o
D
PET Scanner Quality Assurance
Introduction
Quality Control
uc
se od
r
ea p ou
Pl e y
R
k
ot n
N a
Th
Procedures required to ensure that the distribution of radiation
emitted from a patient is accurately reflected in the measured raw
raw
data.
Brad Kemp, PhD
Department of Radiology
Mayo Clinic
Facets
• Quality assurance:
– verify scanner is operating properly
– identify problems prior to scanning patients
• Acquisition of corrections to compensate for known imperfections
in data measurement
WE-A-330D-1
Outline
• Corrections
• System Calibration
• Daily QA
• NEMA NU 22-2001 Performance Standards
System Correction: Normalization
Purpose: Corrects for the variations in efficiency in lines of
response (LOR) in each slice of the sinogram
• A uniformity correction
• For 2D: Direct Measurement using a low activity source
• For 3D: Indirect Measurement using a uniform phantom
• Acquired quarterly or after system maintenance
• Scanner certification
e
1
o
D
System Calibration
Quality Assurance
Quality Assurance Test Requirements
• A phantom of known activity concentration is imaged and
reconstructed with uniform attenuation correction
• Average counts in a slice are calculated
• Well defined regimen of measurements
uc
se od
r
ea p ou
Pl e y
R
k
ot n
N a
Th
Purpose: convert image counts/pixel to activity/volume
→From this a system calibration is calculated
– Quick and easy to conduct
– Sensitive to modes of failure of the scanner
– Preferably quantitative, not qualitative
• Acquired monthly or after system maintenance
Blank Scan
Quantitative Analysis of Blank Scans
• On standstand-alone PET systems, used with transmission sinograms
to create attenuation correction factors
• Acquired daily - good source of QA data
• Emission scan of uniform cylinder
Transmission Sinograms
Good blank
Bad block
Sinogram of a
Uniform Cylinder
• The quantitative analysis of blank scans is an important
aspect of the QA of a PET scanner
• Quantitative analysis of blank scans:
- validates system calibration
- provides information on crystal, block or module (bucket)
efficiencies with respect to system average
- compare to baseline - monitors system stability
- ensures validity of normalization
e
2
o
D
Quantitative Analysis of Blank Scans
Overview: Daily QA
Daily QA analysis calculates a global and slice normalized mean square
error (NMSE) by using a reference scan; the reference scan is defined as the
blank scan acquired immediately after normalization
Time Line
uc
se od
r
ea p ou
Pl e y
R
k
ot n
N a
Th
Normalization
Normalization
Reference Blank
Daily Blank
Has the system drifted?
Is the normalization valid?
Normalized Mean Square Error
Quality Assurance
Pixel to pixel comparison
Reference Blank
Daily Blank
Detector and Electronics Characterization
– Singles mode detector calibration
• Crystal map
• PMT gain adjustment
• Energy map
– Coincidence timing calibration
∑ (Counts − Counts )
NMSE =
∑ (Counts )
i
i
Did the engineer calibrate the scanner properly?
DAY 2
i
REF
i
REF 2
i
Compute global and slice NMSE
e
3
o
D
Verification of Crystal Map
Crystal Map
QA of crystal maps created during calibration of the scanner
Purpose:
• Map the position of the detected event to a specific crystal
• Obtained quarterly or after detector maintenance
uc
se od
r
ea p ou
Pl e y
R
k
ot n
N a
Th
Crystal Map Look Up Table
Count distribution in a block
PMT Gain Adjustment
Coincidence Timing Calibration
Purpose:
Purpose:
• Balances the gain characteristics of the PMTs in a block
• Compensates for PMT gain drift with temperature, age, …
• Acquired weekly (daily) or after detector maintenance
• Adjust for timing delays so events from all blocks are
timestamped equivalently
• Acquired weekly (daily) or after detector maintenance
0.0
Update Gains
Update Gains
Block Bias (%)
-1.0
-2.0
Morning
-3.0
Afternoon
-4.0
γ
γ
-5.0
Coincidence ?
Yes
Register
Event
-6.0
28-Feb
7-Mar
14-Mar
21-Mar
28-Mar
Date
e
4
o
D
Daily Quality Assurance
Example: NMSE and Block Bias
Can the Scanner be Used Today?
QA must be sensitive to modes of failure of the scanner
NMSE: Error with respect to baseline scan
Potential Problems
• System stability, drifts
• Detector module / PMT/ preamp failure
• Loose cables, connectors
• Inoperable gantry motors, source loader
6
0.010
• Automatic PMT gain adjustment
• Blank scan
4
2
0.006
0
-2
0.004
-4
0.002
-6
0.000
• QA will detect but not prevent these problems
Block Bias (%)
Normalized Mean Square Error
uc
se od
r
ea p ou
Pl e y
R
k
ot n
N a
Th
Daily QA Regimen
0.008
Global NMSE
Slice NMSE
Block Bias
10-Dec-02
-8
29-Jan-03
20-Mar-03
9-May-03
28-Jun-03
17-Aug-03
Date
Daily QA: Loose Cable
Loose Cable
Better
QA Schedule
Detector and Electronics
Characterization
– Crystal map
– PMT gain
– Coincidence timing
Frequency
Quarterly
Weekly (daily)
Weekly (daily)
System Corrections
Normalization
– Scanner calibration
– Blank Scan
–
Quarterly
Monthly (weekly)
Daily
e
5
o
D
NEMA NU 2-2001
NEMA NU 2-2001
Acceptance Testing
PET Performance Measurements
Can we use our new scanner?
uc
se od
r
ea p ou
Pl e y
R
k
ot n
N a
Th
National Electrical Manufacturers Association.
NEMA Standards Publication NU 22-2001:
Performance Measurements of Positron
Emission Tomographs
Annual QA
Is the scanner still performing within specification?
Phantoms for NU 2-2001
NEMA NU 22-2001
Performance Measures:
Three phantoms:
Scatter phantom
203x700mm phantom with activity
in line source
NEMA NU 2-2001
Sensitivity phantom
• Spatial Resolution (Transaxial, axial)
• Sensitivity
• Scatter Fraction
Image quality phantom
• Count Losses
• Count Rate Correction Accuracy
• Image Quality
e
6
o
D
Spatial Resolution
Sensitivity
• Spatial resolution of a system represents its ability to distinguish
distinguish
between two points of radioactivity
uc
se od
r
ea p ou
Pl e y
R
k
ot n
N a
Th
• F-18 point sources in air at six locations:
– (0,1), (0,10) and (10,0) cm
– Center of axial FOV and ¼ axial FOV from center
• Sensitivity of a scanner represents its ability to detect
annihilation radiation
• Rate of true coincidence counts per unit radioactivity (expressed
(expressed
in cps/kBq
cps/kBq)) in absence of attenuating media
• Rationale: need material around source to ensure annihilation of
positrons, but this material also attenuates the annihilation
photons
• Based on technique by Bailey DL, Jones T, et al. Eur J Nucl
Med 1991;18:3741991;18:374-379.
• Reconstruct: image pixel < ⅓ expected FWHM
• Profile width ~ 2 times FWHM
• Report FWHM and FWTM in radial, tangential and axial
directions
Sensitivity
Sensitivity
• Successive measurements with a 700 mm line source with a known
amount of FF-18 surrounded by nested, known absorbers
• Measure at radial locations of 0 and 10cm
• Report system sensitivity and slice sensitivity profile
2D
3D
0.05
Count Rate, R (cps)
1.5E+05
R0
1.0E+05
System Sensitivity =
R0
Activity
0.03
0.02
R0
R10
0.01
0.00
0.30
R0
R10
0.20
0.10
0.00
0
5.0E+04
Sensitivity (cps/kBq)
• The count rate with no absorber is extrapolated from these measures
measures
Sensitivity (cps/kBq)
0.40
0.04
10
20
30
Slice
40
50
0
10
20
30
40
50
Slice
0.0E+00
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
Sleeve Thickness (cm)
e
7
o
D
Scatter Fraction
Scatter Fraction
Scatter Fraction vs Slice
• Scatter fraction is a measure of the system sensitivity to scatter
scatter
SF =
Scatter
Scatter + Trues
2D
20
3D
40
uc
se od
r
ea p ou
Pl e y
R
k
ot n
N a
Th
Scatter Fraction (%)
Scatter Fraction (%)
• Use 203 mm diameter polyethylene cylinder of length 700 mm, with
activity located in a line source of diameter 2.3mm that is 4.5mm
4.5mm
off axis
• Measured with low activity (Randoms:Trues
(Randoms:Trues = 1%) to avoid random
coincidences, deadtime and pulse pileup.
35
15
10
5
30
25
20
15
10
5
0
0
0
10
20
30
40
50
0
10
Slice
Count Rate Performance
• Measurement of count rate performance gives an indication of
scanner performance as a function of activity
30
40
50
Slice
Count Rate Performance
• Calculate Noise Equivalent Count Rate
RNEC =
• Use 700 mm long polyethylene cylinder
• Measured with high initial activity of FF-18
3D: 800 MBq;
MBq; 2D: 5 GBq
• Acquire data until randoms and deadtime losses are negligible
(14 to 18 hrs)
20
RTrues
2
RTrues
+ RScatter + k ⋅ RRandoms
RNEC: Figure of merit relating scanner performance to sinogram
SNR after randoms and scatter corrections.
For NEMA, k = 1 (calculated Randoms);
Randoms); RNEC (1R)
• Report peak NEC and effective activity concentration at peak
e
8
o
D
Count Rate Performance
Image Quality Measurement
Peak RTrues: 515 kcps @ 44 kBq/ml
kBq/ml
• Standardized imaging situation that simulates a clinical whole
body imaging condition
• Phantom consists of a torso phantom with hot and cold lesions in
a warm background
• Scatter phantom abutted to image quality phantom
Peak RNEC (1R):118 kcps @ 22 kBq/ml
kBq/ml
Peak RNEC (2R): 80 kcps @ 18 kBq/ml
kBq/ml
NECR (1R)
NECR (2R)
2500
2000
NECR (kcps)
120
3000
Count Rate (kcps)
uc
se od
r
ea p ou
Pl e y
R
k
ot n
N a
Th
140
3500
Prompts
Trues
R+S
1500
1000
500
100
80
60
40
20
0
0
0
10
20
30
40
50
60
Activity Concentration (kBq/cc)
0
10
20
30
Image Quality Measurement
•
•
•
•
•
•
Hot spheres: 10, 13, 17, 22 mm id
Cold spheres: 28, 37 mm id
Lung insert
Activity in hot spheres 8 and 4 times that of background
Activity in background 5.3 kBq/ml
kBq/ml
Simulated acquisition 100cm in 60min
Tacq =
40
50
60
Activity Concentration (kBq/cc)
Image Quality Measurement
• Report image contrast and SNRs for hot and cold lesions,
residual error in lung, variability in background
• Visual inspection for artifacts
60 min
Δz ≈ 7 min
100 cm
• Repeat acquisition three times
• Reconstruct using clinical protocol
e
9
o
D
NEMA and Lutetium-176
• LYSO is inherently radioactive
• Background radiation gives rise to Randoms,
Randoms, some Trues
uc
se od
r
ea p ou
Pl e y
R
k
ot n
N a
Th
• Implications for NEMA: cannot obtain Randoms:Trues ratio of 1%
• For count rate and sensitivity measurements – acquire delayed event
to measure intrinsic randoms rates
• Watson CC, et al. J Nucl Med 2004;45:8222004;45:822-826.
• Erdi YE, et al. J Nucl Med 2004;45:8132004;45:813-821.
e
10