Lecture'3'– CT#Image#Reconstruction#and#Display

! Image%reconstruction%in%CT%is%a%mathematical%
process%that%generates%images%from%X5ray%
projection%data%acquired%at%many%different%
angles%around%the%patient.
! using%a%variety%of%methods,%e.g%:
! filtered)back.projection)technique)– most)common
! algebraic)reconstruction)technique)(ART))– not)common)
! adaptive)statistical)iterative)reconstruction)(ASiR))– new
!Image&reconstruction&impact&on&image&&&&&&&&
quality&and&therefore&on&radiation&dose.

!Reconstruct&images&with&the&lowest&possible&
noise&without&sacrificing&image&accuracy&
and&spatial&resolution.&

!Reconstructions&improve&image&quality.
 Scan;Profile;Data
(from;each;detector)

detectors 10
156
150
126
110
50
x"ray;beam 44
48
61
78
90
108
90
60
55
58
x"ray 89
patient 150
tube 156
130
101
50
22
18
71
26
32
41
59
79
169
162
150
142
139
65
12

plot;of;values;|;values
! Two$major$categories$of$methods$exist :
! analytical$reconstruction
! iterative$reconstruction
! filtering$of$the$scan$profiles$is$required$for$back$
projection$techniques$
! various$different$types$of$filters$can$be$used$to$
enhance:
" edges$– mainly$used$in$imaging$bone
" soft$tissue$– used$for$abdo$/$head$
! iterative'reconstruction,'lower'image'noise'and'
higher'spatial'resolution'compared'with'FBP .

! iterative'reconstruction'can'reduce'image'
artifacts'such'as'beam'hardening'and'metal'
artifacts.
! Back%projection%(FBP)%one%type%of%analytical%
reconstruction%that%is%currently%widely%used%on%
clinical%CT%scanners%because%of%their%
computational%efficiency%and%numerical%
stability.%
! The$central$high$uptake,$from$the$original,$is$now$
distinguishable,$but$it$has$STAR;burst$interference$
pattern
! This$leads$to$Star$artefact
! This$can$be$solved,$but$not$completely$by$just$
accepting$the$high$values.
! How$does$BP$work$?
!

13
14
15
17
Original(signal(profile Convolution(kernel Filtered(signal(profile

!
=

Standard

! =

Smoothing

! =

Sharpening

18
 Simple back projection Filtered back projection

⊗ =
Original Convolution Convolution Filtered profile
profile operator kernel

Simple back projection versus filtered back projection. Simple back projection results in a blurred
image. Filtered back projection removes the blur.
19
! We#will#solve#a#numerical#example#of#simple#
back#projections#to#illustrate#how#the#more#
complicated#filtered#back#projection#technique#
is#performed.
 3 4 7

P1

3
1$ 8 9
4

5 11 P2
P4
4 12
8 P3 1
BLUE arrow
! Ray+sum+for+Left+Hand+Side+(LHS)+
=+4+(3+++1)

! Ray+sum+for+Right+Hand+Side+(RHS)+
=+12+(4+++8)
GREEN Arrow
! Ray+sums+for+45+degrees+
=+4,11(3++8)+and+1
RED$arrow
! Ray$sum$for$Upper$Level$
=$7(3+$4)

! Ray$sum$for$Lower$level$
=$9(1$+$8)
YELLOW Arrow
! Ray-sums-for-135-degrees-
=-3,-5(4-+-1)-and-8-
Add'data'from'the'0'degree'rays'into'the'matrix'
to'obtain'the'first'guess

7 7
(3+4)' (3+4)'
9 9
(1+8)' (1+8)''
2:$Add$the$data$from$the$45$degree$rays$to$
the$first$guess$

18 11
(11$+$7) (4$+$7)
10 20
(1$+$9) (11$+$9)
3:%Add%the%data%from%the%90%degree%rays%
to%the%second%guess

22 23
(18%+%4) (11%+%12)
14 32
(10%+%4) (20%+%12)
4:%%Add%the%data%from%the%135%degree%
rays%to%the%third%guess

25 28
(22$+$3) (23$+$5)
19 40
(14%+%5) (32%+8)
! subtract(a(constant( 9 12
value(from(each(
square. (25%– 16) (28%– 16)
! in(this(case(we(would(
use(16(as(it(is(the(
sum(of(all(of(the(
3 24
numbers(in(the( (19(– 16) (40(– 16)
original(matrix
 Final&Step

! Reduce&the& 3 4
preceding&matrix&to&
a&simple&ratio&by&
(9$/$3) (12$/$3)
using&the&obvious&
common&divisor 1 8
! In&this&case&=&3 (3&/&3) (24&/&3)
! Filter'mask'is'multiplied'by'the'corresponding'
cells'in'the'image'and'the'total'sum'is'placed'in'
the'centre'of'the'matrix'area'in'the'convoluted'
image.
! Then'the'mask'is'shifted'one'column'and'the'
complete'process'repeated'again'
! The$signal$data$from$the$CT$detector$array$are$
corrected
! Then$each$projection$undergoes$the$convolution$
process$before$back:projection
! The$type$of$filter$can$be$selected$by$the$user$to$either$
smooth$the$image$or$enhance$its$edges
What is the approximate pixel size if the
FOV is 48cm and a 1,024 x 1,024 matrix
is used?

a. 0.47 mm x 0.47 mm
b. 0.48 mm x 0.48 mm
c. 0.94 mm x 0.94 mm
d. 2.34 mm x 2.34 mm
e. 4.68 mm x 4.68 mm

Pixel size = FOV / matrix size

iterative reconstruction introduces a correction loop
in the image generation process that cleans up
artifacts and noise in low-dose images. After the initial
reconstruction using traditional filtered back
projection, synthetic measurement data are calculated
from the image (in the so-called image space), as if
the image had been the measurement object in a CT-
scan. These new data are compared to the actually
measured raw data, and a correction image is
calculated.
Filtered Back Projection Filtered Back Projection Theoretical IR
50 % Less Radiation Dose
Statistical IR Initial Raw-data IR Sinogram IR
! Is
a feature that uses the information acquired
during the scan and repeated reconstruction steps
to produce an image with less “noise” or better
image quality (e.g., higher spatial resolution or
decreased artifacts) than is achievable using
standard reconstruction techniques
The use of Iterative Reconstruction by itself may
not decrease CTDIvol; with use of Iterative
Reconstruction, image quality will change and this
may allow a reduction in the CTDIvol by adjusting
the acquisition parameters used for the exam
! GE#Medical#have#released#a#new#reconstruction#
method#called#Adaptive#Statistical#iterative#
Reconstruction#(ASiR)#
! an#iterative#method,#requires#multiple#passes#
hence#only#recently#used#with#now#very#high#
powered#computers
! claims:#low#doses#can#be#used#when#using#ASiR#
with#the#same#image#quality