Zaim S, Peterfy CG, White D, et al. Self-gated 4D multiphase, steady-state imaging
Genant HK. Metallic artefacts in MR imaging: effects with contrast enhancement (MUSIC) using rotating
of main field orientation and strength. Clin Radiol. cartesian K-space (ROCK): validation in children
2003;58(4):322–8. with congenital heart disease. Magn Reson Med.
32. Jungmann PM, Ganter C, Schaeffeler CJ, Bauer JS, 2017;78(2):472–83.
Baum T, Meier R, et al. View-angle tilting and slice- 47. Han F, Zhou Z, Du D, Gao Y, Rashid S, Cao M, et al.
encoding metal artifact correction for artifact reduc- Respiratory motion-resolved, self-gated 4D-MRI
tion in MRI: experimental sequence optimization for using rotating Cartesian K-space (ROCK): initial
orthopaedic tumor Endoprostheses and clinical appli- clinical experience on an MRI-guided radiotherapy
cation. PLoS One. 2015;10(4):e0124922. system. Radiother Oncol. 2018;127(3):467–73.
458 Y. Yang et al.
48. Huang WY, Wen JB, Wu G, Yin B, Li JJ, Geng 55. Shaverdian N, Yang Y, Hu P, Hart S, Sheng K,
DY. Diffusion-weighted imaging for predicting and Lamb J, et al. Feasibility evaluation of diffusion-
monitoring primary central nervous system lym- weighted imaging using an integrated MRI-
phoma treatment response. AJNR Am J Neuroradiol. radiotherapy system for response assessment to
2016;37(11):2010–8. neoadjuvant therapy in rectal cancer. Br J Radiol.
49. Yang Y, Cao M, Sheng K, Gao Y, Chen A, Kamrava 2017;90(1071):20160739.
M, et al. Longitudinal diffusion MRI for treatment 56. Le Bihan D, Breton E, Lallemand D, Grenier P,
response assessment: preliminary experience using an Cabanis E, Laval-Jeantet M. MR imaging of intra-
MRI-guided tri-cobalt 60 radiotherapy system. Med voxel incoherent motions: application to diffusion
Phys. 2016;43(3):1369–73. and perfusion in neurologic disorders. Radiology.
50. Gao Y, Han F, Zhou Z, Cao M, Kaprealian T, Kamrava 1986;161(2):401–7.
M, et al. Distortion-free diffusion MRI using an MRI- 57. Iima M, Le Bihan D. Clinical Intravoxel incoherent
guided Tri-Cobalt 60 radiotherapy system: sequence motion and diffusion MR imaging: past, present, and
verification and preliminary clinical experience. Med future. Radiology. 2016;278(1):13–32.
Phys. 2017;44(10):5357–66. 58. Ingo C, Magin RL, Colon-Perez L, Triplett W, Mareci
51. Lewis B, Guta A, Mackey S, Gach HM, Mutic S, TH. On random walks and entropy in diffusion-
Green O, et al. Evaluation of diffusion-weighted MRI weighted magnetic resonance imaging studies of
and geometric distortion on a 0.35T MR-LINAC neural tissue: random walks and entropy in diffusion-
at multiple gantry angles. J Appl Clin Med Phys. weighted MRI. Magn Reson Med. 2014;71(2):617–27.
2021;22(2):118–25. 59. Karaman MM, Sui Y, Wang H, Magin RL, Li Y, Zhou
52. Nardini M, Capotosti A, Mazzoni LN, Cusumano XJ. Differentiating low- and high-grade pediatric
D, Boldrini L, Chiloiro G, et al. Tuning the optimal brain tumors using a continuous-time random-walk
diffusion-weighted MRI parameters on a 0.35-T diffusion model at high b -values: pediatric brain
MR-Linac for clinical implementation: a phantom tumors using a CTRW diffusion model. Magn Reson
study. Front Oncol. 2022;12:867792. Med. 2016;76(4):1149–57.
53. Gao Y, Ghodrati V, Kalbasi A, Fu J, Ruan D, Cao M, 60. Pham J, Savjani RR, Gao Y, Cao M, Hu P, Sheng
et al. Prediction of soft tissue sarcoma response to K, et al. Evaluation of T2-weighted MRI for visu-
radiotherapy using longitudinal diffusion MRI and alization and sparing of urethra with MR-guided
a deep neural network with generative adversarial radiation therapy (MRgRT) on-board MRI. Cancers.
network-based data augmentation. Med Phys. 2021;13(14):3564.
2021;48(6):3262–372. 61. Pham J, Savjani RR, Yoon SM, Yang T, Gao Y, Cao
54. Gao Y, Kalbasi A, Hsu W, Ruan D, Fu J, Shao J, M, et al. Urethral interfractional geometric and
et al. Treatment effect prediction for sarcoma patients dosimetric variations of prostate cancer patients:
treated with preoperative radiotherapy using radiomics a study using an onboard MRI. Front Oncol.
features from longitudinal diffusion-weighted MRIs. 2022;12:916254.
Phys Med Biol. 2020;65(17):175006.
Image-Based Biomarkers
in Magnetic Resonance-Guided
24
Radiotherapy (MRgRT)
Andrea D’Aviero, Haidy Nasief, Charlotte Robert,

and Davide Cusumano
Introduction application of AI to healthcare has increased in

last years in different settings, from diagnosis to
Artifical intelligence (AI) is a scientific field with outcome prediction [2, 3]. Radiotherapy (RT) in
the main aim to explore relationships and con- particular has been consistently improving by the
nections in datasets with different data formats to introduction of AI, providing new tools to accel-
extract meaningful information. Using complex erate the patient workflow [1], supporting clini-
algorithms, AI is able of performing specific cians and physicists in the manual steps such as
tasks, reducing workload for humans, and play- contouring and treatment planning [4–8], improv-
ing a relevant role in the world of automation [1] ing workflow process and outcome evaluations.
but also providing new knowledge that is at the In this context, radiomics represents one of the
core of healthcare decision support tools. The most advanced applications in radiotherapy, con-
sisting in extracting numerical parameters from
images and then using such parameters to predict
A. D’Aviero
UO Radioterapia Oncologica, Mater Olbia Hospital,
treatment responses or treatment toxicities [9,
Olbia, Italy 10]. Radiomic features aim to describe tumors
e-mail: [email protected] from a genetic and phenotypical point of view
H. G. Nasief through image intensity analysis [11]. In other
Department of Radiation Oncology, Medical College words, in the era of omics-guided medicine,
of Wisconsin, Milwaukee, WI, USA radiomics is intended to support the definition of
image-based biomarkers to be used in treatment
C. Robert (*) decisions, alone or in conjunction with markers
UMR 1030 Radiothérapie Moléculaire et Innovation
Thérapeutique, ImmunoRadAI, Université Paris-
extracted by other “omics,” such as genomics,
Saclay, Institut Gustave Roussy, Inserm, metabolomics, transcriptomics, or proteomics [3,
Villejuif, France 12, 13]. Radiomics can be separated into two cat-
Department of Radiation Oncology, Gustave Roussy, egories: conventional radiomics, which relies on
Villejuif, France handcrafted features, and deep-learning-based
e-mail: [email protected] radiomics, for which the characteristics of inter-
D. Cusumano est are defined by the algorithm itself.
Fondazione Policlinico Universitario “Agostino Conventional radiomics consists in a complex
Gemelli” IRCCS, Rome, Italy
process of extraction of quantitative image fea-
UO Fisica Medica e Radioprotezione, Mater Olbia tures from a region of interest (ROI) followed by
Hospital, Olbia, Italy
460 A. D’Aviero et al.
a model creation correlated to clinical endpoints 5. Extracting radiomic features from reference
[9, 12, 14–16]. and daily images, and calculating delta-
As a subdomain of radiomics, delta-radiomic radiomic features as the relative net change of
strategies aim in particular to quantify the patient radiomic features as compared to the refer-
sensitivity to the treatment using longitudinal ence image.
assessment to study the temporal variation of the 6. Usage of feature selection strategies when
radiomic features during the course of treatment: necessary.
the potentialities of this approach are very rele- 7. Building of the model for the appropriate clin-
vant, and delta radiomics seems able to overcome ical task.
the performance obtainable by radiomic features 8. Evaluation of the model’s generalizability
extracted in single time frames, as demonstrated performances on an internal test set (subopti-
by some experiences reported in more details in mal) or an independent external test set [17].
the next paragraphs [15, 17, 18]. Within imaging
technologies, magnetic resonance imaging (MRI) Unlike computed tomography (CT) or positron-
represents one of the most interesting fields of emission tomography (PET), MR imaging has the
application of radiomics due to the increased soft disadvantage of being a nonquantitative imaging
tissue differentiation and availability of comple- modality for conventional T1- or T2-weighted ana-
mentary image contrasts. The implementation in tomical sequences, resulting in high intra- and inter-
RT of hybrid MR-Linac systems represents a image intensity variability. MR image intensities are
unique opportunity in the development and appli- therefore highly dependent on manufacturer,
cation of delta-radiomic approaches [17]. sequence type, and acquisition parameters such as
The aim of this chapter is to define the efforts magnetic field, repetition and echo times, and voxel
of radiomics in providing reliable biomarkers for size, all of which having a significant impact on the
RT treatment optimization in MRgRT. radiomic indices extracted [19–21]. This observa-
tion makes the development of robust radiomic sig-
natures that can be generalized to a multicenter scale
Radiomic Workflow in the Context even more complex. To solve this problem, several
of MRgRT methods have been proposed, ranging from simple
image preprocessing methods applied in the image
In the context of MRgRT, only conventional domain to methods applied in the feature domain
handcrafted radiomics and its delta-radiomics itself, notably a posteriori statistical method ComBat
variant have been implemented to date. The and its variants [22]. In MRI, although preprocess-
delta-radiomic process can be summarized into ing methods depend on the tumor location, they con-
the following major steps as shown in Fig. 24.1: ventionally include steps for correcting magnetic
field inhomogeneities, spatial resampling, and inten-
1. Acquiring longitudinal images during sity harmonization [10]. More recently, deep-learn-
treatment. ing methods have been proposed as an interesting
2. ROI delineation—At this stage, registering alternative for image harmonization [23]. A final
images acquired at the different time frames step before feature extraction involves gray-level
and populating ROI to subsequent fractions discretization, which consists of grouping similar
can be an option for time saving. intensities into bins, with the aim of minimizing the
3. Contour quality assurance checks. impact of noise on texture features and reducing
4. MRI preprocessing (including in particular computation time.
bias correction, spatial resampling, intensity Traditional radiomics then consists of extract-
standardization)—Use of a posteriori ing hand-created features calculated using pre-
radiomic features harmonization strategies defined mathematical formulae [24]. The features
can be also an option. include semantic and agnostic features. Semantic
24 Image-Based Biomarkers in Magnetic Resonance-Guided Radiotherapy (MRgRT) 461
Fig. 24.1 Major steps for delta-radiomic feature extraction and model building in MRgRT as described in the text
features refer to features conventionally used in 3. Second-order features, also known as textural
the vocabulary of radiologists to describe lesions. features, providing information on the
Agnostic features (Fig. 24.2) can be divided into arrangement of grayscale values in the region
four subgroups: of interest
4. Higher-order features extracted from filtered
1. Shape-based features describing the 3D images such as wavelets, Laplacian of
geometry of the segmented structure Gaussian (LoG) filter, square, square root,
2. First-order features quantifying the distribu- logarithmic, and exponential filters
tion of voxel intensities
a b
Fig. 24.2 Histogram or first-order features reflect voxel zone size matrices) reflect the complex and unique spatial
intensity distribution only (a). Unlike histogram-based arrangement of voxels [24]. (Reprinted from [24], with
features, textural or second-order features (b) derived permission from Elsevier)
from texture matrices (e.g., co-occurrence, run length,
Imaging Biomarkers in MRgRT A preliminary experience focused on prostate

cancer has recently compared radiomic features
The development of MRgRT techniques has been extracted from MR simulator with those extracted
made possible by the spread of hybrid MR-Linac from MR-Linac. For a total of 26 patients, the
units, based on the implementation of radiother- authors acquired 3 MR images: 1 using an
apy delivery systems with different onboard MR MR-simulator and 2 using the 1.5 T MR-Linac
scanners. A summary of the radiomic signatures before the delivery of the first treatment fraction.
related to these units is presented below. They used a 3D T2-w-TSE sequence considering
the gross tumor volume (GTV) and planning tar-
get volume (PTV) as regions of interest for
Elekta’s MR-Linac System (Unity) radiomic analysis and extracted 1023 radiomic
features. The results of this study reported that
Elekta’s hybrid MR-Linac system (Unity) has a only approximately 9% of the radiomic features
field strength of 1.5 T associated with a 7 MV extracted showed high repeatability and repro-
flattening filter-free (FFF) linear accelerator and ducibility [25].
is defined as high-field system by the scientific Very recently, a study was performed using
community as described in Chap. 10. Evidence in T2-weighted MRI scans acquired on a 1.5 T
the field of radiomics of such a system is limited MR-Linac to investigate the ability of MRI
to date, given its recent clinical implementation. radiomics to correctly classify patients according
to their D’Amico score (high risk vs. low/inter- extracted from reference and daily MRIs to pre-
mediate risk). A total of 111 patients were dict the necessity of online adaptation automati-
enrolled for model building (76 patients) and cally and objectively in MRgRT. The correlation
testing (35 patients), resulting in ROC-AUC val- between the features extracted from the reference
ues of 0.718 and 0.750 for the CTV- and PTV- and the daily MRIs calculated in a ROI can indi-
based models in the test cohort [26], cate the presence of anatomical differences. A
respectively. high correlation indicated small anatomic differ-
Another study was performed using combi- ence between the reference and daily images,
nation of wavelet delta-radiomic (WDRFs) implying that ATS is not necessary and that ATP
textures extracted from 90 daily motion-aver- is sufficient for that fraction, while a low correla-
aged MRI sets derived from 4D MRIs acquired tion would indicate substantial anatomic differ-
during MR-guided adaptive radiation therapy ence between the daily and reference images,
along with CA19-9, CEA, metastatic status, suggesting that ATS is necessary for the fraction.
and follow-up data. They developed and dem- To select the optimum ROI for feature extraction,
onstrated the method using 90 datasets isodose surface (IDS) from the reference plan
extracted from 18 patients treated in 5 stereo- was converted into structures, and the region
tactic body radiation therapy fractions with a enclosing the 100–50% IDS was selected as the
1.5 T MR-Linac and showed that incorporat- optimal ROI. Typically, the 100% IDS covers the
ing clinical biomarkers with the significant tumor volume, while the region between 100 and
WDRFs increased the concordance statistics 50% IDSs contains the dose-limiting OAR(s)
equal to 0.65 using clinical biomarkers alone with high-dose gradient (Fig. 24.3). Features
to 0.96 using the combination of delta extracted from this ROI can detect anatomical
radiomics with the clinical biomarkers [27]. changes that can affect the dose-limiting OARs
Delta radiomics has also been utilized recently and be used to guide the decision toward online
to aid in the treatment adaptation process to pre- ATS versus ATP workflow [28].
dict necessity of online adaptive replanning based
on wavelet multiscale delta-radiomic textures
extracted from 119 daily MRIs acquired during ViewRay’s MR-Linac System (MRIdian)
MRgART for 24 pancreatic cancer patients
treated on a 1.5 T MR-Linac [28]. The system Current evidence in the field of radiomics based
offers two adaptive planning techniques to on onboard MRI is mainly based on the deploy-
address the inter-fractional variations based on ment of hybrid MR-Linac units equipped with
the daily MRI: an adapt-to-position (ATP) work- low-field MR scanner (0.35 T). Such systems
flow where an adaptive plan is generated and have in fact been widely used in different centers,
optimized using the original contours from the showing very promising results in terms of feasi-
reference images by shifting the multi-leaf colli- bility and clinical effectiveness [6, 29]. The low-
mators (MLC) based on the co-registration of the field MR-Linac system joins a radiation therapy
daily and the reference image and an adapt-to- delivery system consisting of a 6MV FFF linear
shape (ATS) workflow (online adaptive replan- accelerator (MRIdian® ViewRay Inc. system,
ning), in which a new plan is generated based on Mountain View, California, USA) with a 0.35 T
the anatomy of the day requiring recontouring on MR scanner [30, 31]. The low-field MR system is
the daily MRI to fully correct for inter-fractional described in Chap. 9, and low-field imaging is
variations. The latter however is labor intensive presented in Chap. 23 of this book.
and time-consuming and might not be necessary One of the earliest experiences in this field
for each treatment; thus, there is a necessity to was provided by Boldrini et al. who conducted a
determine when it is needed during online treat- feasibility study of delta-radiomic analysis in
ment adaptation process. They developed a model locally advanced rectal cancer (LARC) [15]. In
based on multiscale wavelet radiomic features this cluster of patients, pathological complete
464
Wavelet Devcomposition Details

Approx. Horizontal Diagonal Vertical
Vertical L1
Multiscale
Texture
Horizontal Features
Wavelet Analysis
MR Date set L4
50–100% IDL Diagonal
Four Decomposition Levels (L1 to L4)
50 50 20 0.8
100 100 40 0.6
150 150 60 0.4
200 200 80 0.2
250 250 100 0

50 100 150 200 250 50 100 150 200 250 20 40 60 80 100
Display Daily image Display Reference MR images Display difference texture map
CLASSIFIER DECISION
ATS
Fig. 24.3 Wavelet multiscale delta-radiomic textures extracted from different decomposition levels from high-dose gradient regions enclosing the 100–50% IDS and classifier
building to automatically determine the necessity of online adaptive replanning using multiscale wavelet-based radiomic features.
A. D’Aviero et al.
response (pCR) is generally associated with bet- 2-year follow-up. The best performing model
ter local control (LC), metastasis-free survival was based on a delta-radiomic feature, analyzing
(MFS), and overall survival (OS) rates [32]. The the variation in the area/surface ratio between the
authors explored the possibility of establishing 54 Gy BED and the simulation time points (AUC
predictive models for clinical complete response of 0.92) [34].
(cCR), with the aim of being able to consider Recent evidence has also demonstrated the
conservative approaches in the clinical manage- validity of an additional biomarker of treatment
ment of patients with LARC. This predictivity response in LARCs called early regression index
was analyzed considering delta-radiomic features (ERITCP). This index models the radiobiological
in a cohort of 16 patients with LARC undergoing behavior of the tumor by considering its volumet-
neoadjuvant MRgRT. ric regression, from simulated treatment (Vpre) to
Radiomic feature extraction was performed mid-therapy (Vmid) [35]. The first theorization of
starting from GTV at different treatment time ERITCP was performed on T2-weighted MR
points, analyzing the change from baseline in images at 1.5 T and was found to be adequate in
terms of ratio, corresponding to the treatment identifying patients with and without pCR with a
simulation. The main result of this hypothesis- good sensitivity (86%) and negative predictive
generating study was the validity of the value (90%) and an area under the receiver oper-
delta-
radiomic features, which showed higher ating characteristic (ROC) curve (AUC) of 0.81
results in terms of predictive performance with (0.69–0.89 as 95% confidence interval, CI). The
respect to radiomic features extracted at the sin- identified cutoff value was found to be 13.1. The
gle time points. In particular, in the study two predictive power of ERITCP was also externally
parameters were shown to be able to fully iden- validated on low-field (0.35 T) MR-Linac sys-
tify complete responders, describing the variation tem, using a different image contrast (T2*/T1)
of a morphological parameter (delta variation of and identifying the BED value of 25 Gy as the
least length, ΔLleast) and a textural one (variation optimal time point to make predictions [36]. The
of gray-level non-uniformity, Δglnu) between the impact of ERITCP has been translated into clini-
second week of treatment and simulation [33], as cal practice through a clinical trial focused on a
associated with cCR. dose escalation protocol based on pCR prediction
The validity of these parameters was then sub- [37]. Patients enrolled in the clinical trial start
sequently evaluated in a multicenter setting, with their MRIgRT treatment with a prescribed dose
data from 43 LARC patients treated with the of 55 Gy in 25 fractions on the GTV plus the cor-
MRgRT technique in 2 different institutions. To responding mesorectum and 45 Gy in 25 frac-
obtain a homogeneous analysis of the extracted tions on the drainage nodes. ERITCP is calculated
data, the radiomic features ΔLleast and Δglnu were at fraction 10: patients showing good prognosis
extracted from MR images acquired at identical (ERITCP <13.1) continue their initial treatment;
biologically effective doses (BED) correspond- the others change the treatment schedule, intensi-
ing to 26.8 Gy. The prediction accuracies of cCR fying the dose on residual GTV through an addi-
in the validation study were 81% and 63% for tional boost, scaling up to 60.1 Gy.
ΔLleast and Δglnu, respectively [33]. The study Additionally, the potentialities of ERITCP
reported also promising results of ΔLleast in pre- have been explored also in the context of cervical
dicting pCR, with an accuracy of 79%. cancer. In a preliminary experience carried out on
With a similar design, a logistic regression 16 patients affected by locally advanced cervical
model was built to predict 2-year disease-free cancer (LACC) and treated using 0.35 T MRgRT,
survival (2yDFS) from multi-institutional data Cusumano et al. showed that ERITCP was able to
including clinical and radiomic variables predict pCR with an area under ROC curve of
extracted from the GTV. In this study, 48 LARC 0.84 [38]. In another experience performed on 88
patients were retrospectively enrolled, 8 of them LACC patients, ERITCP showed significant per-
(16.7%) having developed distant metastases at a formances in predicting not only pCR but also
local control, disease-free survival, and overall geneity of pancreatic neoplasms through
survival after 2 years from the end of treatment extracted radiological features, capable of early
[39]. Interestingly, in this experience, ERITCP identifying treatment responders.
was calculated also starting from diffusion-
weighted (DW) MRI and not only starting from
T2-weighted, and the authors observed that the Future Developments
parameter calculated on DW images showed and Perspectives
higher performance with respect to T2 (AUC of
0.77 and 0.75 for DW and T2-w, respectively, in The development of delta-radiomics algorithms
predicting pCR). and related evidence is opening a new perspec-
Another interesting site where MRgRT tive in the radiation treatment of oncological dis-
reported promising results was the treatment of eases. The variation of radiomic parameters
locally advanced pancreatic cancer (LAPC). during RT treatment as a presumptive index of
The availability of high soft tissue contrast tumor response could lead to the development of
images and the possibility of online adapting response prediction models to be used for accu-
every single fraction have led to different dose rate adaptation of ongoing treatment. This per-
escalation studies with excellent results in terms spective is indeed the focus of randomized
of local control and overall survival [40]. clinical trials that will assess the appropriateness
Pancreatic neoplasms represent a further tumor of dose escalation in patients identified by pre-
site where radiomic analysis has yielded inter- dictive response models as poor responders [44].
esting results, with Simpson et al.’s study being In view of the progressive increase in the
the first to evaluate the possibility of extracting number of centers equipped with hybrid
radiomic features [41]. Analyzing features MR-Linac units, the definition of multicenter
extracted from the GTV of 20 patients with collaborative networks for large-scale validation
LAPCs, machine learning models based on of predictive models is desirable [45]. The inte-
LASSO regression and random forest were pro- gration of image standardization methods into
posed, with an AUC of 0.81 (95% CI 0.59–1) for the radiomic workflow, which have so far been
the best model. A similar study was conducted little applied in this MR-Linac context, could
on 26 patients with locally advanced and bor- become a point of attention to ensure the general-
derline resectable pancreatic adenocarcinoma ization of the models developed in a multicenter
treated with MRgRT of 50 Gy in 5 fractions. setting [46, 47]. A further relevant perspective in
The results showed that linear normalization of defining the future role of AI and radiomics in
the image intensities to the median kidney value MRgRT treatment is that of integrating biomark-
improved the reproducibility of the radiomic ers derived from other omics sciences (pro-
features, with histogram skewness being signifi- teomics, genomics, metabolomics) for the
cantly associated with progression-free survival personalization of each treatment.
(p = 0.005, HR = 2.75) [42]. This chapter has reviewed the current litera-
In LAPC, the 1-year local control rate (1y- ture about radiomics applied to MRgRT, focusing
LC) has also been the focus of studies using on delta-radiomics approach, which is made pos-
radiomic analyses to define predictive models. A sible by the acquisition of longitudinal images at
retrospective study of 35 patients treated with several points during the course of radiotherapy
adaptive MRgRT identified a textural feature treatment. Current knowledge is essentially based
capable of predicting 1y-LC [43]. This feature on low-field MRI, due to its earlier implementa-
was the variation of cluster shade with respect to tion, with promising results for several patholo-
the simulation, calculated when a BED value of gies, i.e., rectal, cervical, and pancreatic cancers.
40 Gy was reached (p = 0.005, AUC = 0.78). The The analysis of regression kinetics, which are
results of these studies would seem to hypothe- ultimately very simple to extract and the opposite
size the possibility of characterizing the hetero- of the black box concept, seems promising for
the introduction of personalized treatments and is 12. Lambin P, Leijenaar RTH, Deist TM, et al. Radiomics:
still being evaluated in clinical trials. Preliminary the bridge between medical imaging and personalized
medicine. Nat Rev Clin Oncol. 2017;14:749–62.
analyses on the 1.5 T MR-Linac are very promis- 13. Bodalal Z, Trebeschi S, Beets-Tan R. Radiomics: a
ing and can play a critical role in online treatment critical step towards integrated healthcare. Insights
adaptation workflows. The integration of delta Imaging. 2018;9:911–4.
radiomics with clinical biomarkers is another 14. Chiloiro G, Rodriguez-Carnero P, Lenkowicz J,
et al. Delta Radiomics can predict distant metastasis
interesting extension that can increase the ability in locally advanced rectal cancer: the challenge to
to create interpretable omics-based models to be personalize the cure. Front Oncol. 2020;10:595012.
developed into invaluable tool to predict treat- 15. Boldrini L, Cusumano D, Chiloiro G, et al. Delta
ment outcomes. radiomics for rectal cancer response prediction with
hybrid 0.35 T magnetic resonance-guided radiother-
apy (MRgRT): a hypothesis-generating study for an
innovative personalized medicine approach. Radiol
References Med. 2019;124:145–53.
16. Gardin I, Grégoire V, Gibon D, et al. Radiomics: prin-
1. Vandewinckele L, Claessens M, Dinkla A, et al. ciples and radiotherapy applications. Crit Rev Oncol
Overview of artificial intelligence-based applica- Hematol. 2019;138:44–50.
tions in radiotherapy: recommendations for imple- 17. Nardone V, Reginelli A, Grassi R, et al. Delta
mentation and quality assurance. Radiother Oncol. radiomics: a systematic review. Radiol Med.
2020;153:55–66. 2021;126:1571–83.
2. Ramesh AN, Kambhampati C, Monson JRT, et al. 18. Cusumano D, Boldrini L, Yadav P, et al. Delta
Artificial intelligence in medicine. Ann R Coll Surg Radiomics analysis for local control prediction in
Engl. 2004;86:334–8. pancreatic cancer patients treated using magnetic res-
3. Limkin EJ, Sun R, Dercle L, et al. Promises and onance guided radiotherapy. Diagnostics. 2021;11:72.
challenges for the implementation of computational 19. Ford J, Dogan N, Young L, et al. Quantitative
medical imaging (radiomics) in oncology. Ann Oncol. radiomics: impact of pulse sequence parameter selec-
2017;28:1191–206. tion on MRI-based textural features of the brain.
4. Casà C, Piras A, D’Aviero A, et al. The impact Contrast Media Mol Imaging. 2018;2018:1729071.
of radiomics in diagnosis and staging of pan- 20. Buch K, Kuno H, Qureshi MM, et al. Quantitative
creatic cancer. Ther Adv Gastrointest Endosc. variations in texture analysis features dependent on
2022;15:26317745221081596. MRI scanning parameters: a phantom model. J Appl
5. Boulanger M, Nunes J-C, Chourak H, et al. Deep Clin Med Phys. 2018;19:253–64.
learning methods to generate synthetic CT from 21. Mayerhoefer ME, Szomolanyi P, Jirak D, et al. Effects
MRI in radiotherapy: a literature review. Phys Med. of MRI acquisition parameter variations and protocol
2021;89:265–81. heterogeneity on the results of texture analysis and
6. Cusumano D, Boldrini L, Dhont J, et al. Artificial intel- pattern discrimination: an application-oriented study.
ligence in magnetic resonance guided radiotherapy: Med Phys. 2009;36:1236–43.
medical and physical considerations on state of art 22. Da-Ano R, Visvikis D, Hatt M. Harmonization strat-
and future perspectives. Phys Med. 2021;85:175–91. egies for multicenter radiomics investigations. Phys
7. Robert C, Munoz A, Moreau D, et al. Clinical imple- Med Biol. 2020;65:24TR02.
mentation of deep-learning based auto-contouring 23. Tixier F, Jaouen V, Hognon C, et al. Evaluation of
tools—experience of three French radiotherapy cen- conventional and deep learning based image harmoni-
ters. Cancer Radiother. 2021;25:607–16. zation methods in radiomics studies. Phys Med Biol.
8. McIntosh C, Conroy L, Tjong MC, et al. Clinical inte- 2021;66:245009.
gration of machine learning for curative-intent radia- 24. Reuzé S, Schernberg A, Orlhac F, et al. Radiomics in
tion treatment of patients with prostate cancer. Nat nuclear medicine applied to radiation therapy: meth-
Med. 2021;27:999–1005. ods, pitfalls, and challenges. Int J Radiat Oncol Biol
9. Lambin P, Rios-Velazquez E, Leijenaar R, et al. Phys. 2018;102:1117–42.
Radiomics: extracting more information from medi- 25. Xue C, Yuan J, Poon DM, et al. Reliability of
cal images using advanced feature analysis. Eur J MRI radiomics features in MR-guided radiother-
Cancer. 2012;48:441–6. apy for prostate cancer: repeatability, reproduc-
10. Dercle L, Henry T, Carré A, et al. Reinventing radia- ibility, and within-subject agreement. Med Phys.
tion therapy with machine learning and imaging bio- 2021;48:6976–86.
markers (radiomics): state-of-the-art, challenges and 26. Zhou Y, Yuan J, Xue C, et al. A pilot study of MRI
perspectives. Methods. 2021;188:44–60. radiomics for high-risk prostate cancer stratification
11. Kumar V, Gu Y, Basu S, et al. Radiomics: the pro- in 1.5 T MR-guided radiotherapy. Magn Reson Med.
cess and the challenges. Magn Reson Imaging. 2023;89:2088–99.
2012;30:1234–48. 27. Nasief HG, Chen X, Erickson BA, et al. Improving
outcomes prediction using a combination of MRI
Wavelet Delta-Radiomic textures and clinical bio- pathological complete response in cervical cancer: a
markers for MR-guided adaptive radiation therapy pilot-study. Appl Sci. 2020;10:8001.
of pancreatic cancer. Int J Radiat Oncol Biol Phys. 39. Cusumano D, Russo L, Gui B, et al. Evaluation of
2022;114:S79–80. early regression index as response predictor in cervi-
28. Nasief HG, Parchur AK, Omari E, et al. Predicting cal cancer: a retrospective study on T2 and DWI MR
necessity of daily online adaptive replanning based on images. Radiother Oncol. 2022;174:30–6.
wavelet image features for MRI guided adaptive radi- 40. Rudra S, Jiang N, Rosenberg SA, et al. Using adaptive
ation therapy. Radiother Oncol. 2022;176:165–71. magnetic resonance image-guided radiation therapy
29. Kurz C, Buizza G, Landry G, et al. Medical physics for treatment of inoperable pancreatic cancer. Cancer
challenges in clinical MR-guided radiotherapy. Radiat Med. 2019;8:2123–32.
Oncol. 2020;15:93. 41. Simpson G, Spieler B, Dogan N, et al. Predictive
30. Klüter S. Technical design and concept of a 0.35 T value of 0.35 T magnetic resonance imaging radiomic
MR-Linac. Clin Transl Radiat Oncol. 2019;18:98–101. features in stereotactic ablative body radiother-
31. Mutic S, Dempsey JF. The ViewRay system: magnetic apy of pancreatic cancer: a pilot study. Med Phys.
resonance-guided and controlled radiotherapy. Semin 2020;47:3682–90.
Radiat Oncol. 2014;24:196–9. 42. Tomaszewski MR, Latifi K, Boyer E, et al. Delta
32. Martin ST, Heneghan HM, Winter DC. Systematic radiomics analysis of magnetic resonance guided
review and meta-analysis of outcomes follow- radiotherapy imaging data can enable treatment
ing pathological complete response to neoadjuvant response prediction in pancreatic cancer. Radiat
chemoradiotherapy for rectal cancer. Br J Surg. Oncol. 2021;16:237.
2012;99:918–28. 43. Cusumano D, Boldrini L, Yadav P, et al. Delta
33. Cusumano D, Boldrini L, Yadav P, et al. Delta Radiomics analysis for local control prediction in
radiomics for rectal cancer response prediction using pancreatic cancer patients treated using magnetic
low field magnetic resonance guided radiotherapy: an resonance guided radiotherapy. Diagnostics (Basel).
external validation. Phys Med. 2021;84:186–91. 2021;11:72.
34. Chiloiro G, Boldrini L, Preziosi F, et al. A predictive 44. Chiloiro G, Cusumano D, Boldrini L, et al.
model of 2yDFS during MR-guided RT neoadjuvant THUNDER 2: THeragnostic Utilities for Neoplastic
chemoradiotherapy in locally advanced rectal cancer DisEases of the rectum by MRI guided radiotherapy.
patients. Front Oncol. 2022;12:831712. BMC Cancer. 2022;22:67.
35. Fiorino C, Gumina C, Passoni P, et al. A TCP-based 45. de Mol van Otterloo SR, Christodouleas JP, Blezer
early regression index predicts the pathological ELA, et al. The MOMENTUM study: an inter-
response in neo-adjuvant radio-chemotherapy of rec- national registry for the evidence-based introduc-
tal cancer. Radiother Oncol. 2018;128:564–8. tion of MR-guided adaptive therapy. Front Oncol.
36. Cusumano D, Boldrini L, Yadav P, et al. External vali- 2020;10:1328.
dation of early regression index (ERITCP) as predic- 46. Orlhac F, Lecler A, Savatovski J, et al. How can we
tor of pathologic complete response in rectal cancer combat multicenter variability in MR radiomics?
using magnetic resonance-guided radiation therapy. Validation of a correction procedure. Eur Radiol.
Int J Radiat Oncol Biol Phys. 2020;108:1347–56. 2021;31:2272–80.
37. Chiloiro G. THeragnostic Utilities for Neoplastic 47. Carré A, Klausner G, Edjlali M, et al. Standardization
DisEases of the rectum by MRI guided radiotherapy. of brain MR images across machines and protocols:
clinicaltrials.gov; 2021. bridging the gap for MRI-based radiomics. Sci Rep.
38. Cusumano D, Catucci F, Romano A, et al. Evaluation 2020;10:12340.
of an early regression index (ERITCP) as predictor of
Epilogue and Outlook
25
Filippo Alongi, Bharat B. Mittal, Indra J. Das,
and Poonam Yadav
Introduction around the world, and the chapters cover every

system available for patient treatment.
MR-Linac is a nascent technology that allows for The general section of the book provides an
simultaneous real-time imaging and precise radi- Introduction (Chap. 1) and then addresses the
ation treatment. It has great potential to impact Role of MRI (Chap. 2), Clinical Necessities
clinical practice and clinical research in radiation (Chap. 3), Education and Training (Chap. 4),
oncology. There have been several review articles Imaging Sequence (Chap. 5), Motion
describing its usage and potential [1, 2]. Management (Chap. 6), Synthetic CT (Chap.
Additionally, there is momentum behind MRI- 7), and Treatment Planning (Chap. 8). The next
only-based radiation treatment [3–5]. There was section describes the existing MR-Linac sys-
a need to gather and summarize the knowledge tems which are being used to treat patients.
and literature on this technology, and this was the These include the ViewRay MRIdian (Chap. 9),
starting point of our thoughts for a book, which Elekta Unity (Chap. 10), and Aurora Magneton
we undertook beginning in the Spring of 2022. (Chap. 11).
The 25 chapters collected here are organized in 4 Since MR-Linac is used for many disease
major sections: General, MR-Linac systems, sites, it was decided that the coverage should be
Clinical, and Miscellaneous futuristic topics. Our given to the most common disease sites, which
view was that chapters should be written by would benefit readers looking to glean from the
experts in each field; hence, our authors represent experience of expert authors and adapt their treat-
a vast number of researchers and clinicians ment options and policies accordingly. Thus, in
the Clinical section, we included chapters on
Oligometastatic Diseases (Chap. 12), Pancreas
(Chap. 13), Liver (Chap. 14), Lung (Chap. 15),
F. Alongi (*)
University of Brescia, Brescia, Italy Prostate (Chap. 16), Breast (Chap. 17),
Gynecological Diseases (Chap. 18), Pediatrics
Advanced Radiation Oncology Department, IRCCS
Sacro Cuore Don Calabria Hospital, (Chap. 19), Brain (Chap. 20), and Head and Neck
Negrar di Valpolicella, Italy (Chap. 21).
B. B. Mittal · I. J. Das · P. Yadav The final section of the book includes chapters
Department of Radiation Oncology, Northwestern which are more futuristic. These chapters cover
Memorial Hospital, Northwestern University the Australian Sydney Project on MR-Linac
Feinberg School of Medicine, Chicago, IL, USA (Chap. 22), Low-Field Imaging (Chap. 23), and
[email protected]; [email protected]
470 F. Alongi et al.
Artificial Intelligence and Image Texture Analysis • 4D MRI option to help manage anatomic
(Chap. 24). motion.
A wealth of information about MR-Linac • Accurate synthetic CT option to reduce depen-
technology and its emerging clinical applications dence on CT imaging.
has been collated here, with up-to-date knowl- • Simplified imaging quality assurance (QA)
edge and insights. Future editions of this book option for low-field MRI.
will include important technological advances, • 3D couch motion (missing in the current Unity
especially autosegmentation and gating tools system).
integrated into the MR-Linac workflow, but more • Curate an image library, to increase clinical
importantly the clinical outcome data from many use in challenging cases.
ongoing trials. • Add an option to check the robustness of the
original plan.
• Technical challenges in the breast due to elec-
Wish List and Future of MR-Linac tron return effect [28, 29].
• Choice of multiple-beam energy.
The future of MR-Linac depends on technologi- • Develop methods for biomarkers and
cal advances but more importantly on the clinical radiomics [30–36].
outcome, which is described in detail by
Rammohan et al. [6]. Currently, a large amount
of clinical data has been reported, but most of it Summary
with a short follow-up and a limited population
size, as shown in the various clinical trials and To conclude, MR-Linac represents one of the
references [5, 7–27]. most advanced medical devices, combining MRI
Some of the potential improvements of this and linear accelerator technology. It unifies the
innovative technology are elaborated below. At capabilities of an MRI scanner and a linear accel-
present, most MR-Linac setups can only accom- erator into one system, enabling real-time imag-
modate 6–8 patients/day, which is not optimal ing and improving radiotherapy delivery during
considering the number of waiting lists and tumor treatment.
treatment requests. To increase the number of The MR-Linac system allows for precise and
treatable patients, several aspects have to be personalized radiotherapy by providing, as was
incorporated: not possible before, high-quality real-time
images of the tumor and surrounding tissue.
• Patient comfort procedures should be Moreover, it allows physicians to track tumor
streamlined. location; account for changes in shape, size, and
• Gantry size of the MR-Linac and field of view position during treatment; and adjust radiation
need to be increased. therapy accordingly. These opportunities and
• Autosegmentation needs to be faster and more challenges have been described by Hehakaya
reliable. et al. [37].
• Dose rate should be modulated in order to MR-Linac can offer several advantages by
minimize treatment time. combining MRI and radiotherapy in one machine.
• Volumetric modulated arc therapy (VMAT) It can improve the precision of treatment and
should be standard to improve dose distribu- reduce potential damage to healthy tissue around
tion and shorten treatment time for each the tumor. Real-time imaging also enables adap-
session. tive treatment planning, allowing radiation to be
• Better image sequencing to enhance morpho- precisely delivered to target areas while minimiz-
logical and functional aspects. ing stress on healthy tissue.
• Online cumulative dose option, to increase the However, MR-Linac technology is relatively
quality of the more complex treatment plans. new, and its evolution is still ongoing. Research
25 Epilogue and Outlook 471
and development work is being continuously car- 6. Rammohan N, Randall JW, Yadav P. History of
ried out to further improve its capabilities and technological advancements towards MR-Linac: the
future of image-guided radiotherapy. J Clin Med.
clinical applications. Promising improvements 2022;11:4730.
include: 7. Kishan AU, Ma TM, Lamb JM, Casado M, et al.
Magnetic resonance imaging-guided vs computed
• Improved targeting, to minimize healthy tis- tomography-guided stereotactic body radiotherapy
for prostate cancer: the MIRAGE randomized clinical
sue damage and increase effectiveness of the trial. JAMA Oncol. 2023;9:365–73.
treatment 8. Michalet M, Bettaïeb O, Khalfi S, Ghorbel A, et al.
• Adaptive treatment, especially for lesions that Stereotactic MR-guided radiotherapy for adrenal
move or change shape during the treatment gland metastases: first clinical results. J Clin Med.
2022;12:291.
sessions, including lung tumors or upper 9. Alongi F, Rigo M, Figlia V, Nicosia L, et al. 1.5T
abdominal sites, affected directly or indirectly MR-guided daily-adaptive SBRT for prostate cancer:
by breathing preliminary report of toxicity and quality of life of the
• Soft tissue visualization, allowing users to first 100 patients. J Pers Med. 2022;12:1982.
10. Nierer L, Eze C, da Silva MV, Braun J, et al.
identify and precisely target lesions difficult to Dosimetric benefit of MR-guided online adaptive
define with standard imaging radiotherapy in different tumor entities: liver, lung,
abdominal lymph nodes, pancreas and prostate.
Indeed, MR-Linac technology has the poten- Radiat Oncol. 2022;17:53.
11. Nicosia L, Trapani G, Rigo M, Giaj-Levra N, et al.
tial capability to advance radiotherapy techniques 1.5 T MR-guided daily adapted SBRT on lymph node
and definitively improve outcomes for cancer oligometastases from prostate cancer. J Clin Med.
patients. Nevertheless, while MR-Linac repre- 2022;11:6658.
sents one of the most innovative steps in modern 12. Chuong MD, Herrera R, Kaiser A, Rubens M, et al.
Induction chemotherapy and ablative stereotactic
radiation therapy, its clinical availability and magnetic resonance image-guided adaptive radiation
application in clinical practice may vary across therapy for inoperable pancreas cancer. Front Oncol.
different countries and facilities. The impact of 2022;12:888462.
MR-Linac on clinical outcomes still remains an 13. Cuccia F, Rigo M, Figlia V, Giaj-Levra N, et al. 1.5T
MR-guided daily adaptive stereotactic body radio-
unresolved question, which will become clearer therapy for prostate re-irradiation: a preliminary
with the many clinical trials that are ongoing. report of toxicity and clinical outcomes. Front Oncol.
Future studies and technological improvements 2022;12:858740.
are awaited. 14. Versteeg AL, Sahgal A, Kawahara N, Rhines LD,
et al. Patient satisfaction with treatment outcomes
after surgery and/or radiotherapy for spinal metasta-
ses. Cancer. 2019;125:4269–77.
References 15. Sayan M, Serbez I, Teymur B, Gur G, et al. Patient-
reported tolerance of magnetic resonance-guided
1. Pollard JM, Wen Z, Sadagopan R, Wang J, et al. The radiation therapy. Front Oncol. 2020;10:1782.
future of image-guided radiotherapy will be MR 16. Tan H, Cheung P, Louie AV, Myrehaug S, et al.
guided. Br J Radiol. 2017;90:20160667. Outcomes of extra-cranial stereotactic body radio-
2. Das IJ, McGee KP, Tyagi N, Wang H. Role and therapy for metastatic breast cancer: treatment indica-
future of MRI in radiation oncology. Br J Radiol. tion matters. Radiother Oncol. 2021;161:159–65.
2019;92:20180505. 17. Weykamp F, Hoegen P, Klüter S, Spindeldreier CK,
3. Owrangi AM, Greer PB, Glide-Hurst CK. MRI-only et al. Magnetic resonance-guided stereotactic body
treatment planning: benefits and challenges. Phys radiotherapy of liver tumors: initial clinical experi-
Med Biol. 2018;63:05tr01. ence and patient-reported outcomes. Front Oncol.
4. Srinivasan S, Dasgupta A, Chatterjee A, Baheti A, 2021;11:610637.
et al. The promise of magnetic resonance imaging 18. Hehakaya C, van der Voort van Zyp JRN, Vanneste
in radiation oncology practice in the management BGL, Grutters JPC, et al. Early health economic
of brain, prostate, and GI malignancies. JCO Glob analysis of 1.5 T MRI-guided radiotherapy for local-
Oncol. 2022;8:e2100366. ized prostate cancer: decision analytic modelling.
5. Chandarana H, Wang H, Tijssen RHN, Das Radiother Oncol. 2021;161:74–82.
IJ. Emerging role of MRI in radiation therapy. J Magn 19. de Mol van Otterloo SR, Christodouleas JP, Blezer
Reson Imaging. 2018;48:1468–78. ELA, Akhiat H, et al. The MOMENTUM study: an
international registry for the evidence-based introduc-
472 F. Alongi et al.
tion of MR-guided adaptive therapy. Front Oncol. treated with partial breast irradiation using the 1.5
2020;10:1328. T MR-linear accelerator. Clin Transl Radiat Oncol.
20. Christiansen RL, Dysager L, Hansen CR, Jensen 2021;27:103–8.
HR, et al. Online adaptive radiotherapy potentially 29. Nachbar M, Mönnich D, Boeke S, Gani C, et al.
reduces toxicity for high-risk prostate cancer treat- Partial breast irradiation with the 1.5 T MR-Linac:
ment. Radiother Oncol. 2022;167:165–71. first patient treatment and analysis of electron return
21. Chuong MD, Bryant J, Mittauer KE, Hall M, et al. and stream effects. Radiother Oncol. 2020;145:30–5.
Ablative 5-fraction stereotactic magnetic resonance- 30. McGee KP, Hwang KP, Sullivan DC, Kurhanewicz
guided radiation therapy with on-table adap- J, et al. Magnetic resonance biomarkers in radiation
tive replanning and elective nodal irradiation for oncology: the report of AAPM task group 294. Med
inoperable pancreas cancer. Pract Radiat Oncol. Phys. 2021;48:e697–732.
2021;11:134–47. 31. Otazo R, Lambin P, Pignol JP, Ladd ME, et al.
22. Krishnan S, Chadha AS, Suh Y, Chen HC, et al. Focal MRI-guided radiation therapy: an emerging para-
radiation therapy dose escalation improves over- digm in adaptive radiation oncology. Radiology.
all survival in locally advanced pancreatic cancer 2021;298:248–60.
patients receiving induction chemotherapy and con- 32. Mata LA, Retamero JA, Gupta RT, García Figueras
solidative chemoradiation. Int J Radiat Oncol Biol R, et al. Artificial intelligence-assisted prostate can-
Phys. 2016;94:755–65. cer diagnosis: radiologic-pathologic correlation.
23. Luterstein E, Cao M, Lamb J, Raldow AC, et al. Radiographics. 2021;41:1676–97.
Stereotactic MRI-guided adaptive radiation therapy 33. Granzier RWY, Verbakel NMH, Ibrahim A, van
(SMART) for locally advanced pancreatic cancer: a Timmeren JE, et al. MRI-based radiomics in
promising approach. Cureus. 2018;10:e2324. breast cancer: feature robustness with respect to
24. Finazzi T, Haasbeek CJA, Spoelstra FOB, Palacios inter-observer segmentation variability. Sci Rep.
MA, et al. Clinical outcomes of stereotactic 2020;10:14163.
MR-guided adaptive radiation therapy for high- 34. Cusumano D, Boldrini L, Dhont J, Fiorino C, et al.
risk lung tumors. Int J Radiat Oncol Biol Phys. Artificial intelligence in magnetic resonance guided
2020;107:270–8. radiotherapy: medical and physical considerations
25. Alongi F, Rigo M, Figlia V, Cuccia F, et al. 1.5 T on state of art and future perspectives. Phys Med.
MR-guided and daily adapted SBRT for prostate 2021;85:175–91.
cancer: feasibility, preliminary clinical tolerability, 35. Gui B, Autorino R, Miccò M, Nardangeli A, et al.
quality of life and patient-reported outcomes during Pretreatment MRI Radiomics based response pre-
treatment. Radiat Oncol. 2020;15:69. diction model in locally advanced cervical cancer.
26. Ugurluer G, Atalar B, Zoto Mustafayev T, Gungor G, Diagnostics (Basel). 2021;11:631.
et al. Magnetic resonance image-guided adaptive ste- 36. Cusumano D, Russo L, Gui B, Autorino R, et al.
reotactic body radiotherapy for prostate cancer: pre- Evaluation of early regression index as response
liminary results of outcome and toxicity. Br J Radiol. predictor in cervical cancer: a retrospective study
2021;94:20200696. on T2 and DWI MR images. Radiother Oncol.
27. Bruynzeel AME, Tetar SU, Oei SS, Senan S, et al. A 2022;174:30–6.
prospective single-arm phase 2 study of stereotactic 37. Hehakaya C, Sharma AM, van der Voort Van Zijp
magnetic resonance guided adaptive radiation therapy JRN, Grobbee DE, et al. Implementation of magnetic
for prostate cancer: early toxicity results. Int J Radiat resonance imaging-guided radiation therapy in rou-
Oncol Biol Phys. 2019;105:1086–94. tine care: opportunities and challenges in the United
28. De-Colle C, Nachbar M, Mӧnnich D, Boeke S, et al. States. Adv Radiat Oncol. 2022;7:100953.
Analysis of the electron-stream effect in patients
Index
A Biologically effective dose (BED), 226, 278

Abdominal organs, 264 Bladder rhabdomyosarcoma, 365
Adaptive radiotherapy (ART), 2, 5, 8, 21, 149, 170, 227, Bo field homogeneity, 160
398 Boolean operations, 357
Adapt-to-position (ATP) workflow, 279, 463 Borderline resectable pancreatic cancer, 245
Adapt-to-shape (ATS) workflow, 279, 463 Borderline resectable tumors, 244
Adjust radiation therapy, 470 Brain metastases, 376, 379–381
Agilent magnet, 428 Brain tumors
Alberta Linac-MR, 205–211 brain metastases, 379–381
Alkaline phosphatase (ALP), 258 contouring and planning, 382–385
Alpha-fetoprotein (AFP), 258 dose constraints, 385–388
American Association for the Study of Liver Diseases glioblastoma, 378–380
(AASLD), 258 on-table adaptation
American Brachytherapy Society, 324 Elekta UnityTM, 388, 390
American College of Radiology, 438 Viewray MRIdianTM, 388
American Joint Committee on Cancer (AJCC) cervical primary and metastatic brain tumors, 375
cancer staging system, 325 traditional approaches, 376, 377
Androgen deprivation therapy (ADT), 285, 286 in United States, 375
Apparent diffusion coefficient (ADC), 278, 399 whole brain radiation therapy (WBRT), 376
Atlas (multi-atlas) approach, 96, 97 Breast radiotherapy
Aurora-RT, historical milestone and development, contouring, 311–313
193–195 diagnosis and treatment, 308
Australian MRI-Linac (AMRL) dose prescription and DVH constraints, 313, 314
adjunct therapeutic equipment, 429 early diagnosis, 308
advantages and disadvantages, 426 image acquisition, 310, 311
beam and dosimetry, 429 incidence of, 307
beam-field interactions, 427 at MR-Linac, 314–316
final component of, 429 oncological outcomes, 308
gadolinium dose enhancement, 429 on-table adaptation, 316
manufactured components, 428 outcome, 317
patient rotation, 430 overcoming challenges, 308, 309
planning and treatment strategies, 429 patient positioning, 310
portable linear accelerator, 426 Breath-hold techniques, 72, 73, 78
protocol development and training, 426 Breath-hold treatments, 71
radiation induction, 427 Breathing fatigue, 30
surgical interventions, 426 Bulk density assignment (BDA), 92, 96
B C
Balanced steady-state free precession (bSSFP) sequence, Carcinoembryonic antigen (CEA), 258
201 Cardiac motion, 70
Beams geometries, 132 Central nervous system (CNS), 375
Biochemical recurrence (BCR), 286 Central pelvic recurrence, 339
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature 473
Switzerland AG 2024
I. J. Das et al. (eds.), A Practical Guide to MR-Linac, https://doi.org/10.1007/978-3-031-48165-9
474 Index
Child-Pugh (CP) scoring system, 259 beam characteristics, 176

Children’s Oncology Group trial ACNS0121, 359 transverse magnetic field, 176
Ciné imaging, 14, 15 Endometrial cancer, 323
Cine MRI with breath-hold technique and target tracking, inoperable endometrial cancer, 341–343
73, 74 European GEC ESTRO, 324
Claustrophobia, 27 External beam radiotherapy (EBRT), 258, 285, 291
Clinical target volume (CTV), 308
Compressed Sensing and Principal Component Analysis
(CS-PCA), 201, 202 F
Computed tomography (CT), 22, 41, 46, 376 Fast Field Echo (FFE) sequences, 280
in radiation therapy, 89 Fat reduction techniques, 55
Conditional GAN (cGAN), 100 Fiducials, 259
Cone beam computed tomography (CBCT), 21–23, 89, Fixed conformal, 153
324, 332 Forced expiratory volume (FEV), 447
Constraint violations, 275 Forced shallow breathing with abdominal compression,
Contrast enhancement, 56–58 71
Conventional fractionation, 286 Fourier transformation, 49
Conventional image-guided radiotherapy, 449 Full width at half maximum (FWHM), 160
Conventional radiomics, 459 Functional biomarkers, 139
Conventional treatment approaches, 361 Fuzzy c-means (FCM) algorithms, 98
Conventionally fractionated radiotherapy, 375
Craniospinal irradiation (CSI), 359
CyberknifeTM, 287, 288 G
Gadolinium contrast enhancement, 310
Gadoxetate, 260
D Gantry isocenter tolerance, 164
Daily dose accumulation, 279 Gaussian mixture models (GMM), 97
Deep inspiratory breath hold (DIBH) technique, 275 Genitourinary rhabdomyosarcoma, 365
Deep-learning methods, 434, 460 Geometric fidelity, 103
Delta-radiomic process, 460, 463, 466 Gibbs/truncation artifact, 63
De-noising techniques, 434 Glioblastoma, 378–380
Dice’s Similarity Index (DSI), 200 Gross tumor volume (GTV), 248, 309
Diffusion weighted imaging (DWI), 53, 54, 60, 278 Gynecological cancer
Diffusion-prepared turbo spin-echo readout (DP-TSE), MR Linacs, role of
455 adaptive MRgRT, 332–335
Diffusion-weighted imaging (DWI), 310, 361, 453, 455, endometrial cancer (see Endometrial cancer)
466 MRgRT, 335–338
Dixon method, 6 recurrent cervical cancer, 337, 339–341
Dose calculation, 14, 134, 135 MRI, 324
electron return effect, 135 cervical and vaginal cancer, 325–328
heterogeneity corrections, 136 endometrial cancer, 328
Dose constraints, 263 pelvic MRI, 329–331
Dose escalation, 262, 263 oligometastatic disease, 346–349
Dose-volume-histograms (DVHs), 224, 225, 261, 262 recurrent endometrial cancer (see Recurrent
Dynamic contrast-enhanced (DCE) MRI, 25, 278, 361 endometrial cancer)
E H
Early regression index (ERITCP), 465 Halbach magnets, 438
Elective para-aortic (PAO) node irradiation, 331 Head and neck neoplasms
Electron density (ED), 128 complex anatomy, 398
CT primary, 128 complex planning, 398
MR primary, 129, 130, 132 constraints, 403, 404
Electron density data, 275 contouring and PTV, 402
Electron return effect (ERE), 135, 150, 151, 177 MRI sequences, 400, 401
Electron streaming effect (ESE), 177 current treatment paradigms for, 396
Elekta Unity MR-Linac, 59 dose de-escalation, 408, 409
4D-MRI, 73 dose escalation
Elekta Unity system, 36, 79, 175, 252, 260, 378, 388, optimal dose, 407, 408
390 epidemiology and risk factors, 395, 396
Index 475
heterogeneity of, 397, 398 L

hypofractionation and SBRT in, 399 Larmor frequency, 104
MRL use in, 399, 400 Liver malignancies, 258
OAR contouring, 403–405 Locally advanced cervical cancer (LACC), 323, 465
on-table adaptation Locally advanced pancreatic cancer (LAPC), 246, 466
CT vs. MRI for reference plan, 411, 412 Locally advanced rectal cancer (LARC), 463
Elekta unity online workflow, 409, 410 Lorentz force, 134, 151, 195
evaluation, 413 Low-field MR imaging
intrafraction motion, 412 advantages of, 443–449
MRI coils, 412, 413 clinical applications of, 449
treatment plan adaptation strategies, 411 clinical purpose, 455
ViewRay MRIdian online workflow, 410, 411 designing coils for, 435–437
outcomes diffusion-weighted MRI (DWI), 453, 455
future perspectives on outcome measurement, 414 engineering challenges of, 435
toxicity reporting, 414 4D MRI, 452–454
tumor control outcome reporting, 414 high-income countries (HICs), 433
patient-specific challenges, 416, 417 high-speed MRI, 451, 453
patterns of failure of, 398, 399 linear SNR dependence, 434, 435
planning challenges, 404, 406 low-field magnet, 449–451
PTV margins, 403 magnetic fields, 434
staffing and time-related challenges, 415, 416 MR signal amplitude, 434
technological challenges, 415 permanent magnet system, 437, 438
treatment, changing landscape of, 397 physical properties of, 435
Hepatic iron deposition, 55 physiological and physical parameters, 433
Hepatocellular carcinoma (HCC), 257 point-of-care and portable MRI, 438, 439, 442
Heterogeneity corrections, 136
High-Z artifacts, 11
Homogeneous magnets, 437 M
Hyperfine, 438, 439 MAESTRO trial, 262
Magnetic resonance fingerprinting, 449, 451
Magnetic resonance for calculating attenuation
I (MRCAT) prostate, 109
Image acquisition, 325 Magnetic resonance imaging (MRI), 6, 24–26, 45, 46,
Image-guided adaptive BT (IGABT), 332 286, 460
Image-guided radiation therapy (IGRT), 5, 6, 21, 35, advantages, 6
149, 287 artifacts, 63, 64
Imaging artifacts, 65 autocontouring and segmentation, 13, 14
Immobilization, planning considerations for, 84, 85 characteristics of, 24
Immunotherapy, 349 ciné imaging, 14, 15
Implanted electromagnetic transponders, 287 contouring, 13
Innovative technology, 470 contrast enhancement, 56–58
Intensity modulated radiation therapy (IMRT), 133, 167, diffusion weighted imaging (DWI), 53, 54
323 dose calculation, 14
Inter- and intra-observer variability, 24 fat reduction techniques, 55
Internal target volumes (ITV), 274 high-Z artifacts, 11, 12
International Federation of Gynecology and Obstetrics imaging protocols, 59
(FIGO), 325 margin, 6–8
Interobserver contouring variation, 40 MR education and training, 14, 15
Intracranial ependymomas, 359 MR simulation, 12, 13
Intrafractional planar imaging, 290 multimodality imaging
Intrahepatic cholangiocarcinoma (ICC), 257 superiority, 9, 10
Intravoxel incoherent motion (IVIM), 455 x-ray imaging, 8, 9
Ion chambers, 195 physics, 47–49
Irradiation margin, 7 problem with, 6
Isocenter placement, 132 proton density (PD), 53
safety, 26, 27
safety considerations, 65
K spin echo and gradient echo, variation in, 52, 53
Kinetic energy released in matter (KERMA), 176 T1-weighted imaging, T2-weighted imaging, and
pulse sequences, 49–52
476 Index
Magnetic resonance imaging with linear accelerators pediatric and adolescent patient
(MR-Linac), 2, 21, 35, 39, 106, 149, 471 abdominal tumors, 363, 364
acceptance testing for, 163, 164 head and neck cancers, 360, 361
architectural and structural planning, 178, 179 modality approach, 355–357
breath-hold techniques, 72, 73 MR-guided linear accelerator treatment workflow,
breathing fatigue, 30 357, 358
cine MRI with breath-hold technique and target neuroblastoma, 364, 365
tracking, 73, 74 patient selection and technical challenges, 367
claustrophobia, 27 pediatric leukemias, 361, 362
commissioning, and routine QA, 184–190 pediatric lymphoma, 362, 363
construction and installation timeline, 179, 180 pelvic tumors in, 365–367
economic analysis and cost, 177, 178 posterior fossa, 359, 360
4D computed tomography, 23 spinal tumors, 359, 360
4D-MRI, 72 supratentorial brain tumors, 358
historical perspective, 1, 2 potential limitations of, 299–301
limitation, 73 preparation, 36, 37
logistics and economics, 157 for prostate cancer (see Prostate cancer)
magnetic resonance imaging (MRI), 24–26 PTV, and OAR, 276–278
safety, 26, 27 radiomic workflow, 460–462
magnetic susceptibility, 30 sample patient plans, 267, 268
metallic implants, 27–29 on table adaptation, 278, 279
in oligometastatic disease (see Oligometastatic on-table adaptive radiotherapy (On-ART), 263, 264
disease (OMD)) training and education during the delivery phase, 39
on-board imaging systems, 21–23 training and education procedures, 40, 41
patient comfort, 29 traditional approaches, 274–276
patient selection, 26 Magnetic resonance relaxometry, 361
planning and dosimetry challenges for, 152 Magnetic susceptibility, 30
reference dosimetry, 165 Magnetic susceptibility artefacts, 106
sequence in, 59 Maximum tolerated dose (MTD), 273
Elekta Unity MR-Linac, 59 Mayneord F factor, 165
Navigator, 60–63 Mean absolute error (MAE), 102
ViewRay MRIdian MR-Linac, 59 Mean centroid displacement (MCD), 200
volumetric TRUFI sequence, 60 Medical linear accelerator, 2
staffing and training, 181 Medulloblastoma, 359
team member roles and responsibilities, 180, 181 Metachronous, 218
treatment planning systems (TPS) (see Treatment Metal implants, 443
planning systems (TPS)) Metal prostheses, 300
thermal effects, 30 Metallic implants, 27–29
wishlist and future of, 470 Metastasis-free interval (MFI), 222
workflow setup and considerations MIRAGE trial, 298
fall back scenarios, 184 Monte Carlo codes, 134–135
implementation of new workflows, 184 Monte Carlo dose calculation algorithms, 135, 357
pre-treatment preparations and online workflow, Monte Carlo (MC) methods, 152
182, 183 Motion
Magnetic resonance-guided radiation therapy (MRgRT), cardiac, 70
6, 35, 36, 123, 178, 449 encompassing methods, 71
advantages, 259, 260 management, 69
for brain tumors (see Brain tumors) physiological, 70
challenges, 280, 281 respiratory, 69
contouring, 260, 276–278 MRIdian machine
dose escalation, 262, 263, 278 acceptance testing for MRI, 160
DVH constraints, 261, 262 commissioning of, 169–172
DVH metrics/evaluation, 279 couch transmission, 165
evaluation, 264, 265 dosimetry consistency with gating, 168
imaging biomarkers in IMRT functionality, 167
Elekta’s MR-Linac system (Unity), 462–464 in-house courses and hands-on training
future developments and perspectives, 466, 467 sessions, 158
ViewRay's MR-Linac system (MRIdian), 463, MLC linearity and positional accuracy, 166
465, 466 MLC transmission and leakage, 166
OARs, 260 MR safety for, 155, 157
outcomes, 265–267, 279, 280 MR tracking latency, 168
MRI-specific training and equipment, 37, 38, 40 practicing contouring and workflow, 158, 159
Index 477
radiation field size, shape, symmetry, and penumbra P

accuracy, 167 Pancreas, 244
receiver coils, SNR and uniformity, 161–163 Pancreatic cancer
system design of, 152–155 Elekta Unity
treatment planning system (TPS), 382 contouring, PTV, and OARs for, 249, 250
ViewRay training, 157 on-table adaptation for, 252
Multichannel endocavitary vaginal applicators, 343 overcoming challenges of radiotherapy in, 246, 247
Multileaf collimator (MLC), 297 plan evaluation, 254
Multimodality approach, 5 symptoms and signs of, 243
Multiparametric (mp) MRI, 139 treatment, 244–246
Multiple image acquisitions with x-rays, 21 ViewRay MRIdian
Multispectral imaging (MSI), 443 contouring, PTV, and OARs for, 248
on-table adaptation for, 252
Pancreatic ductal adenocarcinoma (PDAC), 243
N Pancreatic neoplasms, 466
Nabothian cysts, 45 Partial breast irradiation, 308
Neoadjuvant partial breast irradiation, 307 Pathological complete response (pCR), 463–465
Neoadjuvant partial breast radiotherapy, 308 Patient coaching
Neoadjuvant therapy, 245 feedback mechanism, 85
Net magnetization vector, 48 patient education, 85
Neuroblastoma, 364, 365 Pediatric Hodgkin’s lymphoma, 362
Non-alcoholic steatohepatitis (NASH), 257 Percent depth dose (PDD), 151, 165, 166
Non-Cartesian sampling, 451 Permanent magnet arrays (PMAs), 437
Non-Hodgkin’s lymphoma (NHL), 362 Permanent magnet system, 437, 438
Non-Invasive Intra-Fractional Tumor-Tracked Physiological motion, 70
Radiotherapy (NifteRT), 205–210 Planning target volume (PTV), 308, 333, 356, 462
Nonlinearity corrections, 439 Planning treatment volume (PTV), 273, 274
Non-small cell lung cancer (NSCLC), 220, 273 Portable scanners, 438
Positron emission tomography (PET), 8
Post-operative radiotherapy, 359
O Post-prostatectomy, 299
Offline (inter-fraction) adaptive planning, 137 Post-prostatectomy RT, 289
Oligometastatic disease (OMD), 346–349 Preamp decoupling, 436
classification, 219 Primary liver cancer, 257
contouring, PTV, and OAR, 223, 224 Primary sclerosing cholangitis (PSC), 257
definition of, 217 Prior Data Assisted Compressed Sensing (PDACS), 199, 200
dose escalation, 226, 227 Process failure mode and effect analysis (P-FMEA), 37
DVH constraints, 224, 225 Prospective acquisition correction (PACE) technique, 60
evaluation, 233 Prostate cancer
on table adaptation, 227 dose accumulation, 288
outcome, 233, 234 dose escalation, 295, 296
overcoming challenges with traditional approaches, DVH constraints, 292, 295
222, 223 evaluation of daily setup, 298
patient selection characteristics, 221, 222 exposure to ionizing radiation, 289
SBRT Treatment for, 218, 220, 221 functional response, 289
subgroups, 218 motion management, 287, 288
treatment of, 234 natural history of, 285
Oligopersistence, 218 OAR, 291, 292
Oligorecurrence, 218 ongoing trials in MRIgRT, 301, 303
On-board imaging systems, 21–23 outcome, 298, 299
Online adaptive approach, 138, 278, 279, 333 plan evaluation, 297, 298
Online-adaptive MRgSBRT, 226 PTV, 291
Online MRgRT, 332 target volumes, 290, 291
On-table adaptive radiotherapy (On-ART), 263, 264 ultrahypofractionation in, 289, 290
Optimization process, 134 Unity
Optimized conformal, 153 pre-treatment phase, 296
Organs at risk (OARs), 5, 69, 259, 273, 324 treatment phase, 297
Osteosarcoma, 54 ViewRay
Out-of-field progression, 348 pretreatment phase, 297
Oxygen-enhanced MRI, 448 treatment phase, 297
478 Index
Prostate specific membrane antigen (PSMA)-PET, 218 breath-hold treatments, 71

Prosthetic devices, 11 forced shallow breathing with abdominal
Proton density (PD), 53 compression, 71
Proton therapy, 356 motion encompassing methods, 71
Pseudo CT, 91 respiratory-gated radiation therapy with X-ray
Pulmonary function tests (PETs), 444 approaches, 71
Pulse-coupled neural networks (PCNN), 208 tracking, 72
Pulse sequences, 49, 50, 325 Rotating bi-planar Linac-MR, RE coils for, 197, 198
Q S
Quality assurance, 141 Secondary malignancies, 258
Quantitative Analysis of Normal Tissue Effects in the Seminal trials, 285
Clinic (QUANTEC), 403 Shimming, 439
Simulation
patient positioning and immobilization, 75, 76
R respiratory motion, 76–78
Radiation-induced liver disease (RILD), 258 Single-shot fast spin echo (SSFSE), 52
Radiation oncology, 5 Single-sided magnets, 437
Radiation therapy, 1, 324 Soft tissue differentiation, 290
Radiation Therapy Oncology Group (RTOG) planning Soft tissue visualization, 278
approach, 382 Solid state detectors, 196
Radical prostatectomy (RP), 285 Sørensen-Dice index, 102
Radiomics, 361, 459 Spatial integrity, 162
Radiotherapy (RT), 28, 431, 459 Spiculae, 312
Real-time (dynamic) adaptive planning, 138 Spinal tumors, 359
Real time motion management, 199–203 Standard fractionated adjuvant breast irradiation, 309
accelerated MR acquisition and compressed sensing Stereotactic ablative body radiation (SABR), 219–220
MRI, 199 Stereotactic body radiotherapy (SBRT), 222, 259, 273,
CNN performance, 204 286, 332, 356, 455
coherent versus incoherent undersampling for CNN for oligometastatic disease, 218, 220, 221
reconstruction, 204 Structural similarity metric (SSIM), 102
CS-PCA, 201, 202 Superconducting magnet, 437
PDACS, 199, 200 Superiority of MRI, 9, 10
real-time CNN implementation, 203 Supratentorial brain tumors, 358
time-domain PCA-based real-time reconstruction for Susceptibility artefacts, 105
dynamic undersampled MRI, 202, 203 Sydney project MR-Linac, 3
Real-time tumor tracking, 278 Synthetic CT scans (sCT)
Recurrent endometrial cancer, vaginal vault recurrences, applications of, 91
343, 344, 346 generation from MRI for radiotherapy dose
Region of interest (ROI), 161, 459 calculation, 92
Residual gradient non-linearity-induced distortions, 104, atlas (multi-atlas) approach, 96, 97
105 bulk density assignment (BDA), 92, 96
Resistive electromagnet, 437 voxel-based methods, 97–101
Respiratory-gated radiation therapy with X-ray implementation in MR radiotherapy workflows,
approaches, 71 108–114
Respiratory gating methods, 23 metrics for evaluation, 95, 101
Respiratory motion, 69, 76–78, 275 accuracy of patient repositioning, 103
Respiratory motion management dosimetric impact, 102, 103
MRI-guided linear accelerator techniques for geometric fidelity, 102
breath-hold techniques, 72, 73 intensity-based metrics, 102
cine MRI with breath-hold technique and target MRI-only workflow, 103
tracking, 73, 74 MRI-QA for distortion/artefact-free synthetic CT,
4D-MRI, 72 103–106
limitation, 73
tracking and gating, 79
considerations for, 81, 82 T
multi-target tracking, 82, 83 Tesla MRI scanner, 260
single plane vs. multiplane tracking, 79, 81 Thermal effects, 30
X-ray-and CT-based techniques for, 71 3D conformal radiation therapy (3DCRT), 153
Index 479
Three-dimensional conformal treatment planning, 132 T1-weighted imaging, 49–51

3D gradient-echo (GRE) sequence, 60 T2-weighted imaging, 49–53
3D half-Fourier acquisition single-shot turbo spin-echo 2D cine, 60
(3D HASTE), 455 2D U-Net architecture, 99
Tracheal implant, 28
Traditional radiomics, 460
Transient charged particle equilibrium (TCPE), 151 U
Treatment delivery system (TDS), 172 UCLA SMART constraints, 262
Treatment planning systems (TPS) Ultra-hypofractionation, 228
adaptive planning considerations, 136 Ultrasound (US) surveillance, 258
anatomical changes, 138 University of California, Los Angeles (UCLA), 267
functional changes, 139 Urethral avoidance, 291
reference plans, 136
timescales, 137
beams geometries, 132 V
commissioning and QA, 124 VacQfix Vacuum Cushions (Qfix), 75
contouring, 125 Vaginal vault recurrences, 343, 344, 346
imaging datasets, 125, 126 Vascular stents, 300
margins, 126 Vascular structures, 291
motion management with cine imaging, 127 View-angle tilting (VAT), 443
online adaptive considerations, 127, 128 ViewRay MR 60Co system, 311
dose calculation, 134, 135 ViewRaw MRIdian system, 260
electron return effect, 135 ViewRay MRIdian MR-Linac, 59
heterogeneity corrections, 136 ViewRay MRIdian system, 149, 156, 248, 252, 378, 388
electron density (ED), 128 ViewRay training, 157
CT primary, 128 Volume adaptation, 398
MR primary, 129, 130 Volumetric imaging, 288
hardware and software requirements, 123 Volumetric modulated arc therapy (VMAT) capabilities,
IMRT, 133 280, 335, 470
integration with treatment delivery software, 124 Volumetric TRUFI sequence, 60
isocenter placement, 132 Voxel-based methods, 97–101
optimization process, 134
patient database, 124
plan evaluation W
dose calculation accuracy, 140 Wavelet delta-radiomic (WDRFs) textures, 463
metrics, 140 Whole brain radiation therapy (WBRT), 376
quality assurance, 141 Wilms tumor, 363, 364
summation, 140
three-dimensional conformal treatment planning, 132
True fast imaging with steady state free precession X
(TRUFI) sequence, 248, 310, 357 X-ray-based respiratory motion management, limitation
Tumor microenvironment (TME), 397 of, 72
Tumor-treating fields (TTFields), 375 X-ray imaging, 8, 9
Turbo Field Echo (TFE), 280 X-ray tube technology, 1

2024 A Practical Guide To MR Linac

Uploaded by

Copyright:

Available Formats

2024 A Practical Guide To MR Linac

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2024 A Practical Guide To MR Linac

Uploaded by

Copyright:

Available Formats

A Practical

ISBN 978-3-031-48164-2 ISBN 978-3-031-48165-9 (eBook)

Paper in this product is recyclable

ultrasound and X-ray imaging, electromagnetic localization, or optical

Chicago, IL, USA Indra J. Das

A book like this A Practical Guide to MR-Linac: Technical Innovation and

1 Introduction to MR-Linac�������������������������������������������������������������� 1

About the Editors

Filippo Alongi, MD is Head of Advanced Radiation Oncology Department at IRCCS

Bharat B. Mittal, MD is Willian N. Brand Professor and Chair of Radiation Oncology at

Mohamed Abazeed Department of Radiation Oncology, Northwestern

Filippo Alongi University of Brescia and Advanced Radiation Oncology

Rosa Patricia Castillo Radiology Department, Miller School of Medicine,

John Hayes Department of Radiation Oncology, Northwestern Memorial

Katie N. Lee Harvard Radiation Oncology Program, Brigham and Women’s

Kyle R. Padgett Radiation Oncology Department, Sylvester Comprehensive

Fernanda Villegas Radiotherapy Physics and Engineering, Medical Radiation

Historical Perspective Becquerel in 1896 and Marie and Pierre Currie

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 1

lation, and improved tumor control, represent- References

Introduction location of the tumor is a critical determinant in

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 5

MR-Only Approach radiation therapy to supplement CT data, mainly

Role of Ciné Imaging Dose Calculation

Summary 10. Gayou O, Miften M. Commissioning and clini-

Introduction imaging technology to enhance tumor definition,

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 21

shows an example of motion-related challenges Many radiotherapy procedures are nowadays

as MR safe. If an item has no known hazards in a artifacts on cardiac implantable electronic

Breathing Fatigue tions, ensuring multiple means of communicat-

Introduction MRIdian by Viewray (Fig. 4.1) consist of a

L. Nicosia (*) · R. F. Borgese

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 35

procedures requiring specific training consist of

• MR safety training for all personnel involved

Hersh Sagreiya, Karthik Sundaram,

Introduction sequences can highlight areas of fluid or edema,

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 45

In this equation, ω is the angular momentum In order to localize a particular point in 3D

T2-Weighted Sequences magnetic fields of spinning nuclei—and T2*

dinal magnetization and ultimately T1-weighting. Diffusion-Weighted Imaging

 at Reduction Techniques (Chemical

Fig. 5.9 (continued)

Imaging Protocols Sequences in MR-Linac

sequence, which is a type of balanced steady-­ Navigator

MRI Artifacts strates a decreased MR signal in the center of the

Robust screening precautions are essential in the

Introduction linear accelerators (MR-Linacs) has provided cli-

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 69

and OARs (e.g., pancreatic cancer adjacent to Respiratory Gating

approach include increased inter-fraction vari- 4D-MRI

 imitations of MRI-Guided Motion

Simulation during treatment. Furthermore, patient set-up

Planning Considerations Given the features available with MR-guided

Introduction photon beam RT since it conditions cross sec-

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 89

addition to difficulties/errors in volumes regions [14]. Then, atlas-based and voxel-based

[30] collected ten patients from the ABLATIVE Atlas-Based Approaches

Intensity-Based Metrics Wang 2014 et al. [102], these terms respectively

Chicago, IL, USA Indra J. Das

1 Introduction to MR-Linac�� 1

Historical Perspective Becquerel in 1896 and Marie and Pierre Currie

Introduction location of the tumor is a critical determinant in

MR-Only Approach radiation therapy to supplement CT data, mainly

Role of Ciné Imaging Dose Calculation

Summary 10. Gayou O, Miften M. Commissioning and clini-

Introduction imaging technology to enhance tumor definition,

Breathing Fatigue tions, ensuring multiple means of communicat-

Introduction MRIdian by Viewray (Fig. 4.1) consist of a

Introduction sequences can highlight areas of fluid or edema,

T2-Weighted Sequences magnetic fields of spinning nuclei—and T2*

dinal magnetization and ultimately T1-weighting. Diffusion-Weighted Imaging

at Reduction Techniques (Chemical

Imaging Protocols Sequences in MR-Linac

sequence, which is a type of balanced steady- Navigator

MRI Artifacts strates a decreased MR signal in the center of the

Introduction linear accelerators (MR-Linacs) has provided cli-

and OARs (e.g., pancreatic cancer adjacent to Respiratory Gating

approach include increased inter-fraction vari- 4D-MRI

imitations of MRI-Guided Motion

Simulation during treatment. Furthermore, patient set-up

Planning Considerations Given the features available with MR-guided

Introduction photon beam RT since it conditions cross sec-

[30] collected ten patients from the ABLATIVE Atlas-Based Approaches

Intensity-Based Metrics Wang 2014 et al. [102], these terms respectively

can be positive if the magnetisation is aligned ecent Guidelines from Scientific

Treatment Planning System algorithms that result in intensity modulated radi-

Contouring reatment Planning Imaging

rtifacts and Geometric Distortions

MR Primary sCT. Multi-atlas methods can take into consider-

Planning Geometry MR-Linac treatment platforms. As mentioned

Adaptive planning strategies have traditionally ffline Adaptive Planning

Plan Evaluation organs in which max dose is a clinically impor-

Quality Assurance algorithm requirements in MR-Linac treatment

Introduction What Is an MR-Linac?

lanning and Dosimetry Challenges

MRIdian Acceptance, tems remains intact. Shielding calculations are

Laser Alignment is aligned using the lasers. The vendor recom-

Reference Dosimetry According to the vendor specifications, the PDD

The dosimetric metrics related to lateral beam IMRT Functionality

Introduction 2017 [4], and it proved the geometrical and

Workflow Setup and Considerations After structure delineation (e.g., GTV/CTV,

Implementation of New Sites workflow should be trained with the personnel

Introduction lel B0-field to radiation beam configuration. The

ccelerated MR Imaging for Real-

erformance of Time-Domain PCA

TX). TPR20,10 is converted to equivalent 6. 3D MR imaging is about 3 min, while real-

isease Overview: Oligometastatic

K. N. Lee · M. A. Huynh (*) Defining Oligometastatic Disease

a 0.35 T MR-linac to reduce respiratory motion DVH Constraints

Dose Escalation tion. In a proof of concept, Henke and colleagues

Evaluation expansion into investigations of oligometastatic

cers originating from prostate (60%) or colorec- Future Directions

Contouring, PTV, and OARs and restaging during initial chemotherapy.

n-Table Adaptation for ViewRay

Plan Evaluation mary disease. Incorporating molecular markers

Clinical Evaluaon and Eligibility

Evaluation ticularly in the liver. The biggest one is that it

Summary 2021;7(1):1–28. https://doi.org/10.1038/