Computer Assisted Clinical Medicine. Prof. Dr. Lothar Schad. 12/9/2008 | Page 1.
Physics of Imaging Systems. Basic Principles of Computer Tomography (CT) I.
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 1
Master‘s Program in Medical Physics
Physics of Imaging Systems Basic Principles of Computer Tomography (CT) I Prof. Dr. Lothar Schad
Chair in Computer Assisted Clinical Medicine Faculty of Medicine Mannheim University of Heidelberg Theodor-Kutzer-Ufer 1-3 D-68167 Mannheim, Germany
[email protected] www.ma.uni-heidelberg.de/inst/cbtm/ckm/
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 2
Literature I
Dance et al.: “Diagnostic Radiology Physics” Publisher: International Atomic Energy Agency http://www-pub.iaea.org/books/IAEABooks/8841/ Diagnostic-Radiology-Physics-A-Handbook-forTeachers-and-Students Free download !!!
Seite 1 1
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 3
Literature II
Kalender: “Computertomographie. Grundlagen, Gerätetechnologie, Bildqualität, Anwendungen”, 2006
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 4
Literature III
Dössel: “Bildgebende Verfahren in der Medizin”, Chapter 4, 2000
Seite 2 2
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 5
Computer Tomography Introduction
Computer Tomography Introduction
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 6
Literature
source: Kalender. Computertomographie, Publicis MCD Verlag 2000
Seite 3 3
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 7
Mean Effective Dose per Head in 1997
nuclear diagnostic
radon with derivates other radiation sources (nuclear wappons & fallout, Tschernobyl reactor accident, nuclear units, science, technic, household) food
X-ray diagnostic
terrestic radiation cosmic radiation at see level
Regulla et al. ZMP 2003
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 8
Radiodiagnostic Examinations in 1997 percentage of typical radiodiagnostic examinations to the collective effective dose and their frequency
collective effective dose
thorax skeletal esophagus uretic system Mammography angiography CT else
incidence
Regulla et al. ZMP 2003
Seite 4 4
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 9
Visionary Idea in 1896
„Und läßt man der Phantasie weiter die Zügel schießen, stellt man sich vor, daß es gelingen würde, die neue Methode des photographischen Prozesses mit Hilfe der Strahlen aus den Crookeschen Röhren so zu vervollkommnen, daß nur eine Partie der Weichteile des menschlichen Körpers durchsichtig bleibt, eine tiefer liegende Schicht aber auf der Platte fixiert werden kann, so wäre ein unschätzbarer Behelf für die Diagnose zahlloser anderer Krankheitsgruppen als die Knochen gewonnen.“
Frankfurter Zeitung, 7. Januar 1896
source: Kalender. Computertomographie, Publicis MCD Verlag 2000
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 10
CT History
Seite 5 5
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 11
Computer Tomography 1973
original EMI CT head scanner 1973 (Mayo Clinic, Rochester, USA) and 80 x 80 matrix head CT image obtained with it source: Gray and Orton. Radiology 2000
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 12
Brain Image 1974, Matrix 80 x 80
source: Kalender. Computertomographie, Publicis MCD Verlag 2000
Seite 6 6
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 13
Historical Question …
in the early years of CT, an often-heard remark was „ Why would anyone want a new X-ray technique that when compared with traditional X-ray imaging: • yields
10 times more coarse spatial resolution
• is 1/100 as fast in collecting image data • costs 10 times more “
Hendee and Ritenour. Medical Imaging Physics, 4th ed. Wiley, 2002
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 14
… and Answer ! Ct volume data set, matrix 1024 × 1024
source: Kalender. Computertomographie, Publicis MCD Verlag 2000
Seite 7 7
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 15
X-Ray and CT Contrast
source: Kalender. Computertomographie, Publicis MCD Verlag 2000
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 16
Why Cross-Sections ?
1. no superposition → higher contrast 2. three-dimensional localization
Seite 8 8
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 17
CT Contrast
CT image
X-ray image
high local contrast of soft tissue structure
low soft tissue contrast due to superimposed bone
source: Kalender. Computertomographie, Publicis MCD Verlag 2000
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 18
Computer Tomography Principle
Computer Tomography Principle
Seite 9 9
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 19
CT Measurement: Principle I
- gives system of equations for each projection - generate enough projections to solve system of equations
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 20
CT Measurement: Principle II
case 1: homogeneous object, monochromatic radiation
- intensity of X-ray radiation can be used to calculate the attenuation- or projection-value P which can lead to the absorption coefficient µ in a simple case (homogeneous object) - µ can not be calculated in inhomogeneous objects since µ is a function of (x, y) → tomographic method !
case 2: inhomogeneous object, monochromatic radiation
case 3: inhomogeneous object, polychromatic radiation
source: Kalender. Computertomographie, Publicis MCD Verlag 2000
Seite 10 10
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 21
CT Measurement: Principle III
X-ray tube collimator detector with electronic collimator
intensity profile
attenuation profile = “projection”
in the simplest case object is scanned linearly by a needle beam at different angles to determine the attenuation profile source: Kalender. Computertomographie, Publicis MCD Verlag 2000
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 22
CT Measurement: Principle IV
algebraic reconstruction technique (ART)
source: Kalender. Computertomographie, Publicis MCD Verlag 2000
Seite 11 11
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 23
CT Image Reconstruction and Convolution
X-ray tube
without convolution
back projection
with convolution
collimator collimator detector with electronic
attenuation profile (simplified)
- corresponds to high pass filter
profile cut source: Kalender. Computertomographie, Publicis MCD Verlag 2000
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 24
Hounsfield Values
definition CT-value: CT-values contain the linear absorption coefficients of the underlying tissue in every volume element with respect to the µ-value of water. Using this definition the CT-values of different organs are relatively stable and independent of the X-ray spectrum.
CT-value = (µ - µwater) / µwater x 1000 HU
Seite 12 12
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 25
Hounsfield Scale compact bone
CT-number [HU]
liver blood pancreas
spongy bone water
kidney
fat
lung
air
4096 values → 212 → 12 Bit source: Kalender. Computertomographie, Publicis MCD Verlag 2000
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 26
CT Windowing
CT-number [HU]
bone window
- windowing of Hounsfield values at different evaluations of CT images mediastinum window
- human eye can distinguish only 60-80 grey values !
lung window
source: Kalender. Computertomographie, Publicis MCD Verlag 2000
Seite 13 13
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 27
Radon Transformation I
Θ = 0 - 180° s = smin - smax
source: Dössel. “Bildgebende Verfahren in der Medizin” 2000
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 28
Radon Transformation II
p(Θ,s)-diagram projection pΘ(s) = series of numbers of all line integrals of f(x,y) at constant angle Θ and variable distance s to the coordinate origin
source: Dössel. “Bildgebende Verfahren in der Medizin” 2000
Seite 14 14
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 29
Radon Transformation: Example
inside outside
radius of area
source: Dössel. “Bildgebende Verfahren in der Medizin” 2000
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 30
Fourier-Slice-Theorem I
Seite 15 15
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 31
Fourier-Slice-Theorem II
- given is a function f(x,y) an its 2D Fourier transformation F(u,v) f(x,y)
2D-FT
F(u,v)
- given is also pΘ(s) defined as a projection of f(x,y) (i.e. a line of the Radon transformation with Θ = const.) and PΘ(w) its 1D Fourier transformation pΘ(s)
1D-FT
PΘ(w)
- then PΘ(w) describes the values of F(u,v) on a radial beam at angle Θ - how to get from the Radon transformation p(Θ,s) back to the function f(x,y) ? • evaluate from all projections pΘ(s) the 1D Fourier transformation PΘ(w) • take the values of the radial beam at Θ and put them into the function F(u,v) • find f(x,y) by doing the inverse 2D Fourier transformation
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 32
X-Ray Pencil Beam Through a Body
incoming X-ray intensity outgoing X-ray intensity X-ray absorption coefficient
Seite 16 16
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 33
CT Reconstruction Principle I - idea of backprojection: transform measured µ-values into “line thickness” for all projections and add up !
y
x
in the order of 1000 projections with 1000 channels are acquired per detector slice and rotation
y
courtesy: Kachelriess, Erlangen
x
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 34
CT Reconstruction Principle II
- wanted:
function µ(x,y) „X-ray absorption coefficient as a function of (x,y) in a body slice“
- measured:
all line integrals over µ(x,y) i.e. Radon transformation of the wanted function µ(x,y)
- generation of images: Fourier reconstruction • measurement of as many as possible projections with a large number of points (pΘ(s)) projections pΘ(s) are 1D Fourier transformed into (PΘ(w)) • transformed projections are recorded in a matrix F(u,v) problem: interpolation between known values on radial beams and required values on a cartesian grid (F(u,v) complex) • multiplication of F(u,v) with a filter function • inverse 2D Fourier transformation
Seite 17 17
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad
CT Sinogram I
absorption µ
10/30/2015 | Page 35
x
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 36
CT Sinogram II
sinogram = sum of all projections
- sinogram: = sum of all projections = “Radon transformation” of µ(x,y) - absorption strength = brightness
scanning distance x
Seite 18 18
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 37
CT: Back Projection
CT image
sinogram
reconstruction: back projection
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 38
CT: Filtering
sinogram
filtered sinogram
filtering
filter kernel
convolution kernel
unfiltered profile
= filtered profile
Seite 19 19
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad
CT: Filtered Back Projection
10/30/2015 | Page 39
CT image
filtered sinogram
reconstruction: filtered back projection
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 40
Unfiltered and Filtered Projections
measurement
unfiltered back projection of one measurement
unfiltered back projection of all measurements
measurement
filtered back projection of one measurement
filtered back projection of all measurements
source: Dössel. “Bildgebende Verfahren in der Medizin” 2000
Seite 20 20
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 41
CT: Projections and Filtered Projections
inverse Fourier transformation of
summing up all filtered projections source: Dössel. “Bildgebende Verfahren in der Medizin”, 2000
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 42
CT: Filtered Back Projection Principle
• filtering of all measured projections pΘ(s) with |w| gives ~ pΘ(s) • following all filtered projections ~ pΘ(s) with respect to Θ for the whole matrix and adding up all values of the filtered projections into the imaging matrix
source: Dössel. “Bildgebende Verfahren in der Medizin” 2000
Seite 21 21
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 43
Filter Functions I - filter limiting:
- filter functions cernel from:
even, ≠ 0 odd
a = detector distance
filter functions H(w) source: Dössel. “Bildgebende Verfahren in der Medizin” 2000
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 44
Filter Functions II
source: Dössel. “Bildgebende Verfahren in der Medizin” 2000
Seite 22 22
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 45
CT: Iterative Reconstruction
pencil beam j
- unknown: 512x512 = 262 144 pixel
- measured: 1000 projections with 800 detectors = 800 000 equations
part of area wij
→ can be algebraic solved !
source: Dössel. “Bildgebende Verfahren in der Medizin” 2000
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 46
CT: Iterative Reconstruction Solution • processing of 1. projection data value is distributed in equal shares to involved projection pixels • processing of 2. projection difference between “forward calculated data values” and really measured values is used for correction in equal shares to involved projection pixels • processing of all other projections in the same way about 1h calculation time for 512 x 512 ! source: Dössel. “Bildgebende Verfahren in der Medizin” 2000
Seite 23 23
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 47
CT Scanner Generations 1-2
II.
I. CT scanner 1. generation pencil beam (1970)
CT scanner 2. generation partly fan beam (1975) 30 detectors, fan: 10°
source: Kalender. Computertomographie, Publicis MCD Verlag 2000
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 48
CT Scanner Generation 1: EMI-Scanner
Seite 24 24
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 49
CT Scanner Generation 2 multi-detector-translational-rotating system
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 50
CT Scanner Generation 3 CT scanner 3. generation fan beam (1976) 500 – 800 detectors fan: 40° - 60° (whole body) 1000 projections / sec but cable problems: 360° forward rotation stop 360° backward rotation stop …
rotating detector arch
III. rotation without translation
souirce: Kalender. Computertomographie, Publicis MCD Verlag 2000
Seite 25 25
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 51
CT 3. Generation: Collimation fan beam scanner
trajectory of X-ray tube
collimator lamellae
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 52
CT Scanner Generation 3: Principle tube 0°
tube beam at tube 180°
rotation
beam at tube 0°
detectors
III.
tube 180° beam at tube 180° and 0°
„jumping focus“
source: Dössel. “Bildgebende Verfahren in der Medizin” 2000
Seite 26 26
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 53
CT Scanner Generation 3: Jumping Focus
C cathode B
C
anode
cooling oil
straton tube courtesy: Kachelriess, Erlangen
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 54
III.
CT Scanner Generation 3-4
fancy beam unit
IV.
ring detector unit
rotation axis collimator lamellae
Xe-chamber
electrodes rotating detector arch (Xe high-pressure chamber)
fix detector arch (scintillators)
Seite 27 27
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 55
CT Scanner Generation 4
fix detector ring
CT scanner 4. generation fan beam up to 5000 detectors problem: no collimation possible very high scatter radiation very expensive
IV.
souirce: Kalender. Computertomographie, Publicis MCD Verlag 2000
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 56
CT Scanner Generation 3-4
“Mayo – Monster”
Seite 28 28
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 57
Typical CT Examination Room
souirce: Kalender. Computertomographie, Publicis MCD Verlag 2000
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 58
Contact Ring Technique
X-ray tube
detector detector electronic
high voltage
projection data
schematic illustration of the contact ring technique for electric energy supply of the X-ray components and for signal transfer from the detectors to computer souirce: Kalender. Computertomographie, Publicis MCD Verlag 2000
Seite 29 29
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 59
Contact Ring Mounting and Testing
souirce: Kalender. Computertomographie, Publicis MCD Verlag 2000
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 60
CT Components Geometry
X-ray tube form filter fix aperture adjustable aperture
gantry opening
field-of-view
rotation center
scatter radiation collimator
adjustable aperture
detector array
fix aperture
souirce: Kalender. Computertomographie, Publicis MCD Verlag 2000
Seite 30 30
RUPRECHT-KARLSUNIVERSITY HEIDELBERG Computer Assisted Clinical Medicine Prof. Dr. Lothar Schad 10/30/2015 | Page 61
Subsecond CT
1 sec. rotation time
0.5 sec. rotation time
- mechanic forces: acceleration ~ 10-20 g, X-ray tube ~ 200 kg → 20-40 000 N ! - mechanic stability: 0.1 mm !
Seite 31 31