6/30/2008
Factors Affecting Accurate
Quantification In PET/CT
Osama Mawlawi, Ph.D.
Department of Imaging Physics
MD Anderson Cancer Center
53 year old man with lymphoma with increased FDG uptake in the
spleen and multiple intra and extra-thoracic nodes consistent with
Multi-focal malignancy.
O. Mawlawi MDACC
Power of PET : Quantification
O. Mawlawi MDACC
Quantification: Power of PET
Measured Data
Random subtract
Normalize
Dead time
Correct Geometry
Calculate/subtract scatter
Correct Attenuation
FBP or IR reconstruction
O. Mawlawi MDACC
calibration
Ideal measured data
O. Mawlawi MDACC
1
6/30/2008
SUV
SUV =
SUV Modifiers
decay-corrected dose/ml of tumor
injected dose/patient weight in grams
• SUVbw (gm/ml)
• The standardized uptake
p
value ((SUV)) is
one means of making PET results more
quantitative.
• Tracer kinetic modeling is used
extensively in quantitative PET (may
include blood samples and dynamic
imaging).
• SUVBSA (m2/ml)
• SUVLBM (gm/ml)
O. Mawlawi MDACC
O. Mawlawi MDACC
Quantitative PET Performance
SUV Modifiers
3 categories that “might” affect SUV measurements
• Patient Compliance
• SUVig (gm/ml)
–
–
•
Scan Conditions
–
–
–
–
•
Scan time post injection
Patient anxiety/comfort during uptake (room temperature, etc.)
Patient motion during acquisition
PET/CT vs dedicated PET
Intrinsic System Parameters and Capability
–
–
–
–
O. Mawlawi MDACC
Fasting
Blood Glucose levels
Calibration
QA – Maintenance of operating parameters
Performance characteristics of scanners – Partial Volume
effects
Image processing algorithms
O. Mawlawi MDACC
2
6/30/2008
Improper Prep. Non fasting
Fast Track Insulin Scan
• High protein/low carbs.
• 6-8 glasses of water
• 6 hrs fasting
• Avoid strenuous exercise
• 70 ––150 mg/dL – Ideal
• 150 ––200 mg/dL - clinical decision
• > 200 mg/dL - cancel
O. Mawlawi MDACC
O. Mawlawi MDACC
Muscle Uptake
Quantitative PET Performance
3 Categories that “might” affect SUV measurements
• Patient Compliance
–
–
•
Scan Conditions
–
–
–
–
•
Crutch Utilized Left Side
Scan time post injection
Patient anxiety/comfort during uptake (room temperature, etc.)
Patient motion during acquisition
PET/CT vs Dedicated PET
Intrinsic System Parameters and Capability
–
–
–
–
O. Mawlawi MDACC
Fasting
Blood Glucose levels
Calibration
QA – Maintenance of operating parameters
Performance characteristics of scanners – Partial Volume
Effects
Image processing algorithms
O. Mawlawi MDACC
3
6/30/2008
Effect of Scan time post injection: Beast Cancer
PET-CT Normal “Brown” Fat
N=20
Max SUV 15.7
Ax CT
Fused
Cor MIP
Beaulieu S et al JNM 2003
Motion Effects
Coronal
O. Mawlawi MDACC
Phantom Study
Sagital
Ax PET
O. Mawlawi MDACC
Patient Movement
Mip
Axial
Sphe re (volume )
1 (16.5 cc)
Gate d
31624
Static
33546
2 (8.4 cc)
31316
21872
30.2
3 (4 cc)
4 (2 cc)
5 (1 cc)
6 (0.5 cc)
29919
22857
19087
11897
20607
13643
8264
5143
31.1
40.3
56.7
56.8
Patient
movement
occurred
after the CT
image was
acquired.
Error (%)
-6.1
O. Mawlawi MDACC
O. Mawlawi MDACC
4
6/30/2008
Impact of Whole-body
Respiratory Gated PET/CT
PET/CT vs Dedicated PET
Disadvantages of dedicated PET
imaging techniques
• Transmission
– Noise due to low g
gamma ray
y flux from rod source
– Transmission is contaminated by emission data
• Scan duration
– Time consuming (emission & transmission )
– Increased patient movement (image blurring)
Static wholebody
Single respiratory phase
(1 of 7, so noisier)
1 cc lesion on CT
The max SUV of the lesion goes from 2 in the static image to 6 in the
respiratory-gated image sequence
Courtesy of P Kinahan, UW
• Efficiency
– Decreased patient throughput
– Difficulty in accurately correlating images to other
diagnostic modalities
O. Mawlawi MDACC
O. Mawlawi MDACC
Short duration, low noise CT-based attenuation correction
Attenuation Measurement
Rod or Pin Source (Ge-68, T½= 271 days)
Patient Transmission Scan
O. Mawlawi MDACC
O. Mawlawi MDACC
5
6/30/2008
PET-CT
Converting CT Numbers to
Attenuation Values
Attenuation
Correction
• The green line shows the
effect of using water scaling
for all materials
Inherent
Image
Registration
0.200
Attenua
ation at 511keV
• For CT values < 0 ,
materials are assumed to
have an energy dependence
similar to water
• For CT values > 0, material
is assumed to have an
energy dependence similar
to a mixture of bone and
water
0.150
Bone/Water
0.100
0.050
0.000
-1000
Water/air
-500
0
500
1000
CT number measured at 140kVp
O. Mawlawi MDACC
O. Mawlawi MDACC
Breathing ArtifactCurvilinear Cold Areas
PET/CT imaging artifacts are due to :
Photopenic Artifact
„
„
„
„
Respiratory motion
Metal implants
p
Truncation
Contrast media
CT
O. Mawlawi MDACC
PET
FUSEDIMAGE
NON AC
O. Mawlawi MDACC
6
6/30/2008
Breathing Artifact
Mismatch between PET and CT – Cardiac Application
Mismatch of lesion location between helical CT and PET
Inaccurate attenuation correction Î inaccurate quantification
Courtesy of J. Brunetti, Holly name
O. Mawlawi MDACC
O. Mawlawi MDACC
Clinical Studies
Average CT – 4D CT
X-ray on
off
1st table position
(0 - 2cm)
2nd table position
(2 - 4 cm)
EI 0%
3rd table position
(4 - 6 cm)
EI 0%
Mismatch 1:
Respiratory singal
EE 50%
EE 50%
CT diaphragm
position lower
than PET
+57%
CT iimages
4D-CT images
0
Helical CT
10
20
Helical CT
without thorax
30
40
50
60
70
80
Mismatch 2:
90%
Combined helical and 4DCT averaging
respiration-averaged CT
CT diaphragm
position higher
than PET
1
Image averaging
Respiration-averaged CT
Courtesy of Tinsu Pan, MDACC
2
O. Mawlawi MDACC
Courtesy of Tinsu Pan, MDACC
O. Mawlawi MDACC
7
6/30/2008
4D PET/CT
4D-PET/CT
O. Mawlawi MDACC
Phantom Study
METAL & CONTRAST artifacts
„
Coronal
Sagital
Mip
Axial
„
„
Sphe re (volume )
1 (16.5 cc)
Gate d
31624
Static
33546
Error (%)
-6.1
2 (8.4 cc)
31316
21872
30.2
3 (4 cc)
4 (2 cc)
5 (1 cc)
6 (0.5 cc)
29919
22857
19087
11897
20607
13643
8264
5143
31.1
40.3
56.7
56.8
For CT values < 0 , materials
are assumed to have an
energy dependence similar to
water
For CT values > 0, material
is assumed to have an
energy dependence similar to
a mixture of bone and water
The green line shows the
effect of using water scaling
for all materials
0.200
Attenuattion at 511keV
Motion Effects
O. Mawlawi MDACC
0.150
error
Bone/Water
0.100
Water/air
0.050
0.000
-1000
-500
0
500
1000
CT number measured at 140kVp
O. Mawlawi MDACC
O. Mawlawi MDACC
8
6/30/2008
Metal Artifacts
Metal Artifacts
Cold spot
Artifact
Artifact
CT SCAN
CT
PET CTAC
PET AC
NO AC
FUSED IMAGE
O. Mawlawi MDACC
O. Mawlawi MDACC
Contrast ArtifactArtifact- Intravenous
Contrast Effects
Contrast
injected
Transaxial
contrast bias
Coronal
Normal CT
Contrast enhanced CT
Contrast: Optiray 320, 100cc, 3cc/sec
CT
O. Mawlawi MDACC
PET CTAC
FUSED IMAGE
O. Mawlawi MDACC
9
6/30/2008
With contrast
Without contrast
SUV change
1.23 to 1.90
54% inc
61 year male with history of esophageal cancer
O. Mawlawi MDACC
O. Mawlawi MDACC
Truncation Artifacts
CT and PET Fields of View
„
X-ray focal spot
PET Detector
Occur when objects extend outside the CT field
of view (50cm), but are in the PET field of view
(70cm).
CT
FOV
PET 70cm image FOV
CT 50cm Fully sampled FOV
PET
FOV
CT Detector
„Truncation
O. Mawlawi MDACC
artifacts are observed in very large patients or patients
scanned with arms down.
O. Mawlawi MDACC
10
6/30/2008
SUVmax changed from 3.25 to 6.05
O. Mawlawi MDACC
CT
PET
FUSED
O. Mawlawi MDACC
Quantitative PET Performance
Attenuation map
3 Factors that may affect SUV measurements
• Patient Compliance
Truncated
–
–
•
Scan Conditions
–
–
–
–
Truncated
corrected
•
54 yrs. male with history of metastatic melanoma of
O. Mawlawi MDACC
the skin. SUV changed from 3.25 to 6.05
Fasting
Blood Glucose levels
Scan time post injection
Patient anxiety/comfort during uptake (room temperature, etc.)
Patient motion during acquisition
PET/CT vs dedicated PET
Intrinsic System Operating Parameters
–
–
–
–
Calibration
QA – Maintenance of operating parameters
System performance characterization – Partial Volume
Image processing algorithms
O. Mawlawi MDACC
11
6/30/2008
Effects of Bad Blocks
Coincidence Calibrations
• Well Counter Correction (WCC)
O. Mawlawi MDACC
Effects of Bad Blocks
O. Mawlawi MDACC
Effects of Bad Blocks
O. Mawlawi MDACC
O. Mawlawi MDACC
12
6/30/2008
Normalization effects
Normalized to Baseline for each Sphere
105
sphere
37 mm
100
sphere
28 mm
95
sphere
22 mm
85
sphere
h
17 mm
80
75
sphere
13 mm
70
sphere
10 mm
5
4,
5
k4
lk
bl
6b
22
M
4,
4,
5
5
M
5
4,
k4
,
lk
bl
lk
5b
Good Norm
M
9b
22
M
5
lk
4,
lk
5b
5b
lk
5b
M
M
4
M
5
4,
5
4,
lk
5b
M
3
k2
,
bl
M
22
M
5
4,
lk
lk
5b
M
5b
M
M
4
6b
lk
4,
lk
9b
5b
M
M
5
4,
M
5
lk
5b
M
lk
5
3
lk
5b
22
4,
lk
5b
M
2,
4
blk
e
lk
5b
M
M
ba
fi n
al
5
e
li n
se
l in
se
5
65
ba
pe
ercent
90
Bad Norm
O. Mawlawi MDACC
Effect of bad Normalization
O. Mawlawi MDACC
Effect of wrong WCC
SUVmax=4.5
Increased
activity due to
error in
normalization
O. Mawlawi MDACC
O. Mawlawi MDACC
13
6/30/2008
Effect of Reconstruction Algorithm
Effect on Partial voluming
13
All spheres contain
the same activity
concentration
c
e
b
d
f
22
10
Profile (10 mm)
28
37
Recovery (%)
a
17
100
22
37
28
(a) 2D-FBP; (b) 2D-OSEM; (c) 3D-RP; (d) 3D-IR; (e) 3D-FORE+FBP; (f) 3D-FORE+OSEM.
17
50
AC max under different reconstruction algorithms (Bq/cc)
13
0
10
Standard
8 x 8 detector
HI-REZ
13 x 13 detector
Sphere diameter
Recovery coefficients
(x104)
Sphere 1
Sphere 2
Sphere 3
Sphere 4
Sphere 5
Algorithm a
3.03
2.74
2.52
2.30
1.75
1.12
Algorithm b
3.14
3.14
2.82
2.90
2.14
1.74
Algorithm c
2.83
2.76
2.62
2.41
1.60
1.07
Algorithm d
2.93
2.92
2.98
3.21
2.19
1.33
Sphere 6
Algorithm e
2.78
2.75
2.64
2.40
1.59
1.10
Algorithm f
2.85
2.74
2.71
2.52
1.65
1.09
Courtesy
of Siemens
O. Mawlawi
MDACC
O. Mawlawi MDACC
A
B
A
B
3D IR
2it 20sub
3D FORE
2it 5sub
2D
OSEM
2it 20sub
2D
OSEM
2it 28sub
C
D
C
3D FORE
5it 28sub
3D FORE
FBP
2D
OSEM
4it 28sub
Max SUVbw
in different
spheres
A
1
2
3
4
5
6
10.8
10.3
9.9
8.9
6.5
3.8
B
8.6
7.6
6.6
5.5
3.9
2.8
C
10.3
9.7
9.1
8.3
6.3
4.7
D
10.4
9.3
9.3
7.7
5.9
4.8
Max SUVbw
in different
spheres
O. Mawlawi MDACC
1
2
3
4
5
6
A
12.7
11.9
11.7
12.7
7.4
4.1
B
15.0
12.3
11.7
13.4
9.6
5.4
C
16.0
13.6
11.9
14.7
10.3
6.3
O. Mawlawi MDACC
14
6/30/2008
Effect of acquisition time on convergence of reconstruction algorithm (2it 21ss)
O. Mawlawi MDACC
Effect of scan duration and count density
on maximum AC
2D
4
x 10
Max d37
12
10
8
6
x 10
scan1
scan2
scan3
scan4
8
7
6
5
4
3
4
2
2
0
0
Effect of smoothing on measured AC
3D
4
9
scan1
scan2
scan3
scan4
14
Max d37
16
O. Mawlawi MDACC
20
40
60
80
100
120
Duration(sec)
140
160
180
1
0
20
40
60
80
100
120
140
160
180
Duration(sec)
O. Mawlawi MDACC
O. Mawlawi MDACC
15
6/30/2008
Effect of ROI type
Conclusion
• SUV is reproducible
• SUV is reliable as long as mediating factors
are accounted for
• Standardization across institutions might be a
more difficult task
Courtesy of P. Kinahan UW
O. Mawlawi MDACC
O. Mawlawi MDACC
16