6/4/2015

Contents

Lectures: 6 and 7
Computer Tomography - CT

Computer Tomograph CT Introduction

CT - Introduction

Computed tomography (CT) = creates two-dimensional

cross sectional images from three-dimensional body
structures.
Computed tomography utilizes a mathematical technique
called reconstruction to accomplish this task => CT is a
mathematical process.
In a basic sense, a CT image = the result of "breaking apart"
a three-dimensional structure and mathematically putting
it back together again and displaying it as a twodimensional image on a screen.

University Politehnica of Bucharest

Introduction
Historic
CT Principle
Equipment
Image Formation & Reconstruction
Applications

The primary goal of any CT system is

to accurately reproduce the internal
structures of the body as twodimensional cross-sectional images.
This goal is accomplished by computed
tomography's superior ability to
overcome superimposition of
structures and demonstrate slight
differences in tissue contrast.
It is important to realize that collecting
many projections of an object and
heavy filtration of the x-ray beam play
important roles in CT image formation.
Each component of a CT system plays
a major role in the accurate formation
of each CT image it produces.

University Politehnica of Bucharest

6/4/2015

X rays vs CT

Principiul de baza

In conventional radiography, subtle differences of less than about 5 percent in

subject contrast (i.e., x-ray attenuation in the body) are not visible in the image.
Three Reasons:

The projection of three-dimensional anatomic information onto a two dimensional image

receptor obscures subtle differences in x-ray transmission through structures aligned
parallel to the x-ray beam.
Conventional image receptors (i.e., film, intensifying and fluoroscopic screens) are not able
to resolve small differences (e.g., 2%) in the intensity of incident radiation.
Large-area x-ray beams used in conventional radiography produce considerable scattered
radiation that interferes with the display of subtle differences in subject contrast.

each of these difficulties is eliminated in computed tomography (CT)

differences of a few tenths of a percent in subject contrast are revealed in the CT image.

Although the spatial resolution of a millimeter or so provided by CT is notably poorer than

that provided by conventional radiography, the superior visualization of subject contrast,
together with the display of anatomy across planes (e.g., cross-sectional) that are not
accessible by conventional imaging
techniques, make CT exceptionally useful for visualizing
University Politehnica of Bucharest
5
anatomy in many regions of the body.

University Politehnica of Bucharest

Overview

Contents

University Politehnica of Bucharest

Introduction
Historic
CT Principle
Equipment
Image Formation & Reconstruction
Applications

6/4/2015

Historic
Computed tomography was introduced as a method in the early of
70s and has been a revolution in medical imaging.
radiography: simple projection of the patient's body on a
detector (film)
CT: multiple projections of the body, planar projections ->
recombined mathematically by computer > provide image sections of the body
While simple radiography -> a summed image properties of the
crossed body, CT is able to discriminate spatial investigated body =>
different spatial properties of the body could now be located.
Since a computer is needed to provide section images => the
technique called computed tomography (tomein (Gr) - to cut).

MOMENTE IMPORTANTE

Historic
first fully electronicdigital imaging modality
although it is recognized that
the name of the medical investigation for
diagnosis using X-ray
=> tomography principle is applied
on other medical techinques:
ultrasound, magnetic resonance, etc..
It is
possible to reconstruct coronal and sagital
sections from the cross sections,
their resolution is, of course, lower than
the cross-sections (the
lowest resolution is oriented in the
direction perpendicular to
the original section).

Projection of a material property

1917 Johan Radon fundamenteaza matematic procedeul

reconstructiei sectiunilor din proiectii
1963 Alan Cormack (Universitatea Cape Town) publica un articol
in care prezinta teoretic tomografia
computerizata cu raze X. Singura reactie vine
de la Centrul Elvetian pentru Studiul Avalanselor !?!

1971 Godfrey Hounsfield (EMI Group) realizeaza primul

tomograf cu raze X. EMI Group se ocupa nu
numai cu muzica ci si cu radar, rachete, calculatoare

1979 Cormack & Hounsfield primesc Premiul Nobel

Cei 50 de ani au fost necesari pentru ca tehnologia sa prinda din
urma conceptele teoretice.

6/4/2015

Project line integral Radon

transform
All of the materials properties along a
straight path are used to form the single
projection value
For convenience, we refer to the various
positions along the line

X-ray transmission measurements

What is a Hounsfield unit or CT

number?
Each pixel is assigned a numerical value (CT number), which is the
average of all the attenuation values contained within the
corresponding voxel.
This number is compared to the attenuation value of water and
displayed on a scale of arbitrary units named Hounsfield units (HU)
after Sir Godfrey Hounsfield.
This scale assigns water as an attenuation value (HU) of zero.
The range of CT numbers is 2000 HU wide although some modern
scanners have a greater range of HU up to 4000. Each number
represents a shade of grey with +1000 (white) and 1000 (black) at
either end of the spectrum
Tissues or materials with attenuation (density) greater than water will
have positive CT numbers. Those that are less dense will have negative CT
numbers.
X-ray attenuation depends on both the density and atomic number (Z) of
materials and the energy of the x-ray photons.

6/4/2015

CT systems
Since their introduction in the '70s, CT systems have
experienced continuous improvement,
distinguished several stages in their
evolution technique - the 4 generations
of CT systems, followed by the appearance
of spiral and multi-sections CT system
Low-contrast-resolution flat-panel volume CT image of a cone-beam
phantom (120 kV, 35 mA, 10-mm section thickness). The CT numbers are
obtained by using a multidetector CT unit to compare contrast resolution.
The flat-panel volume CT scanner can differentiate inserts with a
difference in attenuation of 5 HU from the background; the 3-HU insert is
barely visible. A clinical multidetector CT scanner can detect a 3-HU insert
without difficulty.

First CT system generation

The first generation used a parallel beam X-ray detector and a
single synchronous translational and rotational movements.

Second CT system generation

The second generation is characterized by using fan shaped X -ray
beam (fan-beam).
Increased number of detectors and the number of low rotation,
leading to a scan time of approximately 20 seconds

After translational motion is per

formed along
the investigated object, sourcedetector system is rotated by 1
degree for the next scan.
The time required to
obtain projections to
reconstruct a section was a
few minutes.

6/4/2015

Third CT system generation

The fourth phase

A fixed detector ring (600) is inserted
Only radiation generator is also rotating

A third generation uses divergent

beams with
openings larger enough to cover
the whole subject examined.
The translational motion is
removed.
A section becomes possible to
acquire in 1-3 seconds.

Spiral (helical) CT

Increased speed of
these scanners has
allowed to obtain
information in real time,
allowing real-time capture
physiological processes.

Spiral (helical) CT

Detection system is arranged in a fixed ring, the X-ray tube is rotating with a continuous
speed motion while the patient table has a slow motion of translation.
Relative to patient Xray tube system detectors have a spiral motion.
Projection of a slice will not
be scanned located in a plan; in this
sense interpolation algorithms are used.

This geometry allows rapid accumulation of

data from multiple sections and a high
resolution in the plane perpendicular to the
original section (the original transverse plane).

6/4/2015

Multi-section CT system
It's just like a spiral system only that it has more fixed detectors rings.
At first only two, now 64 with increasing velocity (from 1/sec. to 4/sec.).
The 2007 models voxel resolution of 0.35 mm isotropic, with a z-axis scan of 16 cm / sec.
The major advantage of multi-section CT is the increased speed of the scanned volume.

METODE DE ACHIZITIE A DATELOR

Tub RX
Rotaie

Fascicol divergent.
Necesita corectii geometrice in
vederea reconstructiei imaginii.
Viteza mare de achizitie a datelor
( < 1 sec./sectiune)
Rezolutie mare
(nr. detectorilor 800)

Arie de detectori

1973 - First commercially available clinical CT

head scanner on market (EMI)

Posibilitatea achizitiei simultane

a mai multor sectiuni (multislicing)

ARHITECTURA UNUI TOMOGRAF COMPUTERIZAT

GANTRY

One of the first EMI head CT scanners in the US was installed at Henry Ford Hospital
(Detroit) in 1973.
The CT image shown was obtained at the Cleveland Clinic in 1974. A large meningioma
has been enhanced by iodinated contrast material.

Consola operatorului

6/4/2015

ARHITECTURA UNUI TOMOGRAF COMPUTERIZAT

ARHITECTURA UNUI TOMOGRAF COMPUTERIZAT

Tub RX

CPU

- +

G.I.T.
C.R.I.
C.R.

MEM
I.O.T.

Arie detectori

ITF DM
CONSOLA

CD / DVD
HDD

Interne
t

C.A.D.
C.C.R.

GANTRY

ARHITECTURA UNUI TOMOGRAF COMPUTERIZAT

ARHITECTURA UNUI TOMOGRAF COMPUTERIZAT

Explicatii:
Gantry initial aceasta componenta constructiva a unui tomograf a fost conceputa
doar ca un suport mecanic pe care sa se poata roti ansamblul tub RX detectori.
Piesa centrala a unui gantry o reprezinta un inel circular care se poate roti in jurul
unui punct central numit centru de rotatie - CR
Pe acest inel se afla dispuse diametral opus tubul RX si aria de detectori
precum si cele doua componente aferente ale acestora: generatorulu de inalta
tensiune (GIT) si circuitul de achizitie al datelor (CAD).
Transmisia catre consola a datelor achizitionate de la pacient precum si
transmisia de la consola la generator se deruleaza prin intermediul unei
interfete de transmisie optica IOT ce permite atingerea unei viteze de
transmisie a datelor extrem de mare si confera o imunitate crescuta la zgomot.
Miscarea de rotatie este asigurata de un bloc specializat circuitul de control
al rotatiei (CCR).

MASA DE EXAMINARE

Consola: intregul proces de calcul necesar reconstructiei este realizat in consola

care ocupa o mica parte din pupitrul de lucru al operatorului.
CPU Procesorul central care tine sub control intregul sistem de tomografie si
asigura comunicarea cu operatorul uman.
Afisarea unei imagini tomografice este asigurata de un monitor alb-negru a carui
rezolutie poate sa ajunga la 1024 de linii, fiecare pixel fiind codat pe 8 biti.
Volumul de date necesar reconstructiei unei singure sectiuni tomografice este de
ordinul a 1.5 MB
Manipularea in timp rezonabil a volumului de date impune existenta unei interfete
de disc magnetic (ITF DM) foarte performante.
Adaptata acesteia este si memoria RAM a carei capacitate este de 64-128 MB;
trasnferul datelor intre aceasta si discul magnetic fiind practic un proces continuu.
Cea mai importanta componenta a consolei este CRI circuitul de reconstructie
a imaginii menit a genera imaginea sectiunii investigate pornind de la setul de
proiectii achizitionat.
Evolutia acestui modul a fost in stransa legatura cu dezv tehnologiei calculatoarelor
astfel incat astazi timpul necesar reconstructiei unei imagini este de 2-3 sec.

6/4/2015

MASA PACIENTULUI

CONSOLA
CPU 32 biti min.
RAM > 4 8 Gby

P1

P2

P3

HDD > 500 GBy

C.R.I. - cea mai importanta componenta (2 - 3 sec./imagine)
- grad mare de paralelizare a calculelor

Miscari variate si foarte precise

Traiectoria elicoidala necesita o viteza constanta

g(s, ) = { g(si,j) | i = 0, N-1 ; j = 0, M-1 }

g j (si) = { g(si,j) | i = 0, N-1 ; j = fixat }

datele initiale
date trimise spre
unul din filtre

Retroproiectia este o suma ordinrea de sumare nu conteaza

Radon Transform

Tomography is performed in 2 steps:

1st step= data acquisition(record of projections)
The result is a set of angular projections.
The set of projections of a single slice is called
sinogram.
2nd step= image reconstruction from projections
There are 2 groups of reconstruction methods:
analytic (e.g. FBP = filtered back projection) and
iterative (e.g. ART = algebraic reconstruction
techniques).

PROIECTIA
Nu reprezinta umbra lasata de corp pe
o
suprafata ci este o masura a modului in
care este structurat interiorul corpului
Pentru a investiga interiorul corpului este
necesara utilizarea unei forme de energie
care poate penetra corpul studiat
Forme de energie utilizate: RX, unde mecanice
Informatia obtinuta este rezultatul
interferentei dintre energia injectata in corp si
materia din interiorul lui.
Efectele sunt de tip cumulativ astfel incat se
pierde rezolutia pe axa propagarii energiei

6/4/2015

PROIECTIA

Operatorul Proiectiei Transformata Radon

RX1 RX2

I0

RX

dI
~I
dl

dI
= I
dl

I (l) = I 0 exp( l)
s

g(s,)

I (l) = I 0 exp f ( x, y ) du

( )

I
0
g = ln = f ( x, y )du = g ( s, )
I ( )

Profilul in lungul RX1

Profilul in lungul RX2

Folosind pentru () descrierea

g1

g2

()

Valoarile scalare g1,2 sunt afectate de structurile traversate dar nu

inglobeaza informatia legata de pozitia neomogenitatilor intalnite
ci doar efectul cumulativ al lor

s = x cos( ) + y sin( )

g ( s , )

NOT

f(x,y)

DEF

f ( x, y ) ( x cos + y sin s) dxdy

Operatorul Proiectiei

Transformata RADON

Transformata Radon
s

= 720
y

Transformata Radon - Proprietati

proiectia pe o directie data nu ofera
rezolutie spatiala in lungul dreptei de
integrare ()

= 00

Daca g1,2(s,)= f1,2(x,y) avem:

1o {af1(x,y)+b f2(x,y)} = ag1(s,)+bg2(s,)

colectarea mai multor proiectii pe

directii diferite completeaza informatia
referitoare la structura interna a corpului
(
)

= 900

= 2700

3o g(s,)= g(s,+2k)

unde k

P(r,)
x

unde a,b sunt constatnte reale

=> g(s,) = 0 pentru

s > D 2
2

1
g ( as, )
a

7o fp(r,+o) = g(s, + o)

8o

>

5o f(x-xo,y-yo) = g(s - xocos - yosin, )

Sinograma
s

g(s,)

D
y>

6o f(ax,ay) =

Graficul 3-D al lui g(s,)

D,
2

2o f ( x, y ) = 0 pentru

4o g(s,)= g(-s, )

odata cunoscuta, functia g poate fi

vizualizata in mai multe moduri:

= 1800

s (-, ) i [0,]

s = r cos( )

M=

f ( x, y)dxdy = g (s, )ds

Verificarea proprietatilor de mai sus este imediata, necesitand cel mult o schimbare
de variabila la integrare.

10

6/4/2015

Transformata Radon Operatorul Retroproiectiei

Transformata Radon Operatorul Retroproiectiei (cont)

Descrie operatiunea inversa proiectiei

Intrebare:

Retroproiectie = Transformata Radon Inversa

???

b(x,y)

NOT

g(s,)

DEF

bp(r,) =

g ( x cos + y sin , )d

in coordonate carteziene (x,y)

g (r cos( ), )d

Operatorul Retroproiectiei

f (u , v) [(u x) cos + (v y ) sin )]d dudv

valoarea lui b(x,y) pentru un punct oarecare se obtine prin sumarea tuturor
razelor (proiectiilor) care trec prin punctul respectiv

[ ( )] =

()

' ( )

( k )

cu k= radacinile lui

1
= f (u , v )
dudv
' (1 )

' (1 ) = ( x u ) 2 + ( y v) 2

f (u, v)

Raspuns:

Transformata Radon Operatorul Retroproiectiei (cont)

Transformata Radon

1
( x u ) 2 + ( y v) 2

dudv = f ( x, y )

1
x2 + y2

Retroproiectie Transformata Radon Inversa

!!!

Transformata Radon Teorema Proiectiei

(1/x)

Prin retroproiectia transformatei Radon se

ajunge la o imagine filtrata a originalului,
filtrul avand o caracteristica de filtru trece jos

Imaginea originala

s = x cos( ) + y sin( )

f (u, v) (u cos + v sin s)dudvd

in coordonate polare (r,)

unde

g(s,) = {f(u,v)}

Daca g(s,)= f(x,y) atunci Transformata Fourier in raport cu variabila s

a proiectiei g(s,o) este egala cu sectiunea centrata pe zero si inclinata cu
unghiul o din Transformata Fourier bidimensionala in coordonate polare
a functiei f(x,y).

Retroproiectia
Transformatei Radon

g
s

f(x,y)

s

g(s,*)

g(s,)

*

f(x,y)

Pentru a obtine o reconstructie exacta trebuie anihilat efectul filtrarii trece jos
ceea ce presupune o filtrare cu un filtru trece sus a fiecarei proiectii inainte de
a efectua retroproiectia.

Fp(1,2)

Teorema ofera o solutie alternativa pentru reconstructia imaginilor in

tomografia RMN unde semnalul receptionat reprezinta spectre de frecventa

11

6/4/2015

Transformata Radon Teorema Proiectiei (demonstratie)

y

g ( s, )

f ( x, y)du

DEF

( )

folosind relatiile de transformare (x,y) in (s,u)

x

x = scos - usin
y = ssin + ucos

s = xcos + ysin
u = -xsin + ycos

g ( s, ) =

f (s cos u sin , s sin + u cos )du

Atunci

s{g(s, )}

g ( s, ) e

js

ds =

f (s cos u sin , s sin + u cos ) e

j s

dsdu

si tinand cont ca s = xcos + ysin putem scrie

s{g(s, )} =

f ( s cos u sin , s sin + u cos ) e

j ( x cos + y sin )

dsdu = (cos,sin)

12