Principle of Computed

tomography

嘉義長庚放射科　　廖書柏
 Introduction
 In 1950 ， Allan M. Cormack develop the theoretical and
mathematical methods used to reconstruct CT images.

 In 1972 Godfrey N. Hounsfield and colleagues of EMI Central

Research Laboratories built the first CAT scan machine,
taking Cormack's theoretical calculation into a real
application.

 For their independent efforts, Cormack and Hounsfield shared

the Nobel Prize in medicine and physiology in 1979.
What is CT scanner?
 A X-ray device capable of cross-section imaging
-creates images of “slice” through patient
Advantages of CT scanning

 Ability of differentiate overlying structure

 Excellent contrast
-overlying structure don’t decrease contrast
-digital images, so variable window settings
X-ray Source and detectors
 Source
-rotating anode disk
-small focus spot (down to 0.6 mm)
-polychromatic beam
 Detectors
-xenon
-solid-state:
NaI(Tl) 、 CsI scintillaton crystals 、 ceramic
materials containing rare-earth oxides 、 BGO and
CdWO4
xenon
xenon

Pressured xenon gas

ionization

Electrical signal
Solid-state
Solid state
Ceramic or crystal scintillatior

Photon capture
Light

Photo-diode

Electrical signal
Collimators
 Pre-patient collimator- control slice thickness
 Pre-detector collimator-reduce scattered radiation
History of CT
 Variations in scanner design based on :
 X-ray tube and detector movement
 Detector arrangement
 Rotating mechanism
First-generation ~1972
 single X-ray tube and one
detector element
 Pencil beam
 about 5 minutes per slice
from 180 degrees rotation.

Translate-rotate movement
EMI CT1000
Second-generation ~1975

 Single X-ray tube and multiple

detector elements
 Narrow fan beam(~10 。 )
 About one minute per slice

Translate-rotate movement
Third-generation ~1975
 Single X-ray tube, rotating
movement
 Multiple detectors in
curvilinear design,
rotating movement
 Fan beam(~30 。 )
 Several seconds per slice

Rotate-rotate movement
Fourth-generation ~1976
 Single X-ray tube, rotating
movement
 Fixed ring as many as
8000 detectors inside of
gantry
 1-s scan time
 Avoiding ring artifact
problem of 3rd generation
scanner
Rotate-stationary movement
Fifth-generation ~1984
 four semicircular tungsten target rings spanning 210
degrees about the patient
 Multiple detectors of two banks, fixed inside of the
gantry
 no mechanical movement
 By using four target rings and two detector banks,
eight slices of the patient may be imaged without
moving patient.
EBCT( electron beam CT)

A sub-second scanner, called “Imatron”

Each sweep of a target ring requires 50 ms and 8 ms delay to reset the beam.
eight parallel slices (scanned two per sweep) requires approximately 224
milliseconds to complete
Sixth-generation ~1989
 Helical /spiral CT was introduced in 1989, based on Generation
Three
 Single X-ray tube and single-row detector
 Never-stop and one-direction rotating X-ray tube, detectors
 Capability to achieve one second image acquisition, or even
sub-second
 Slip ring replaced with the x-ray tube voltage cables enable
continual tube rotation.
Slip ring technology in 1985

Slip ring allow continuous gantry rotation

Conventional mechanism
Seventh-generation ~1998
 Single X-ray tube ,Multiple-row detector, rotating
movement
 Allow simultaneous acquisition of multiple slice in a
single rotation
 Half-second rotation(0.5 s)
 Sub-second scanner
 The Basic CT Term
 Image matrix
 Linear attenuation coefficient
 CT numbers
Image matrix
 Every CT slice is subdivided into a matrix of up to
1024X1024 volume element (voxel)
 The viewed image is then reconstructed as a corresponding
matrix of picture element (pixel)
 Each pixel is assigned a numerical value (CT number),
which is the average of all the attenuation values contained
within the corresponding voxel.
Voxel size= pixel size X slice
thickness

The diameter of image

reconstruction is called the
field of view (FOV).

Pixel size=FOV/matrix size

Linear Attenuation Coefficient (μ)
 Basic property of matter
 Depends on x-ray energy and atomic number (Z) of
materials.
 Attenuation coefficient reflects the degree to which x-ray
intensity is reduced by a material

I0 x I

I = I0 e -μx
 I0 x1 x2 x3 I

I = I0 e -(μ1x1+μ2x2)

I0 x1 xn I
n
                                                                                       
I = I0 e Σ
- i=1μixi

μ(x, y) is the linear attenuation coefficient for the material in

the slice
CT numbers
 The precise CT number of any given pixel is calculated
from the X-ray attenuation coefficient of the tissue
contained in the voxel.
 CT number ranged from -1000~3095(12 bit)

When k=1000, the CT numbers

are Hounsfield units
 Tissue μ(cm-1)
 CT numbers normalized in this
manner provide a range of Bone 0.528
several CT numbers for 1% Blood 0.208
change in attenuation Gray matter 0.212
coefficient.
White matter 0.213

CSF 0.207
Water 0.206
Fat 0.185
Air 0.0004
Linear attenuation coefficient of various body tissues for 60 keV x-ray
The hounsfield scale of CT numbers
Image reconstruction
 Image reconstruction
 The image is reconstructed from projections by a
process called Filtered Backprojection .
 "Filtered" refers to use digital algorithms called
convolution to improve image quality or change certain
image quality characteristics, such as detail and noise
 "Backprojection" is the actual process used to produce or
"reconstruct" the image. 
The filtered backprojection process involves the
following steps:
 generating a sinogram from a set of N projections
 filtering the data to compensate for blurring
 Backprojecting the data
.
Projection and sinogram
 Ray: the X-ray read by every one detector within a short
time interval.
 Projection:yall rays sum in a direction
 Sinogram: all projections 
P(t)
t 

 x

μ(x,y)
t
X-rays Sinogram
P(t) =
Filter
 a de-blurring function is combined (convolved) with the
projection data to remove most of the blurring before the
data are backprojected.
 A high-frequency filter reduces noise and makes the
image appear “smoother.”
 A low-frequency filter enhances edges and makes the
image “shaper.”

A low-frequency filter may be referred to as a “high-pass” filter because it

suppresses low frequencies and allows high frequencies to pass.
Backprojection
 Projection data (in Sinogram)  1D-FT  filled in k-
space  central slice projection theorem  2D-
inverse FT  CT images

中央切面投影理論
(Central Slice Projection Theorem, CSPT) ：
If a 1D Fourier Transform is performed on a projection of
an object of some angle, the result will be identical to one
line on 2D Fourier Transform of that object and at that
angle.
Central Slice Projection Theorem

y ky
P(t)
t
F[P(t)]
 x  kx

μ(x,y) F(kx,ky)
CSPT can relate the Fourier transform of the projection to one line in the 2D K
space formed by the 2D Fourier transform of μ(x,y)
 y
ky
F-1[F(kx,ky)] v

x u
kx

u(x,y) F(kx,ky)

Inverse 2D-FT
object 1D-FT

image

Inverse 2D-FT
Backprojection
Filtered backprojection

Filtered backprojection removes the star-like blurring seen in simple

backprojection.
Filtered backprojection
Image display
Image manipulation

 Image manipulation belongs to the domain of

digital image processing.
 Window width and level
 The window width covers CT numbers of all the tissue of
interest that is displayed as shades of gray, ranging from black
to white. Thus width controls the contrast in the displayed
image. 
 The level control adjust the center of the window and identifies
the type of tissue to be imaged.
Reducing window width increases the displayed image contrast among the tissues
WL= 40 WW= 350 WL= -600 WW=1500
Spiral CT technology
Pitch
 Pitch is defined as the patient couch movement per
rotation divided by the slice thickness.
 Pitch= couch movement per rotation
beam collimation

Couch movement 5 mm/rot 10 mm/rot

Slice thickness 5 mm 5 mm

pitch 5/5=1 10/5=2

 effects of increasing pitch
 Faster scan time for a specific
volume body.
 Dose is reduced because radiation is
less concentrated
 Image resolution might be reduced

when the pitch is increased,

table appears to move faster along the patient's body
Reconstruction interval/increment

 The RI determines the degree of sectional

overlap to improve image quality.
 As RI decreases, image quality increases” but
with trade-offs of increase image processing time,
data storage requirements, and physician time for
image review”
Slice
thickness=5 mm
RI=2.5 mm

overlaping 50%

RI=2 mm
overlaping (5-
2)/5=60%
Image quality
Spatial resolution in CT
 focal spot size
 detector dimensions
 Slice thickness
 Pixel size Pixel size= FOV/matrix size

 Pitch
 artifact
Image Artifact in CT
 Image Artifact
 Artifacts are any discrepancy between the CT numbers
represented in the image and the expected CT numbers
 Common artifacts
 Beam hardening
 Partial volume effect
 bad detector(3th scanner)
 Metal
 Patient motion
 Beam hardening effect
 Linear attenuation coefficients vary with photon energy.
 After passing through a given thickness of tissue , lower-energy
x-rays are attenuated to a greater extent than high-energy x-rays
are.
 artifacts such as a reduced attenuation toward the center of
tissue (cupping) and streaks that connect tissues with strong
attenuation.
polychromatic spectrums
Cupping Artifact
Means for suppressing beam hardening effect
 pre-filtering X-rays
 avoiding high X-ray absorbing regions if
possible
 applying appropriate algorithms
Partial volume effect
 Partial volume artifacts are the result of a variety of
different tissue types being contained within a single voxel
 Measured attenuation coefficient are averaged by all
components
 use thinner slice to reduce
bad detector(3th scanner)

 Each detector views a separate

ring of an anatomy.
 any single detector or a bank of
detectors malfunctions will
produce ring artifact
Metal artifact
 Metal materials can cause the streaking artifacts due
to block parts of projection data
ex: Dental fillings
Prosthetic devices
Surgical clip
 Remove the metal material as possible to reduce the
artifact
Patient motion
 Voluntary and involuntary motion
can cause streaking artifacts in
the reconstructed image.
 Reduce motion:
-Shorter scan time
-Immobilization and
positioning aid
Effect of reducing projections

96 24 12
the number of views (projections)
Effect of reducing rays

200 50 25

The numbers of the data point (rays) per projection

Thanks for your attention~