Accepted Manuscript

Title: Dual-layer detector CT of the head: initial experience in

visualization of intracranial hemorrhage and hypodense brain
lesions using virtual monoenergetic images

Authors: Simon Lennartz, Kai Roman Laukamp, Victor

Neuhaus, Nils Große Hokamp, Markus Le Blanc, Volker
Maus, Christoph Kabbasch, Anastasios Mpotsaris, David
Maintz, Jan Borggrefe

PII: S0720-048X(18)30316-4
DOI: https://doi.org/10.1016/j.ejrad.2018.09.010
Reference: EURR 8303

To appear in: European Journal of Radiology

Received date: 15-2-2018

Revised date: 7-9-2018
Accepted date: 10-9-2018

Please cite this article as: Lennartz S, Laukamp KR, Neuhaus V, Große Hokamp N,
Le Blanc M, Maus V, Kabbasch C, Mpotsaris A, Maintz D, Borggrefe J, Dual-layer
detector CT of the head: initial experience in visualization of intracranial hemorrhage
and hypodense brain lesions using virtual monoenergetic images, European Journal of
Radiology (2018), https://doi.org/10.1016/j.ejrad.2018.09.010

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
Dual-layer detector CT of the head: initial experience in visualization of

intracranial hemorrhage and hypodense brain lesions using virtual

monoenergetic images

Simon Lennartz, MD1,3,4, Kai Roman Laukamp, MD1,4, Victor Neuhaus, MD1, Nils Große
Hokamp, MD1, Markus Le Blanc, MD1, Volker Maus, MD1, Christoph Kabbasch, MD1,
Anastasios Mpotsaris, MD 2, David Maintz, MD1, Jan Borggrefe, MD1

T
1
University Hospital of Cologne, Institute for Diagnostic and Interventional Radiology,
Kerpener Straße 62, 50937, Cologne, Germany

IP
2University
Hospital of Aachen, Clinic for Diagnostic and Interventional Radiology,
Pauwelsstraße 30, 52074 Aachen

R
3
Telephone number: +4915773520225, Fax number: +4922147882384, Email address:

SC
[email protected]

4 contributed equally

U
N
Corresponding author:
Simon Lennartz, M.D.
A
Institut für Diagnostische und Interventionelle Radiologie, Uniklinik Köln
Kerpener Str. 62, 50937 Köln, Germany
M

Fax number: +49-221-478-82384, telephone number: +49157-73520225

E-mail: [email protected]
ED
PT

Abstract
E
CC

Purpose
A

Retrospective comparison of diagnostic quality of virtual monoenergetic images (VMI)

and conventional images (CI) reconstructed from dual-layer detector CT (DLCT)

regarding intraparenchymal hemorrhage (IPH) and hypodense parenchymal lesions

(HPL) of the brain.

1
Methods

58 patients underwent unenhanced DLCT of the head. CI and VMI ranging from 40 to

120 keV were reconstructed. Objective image quality was assessed using ROI-based

measurements within IPH, HPL, grey matter, white matter and cerebrospinal fluid, from

which contrast to noise ratio (CNR) was calculated. Two radiologists assessed IPH,

T
HPL, artifacts and image noise on a 5-point Likert-scale. Statistical significance was

IP
determined using Wilcoxon rank sum test.

R
Results

SC
In comparison to conventional images, CNR of HPL to white matter was significantly

increased in VMI at 120 keV (p≤0.01), whereas at 40 keV, CNR to grey matter was
U
enhanced (p≤0.0001). Contrary, CNR of IPH to white matter was increased at 40 keV
N
(p≤0.01), while CNR to grey matter was improved at 120 keV (p≤0.01). Subjective
A
M

readings confirmed best delineation of IPH within grey matter at 120 keV. Both readers

detected four additional hyperdense lesions within white and one within grey matter at
ED

40 keV.
PT

Conclusions

VMI obtained with DLCT can improve depiction of hypodense parenchymal lesions and
E

intraparenchymal hemorrhage. The initial data show a great dependency on the type
CC

of pathology and on its location: hypodense lesions in white matter and hyperdense
A

lesions in grey matter are better visualized in higher keV reconstructions, while

hyperdense lesion in white matter and hypodense lesions in grey matter are better

visualized at low keV values.

Keywords

2
Intracranial Hemorrhages, Dual Energy Computed Tomography, Cerebral Infarction,

Neuroimaging, Brain Edema

Introduction

For diagnosis of intracranial hemorrhage and ischemia, unenhanced CT of the head is

the method of choice, combining fast image acquisition and a widespread availability

T
[1, 2]. Dual energy CT (DECT) has been shown to improve image quality and lesion

IP
characterization in neuroradiological imaging using virtual monoenergetic

R
reconstructions as well as material decomposition [3–8], whereas cerebral lesion

SC
assessment with the novel dual-layer CT (DLCT) has not been thoroughly examined

U
yet. DECT is based on the acquisition of a high- and a low-energy dataset of the same
N
anatomical region using a polyenergetic x-ray spectrum. To obtain virtual
A
monoenergetic images (VMI), balanced combinations of these two acquired datasets
M

can be calculated, allowing to reconstruct images that simulate what an acquisition

using a monoenergetic beam at a given energy could look like [9]. As x-rays with lower
ED

photon energies are attenuated to a high extent, VMI at low keV levels provide an

increased soft tissue contrast. In contrast, higher-energetic x-rays are attenuated less,
PT

so that beam hardening artifacts close to dense structures (e.g. bone, metal implants)

can be reduced. Established DECT systems operate with different tube- based
E
CC

techniques.[10]. In contrast, the recently introduced detector-based approach consists

of a single-source system working with a dual-layer-detector, in which the upper,

A

yttrium-based layer detects lower-energetic photons while the subjacent layer which is

made of gadolinium-oxysulphide detects higher-energetic photons [11]. This setup

allows for a simultaneous measurement of both datasets in consistent spatial and

angular location [12–15]. Therefore DLCT allows post processing within the projection

3
domain, whereas in tube-based DECT, the post processing either takes place within

the image domain or requires previous angular and temporal interpolation [5]. Another

benefit of the detector-based acquisition method is that dual-energy datasets are

obtained with every clinical scan.

This study aims to retrospectively evaluate subjective and quantitative image quality

T
parameters of intraparenchymal hemorrhage (IPH) and hypodense parenchymal

IP
lesions (HPL) in unenhanced dual layer detector CT of the head comparing VMI

ranging from 40 to 120 keV and CI reconstructed from the same scans. For this

R
SC
purpose, quantitative contrast-to-noise ratio (CNR) between the different VMI levels

and CI was compared and delineation of IPH and HPL to the surrounding tissue,

U
image noise, artifacts and the detection of additional lesions were evaluated
N
subjectively.
A
M

Materials and Methods

ED

This retrospective study was conducted with institutional review board approval while
PT

informed consent was waived. Patients (n=58) were consecutively selected and

included in two subsets of 29 each: inclusion criteria were the presence of IPH (group
E
CC

1) or HPL including both perifocal edema and postischemic areas (group 2). Inclusion

of patients was based on the diagnoses mentioned in the radiological report which
A

were confirmed by a senior neuroradiologist. In all patients of group 2, HPL were

located within white matter (WM). A total of 8 patients with severe artifacts due to

movement or implants were excluded. Of the 58 patients included in the study, 31 were

men and 27 were women. The average age was 56.8±16.9 years.

4
Acquisition parameters

The CT scans were performed for clinical indications between June and September

2016 using the dual-layer detector CT (IQon® Spectral CT, Philips Healthcare,

Amsterdam, The Netherlands). CI at 120 kV as well as VMI were retrospectively

reconstructed from the raw data of the same CT scan. A consistent clinical scan

T
protocol was applied: unenhanced image acquisition, slice thickness 1 mm

IP
(reconstructed 5 mm), tube voltage 120 kV, tube current 320 mAs, CT dose indexvol of

55 mGy, pitch 0.36, rotation 0.33s, collimation 64 x 0.625, kernel: brain standard. VMI

R
SC
in the range of 40 keV and 120 keV with 10 keV increment as well as CI were

reconstructed using the same, dedicated spectral image reconstruction algorithm

U
comparable to hybrid iterative reconstruction methods as well as a consistent kernel
N
(Spectral B, denoising level 3, Philips Healthcare, Amsterdam, The Netherlands).
A
Quantitative analysis
M

For objective image analysis, VMI and CI were reviewed by two independent
ED

radiologists with 2 years of experience in head CT assessment using a proprietary

vendor console (Spectral Diagnostics Suite (SpDS), Philips, Amsterdam, NL).

PT

Beforehand, both reviewers were trained in ROI placement and data acquisition on
E

independent test data by a senior neuroradiologist. Axial images were reviewed in the
CC

brain window (width/center: 70/35 HU). Objective image quality assessment was

performed using ROI-based measurements of attenuation and standard deviation (SD)

A

in IPH and HPL, grey matter (GM), white matter (WM) and cerebrospinal fluid (CSF).

For this purpose, the respective ROIs were placed within these 5 target areas using

the CI and remained equal when switching to VMI at different keV levels. The size of

the ROIs placed in the grey and the white matter as well as the thalamus and the

5
internal capsule were kept constant at 25 mm 2 and only adjusted to prevent inclusion

of tissues of different attenuations Image noise was determined as the standard

deviation of each ROI in Hounsfield units (HU). Contrast to noise ratio was calculated

in accordance with prior studies on DECT of the head [8, 3]. Before statistical analysis,

quantitative measurements were reviewed by a senior neuroradiologist.

T
Subjective analysis

IP
Subjective analysis of VMI and CI was performed independently by two unblinded

R
radiologists with 3 and 5 years of experience in head CT interpretation. To compare CI

SC
and VMI, the following keV levels were set within the reviewing process: 40, 55, 70,

85, 100, 120 keV. The following 5-point Likert scales were used [8]: differentiation

U
between IPH/HPL and the surrounding tissue (1 = difficult, uncertain diagnosis; 2 =
N
poor, limited diagnosis; 3 = decent, diagnostic; 4 = good, diagnostic, 5=excellent,
A
certain diagnosis), artifacts within the subcalvarial space and image noise
M

(1=excessive, 2=severe, 3=noticeable, 4=barely noticeable, 5=none). Moreover,

ED

additional lesions detected in VMI at 40 keV compared to CI were reported and

reviewed by a senior neuroradiologist.

PT

Statistical analysis
E

For statistical evaluation, the acquired datasets were analyzed using JMP Software
CC

(V12, SAS Institute, Cary, USA). Wilcoxon rank sum test was performed to determine

significance of differences between results of both quantitative and subjective analysis.

A

Statistical significance was defined as p≤0.05. Further results are summarized as

mean ± standard deviation. Inter-rater reliability was determined using the intraclass

correlation (ICC) with the following scale: >0.8: excellent, >0.6: good, >0.4: moderate,

<0.4: poor [16]. ICC was differentiated according to lesion type (HPL/IPH) and

6
summarized with regards to assessed reconstructions (CI, VMI from 40 to 120 keV)

and lesion localizations (GM/WM).

Results

T
Intraparenchymal hemorrhage

IP
Objective analysis

CNR of IPH to GM increased successively with higher energy levels leading to an

R
optimal CNR of 4.3 ± 2.8 at 120 keV (CI: 2.5 ± 1.4, p ≤ 0.01; fig. 1 A).

SC
In contrast, VMI at low keV levels showed a significantly enhanced CNR of IPH to WM

U
compared to 120 keV and CI (CNR40keV: 5.4 ± 3.1 vs. 3.7 ± 1.6 in CI, p ≤ 0.01), marking
N
an increase of 45.9% (fig. 1 C).
A

Subjective analysis
M

Delineation of IPH adjacent to GM was gradually improved with rising keV levels. VMI
ED

from 100-120 keV received significantly increased ratings with a median of 4 compared

to CI (p ≤ 0.0001) (fig. 1 B). For the assessment of IPH within WM, VMI in general
PT

received higher Likert scores than CI, yet without statistical significance (fig. 1 D). In
E

five different patients with IPH, both observers detected the same additional
CC

hyperdense lesions in VMI at 40 keV which could not be detected in CI. Of those five

lesions, four were located within WM, whereas one was found within GM (fig. 2). ICC
A

for the assessment of IPH was good (0.69).

Hypodense parenchymal lesions

Objective analysis

7
CNR of HPL to GM was significantly increased with gradually decreasing keV resulting

in a maximum CNR of 4.4 ± 1.5 at 40 keV compared to 2.5 ± 1.0 in CI (p ≤ 0.0001; fig.

3 A). Opposingly, VMI at high keV showed significantly improved CNR of HPL to WM

in comparison to both VMI at low keV levels and CI, gradually increasing from 40 keV

to 120 keV. VMI at 120 keV yielded a CNR of 2.6 ± 1.4 compared to 1.4 ± 0.8 in CI,

which marks an increase of 85.7% (p ≤ 0.01) (fig. 3 B).

T
IP
Subjective analysis

Delineation of HPL to surrounding WM was rated best with a median score of 4 in VMI

R
at 100 keV compared to a median score of 3 in CI (p ≤ 0.0001, fig. 3 C). VMI at 40-65

SC
keV received lower Likert scores than CI (VMI 40 keV: 2, p ≤ 0.0001). For the

assessment of HPL, ICC was moderate (0.45).

U
N
General image parameters
A
M

CNR of grey and white matter showed an incremental rise with decreasing keV levels

(fig. 4). In VMI at 40-60 keV, CNR was significantly higher than in CI (2,8 ± 1,2; p ≤
ED

0.0001) while VMI at 40 keV yielded an optimal CNR of 11.1 ± 4.1. In VMI from 50 keV

to 120 keV, quantitative image noise of white matter was significantly lower than in CI
PT

(p ≤ 0.0001), gradually decreasing with rising VMI levels and reaching a minimum in

VMI at 120 keV (3.5 ± 0.8 compared to 5.2 ± 1.0 in CI). Pertaining to grey matter, image
E
CC

noise decreased from 5.1 ± 1.0 in CI to 3.7 ± 1.3 in VMI at 120 keV. Accordingly, VMI

in the range of 70-120 keV received superior Likert scores for subjective image noise
A

compared to CI (p ≤ 0.0001), while VMI at 120 keV were rated best (figure 5 A). A

similar trend was observed in subjective assessment of subcalvarial beam hardening

artifacts: VMI ranging from 85 to 120 keV were rated significantly higher than CI (p ≤

0.0001) regarding artifact reduction, VMI at 100 and 120 keV received median scores

8
of 5 compared to 3 in CI (fig. 5 B). ICC regarding image noise and subcalvarial artifacts

was good (0.64).

Discussion

In this study, we compared the image quality of virtual monoenergetic images to

conventional images reconstructed from unenhanced spectral detector computed

T
tomography scans of the head and assessed optimal energy levels of VMI for the

IP
delineation of both intraparenchymal hemorrhage as well as hypodense brain lesions.

R
SC
Previous studies showed that VMI improve image quality of unenhanced head CT due

to increased grey-white matter differentiation and reduced beam hardening artifacts of

U
the skull [8, 3, 17–19]. For instance, in a study in which the same scanner and image
N
parameters were used, Neuhaus et al. stated that image noise as well as subcalvarial
A
artifacts significantly decreased in VMI at 120 keV [8]. These findings could be
M

confirmed by our study, in which VMI at 120 keV yielded lowest artifacts and image

noise as well: White matter image noise decreased from to 5.2 ± 1.0 (CI) to 3.5 ± 0.8
ED

(VMI120keV) while grey matter image noise decreased from 5.1 ± 1.0 to 3.7 ± 1.2.
PT

Regarding grey and white matter differentiation, Neuhaus et al. observed highest CNR

values in VMI at 40 keV (7.1 ± 1.1) which could also be verified in this study, while
E

CNR values at this keV level were even higher (11.1 ± 4.1). In another study by Cho
CC

et al. which investigated the DLCT-angiography derived spectral reconstructions for

A

patients with acute intracerebral hemorrhage, low keV were found to improve

visualization of these hemorrhagic lesions [20]. Following these findings, it could be

expected that CNR of IPH adjacent to both GM and WM would increase in VMI at low

keV as at lower photon energies, the absorption coefficient increases due to higher

9
cross section of photoelectric effect leading to an enhanced soft tissue contrast [11].

However, this was the case only for lesions with an intrinsic high contrast: HPL in GM

and IPH in WM yielded an optimal CNR to circumjacent tissue in VMI at 40 keV. In

contrast, we found that CNR as well as subjective delineation of low-contrast lesions

(IPH to GM and HPL to WM) were optimal in VMI at 120 keV. Regarding IPH within

GM, the most plausible explanation for these results appears to be the significant

T
decrease of beam hardening artifacts near the skull. For HPL within WM, the contrast

IP
enhancement in high keV reconstructions is most likely due to the markedly decreased

R
image noise in these reconstructions which increases CNR and subjective

SC
differentiation to surrounding tissue.

U
As already alluded in some of the DECT studies mentioned earlier, the results of this
N
study suggest that there is not one single optimal keV level for diagnostic assessment
A
of brain lesions. In contrast to source-based DECT, DLCT provides the option to
M

redefine keV levels within the clinical reporting process which appears to be relevant

in order to further improve lesion assessment. Based on our results, higher keV levels
ED

(120 keV) should be chosen when aiming to detect discrete subarachnoid, subdural

and epidural hemorrhages, as beam hardening artifacts and noise in the subcalvarial
PT

space are smallest [3]. Previous studies revealed that intermediate keV settings (65
E

keV) allow for an optimal grey-white-matter differentiation [8]. Low keV settings (40
CC

keV) might allow a more precise localization of smaller lesions within the basal ganglia

and have shown to enhance sensitivity for small petechial hemorrhages that were not
A

detectable in CI or higher VMI reconstructions; however, the actual quality of those

additional lesions needs to be verified in further studies.

For HPL, maximum CNR to GM was yielded in VMI at 40 keV while CNR of those

lesions to WM was significantly increased in VMI at 120 keV. In accordance,

10
delineation of HPL to surrounding WM received best mean rating scores in VMI at 100

keV. This could have relevant implications for clinical diagnostics, as the amount of

ischemic hypoattenuation in unenhanced head CT is an important predictor of clinical

outcome in acute ischemic stroke [21–24]. Thus, increased image quality as well as

improved delineation of HPL, as shown in this study, could be beneficial in the

diagnostic management of stroke patients. Moreover, the detection of cerebral edema

T
could be further improved at low keV, due to higher contrast between HPL and GM at

IP
low keV shown in this study. Thus, regarding hypodense lesions, the optimal keV

R
settings also depend on the suspected pathologies.

SC
This study has several limitations. The retrospective study design generally includes

U
the risk of selection or information bias. MRI was not available for all eligible study
N
subjects and therefore not chosen as a mandatory standard of reference. However,
A
all included lesions were unambiguously diagnosed in the radiological CT report,
M

correlated with the clinical presentation and corroborated in follow-up CT. The

detection of additional hyperdense lesions in VMI at 40 keV by both observers is a

ED

finding of clinical interest. However, as the absolute truth for those lesions is not known,

their significance and clinical impact remains vague and needs to be investigated in
PT

future studies with a larger cohort correlating the findings with another modality such
E

as MRI. Another limitation pertaining to subjective assessment is that true blinding

CC

regarding image reconstructions was not possible and therefore waived since the

different reconstructions could easily be identified e.g. due to noise perception or

A

subjective image impression. Lastly, ICC for the assessment of HPL was only

moderate compared to the assessment of IPH and general image parameters such as

noise and artifact reduction for which the ICC was good (overall ICC: 0.61).

11
In conclusion, VMI provided by DLCT improve image quality for the assessment of HPL

and IPH in unenhanced CT of the head by enhancing contrast between brain

parenchyma and respective lesions in comparison to CI. As the optimal keV settings

for clinical use depend on the suspected pathologies and their location, it is of great

advantage that images can be reconstructed with appropriate keV in real time. VMI at

low keV improve image quality for the depiction of IPH within WM and HPL within GM

T
further allowing the detection of additional hyperdense lesions. Due to artifact reduction

IP
and increased contrast, VMI at high keV yield an enhanced image quality for depiction

R
of IPH within GM near the calvarium. Thus, when available, VMI of unenhanced

SC
spectral detector CT of the head should be considered for reporting. As standard

reconstructions, VMI at 120 keV seem beneficial for low-contrast lesions (e.g. white
U
matter edema) whereas VMI at 40 keV are optimal for lesions with an intrinsic high
N
contrast (e.g. basal ganglia infarction).
A
M
ED

Compliance with ethical standards:

PT

Guarantor:
E
CC

The scientific guarantor of this publication is Jan Borggrefe.

A

Conflict of interest:

The authors of this manuscript declare no relationships with any companies, whose products
or services may be related to the subject matter of the article.

12
Funding:

The authors state that this work has not received any funding.

T
IP
Statistics and biometry:

R
No complex statistical methods were necessary for this paper.

SC
Informed consent: U
N
A
Written informed consent was waived by the Institutional Review Board.
M
ED

Ethical approval:
PT

Institutional Review Board approval was obtained.

E
CC

Study subjects or cohorts overlap:

No overlaps.
A

Methodology:

 retrospective
 diagnostic study

13
  performed at one institution

References

[1] Hemphill JC3, Greenberg SM, Anderson CS, Becker K, Bendok BR, Cushman M, et al.
Guidelines for the Management of Spontaneous Intracerebral Hemorrhage: A Guideline
for Healthcare Professionals From the American Heart Association/American Stroke
Association. Stroke 2015;46:2032–60. doi:10.1161/STR.0000000000000069.
[2] Report of the Joint Committee for Stroke Facilities. XII. Computed tomography in the

T
management of cerebrovascular disease. Stroke 1975;6:103–7.

IP
[3] Pomerantz SR, Kamalian S, Zhang D, Gupta R, Rapalino O, Sahani DV, Lev MH. Virtual
monochromatic reconstruction of dual-energy unenhanced head CT at 65-75 keV

R
maximizes image quality compared with conventional polychromatic CT. Radiology
2013;266:318–25. doi:10.1148/radiol.12111604.

SC
[4] Tawfik AM, Kerl JM, Bauer RW, Nour-Eldin N-E, Naguib NNN, Vogl TJ, Mack MG. Dual-
energy CT of head and neck cancer: average weighting of low- and high-voltage
acquisitions to improve lesion delineation and image quality-initial clinical experience.

U
Investigative radiology 2012;47:306–11. doi:10.1097/RLI.0b013e31821e3062.
[5] Yu L, Leng S, McCollough CH. Dual-energy CT-based monochromatic imaging. AJR.
N
American journal of roentgenology 2012;199:S9-S15. doi:10.2214/AJR.12.9121.
[6] Shinohara Y, Sakamoto M, Iwata N, Kishimoto J, Kuya K, Fujii S, et al. Usefulness of
A
monochromatic imaging with metal artifact reduction software for computed tomography
angiography after intracranial aneurysm coil embolization. Acta radiologica (Stockholm,
M

Sweden : 1987) 2014;55:1015–23. doi:10.1177/0284185113510492.

[7] Guggenberger R, Winklhofer S, Osterhoff G, Wanner GA, Fortunati M, Andreisek G, et
al. Metallic artefact reduction with monoenergetic dual-energy CT: systematic ex vivo
ED

evaluation of posterior spinal fusion implants from various vendors and different spine
levels. European radiology 2012;22:2357–64. doi:10.1007/s00330-012-2501-7.
[8] Neuhaus V, Abdullayev N, Grosse Hokamp N, Pahn G, Kabbasch C, Mpotsaris A, et al.
Improvement of Image Quality in Unenhanced Dual-Layer CT of the Head Using Virtual
PT

Monoenergetic Images Compared With Polyenergetic Single-Energy CT. Investigative

radiology 2017. doi:10.1097/RLI.0000000000000367.
[9] Alvarez RE, Macovski A. Energy-selective reconstructions in X-ray computerized
E

tomography. Physics in medicine and biology 1976;21:733–44.

CC

[10] Silva AC, Morse BG, Hara AK, Paden RG, Hongo N, Pavlicek W. Dual-energy (spectral)
CT: applications in abdominal imaging. Radiographics : a review publication of the
Radiological Society of North America, Inc 2011;31:1031-46; discussion 1047-50.
doi:10.1148/rg.314105159.
A

[11] McCollough CH, Leng S, Yu L, Fletcher JG. Dual- and Multi-Energy CT: Principles,
Technical Approaches, and Clinical Applications. Radiology 2015;276:637–53.
doi:10.1148/radiol.2015142631.
[12] Kalender WA, Klotz E, Suess C. Vertebral bone mineral analysis: an integrated approach
with CT. Radiology 1987;164:419–23. doi:10.1148/radiology.164.2.3602380.

14
[13] Kalender WA, Perman WH, Vetter JR, Klotz E. Evaluation of a prototype dual-energy
computed tomographic apparatus. I. Phantom studies. Medical physics 1986;13:334–9.
doi:10.1118/1.595958.
[14] Flohr TG, McCollough CH, Bruder H, Petersilka M, Gruber K, Suss C, et al. First
performance evaluation of a dual-source CT (DSCT) system. European radiology
2006;16:256–68. doi:10.1007/s00330-005-2919-2.
[15] Johnson TRC, Krauss B, Sedlmair M, Grasruck M, Bruder H, Morhard D, et al. Material
differentiation by dual energy CT: initial experience. European radiology 2007;17:1510–
7. doi:10.1007/s00330-006-0517-6.
[16] Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability.
Psychological bulletin 1979;86:420–8.

T
[17] Hwang WD, Mossa-Basha M, Andre JB, Hippe DS, Culbertson S, Anzai Y. Qualitative

IP
Comparison of Noncontrast Head Dual-Energy Computed Tomography Using Rapid
Voltage Switching Technique and Conventional Computed Tomography. Journal of
computer assisted tomography 2016;40:320–5. doi:10.1097/RCT.0000000000000350.

R
[18] Hixson HR, Leiva-Salinas C, Sumer S, Patrie J, Xin W, Wintermark M. Utilizing dual

SC
energy CT to improve CT diagnosis of posterior fossa ischemia. Journal of
neuroradiology. Journal de neuroradiologie 2016;43:346–52.
doi:10.1016/j.neurad.2016.04.001.
[19] Lin XZ, Miao F, Li JY, Dong HP, Shen Y, Chen KM. High-definition CT Gemstone

U
spectral imaging of the brain: initial results of selecting optimal monochromatic image for
N
beam-hardening artifacts and image noise reduction. Journal of computer assisted
tomography 2011;35:294–7. doi:10.1097/RCT.0b013e3182058d5c.
A
[20] Cho SB, Baek HJ, Ryu KH, Moon JI, Choi BH, Park SE, et al. Initial clinical experience
with dual-layer detector spectral CT in patients with acute intracerebral haemorrhage: A
M

single-centre pilot study. PLOS ONE 2017;12:e0186024.

doi:10.1371/journal.pone.0186024.
[21] Hungerford JP, Hyer M, Turk AS, Turner RD, Chaudry MI, Fargen KM, Spiotta AM.
ED

Impact of ASPECT scores and infarct distribution on outcomes among patients

undergoing thrombectomy for acute ischemic stroke with the ADAPT technique. Journal
of neurointerventional surgery 2016. doi:10.1136/neurintsurg-2016-012528.
[22] Hacke W. Intravenous Thrombolysis With Recombinant Tissue Plasminogen Activator for
PT

Acute Hemispheric Stroke. JAMA 1995;274:1017.

doi:10.1001/jama.1995.03530130023023.
[23] Kummer R von, Allen KL, Holle R, Bozzao L, Bastianello S, Manelfe C, et al. Acute
E

stroke: usefulness of early CT findings before thrombolytic therapy. Radiology

CC

1997;205:327–33. doi:10.1148/radiology.205.2.9356611.
[24] Pexman JHW, Barber PA, Hill MD, Sevick RJ, Demchuk AM, Hudon ME, et al. Use of
the Alberta Stroke Program Early CT Score (ASPECTS) for Assessing CT Scans in
Patients with Acute Stroke. AJNR Am J Neuroradiol 2001;22:1534–42.
A

15
 Tables
CI VMI

40 keV 50 keV 60 keV 70 keV 80 keV 90 keV 100 keV 110 keV 120 keV

Attenuation

GM 37.6±3.9 72.8±16.8 54.8±10.0 44.2±6.1 38.0±4.0 34.0±3.0 31.3±2.7 29.6±2.6 28.3±2.7 27.4±2.8

WM 27.3±2.7 30.3±7.4 28.7±4.8 27.8±3.3 27.2±2.5 26.8±2.1 26.6±2.0 26.4±1.9 26.3±1.8 26.3±1.8

IPH 57.8±9.6 74.3±23.7 65.8±15.9 60.7±12.0 57.7±10.3 55.9±9.6 54.6±9.3 53.8±9.2 53.2±9.2 52.7±9.2

HPL 16.3±6.3 24.8±8.2 20.1±5.8 17.4±5.4 15.9±5.6 14.8±5.9 14.2±6.3 13.7±6.5 13.3±6.7 13.2±6.7

Noise

T
GM 5.1±1.1 7.7±2.9 5.6±1.8 4.6±1.3 4.1±1.1 3.9±1.1 3.8±1.2 3.8±1.2 3.7±1.3 3.7±1.3

IP
WM 5.2±1.0 4.6±0.9 4.1±0.8 3.8±0.7 3.6±0.8 3.6±0.8 3.6±0.8 3.5±0.8 3.5±0.8 3.5±0.8

IPH 6.9±2.1 7.5±3.6 6.8±3.1 6.4±2.9 6.1±2.7 6.0±2.7 5.9±2.6 5.9±2.6 5.8±2.6 5.8±2.6

R
HPL 5.6±1.0 5.2±1.6 4.6±1.2 4.3±1.0 4.1±0.9 4.0±0.9 3.9±0.9 3.9±0.8 3.9±0.8 3.9±0.8

CNR

SC
IPH/GM 2.5±1.4 0.2±2.4 1.4±2.0 2.4±1.9 3.1±2.1 3.5±2.4 3.9±2.6 4.0±2.6 4.2±2.7 4.3±2.8

IPH/WM 3.7±1.6 5.4±3.1 5.2±2.7 5.0±2.5 4.8±2.4 4.7±2.4 4.6±2.4 4.5±2.3 4.5±2.3 4.4±2.4

HPL/GM 2.5±1.0 4.4±1.5 4.1±1.4 3.8±1.2 3.4±1.2 3.1±1.2 2.9±1.3 2.8±1.3 2.6±1.3 2.5±1.3

HPL/WM 1.4±0.8 0.7±1.3 1.3±1.2 1.8±1.2 2.1±1.2

U
2.3±1.3 2.4±1.3 2.5±1.4 2.6±1.4 2.6±1.4
N
GM/WM 28±1.2 11.1±4.1 7.7±2.8 5.2±1.9 3.5±1.3 2.3±1.0 1.5±1.0 1.0±1.0 0.6±1.0 0.3±1.0
A
Table 1: Quantitative values for attenuation, image noise and contrast-to-noise ratio
M

(CNR). Results are indicated as mean ± SD.

ED

CI VMI
PT

40 keV 55 keV 70 keV 85 keV 100 keV 120 keV

Differentiation

IPH/GM 4 (3-4) 1 (1-2) 2 (2-3) 4 (3-4) 4 (4-5) 4 (4-5) 4 (4-5)

E

IPH/WM 4 (4-5) 5 (3-5) 5 (4-5) 5 (4-5) 5 (4-5) 5 (4-5) 5 (4-5)

HPL/WM 3 (2-4) 2 (1-3) 3 (2-4) 4 (3-5) 4 (3-5) 4 (4-5) 4 (4-5)

CC

Subjective Image noise 3 (2-3) 2 (2-2) 3 (2-3) 3 (3-4) 4 (3-5) 4 (4-5) 4 (4-5)

Subcalvarial artifacts 3 (3-3) 1 (1-1) 2 (2-3) 4 (3-4) 5 (4-5) 5 (5-5) 5 (5-5)

A

Table 2: Results of the qualitative assessment regarding differentiation of IPH/HPL to

surrounding brain parenchyma, subjective image noise and subcalvarial artifacts.
Results are median and range.

16
Figure legends

T
R IP
SC
U
N
A
M
ED

Fig. 1: (A) Mean CNR (+ SD) of IPH to GM; at 120 keV, CNR is optimal corresponding
PT

to highest median Likert scores for subjective differentiation of IPH to surrounding GM

at 120 keV (B). (C) Mean CNR (+SD) of IPH within white matter being highest at 40
keV while median subjective scores for differentiation of IPH/WM did not significantly
E

differ between the tested reconstructions. Significance levels are indicated with stars:
p≤0.05*, ≤0.01**, ≤0.001***. Grey dots indicate statistical outliers.
CC
A

17
 T
R IP
SC
U
Fig. 2: Additional hyperdense lesions (arrows) at 40 keV (B, D, F, H) that initially were
not detected in CI (A, E, C, G).
N
A
M
ED
PT

Fig. 3: Mean CNR (+ SD) of HPL to GM (A) and WM (B). Corresponding to optimal
CNR of HPL/WM at 120 keV, median Likert scores regarding differentiation of IPH to
the WM were the highest at the same keV level (C). Significance levels are indicated
with stars: p≤0.05*, ≤0.01**, ≤0.001***. Grey dots indicate statistical outliers.
E
CC
A

18
 T
R IP
SC
U
N
A
M

Fig. 4: Mean CNR (+SD) of GM to WM which was superior in VMI at 40 keV.

Significance levels are indicated with stars: p≤0.05*, ≤0.01**, ≤0.001***.
ED
E PT
CC
A

Fig. 5: Median rating scores for subjective image noise (A) and artifact reduction (B)
clearly favoring VMI at high keV levels. Significance levels are indicated with stars:
p≤0.05*, ≤0.01**, ≤0.001***. Grey dots indicate statistical outliers.

19
 T
R IP
SC
U
N
A
M
ED
E PT
CC
A

Fig. 6: Image examples for improved delineation of HPL in VMI at 120 keV (B)
compared to CI (A) as well as reduction of subcalvarial artifacts (C/D) in VMI at 120
keV.

20