7/21/2014
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
1
7/21/2014
Credits
Jeff Siewerdsen, Johns Hopkins
Grace Gang, Johns Hopkins
Guanghong Chen, UW
Ke Li, UW
Duke CIPG
Clinical Imaging Physics Group
4
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
2
7/21/2014
Optimization
framework
n
Be
it
ef
Quality indices
(d’, Az, …)
Different patients
and indications
e
gim
re
Optimization regime
Scan factors
(mAs, kVp, pitch, recon,
kernel, …)
sk
Ri
e
gim
re
Safety indices
(organ dose,
eff. dose, …)
Outline
1. Quality Characterization
2. Quality assessment
3. Quality implementation
3
7/21/2014
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
ASiR (GE)
• Limited utility of prior quality metrics
Veo (GE)
• Need for nuanced implementationIRISfor
(Siemens)
effective improvement in patientSAFIRE
care(Siemens)
AIDR (Toshiba)
iDose (Philips)
…
FBP
Reconstruction
Iterative
Reconstruction
Noise
texture?
IR
Images courtesy of Dr de Mey and Dr Nieboer, UZ Brussel, Belgium
FBP
Low Dose CT @ 114 DLP, 1.9 mSv
4
7/21/2014
FBP
Reconstruction
Iterative
Reconstruction
courtesy of University of Erlangen, Germany
FBP
Reconstruction
Iterative
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
5
7/21/2014
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
sB
Parameters that affect IQ
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
Contrast
Lesion size
Lesion shape
Lesion edge profile
Resolution
Viewing distance
Display
Noise magnitude
Noise texture
Operator noise
Feature of
interest
Image details
Distractors
Parameters that are measured
by CNR
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
Contrast
Lesion size
Lesion shape
Lesion edge profile
Resolution
Viewing distance
Display
Noise magnitude
Noise texture
Operator noise
1. Contrast
Feature of
interest
Image details
?
8. Noise magnitude
Distractors
6
7/21/2014
Why CNR is not enough:
Noise texture
FBP
IR
Solomon, AAPM 2012
Resolution and noise, eg 1
Lower noise but
different texture
Comparable
resolution
x 10
1
-FBP
-IRIS
-SAFIRE
0.8
-6
-FBP
-IRIS
-SAFIRE
2.5
2
NPS
MTF
0.6
1.5
0.4
1
0.2
0
0.5
0
0.2
0.4
0.6
Spatial frequency
0
0.8
0
0.2
0.4
0.6
Spatial frequency
0.8
Resolution and noise, eg 2
Lower noise but
different texture
Higher resolution
x 10-5
1
0.5
0.9
0.45
0.8
0.4
0.7
0.35
0.3
0.5
NPS
MTF
0.6
-MBIR
-ASIR
-FBP
0.4
0.3
0.25
0.2
0.15
0.2
0.1
0.1
0.05
0
-MBIR
-ASIR
-FBP
0
0.2
0.4
0.6
Spatial frequency
0.8
0
0
0.2
0.4
Spatial frequency
0.6
0.8
7
7/21/2014
NPS vs dose
Nonlinearity
NPS (Veo)
Normalized NPS (Veo)
800
6
500
2
2
NPS (HU mm )
600
400
Normalized NPS (A.U.)
5 mAs
12.5 mAs
25 mAs
50 mAs
100 mAs
200 mAs
300 mAs
700
300
200
100
0
0.2
0.4
0.6
4
3
2
1
0
0.8
0
0.2
0.4
fxy (mm-1)
0.6
0.8
fxy (mm-1)
Li, Tang, and Chen, “Statistical Model Based Iterative Reconstruction (MBIR) in clinical CT
systems: Experimental assessment of noise performance,” Med. Phys. (2014)
NPS peak frequency vs dose
FBP
FBP (fitting)
Veo
Veo (fitting)
-1
Peak frequency (mm )
0.8
0.4
0.2
0.1
fpeak  (mAs)0.16
0.05
0.5
1.25
2.5
5
10
20
30
mAs
Li, Tang, and Chen, “Statistical Model Based Iterative Reconstruction (MBIR) in clinical CT
systems: Experimental assessment of noise performance,” Med. Phys. (2014)
Resolution’s joint dependence on contrast and dose
1
Veo
0.8
PSF width (mm)
0
5 mAs
12.5 mAs
25 mAs
50 mAs
100 mAs
200 mAs
300 mAs
5
0.8
0.7
FBP
0.6
0.6
0.5
0.4
0.4
0.2
16
100%
33
75%
62 99
224 346
Contrast (HU)
50%
814 1710
25%
Dose
8
7/21/2014
Detectability index
Resolution and
contrast transfer
Attributes of image
feature of interest
×
Image noise magnitude
and texture
(d )
2
'
NPWE
[ òò MTF (u,v)W
2
=
òò MTF (u,v)W
2
2
Task
2
Task
(u,v)E 2 (u,v)dudv
]
2
2
(u,v)NPS(u,v)E 4 (u,v) + MTF 2 (u,v)WTask
(u,v)N i 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)
(d )
2
'
FH
=
òò
2
MTF 2 (u,v)WTask
(u,v)
dudv
NPS(u,v)
Non-prewhitening observer (NPW)
'
(dNPW
) =
2
[ òò MTF
òò MTF
2
2
2
(u,v)WTask
(u,v)dudv
]
2
2
(u,v)WTask
(u,v)NPS(u,v)dudv
NPW observer with eye filter (NPWE)
(d )
2
'
NPWE
[ òò MTF (u,v)W
2
=
òò MTF (u,v)W
2
2
Task
2
Task
(u,v)E 2 (u,v)dudv
]
2
2
(u,v)NPS(u,v)E (u,v) + MTF (u,v)WTask
(u,v)N i dudv
4
2
2
Task function:
ær ö
C (r ) = C peak (1 - çç ÷÷ )n
èR ø
Task characteristics for detection and estimation
Iodine concentration
Size
F
Chen, SPIE 2013
9
7/21/2014
GE MBIR Dose Reduction Potential
Large feature
detection task
1
20%
0.9
0.85
AUC (NPWE)
ACR Acrylic insert
MBIR
0.95
78%
0.8
FBP
0.75
0.7
0.65
0.6
0.55
0.5
0
1
2
3
4
5
Eff dose (mSv)
6
7
3.7
mSv
0.2
mSv
Richard, Li, Samei, SPIE 2011
GE MBIR Dose Reduction Potential
Small feature
detection task
1
29%
0.9
AUC (NPWE)
ACR 0.5 lp/mm bar pattern
MBIR
0.95
0.85
64%
0.8
0.75
FBP
0.7
0.65
0.6
0.55
0.5
0
1
2
3
4
5
Eff dose (mSv)
6
7
3.7
mSv
0.2
mSv
Richard, Li, Samei, SPIE 2011
Christianson, AAPM 2012
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
10
7/21/2014
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?
Default clinical protocol at 12 mGy
23% dose reduction
11
7/21/2014
How much can IR reduce dose?
Default clinical protocol at 12 mGy
6% dose reduction
How much can IR reduce dose?
Default clinical protocol at 12 mGy
33% dose reduction
CNR vs observer
performance
12
7/21/2014
d’ vs observer performance
Our contrast detail phantom
64.5 mm
25%
2 mm
48.75 mm
33 mm
20 mm
30 mm
30 mm
15 mm
18%
4 mm
6 mm
36%
25%
HU
100
90
36% 80
70
60
50
40
30
20
10
0
18%
8%
13%
Bla
c
DM k +
8
DM 530
8
DM 525
8
DM 520
8
DM 515
8
DM 510
Ve 850
5
ro
W
DM hite
8
DM 430
8
DM 425
9
DM 795
9
DM 785
9
DM 770
97
DM 6 0
9
DM 750
97
Ta 40
n
Ve go
+
ro
B
Ve lack
ro
Ve Blue
ro
Ve Gre
y
ro
Ta Cle
ng
ar
o
RG Gre
y
D
51
60
F
Du C 7
2
ru
sW 0
hit
e
165 mm
Ta
ng
o
TangoPlus
CT images of the low-contrast phantom
B31s
I31s-5
13
7/21/2014
Acquired CT data
•
•
•
•
Siemens SOMATOM Force
4 doses (0.7, 1.4, 2.9, 5.8 mGy)
FBP, STD = 36 HU
ADMIRE−3, STD = 24 HU
2 slice thicknesses (0.6, 5 mm)
4 Recons (FBP, ADMIRE 3-5)
ADMIRE−4, STD = 20 HU
ADMIRE−5, STD = 15 HU
(a)
FBP, STD = 36 HU
NPS (mm2HU2)
1500
ADMIRE−3, STD = 24 HU
FBP
ADMIRE−3
ADMIRE−4
ADMIRE−5
1000
500
TTF and 0NPS
across recons
0
0.1 0.2 0.3 0.4 0.5 0.6
0.7 0.8 0.9
1
Spatial Frequency (1/mm)
ADMIRE−4, STD = 20 HU
ADMIRE−5, STD = 15 HU
(b)
TTF
1
FBP
ADMIRE−3
ADMIRE−4
ADMIRE−5
0.5
0.1
(a)
0
0.1
0.2
0.3 0.4 0.5 0.6 0.7
Spatial Frequency (1/mm)
0.8
0.9
1
NPS (mm2HU2)
(c) 1500
FBP
ADMIRE−3
ADMIRE−4
ADMIRE−5
1000
500
0
0
(b)
0.1
0.2
0.3 0.4 0.5 0.6 0.7
Spatial Frequency (1/mm)
0.8
0.9
1
TTF
1
FBP
ADMIRE−3
ADMIRE−4
ADMIRE−5
0.5
0.1
0
0.1
0.2
0.3 0.4 0.5 0.6 0.7
Spatial Frequency (1/mm)
0.8
0.9
1
(c)
CT images across dose and recon
IR Strength
Dose
14
7/21/2014
Insert Counting Experiment
• Visible groups across:
– Dose (0.74, 1.4, 2.9, 5.8
mGy)
– Slice Thickness (0.6,
1.8, 5 mm)
– Recon (FBP, ADMIRE 35)
Total number of discernable objects
0.8
0.8
2.0
1.9
0.74
0.74
0.0
0.2
0.1
6.2
6.6
6.6
5.2 6.9
5.6
6.1
5.2
5.6
6.1
4.4
4.4
2.4
2.6
2.9
2.4 VIG score
2.6
2.9
1.6
1.6
6.0
7.0
3.6
5.4
VIG score
6.0
4.6
5.2
2.0
2.2
3.2
3.6
3.1
1.1
0.6
4.4
5.8
6.0
7.0
score
0.6
5.3
VIG score
6.0
5.4
VIG
2.4
2.6
2.9
1.4
0.74 2.91.4
5.82.9
CTDIvol (mGy)
CTDIvol (mGy)
2.0
2.2
3.2
1.1
3.1
4.6
5.2
3.6
0.0
0.2
0.1
0.0
0.0
0.0
0.2
1.6
3.1
score
FBP
ADMIRE−3
ADMIRE−4
ADMIRE−5
SliceThicknes :5m
0.0
0.8
0.0 0.8
0.0
2.0
0.0
0.2
0.4
1.9
0.0
3.0
0.8
3.8
0.8
2.0
1.4
3.2
0.4
3.8
1.9
5.3
3.0
3.8
VIG score
5.8
2.0
2.2
3.2
5.4
6.0
4.6
7.0
5.2
6.0
5.3
VIG
1.1
0.6
5.2
5.6
6.1
6.2
6.6
6.6
6.9
8.8
9.4
9.8
10.4
1.4
0.74 2.91.4
5.82.9
CTDIvol (mGy)
CTDIvol (mGy)
5.8
0.74
1
1.4
0.74 2.91.4
5.82.9
CTDIvol (mGy)
CTDIvol (mGy)
5.8
1
FBP
ADMIRE3
y=0.938*x+0.101, R2 = 0.116
0.9
0.9
FBP
ADMIRE3
y=1.06*x+−0.182, R 2 = 0.706
Observer Performance
Observer Performance
IQ metrics vs human performance
0.8
0.7
0.6
0.5
0.55
0.7
0.75
0.4
0.4
0.65
0.55
0.9
0.95
1
0.64
0.66
0.68
2
NPWE
FBP
ADMIRE−3
y=2.91*x+−1.05, R2 = 0.77
0.8
0.7
0.6
0.5
0.7
0.6
0.75 0.650.8
0.85
0.7
AUCSNR
AUC
0.90.75 0.95
0.81
0.64
0.68
NPW
1
0.4
0.54
0.56
0.58
0.6
0.62
AUCNPWE
1
FBP
ADMIRE3
y=1.84*x+−0.494, R 2 = 0.725
Observer Performance
0.9
0.8
0.7
0.6
0.5
0.4
0.55
0.9
0.5
0.5
0.6
0.65
AUCCNR
0.8
0.85
AUCSNR
1
FBP
FBP
ADMIRE3
ADMIRE3
2
y=1.84*x+−0.494,
y=1.06*x+−0.182,RR 2= =0.725
0.706
0.6
0.6
0.55
0.75
d'
0.7
0.7
0.5
0.7
2
0.8
0.8
0.6
0.6
NPW
Observer Performance
0.9
0.9
0.7
0.7
0.4
0.65
0.75
Observer Performance
Observer Performance
Observer Performance
FBP
ADMIRE3
y=0.938*x+0.101, R2 = 0.116
0.8
0.9
0.7
11
1
0.4
0.5
0.6
0.65
AUCCNR
d'
CNR
0.9
0.8
0.5
0.4
0.5
Observer Performance
1.4
2.9
CTDIvol (mGy)
3.0
3.8
3.2
3.8
FBP
ADMIRE−3
ADMIRE−4
ADMIRE−5
SliceThicknes :1.8m
VIG score
1.4
0.0
0.2
0.1
VIG score
2.4
2.6
2.9
0.4
3.2
3.8
0.0
1.4
4.4
Slice Thickness:
Slice5Thickness:
mm
5 mm
score
8.8
9.4
9.8
10.4
Slice Thickness:
Slice1.8
Thickness:
mm
1.8 mm
VIG
8.8
9.4
9.8
10.4
Slice Thickness:
Slice0.6
Thickness:
mm
0.6 mm
0.0
0.0
0.0
0.2
SliceThicknes :0.6m
0.74 1.4 2.9 5.8 0.74 1.4 2.9 5.8 0.74 1.4 2.9 5.8
0.74
FBP
ADMIRE−3
ADMIRE−4
ADMIRE−5
6.2
6.6
6.6
6.9
6.2
6.6
6.6
6.9
5.2
5.6
6.1
8.8
9.4
9.8
10.4
Slice Thickness: 5 mm
FBP
ADMIRE−3
y=2.91*x+−1.05, R2 = 0.77
0.8
0.7
0.6
0.5
0.6
0.65
0.7
AUCNPW
0.75
0.8
0.4
0.54
0.56
0.58
0.6
0.62
AUCNPWE
0.66
15
7/21/2014
Task-Based Detectability
Extended to quadratic PL image reconstruction
Line Pair
Detection Task
0.6
Detectability Map
d’(x,y)
0.4
0.2
0
G. Gang et al. Med Phys 41 (2014)
Justification of Assumptions:
Local Stationarity and Linearity
% difference between Fourier
and spatial domain calculation of d’ Cylinde
r
Ellipse phantom
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.
FBP
PL
G. Gang et al. Med Phys 41 (2014)
Estimability index (e’): prediction of the
quantification precision capturing the interaction
between the imaging system, the segmentation
software, and the lesion
2
 TTF 2 (u ,v ,w ;C , N )  W (u ,v ,w ) 2 dudvdw 
task
 

e' 
2
[
NPS
(
u
,
v
,
w
;
N
)

N
(
u
,
v
,
w
)]  Wtask (u ,v ,w ) dudvdw
i

2
Noise
3D Noise power spectrum
(NPS)
Texture and magnitude of the
noise
Internal noise (Ni)
Inconsistency of the
quantification software due
to placement of random
seeds
Resolution
3D Task transfer
function (TTF)
System resolution
with respect to the
contrast of the
object and image
noise
Tumor characteristics
3D Task function (Wtask)
Size, shape, and contrast of
the nodule being quantified
16
7/21/2014
Empirical precision measurement
LungVCAR
GE Healthcare
LUNGMAN,
Kyoto Kagaku, Japan
Predicted
precision matches
empirical
precision!
PRC j  1.96 2Wj2 /Vtrue
 2.77ˆWj /Vtrue
_
 2.77
 ni  Kk
1
(V ijk V ij )2
1
n (K  1)
/Vtrue
Outline
1. Quality Characterization
2. Quality assessment
3. Quality implementation
17
7/21/2014
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
120 mm
112 mm
185 mm
177 mm
230 mm
220 mm
300 mm
290 mm
370 mm
355 mm
Abdomen
Chest
Head
Age
Percentile
Age
Percentile
0
12
0
50
7
5
5.5
5
Age
Percentile
0
5
6
50
10
50
3
95
15
5
16
5
12
50
3
95
8
95
12
50
-
-
21
5
16
50
12
95
21
50
19
95
-
-
20
95
-
-
-
-
57
18
7/21/2014
Design: Size
Adult representation percentages
MP 3.0
section
Water size
equivalent
M
Abdomen
F
M
Chest
F
M
Head
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
-
58
Design: Resolution, HU, noise
•
•
•
•
•
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
59
Non-Stationary Noise and Resolution
NPS
The Local NPS and MTF
2.5
x10-6
0
MTF
1
0
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)
19
7/21/2014
Design: Resolution, HU, noise
72°
72°
Air
-1000 HU
144°
Water
0 HU
Air
-1000 HU
144°
1.02”
Water
0 HU
1.15”
0.53”
0.56”
Iodine 0°
8.5 mgI/cc
225 HU
Iodine 0°
8.5 mgI/cc
225 HU
Bone
950 HU
1.0”
Polystyrene
-40 HU
216°
216°
Bone
950 HU
Polystyrene
-40 HU
288°
288°
Sections 2-5
Smallest Section 1
imQuest
(image quality evaluation software)
HU, Contrast, Noise, CNR, MTF, NPS, and d’ per patient size, mA modulation profile
% Noise Reduction From IR
Noise texture effect
Water
Water + Sponge
Water + Acrylic
80%
70%
60%
50%
40%
30%
20%
10%
0%
10
20
30
40
mAs
50
100
IR noise reduction:
65% in uniform bkgd
20% in textured bkgd
150
Solomon, SPIE 2012
20
7/21/2014
Mercury phantom 3.0
Added contrast detail and texture
Textured lung phantom
Generation 6
1
2
3
4
5
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
21
7/21/2014
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
Anatomically informed
heterogeneity phantoms
30 mm
30 mm
165 mm
165 mm
68
Lung texture phantom
30 mm
165 mm
69
22
7/21/2014
Soft-tissue texture phantom
30 mm
165 mm
70
How was the data analyzed?
=
+
=
STD
FBP
Noise in the uniform phantom
0
10
20
30
20
30
IR
STD (HU)
0
10
STD (HU)
0
STD (HU)
30
23
7/21/2014
FBP
Noise in the soft-tissue phantom
0
10
20
30
20
30
20
30
20
30
IR
STD (HU)
0
10
STD (HU)
0
STD (HU)
30
FBP
Noise in the lung phantom
0
10
IR
STD (HU)
0
10
STD (HU)
0
STD (HU)
30
FBP
Comparing noise across textures
0
10
20
30
0
10
20
30
0
10
STD (HU)
20
30
20
30
STD (HU)
IR
STD (HU)
0
10
20
STD (HU)
30
0
10
20
STD (HU)
30
0
10
STD (HU)
24
7/21/2014
What about noise texture?
Lung Texture NPS
FBP
300
1.4
2
2 2)
NPS
(HU(mm
*mm
nNPS
)
IR
FBP: Red Pixels
IR: Red Pixels
IR: Blue Pixels
1.2
250
1
200
0.8
150
0.6
100
0.4
50
0.2
0
0
0.2
0.4
0.6
0.8
Spatial Frequency (cycles/mm)
1
Outline
1. Quality Characterization
2. Quality assessment
3. Quality implementation
Components of quality assurance
System performance assessment
Quality by inference
Prospective protocol definition
Quality by prescription
Retrospective performance assessment
Quality by outcome
25
7/21/2014
kV IR optimization
ACR phantom
Task
Optimal technique
% dose reduction
(wrt 120 kVp FBP)
No Iodine
80/100 kVp with IRIS
36%
With Iodine
80 kVp with IRIS
40%
Feature with iodine
Feature with no iodine
1
0.9
0.9
0.8
0.8
AZ
AZ
1
0.7
0.7
0.6
0.5
0.6
0
2
4
6
8
10
Eff dose (mSv)
0.5
0
2
4
6
8
10
Eff dose (mSv)
Detectability trends
with dose/size
Samei, Richard, Med Phys, in press, 2014
Smitherman, AAPM, 2014
80
Protocol optimization
• Setting dose to achieve a targeted task
performance for a given size patient
Smitherman, AAPM, 2014
26
7/21/2014
Protocol optimization
• Setting dose to achieve a targeted task
performance for a given size patient
Smitherman, AAPM, 2014
Quality-dose dependency
Quantitative volumetry via CT
PRC: Relative difference between any two repeated
quantifications of a nodule with 95% confidence
Noise texture vs kernel
GE
Siemens
Solomon, Samei, Med Phys, 2012
27
7/21/2014
Texture similarity
Sharpness
Solomon, Samei, Med Phys, 2012
CT image quality monitoring
GNI
100
GNI
80
y=x
R2 = 0.998
60
40
20
0
Christianson, AAPM, 2014
FBP
0 20 40 60 80 100
Subtraction Method
Noise per slice
ASiR
28
7/21/2014
Trends in dose and noise
Abdomen Pelvis Exams (n=2358)
50
Scanner 1
45
Scanner 2
Scanner 1
45
Scanner 2
40
Scanner 3
35
35
30
30
GNL (HU)
SSDE (mGy)
40
50
25
20
25
20
15
15
10
10
5
5
0
Scanner 3
0
20
30
40
Effective Diameter (cm)
20
30
40
Effective Diameter (cm)
Christianson, AAPM, 2014
Trends in dose and noise
Abdomen Pelvis Exams (n=2358)
50
Scanner 1
45
Scanner 2
Scanner 1
45
Scanner 2
40
Scanner 3
35
35
30
30
GNL (HU)
SSDE (mGy)
40
50
25
20
15
Scanner 3
25
20
15
10
10
5
SSDE varies with size
0
20
5
0
30
40
Effective Diameter (cm)
20
30
40
Effective Diameter (cm)
Christianson, AAPM, 2014
Variability in dose and noise
Abdomen Pelvis Exams (n=2358)
30
ACR DIR
25th 75th
25
25
Target noise level
20
GNL (HU)
SSDE (mGy)
20
30
15
15
10
10
5
5
0
0
Scanner 1
Scanner 2
Scanner 3
Scanner 1
Scanner 2
Scanner 3
Christianson, AAPM, 2014
29
7/21/2014
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!
30