AAPM 2014 Challenges and opportunities in assessment of...
Transcript of AAPM 2014 Challenges and opportunities in assessment of...
7/21/2014
1
AAPM 2014
Challenges and
opportunities in
assessment of
image quality Ehsan Samei
Department of Radiology
Duke University Medical Center
Disclosures
• Research grant: NIH R01 EB001838
• Research grant: General Electric
• Research grant: Siemens Medical
• Research grant: Carestream Health
7/21/2014
2
Credits
Jeff Siewerdsen, Johns Hopkins
Grace Gang, Johns Hopkins
Guanghong Chen, UW
Ke Li, UW
4 Clinical Imaging Physics Group
Duke CIPG
Imaging quality and safety
• Quality: Imaging provides a clinical benefit
– Diagnostic information
– Image quality: Universally-appreciated
– Enhancement dependent on illusive criteria
• Safety: Imaging involves a level of “cost”
– Monitory cost
– Information excess
– Radiation cost
Optimization involves 4 steps:
1. Reasonable measures of risk
2. Reasonable measures of image quality
3. Justified balance between the two by targeted adjustment of system parameters
4. Consistent implementation
We cannot optimize imaging without quantification
7/21/2014
3
Optimization framework
Safety indices (organ dose,
eff. dose, …)
Quality indices (d’, Az, …)
Scan factors (mAs, kVp, pitch, recon,
kernel, …)
Benefit re
gim
e
Risk re
gim
e
Different patients
and indications
Optimization regime
Outline
1. Quality Characterization
2. Quality assessment
3. Quality implementation
7/21/2014
4
Model-based recons +s
• Decades in the making
• Enabled by high-powered computers
• Significant potential for dose reduction
• Potential for improved image quality
Model-based recons -s
• Speed
• Increased vendor-dependence
• Unconventional image appearance
• Limited utility of prior quality metrics
• Need for nuanced implementation for effective improvement in patient care
ASiR (GE)
Veo (GE)
IRIS (Siemens)
SAFIRE (Siemens)
AIDR (Toshiba)
iDose (Philips)
…
Noise texture?
Low Dose CT @ 114 DLP, 1.9 mSv Images courtesy of Dr de Mey and Dr Nieboer, UZ Brussel, Belgium
FBP Reconstruction
Iterative Reconstruction
FBP IR
7/21/2014
5
courtesy of University of Erlangen, Germany
Iterative
Reconstruction FBP
Reconstruction
courtesy of Mayo Clinic Rochester, MN
FBP
Reconstruction
Iterative
Reconstruction
Low Dose CT @ 338 DLP, 1.8 mSv Images courtesy of Dr de Mey and Dr Nieboer, UZ Brussel, Belgium
FBP
Reconstruction Iterative
Reconstruction
7/21/2014
6
Qualimetrics 1.0: CNR
• 1st order approximation of image quality
• Related to detectability for constant resolution and noise texture (Rose, 1948)
• Task-generic
CNR =QL -QB
s B
Parameters that affect IQ
1. Contrast
2. Lesion size
3. Lesion shape
4. Lesion edge profile
5. Resolution
6. Viewing distance
7. Display
8. Noise magnitude
9. Noise texture
10. Operator noise
Feature of
interest
Image details
Distractors
Parameters that are measured
by CNR
1. Contrast
2. Lesion size
3. Lesion shape
4. Lesion edge profile
5. Resolution
6. Viewing distance
7. Display
8. Noise magnitude
9. Noise texture
10. Operator noise
Feature of
interest
Image details
Distractors
1. Contrast
8. Noise magnitude
?
7/21/2014
7
Why CNR is not enough: Noise texture
FBP IR
Solomon, AAPM 2012
Resolution and noise, eg 1
Comparable resolution
Lower noise but different texture
0 0.2 0.4 0.6 0.8 0
0.2
0.4
0.6
0.8
1
Spatial frequency
MT
F
0 0.2 0.4 0.6 0.8 0
0.5
1
1.5
2
2.5
x 10 -6
Spatial frequency
NP
S
-FBP
-IRIS
-SAFIRE
-FBP
-IRIS
-SAFIRE
Higher resolution Lower noise but different texture
-MBIR
-ASIR
-FBP
0 0.2 0.4 0.6 0.8 0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Spatial frequency
MT
F
0 0.2 0.4 0.6 0.8 0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Spatial frequency
NP
S
x 10-5
Resolution and noise, eg 2
-MBIR
-ASIR
-FBP
7/21/2014
8
NPS vs dose
0 0.2 0.4 0.6 0.80
100
200
300
400
500
600
700
800
fxy
(mm-1)
NP
S (
HU
2m
m2)
NPS (Veo)
5 mAs
12.5 mAs
25 mAs
50 mAs
100 mAs
200 mAs
300 mAs
Li, Tang, and Chen, “Statistical Model Based Iterative Reconstruction (MBIR) in clinical CT
systems: Experimental assessment of noise performance,” Med. Phys. (2014)
0 0.2 0.4 0.6 0.80
1
2
3
4
5
6
fxy
(mm-1)
Norm
aliz
ed N
PS
(A
.U.)
Normalized NPS (Veo)
5 mAs
12.5 mAs
25 mAs
50 mAs
100 mAs
200 mAs
300 mAs
Nonlinearity
NPS peak frequency vs dose
0.5 1.25 2.5 5 10 20 300.05
0.1
0.2
0.4
0.8
mAs
Peak fre
quency (
mm
-1)
FBP
FBP (fitting)
Veo
Veo (fitting)
fpeak
(mAs)0.16
Li, Tang, and Chen, “Statistical Model Based Iterative Reconstruction (MBIR) in clinical CT
systems: Experimental assessment of noise performance,” Med. Phys. (2014)
16 33 62 99 224 346 814 1710 25% 50%
75% 100%
0.2
0.4
0.6
0.8
1
DoseContrast (HU)
PS
F w
idth
(m
m)
0.4
0.5
0.6
0.7
0.8
Veo
FBP
Resolution’s joint dependence on contrast and dose
7/21/2014
9
Detectability index
Resolution and
contrast transfer
Attributes of image
feature of interest
Image noise magnitude
and texture
×
dNPWE'( )
2
=MTF 2(u,v)WTask
2 (u,v)òò E 2(u,v)dudv[ ]2
MTF 2(u,v)WTask
2 (u,v)òò NPS(u,v)E 4 (u,v) + MTF 2(u,v)WTask
2 (u,v)Ni dudv
Richard, and E. Samei, Quantitative breast tomosynthesis: from detectability to estimability. Med Phys, 37(12), 6157-65 (2010).
Chen et al., Relevance of MTF and NPS in quantitative CT: towards developing a predictable model of quantitative... SPIE2012
Task-based quality index
Fisher-Hotelling observer (FH)
Non-prewhitening observer (NPW)
NPW observer with eye filter (NPWE)
dFH'( )
2
=MTF 2(u,v)WTask
2 (u,v)
NPS(u,v)òò dudv
dNPW'( )
2
=MTF 2(u,v)WTask
2 (u,v)òò dudv[ ]2
MTF 2(u,v)WTask
2 (u,v)òò NPS(u,v)dudv
dNPWE'( )
2
=MTF 2(u,v)WTask
2 (u,v)òò E 2(u,v)dudv[ ]2
MTF 2(u,v)WTask
2 (u,v)òò NPS(u,v)E 4 (u,v) + MTF 2(u,v)WTask
2 (u,v)Ni dudv
F
Iodin
e c
once
ntr
ation
Size
Task function: Task characteristics for detection and estimation
C (r ) =Cpeak(1-
r
R
æ
èçç
ö
ø÷÷
2
)n
Chen, SPIE 2013
7/21/2014
10
GE MBIR Dose Reduction Potential
Large feature
detection task
FBP
0 1 2 3 4 5 6 7 0.5
0.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
Eff dose (mSv)
AU
C (
NP
WE
)
0.2
mSv
3.7
mSv
ACR Acrylic insert MBIR
78%
20%
Richard, Li, Samei, SPIE 2011
FBP
0 1 2 3 4 5 6 7 0.5
0.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
Eff dose (mSv)
AU
C (
NP
WE
)
Small feature
detection task
0.2
mSv
3.7
mSv
ACR 0.5 lp/mm bar pattern
64%
29%
MBIR
Richard, Li, Samei, SPIE 2011
GE MBIR Dose Reduction Potential
Duke-UMD-NIST study
Parameter GE Discovery CT 750 HD
Siemens Flash Philips iCT
kVp 120 120 120
Rotation time 0.5 0.5 0.5
SFOV 50 50 50
DFOV 25 25 25
Recon Algorithm FBP Standard / ASIR50
B31f / I30f Safire 3
B / iDose5
Recon Mode Helical Helical Helical
Collimation 0.625 0.6 0.625
Pitch 0.984 1 0.93
Slice Thickness 1.25 1 1
Dose levels: 20, 12, 7.2, 4.3, 1.6, 0.9 mGy
For anonymity will be referred to as vendor A, B, and C
Christianson, AAPM 2012
7/21/2014
11
Observer study design
Observer study design 3 scanner models
3 dose levels
2 reconstruction algorithms
10 slices
x 5 repeated exams
Total of 900 images
12 expert observers from two institutions
How much can IR reduce dose?
23% dose reduction
Default clinical protocol at 12 mGy
7/21/2014
12
How much can IR reduce dose?
6% dose reduction
Default clinical protocol at 12 mGy
How much can IR reduce dose?
33% dose reduction
Default clinical protocol at 12 mGy
CNR vs observer performance
7/21/2014
13
d’ vs observer performance
30 mm
15 mm
20 mm30 mm
33 mm
48.75 mm
64.5 mm
165 mm
6 mm4 mm
2 mm
8% 13%18%
25%
36%36%
25%
18%
TangoPlus
Our contrast detail phantom
0
10
20
30
40
50
60
70
80
90
100
HU
Tan
go B
lack
+
DM 853
0
DM 852
5
DM 852
0
DM 851
5
DM 851
0
DM 850
5
Vero W
hite
DM 843
0
DM 842
5
DM 979
5
DM 978
5
DM 977
0
DM 976
0
DM97
50
DM 974
0
Tan
go +
Vero Black
Vero Blue
Vero Grey
Vero Clear
Tan
go G
rey
RGD 516
0
FC 720
Durus
White
CT images of the low-contrast phantom
B31s
I31s-5
7/21/2014
14
• Siemens SOMATOM Force
• 4 doses (0.7, 1.4, 2.9, 5.8 mGy)
• 2 slice thicknesses (0.6, 5 mm)
• 4 Recons (FBP, ADMIRE 3-5)
Acquired CT data
FBP, STD = 36 HU
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
500
1000
1500
Spatial Frequency (1/mm)
NP
S (
mm
2H
U2)
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0.1
0.5
1
Spatial Frequency (1/mm)
TT
F
ADMIRE−3, STD = 24 HU
ADMIRE−4, STD = 20 HU ADMIRE−5, STD = 15 HU
FBP
ADMIRE−3
ADMIRE−4
ADMIRE−5
FBP
ADMIRE−3
ADMIRE−4
ADMIRE−5
(a)
(b)
(c)
TTF and NPS across recons
FBP, STD = 36 HU
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
500
1000
1500
Spatial Frequency (1/mm)
NP
S (
mm
2H
U2)
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0.1
0.5
1
Spatial Frequency (1/mm)
TT
F
ADMIRE−3, STD = 24 HU
ADMIRE−4, STD = 20 HU ADMIRE−5, STD = 15 HU
FBP
ADMIRE−3
ADMIRE−4
ADMIRE−5
FBP
ADMIRE−3
ADMIRE−4
ADMIRE−5
(a)
(b)
(c)
CT images across dose and recon Dose
IR S
trength
7/21/2014
15
• Visible groups across:
– Dose (0.74, 1.4, 2.9, 5.8 mGy)
– Slice Thickness (0.6, 1.8, 5 mm)
– Recon (FBP, ADMIRE 3-5)
Insert Counting Experiment
Total number of discernable objects
0.74 1.4 2.9 5.8
0.0
0.0
0.4
1.4
0.0
0.8
1.9
3.2
0.0
0.8
3.0 3.8
0.2
2.0
3.8
5.3
CTDIvol
(mGy)
VIG
sco
re
Slice Thickness: 0.6 mm
0.74 1.4 2.9 5.8
0.0
0.6
3.1
3.6
0.2
2.0
4.6 5
.4
0.1
2.2
5.2
6.0
1.1
3.2
6.0
7.0
CTDIvol
(mGy)
VIG
sco
re
Slice Thickness: 1.8 mm
0.74 1.4 2.9 5.8
1.6
4.4
6.2
8.8
2.4
5.2
6.6
9.4
2.6
5.6 6
.6
9.8
2.9
6.1 6
.9
10
.4
CTDIvol
(mGy)
VIG
sco
re
Slice Thickness: 5 mm
FBP
ADMIRE−3
ADMIRE−4
ADMIRE−5
0.74 1.4 2.9 5.8
0.0
0.0
0.4
1.4
0.0
0.8
1.9
3.2
0.0
0.8
3.0 3.8
0.2
2.0
3.8
5.3
CTDIvol
(mGy)
VIG
sco
re
Slice Thickness: 0.6 mm
0.74 1.4 2.9 5.8
0.0
0.6
3.1
3.6
0.2
2.0
4.6 5
.4
0.1
2.2
5.2
6.0
1.1
3.2
6.0
7.0
CTDIvol
(mGy)
VIG
sco
re
Slice Thickness: 1.8 mm
0.74 1.4 2.9 5.8
1.6
4.4
6.2
8.8
2.4
5.2
6.6
9.4
2.6
5.6 6
.6
9.8
2.9
6.1 6
.9
10
.4
CTDIvol
(mGy)
VIG
sco
re
Slice Thickness: 5 mm
FBP
ADMIRE−3
ADMIRE−4
ADMIRE−5
0.74 1.4 2.9 5.8
0.0
0.0
0.4
1.4
0.0
0.8
1.9
3.2
0.0
0.8
3.0 3.8
0.2
2.0
3.8
5.3
CTDIvol
(mGy)
VIG
sco
re
Slice Thickness: 0.6 mm
0.74 1.4 2.9 5.8
0.0
0.6
3.1
3.6
0.2
2.0
4.6 5
.4
0.1
2.2
5.2
6.0
1.1
3.2
6.0
7.0
CTDIvol
(mGy)
VIG
sco
re
Slice Thickness: 1.8 mm
0.74 1.4 2.9 5.8
1.6
4.4
6.2
8.8
2.4
5.2
6.6
9.4
2.6
5.6 6
.6
9.8
2.9
6.1 6
.9
10
.4
CTDIvol
(mGy)
VIG
sco
re
Slice Thickness: 5 mm
FBP
ADMIRE−3
ADMIRE−4
ADMIRE−5
0.74 1.4 2.9 5.8
0.0
0.0
0.4
1.4
0.0
0.8
1.9
3.2
0.0
0.8
3.0
3.8
0.2
2.0
3.8
5.3
CTDIvol (mGy)
VIG score Slice Thickness: 0.6 mm
0.74 1.4 2.9 5.8
0.0
0.6
3.1
3.6
0.2
2.0
4.6
5.4
0.1
2.2
5.2
6.0
1.1
3.2
6.0
7.0
CTDIvol (mGy)
VIG score Slice Thickness: 1.8 mm
0.74 1.4 2.9 5.8
1.6
4.4
6.2
8.8
2.4
5.2
6.6
9.4
2.6
5.6
6.6
9.8
2.9
6.1
6.9
10.4
CTDIvol (mGy)
VIG score
Slice Thickness: 5 mm
IQ metrics vs human performance
0.5 0.55 0.6 0.65 0.7 0.750.4
0.5
0.6
0.7
0.8
0.9
1
AUCCNR
Ob
serv
er
Pe
rfo
rman
ce
FBP
ADMIRE3
y=0.938*x+0.101, R2 = 0.116
0.65 0.7 0.75 0.8 0.85 0.9 0.95 10.4
0.5
0.6
0.7
0.8
0.9
1
AUCSNR
Ob
serv
er
Pe
rfo
rman
ce
FBP
ADMIRE3
y=1.06*x+−0.182, R2 = 0.706
0.55 0.6 0.65 0.7 0.75 0.80.4
0.5
0.6
0.7
0.8
0.9
1
AUCNPW
Ob
se
rve
r P
erf
orm
an
ce
FBP
ADMIRE3
y=1.84*x+−0.494, R2 = 0.725
0.54 0.56 0.58 0.6 0.62 0.64 0.66 0.680.4
0.5
0.6
0.7
0.8
0.9
1
AUCNPWE
Ob
se
rve
r P
erf
orm
an
ce
FBP
ADMIRE−3
y=2.91*x+−1.05, R2 = 0.77
CNR NPWd '2
NPWEd '2
0.5 0.55 0.6 0.65 0.7 0.750.4
0.5
0.6
0.7
0.8
0.9
1
AUCCNR
Ob
serv
er
Pe
rfo
rman
ce
FBP
ADMIRE3
y=0.938*x+0.101, R2 = 0.116
0.65 0.7 0.75 0.8 0.85 0.9 0.95 10.4
0.5
0.6
0.7
0.8
0.9
1
AUCSNR
Ob
serv
er
Pe
rfo
rman
ce
FBP
ADMIRE3
y=1.06*x+−0.182, R2 = 0.706
0.55 0.6 0.65 0.7 0.75 0.80.4
0.5
0.6
0.7
0.8
0.9
1
AUCNPW
Ob
se
rve
r P
erf
orm
an
ce
FBP
ADMIRE3
y=1.84*x+−0.494, R2 = 0.725
0.54 0.56 0.58 0.6 0.62 0.64 0.66 0.680.4
0.5
0.6
0.7
0.8
0.9
1
AUCNPWE
Ob
se
rve
r P
erf
orm
an
ce
FBP
ADMIRE−3
y=2.91*x+−1.05, R2 = 0.77
7/21/2014
16
Task-Based Detectability Extended to quadratic PL image reconstruction
G. Gang et al. Med Phys 41 (2014)
Detectability Map d’(x,y)
0
0.4
0.6
0.2
Line Pair Detection Task
G. Gang et al. Med Phys 41 (2014)
Cylinde
r
Ellipse Thorax
Similar levels of d’ computed via Fourier
and spatial domain approach (~5-15%
agreement)
Applicable to various objects and imaging
tasks.
Local Fourier metrology applicable to
quadratic PL reconstruction, gives a useful
framework for system design and
optimization.
% difference between Fourier
and spatial domain calculation of d’
Ellipse phantom
Justification of Assumptions:
Local Stationarity and Linearity
FBP
PL
Resolution Tumor characteristics
3D Task function (Wtask)
Size, shape, and contrast of
the nodule being quantified
Noise
3D Noise power spectrum
(NPS)
Texture and magnitude of the
noise
3D Task transfer
function (TTF)
System resolution
with respect to the
contrast of the
object and image
noise
222
2
2
( , , ; , ) ( , , )'
[ ( , , ; ) ( , , )] ( , , )
task
i task
TTF u v w C N W u v w dudvdwe
NPS u v w N N u v w W u v w dudvdw
Estimability index (e’): prediction of the
quantification precision capturing the interaction
between the imaging system, the segmentation
software, and the lesion
Internal noise (Ni)
Inconsistency of the
quantification software due
to placement of random
seeds
7/21/2014
17
LUNGMAN,
Kyoto Kagaku, Japan
Empirical precision measurement
LungVCAR
GE Healthcare
Predicted
precision matches
empirical
precision! 2
_2
1 1
1.96 2 /
2.77 /ˆ
( )2.77 /
( 1)
j Wj true
Wj true
ijk ijn Ki k true
PRC V
V
V VV
n K
Outline
1. Quality Characterization
2. Quality assessment
3. Quality implementation
7/21/2014
18
Task-based assessment metrology Mercury Phantom 3.0
• Diameters matching population cohorts
• Depths consistent with cone angles
• Straight-tapered design enabling evaluation of AEC response to discrete and continuous size transitions
55
Design: Base material
• Polyethylene
– 80 HU @ 120 KVp
– Near patient equivalent
– Affordable
– Easy to machine
56
Design: Size Pediatric representation percentages
MP 3.0 section
Water size equivalent
Abdomen Chest Head
Age Percentile Age Percentile Age Percentile
120 mm 112 mm 0 12 0 50
0 5 7 5 5.5 5
185 mm 177 mm 6 50 10 50 3 95
15 5 16 5 12 50
230 mm 220 mm
3 95 8 95
- - 12 50 16 50 21 5
300 mm 290 mm 12 95
19 95 - - 21 50
370 mm 355 mm 20 95 - - - - 57
7/21/2014
19
Design: Size Adult representation percentages
58
MP 3.0 section
Water size equivalent
Abdomen Chest Head
M F M F M F
120 mm 112 mm - - - - - -
185 mm 177 mm - - - - 25 75
230 mm 220 mm 0.4 9 0.06 1.4 - -
300 mm 290 mm 27.1 61 14 48 - -
370 mm 355 mm 80 90.3 60 87 - -
59
• Representation of abnormality-relevant HUs
• Sizes large enough for resolution sampling
• Maximum margin for individual assessment
• Iso-radius resolution properties
• Matching uniform section for noise assessment
Design: Resolution, HU, noise
0
1
Non-Stationary Noise and Resolution
MTF
0
2.5 x10-6
NPS The Local NPS and MTF
Noise and resolution model for Penalized Likelihood (PL) model-based reconstruction.* Predictive framework for NPS, MTF, and detectability index (d’) enables task-based design and optimization of new systems using iterative reconstruction.
*Fessler et al. IEEE-TIP (1996) G. Gang et al. Med Phys 41 (2014)
7/21/2014
20
0°
72°
144°
216°
288°
0°
72°
144°
216°
288°
1.15”
1.02” Water
0 HU
Polystyrene
-40 HU
Bone
950 HU
Iodine
8.5 mgI/cc
225 HU
Air
-1000 HU
0.56” 0.53”
Sections 2-5 Smallest Section 1
Water
0 HU
Polystyrene
-40 HU
Bone
950 HU
Iodine
8.5 mgI/cc
225 HU
Air
-1000 HU
Design: Resolution, HU, noise
1.0”
imQuest
HU, Contrast, Noise, CNR, MTF, NPS, and d’ per patient size, mA modulation profile
(image quality evaluation software)
Noise texture effect
0%
10%
20%
30%
40%
50%
60%
70%
80%
10 20 30 40 50 100 150
% N
ois
e R
ed
uc
tio
n F
rom
IR
mAs
Water Water + Sponge Water + Acrylic IR noise reduction:
65% in uniform bkgd
20% in textured bkgd
Solomon, SPIE 2012
7/21/2014
21
Mercury phantom 3.0 Added contrast detail and texture
Generation 1Generation 2Generation 3Generation 4Generation 5Generation 6
Textured lung phantom
Textured soft-tissue phantom: clustered lumpy background*
*Bochud, F., Abbey, C., & Eckstein, M. (1999). Statistical texture synthesis of mammographic images with super-blob
lumpy backgrounds. Optics express, 4(1), 33–42. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/19396254
7/21/2014
22
*Bochud, F., Abbey, C., & Eckstein, M. (1999). Statistical texture synthesis of mammographic images with super-blob
lumpy backgrounds. Optics express, 4(1), 33–42. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/19396254
Textured soft-tissue phantom: clustered lumpy background*
Anatomically informed heterogeneity phantoms
30 mm
165 mm
30 mm
165 mm
68
Lung texture phantom
30 mm
165 mm
69
7/21/2014
23
Soft-tissue texture phantom
30 mm
165 mm
70
How was the data analyzed?
= +
= STD
Noise in the uniform phantom
0 30 STD (HU)
0 10 20 30
STD (HU)
0 10 20 30
STD (HU)
FBP
IR
7/21/2014
24
Noise in the soft-tissue phantom
0 30 STD (HU)
0 10 20 30
STD (HU)
0 10 20 30
STD (HU)
FBP
IR
Noise in the lung phantom
0 30 STD (HU)
0 10 20 30
STD (HU)
0 10 20 30
STD (HU)
FBP
IR
Comparing noise across textures
0 10 20 30
STD (HU)
0 10 20 30
STD (HU)
0 10 20 30
STD (HU)
0 10 20 30
STD (HU)
0 10 20 30
STD (HU)
0 10 20 30
STD (HU)
FBP
IR
7/21/2014
25
What about noise texture?
FBP IR
0 0.2 0.4 0.6 0.8 10
50
100
150
200
250
300
Spatial Frequency (cycles/mm)
NP
S (
HU
2*m
m2)
Lung Texture NPS
FBP: Red Pixels
IR: Red Pixels
IR: Blue Pixels
0 0.2 0.4 0.6 0.8 10
0.2
0.4
0.6
0.8
1
1.2
1.4
Spatial Frequency (cycles/mm)
nN
PS
(m
m2)
Lung Texture NPS
FBP: Red Pixels
IR: Red Pixels
IR: Blue Pixels
Outline
1. Quality Characterization
2. Quality assessment
3. Quality implementation
Components of quality assurance
Retrospective performance assessment Quality by outcome
Prospective protocol definition Quality by prescription
System performance assessment Quality by inference
7/21/2014
26
0 2 4 6 8 10 0.5
0.6
0.7
0.8
0.9
1
Eff dose (mSv)
AZ
Feature with no iodine
kV IR optimization
0 2 4 6 8 10 0.5
0.6
0.7
0.8
0.9
1
Eff dose (mSv)
AZ
Feature with iodine
Task Optimal technique % dose reduction (wrt 120 kVp FBP)
No Iodine 80/100 kVp with IRIS 36%
With Iodine 80 kVp with IRIS 40%
ACR phantom
Samei, Richard, Med Phys, in press, 2014
De
tect
abili
ty t
ren
ds
w
ith
do
se/s
ize
80 Smitherman, AAPM, 2014
Protocol optimization
• Setting dose to achieve a targeted task performance for a given size patient
Smitherman, AAPM, 2014
7/21/2014
27
Protocol optimization
• Setting dose to achieve a targeted task performance for a given size patient
Smitherman, AAPM, 2014
PRC: Relative difference between any two repeated
quantifications of a nodule with 95% confidence
Quality-dose dependency Quantitative volumetry via CT
Noise texture vs kernel
GE Siemens
Solomon, Samei, Med Phys, 2012
7/21/2014
28
Texture similarity
Sharpness Solomon, Samei, Med Phys, 2012
CT image quality monitoring
GNI
0
20
40
60
80
100
0 20 40 60 80 100
GN
I
Subtraction Method
y = x R2 = 0.998
Christianson, AAPM, 2014
Noise per slice FBP
ASiR
7/21/2014
29
0
5
10
15
20
25
30
35
40
45
50
20 30 40
GN
L (H
U)
Effective Diameter (cm)
Scanner 1
Scanner 2
Scanner 3
0
5
10
15
20
25
30
35
40
45
50
20 30 40
SSD
E (m
Gy)
Effective Diameter (cm)
Scanner 1
Scanner 2
Scanner 3
Trends in dose and noise
Abdomen Pelvis Exams (n=2358)
Christianson, AAPM, 2014
0
5
10
15
20
25
30
35
40
45
50
20 30 40
GN
L (H
U)
Effective Diameter (cm)
Scanner 1
Scanner 2
Scanner 3
0
5
10
15
20
25
30
35
40
45
50
20 30 40
SSD
E (m
Gy)
Effective Diameter (cm)
Scanner 1
Scanner 2
Scanner 3
Trends in dose and noise
Abdomen Pelvis Exams (n=2358)
SSDE varies with size
Christianson, AAPM, 2014
0
5
10
15
20
25
30
Scanner 1 Scanner 2 Scanner 3
SSD
E (m
Gy)
0
5
10
15
20
25
30
Scanner 1 Scanner 2 Scanner 3
GN
L (H
U)
Variability in dose and noise
Target noise level
ACR DIR
25th -
75th
Abdomen Pelvis Exams (n=2358)
Christianson, AAPM, 2014
7/21/2014
30
Conclusions
• New technologies necessitate an upgrade to
performance metrology towards higher degrees of
clinical relevance:
• Physical surrogates of clinical performance
• “Taskful” metrology and dependencies
• Incorporation of texture into image quality
estimation
• Extension of quality metrology to quality
monitoring and quality analytics
Thank you!