Page 1
Data Acquisition Electronicsfor
Positron Emission Tomography
Data Acquisition Electronicsfor
Positron Emission Tomography
William W. MosesLawrence Berkeley National Laboratory
May 24, 2010
• PET Overview• PET Electronics Requirements & Trends• OpenPET Electronics
Outline:
• This work was supported in part by the U.S. DOE (contract No. DE-AC02-05CH11231) and in part by the NIH (NIBIB grant No. R01-EB006085).
• Thanks to S. Choong, C. Vu, and Q. Peng of LBNL, C. Jackson, S. Buckley, and N. Pavlov of SensL, M. Casey and J. Young of Siemens Medical Solutions, D. McDaniel of GE Medical Systems, and J. Karp of U. Penn.
Page 2
Nuclear MedicineNuclear Medicine
• Patient injected with small amount of radioactive drug.
• Drug localizes in patient according to metabolic properties of that drug.
• Radioactivity decays, emitting gamma rays.
• Gamma rays that exit the patient are imaged.
•Well Established Clinical Technique•10 Million Studies Annually
•Well Established Clinical Technique•10 Million Studies Annually
Gamma Camera
Page 3
Ring of PhotonDetectors • Radionuclide decays by
emitting a positron (+ ).
• + annihilates with e– from tissue, forming back-to-back511 keV photon pair.
• 511 keV photon pairs detected via time coincidence.
• Positron lies on line defined by detector pair.
• Detects Pairs of Back-to-Back 511 keV Photons• Use Computed Tomography to Form 2-D Image
• Detects Pairs of Back-to-Back 511 keV Photons• Use Computed Tomography to Form 2-D Image
Positron Emission Tomography (PET)Positron Emission Tomography (PET)
Page 4
MRI & PET Images of EpilepsyMRI & PET Images of Epilepsy
• MRI “Sees” Structure with 0.5 mm Resolution
• PET “Sees” Metabolism with 4.0 mm Resolution
MRI PET
Page 5
Combine PET & CT (Oncology)Combine PET & CT (Oncology)CT Image PET Image Fused Image Post-Therapy
•Anatomy from X-Ray CT, Function from PET•Current “Standard of Care”
•Anatomy from X-Ray CT, Function from PET•Current “Standard of Care”
Images courtesy of Stig Larsson, Karolinska Institute
Page 6
Combine PET & CT (Cardiac)Combine PET & CT (Cardiac)
Image courtesy ofGustav von Schulthess, University Hospital, Zurich
Anatomy from X-Ray CT, Function from PETAnatomy from X-Ray CT, Function from PET
Page 7
What Is A “PET Event”?(What Do You Need To Record?)
What Is A “PET Event”?(What Do You Need To Record?)
•Non-TOF PET: A Chord — Defined By Two End Points•Time-of-Flight PET: A Chord and t
•Non-TOF PET: A Chord — Defined By Two End Points•Time-of-Flight PET: A Chord and t
Page 8
What Is The Electronics Concept?What Is The Electronics Concept?
• Identify “Singles Events”•Find Time Coincidences Between Singles Events w/ t
•“Coincident Event” = Pair of Singles Events (Chord)
• Identify “Singles Events”•Find Time Coincidences Between Singles Events w/ t
•“Coincident Event” = Pair of Singles Events (Chord)
• Position (crystal of interaction)
• Time Stamp (arrival time)
• Energy Validation (=511 keV?)
“Singles Event”
• Position (crystal of interaction)
• Time Stamp (arrival time)
• Energy Validation (=511 keV?)
“Singles Event”t
Page 9
NumbersNumbers
• 150-300 Detector Modules (25 cm2 each) per Camera
• 100 kHz Typical Singles Rate per Module
• 1 MHz Maximum Singles Rate per Module
• 1 MHz Typical Coincidence Rate per Camera
• 4 MHz Maximum Coincidence Rate per Camera
• 1 ns fwhm Timing Resolution for Conventional PET
• 250 ps fwhm Timing Resolution for TOF PET
•Low Coincidence / Singles Ratio•Low Occupancy in Each Detector Module
Multiplex Singles Events
•Low Coincidence / Singles Ratio•Low Occupancy in Each Detector Module
Multiplex Singles Events
Page 10
PET Detector ModulePET Detector Module
0500
1000150020002500300035004000
Counts
Decode Crystals Using Anger Logic (Light Sharing)Decode Crystals Using Anger Logic (Light Sharing)
X-Ratio
Y-Ratio
ProfilethroughRow 2
Array of Scintillator Crystals
Photomultiplier Tubes
Page 11
1 2 3 4 5 6 7
8 9 10 11 12 13 14
15 16 17 18 19 20 21
22 23 24 25 26 27 28
29 30 31 32 33 34 35
36 37 38 30 40 41 41
43 44 45 46 47 48 49
50 51 52 53 54 55 56
Position IdentificationPosition Identification
1 2 3 4 5 6 7
8 9 10 11 12 13 14
15 16 17 18 19 20 21
22 23 24 25 26 27 28
29 30 31 32 33 34 35
36 37 38 39 40 41 42
43 44 45 46 47 48 49
50 51 52 53 54 55 56
Y
X
E=A+B+C+DY=(A+B)/EX=(B+D)/E
C
A
D
B
•Identify Crystal of Interaction Using Lookup Table•Position Given by Crystal ID
•Identify Crystal of Interaction Using Lookup Table•Position Given by Crystal ID
Page 12
Energy ValidationEnergy Validation
•Yes/No Decision Based on Measured Energy•Individual Thresholds for Each Crystal
•Implemented Using Lookup Table
•Yes/No Decision Based on Measured Energy•Individual Thresholds for Each Crystal
•Implemented Using Lookup Table
0
2000
4000
6000
8000
10000
12000
14000
0 100 200 300 400 500
Counts per Bin
Pulse Height Bin
YesNo No
E
Page 13
Time StampTime Stamp
• “Raw” Time Stamp is Digitized Time Since Last Time Slice Boundary
• Each Crystal Assumed to Have aDifferent (But Constant) Propagation Delay
• Propagation Delays Measured w/ Positron Sourceand Stored in a Lookup Table
• Crystal by Crystal Correction Factor Added to “Raw” Time
• “Raw” Time Stamp is Digitized Time Since Last Time Slice Boundary
• Each Crystal Assumed to Have aDifferent (But Constant) Propagation Delay
• Propagation Delays Measured w/ Positron Sourceand Stored in a Lookup Table
• Crystal by Crystal Correction Factor Added to “Raw” Time
TDC
CFD Master Clock
Time Slice Boundary
Clock
Start / Reset
Stop
From Central Timing BoardDetector Module
Page 14
Identifying Time CoincidencesIdentifying Time Coincidences
•Break Time Into Slices (100–250 ns / slice)•Search for Singles Within tmax (4–12 ns) in Each Slice
•Greatly Reduces Combinatorics
•Break Time Into Slices (100–250 ns / slice)•Search for Singles Within tmax (4–12 ns) in Each Slice
•Greatly Reduces Combinatorics
Time
Slice 1 Slice 2 Slice 3 Slice 4 Slice 5 Slice 6
Single
Single
Single
Single
Single
tmax
Coincidence
Single
Single
Coincidence
Single
Single
Single
Single
Single
Single
SingleSingle
Single
Single SingleCoincidence
Coincidence
Single
Single
SingleSingle
CoincidenceNot Coincidence
Page 15
Siemens & GE Detector TopologySiemens & GE Detector Topology
•4 PMTs to Decode Event Position• Independent Modules Operating in Parallel
•4 PMTs to Decode Event Position• Independent Modules Operating in Parallel
ScintillatorCrystal Array
Page 16
Chord
Siemens & GE System ArchitectureSiemens & GE System Architecture
DetectorBoard
Detectors
MultiplexerBoard
Coincidence Host PC
• Position, Time, & Energy Computed for Each Trigger• All Data Flow is Forward No Handshaking
• Position, Time, & Energy Computed for Each Trigger• All Data Flow is Forward No Handshaking
DetectorBoard
MultiplexerBoard
Position
Time
Position
Time
Position
Time
Energy
Position
TimeEnergy
Page 17
Philips Detector TopologyPhilips Detector Topology
•7–20 PMTs to Decode Event Position•Highly Coupled—Each PMT Used in Many “Modules”
•7–20 PMTs to Decode Event Position•Highly Coupled—Each PMT Used in Many “Modules”
ScintillatorCrystal Array
Page 18
Chord
Philips System ArchitecturePhilips System Architecture
DetectorBoard
Detectors CoincidenceBoard
ProcessingBoard
Host PC
• Position & Energy Computed Only for Coincidences• Bidirectional Data Flow Handshaking
• Position & Energy Computed Only for Coincidences• Bidirectional Data Flow Handshaking
DetectorBoard
Time
Time
Coincidence
Coincidence
ADC Output
ADC Output
PMT ID
Position
Energy
Position
Energy
Page 19
Typical Front End ArchitectureTypical Front End ArchitecturePMTs
Corrections Done In Real Time, Outputs Singles EventsCorrections Done In Real Time, Outputs Singles Events
B
Gain Adjust,Integrate
Sum
Gain Adjust,Integrate
Gain Adjust,Integrate
Gain Adjust,Integrate
Sum
Sum
ADCVin
Vref
ADCVin
Vref
A
C
DADC
Vin
Vref
Crystal Lookup(X & Y ID)
EnergyValidation
(E & ID 511?)
Time Stamp(T & ID T Stamp)
Event FormattingCFD TDC
+5V
A+B
B+D
Y
X
E
T
Logic
Page 20
Siemens Implementation (~1995)Siemens Implementation (~1995)PMTs
Analog Done w/ Discretes, Digital Done w/ Custom ASICAnalog Done w/ Discretes, Digital Done w/ Custom ASIC
B
Gain Adjust,Integrate
Sum
Gain Adjust,Integrate
Gain Adjust,Integrate
Gain Adjust,Integrate
Sum
Sum
ADCVin
Vref
ADCVin
Vref
A
C
DADC
Vin
Vref
Crystal Lookup(X & Y ID)
EnergyValidation
(E & ID 511?)
Time Stamp(T & ID T Stamp)
Event FormattingCFD TDC
+5V
A+B
B+D
Y
X
E
T
Custom ASIC
Page 21
Siemens Implementation (~2000)Siemens Implementation (~2000)PMTs
Analog Done w/ Custom ASIC, Digital Done w/ FPGAAnalog Done w/ Custom ASIC, Digital Done w/ FPGA
B
Gain Adjust,Integrate
Sum
Gain Adjust,Integrate
Gain Adjust,Integrate
Gain Adjust,Integrate
Sum
Sum
ADCVin
Vref
ADCVin
Vref
A
C
DADC
Vin
Vref
Crystal Lookup(X & Y ID)
EnergyValidation
(E & ID 511?)
Time Stamp(T & ID T Stamp)
Event FormattingCFD TDC
+5V
A+B
B+D
Y
X
E
T
FPGA & Memory
Custom ASIC
Page 22
Siemens Implementation (~2005)Siemens Implementation (~2005)PMTs
Op-AmpFree Running ADC, More Digital Done w/ FPGAOp-AmpFree Running ADC, More Digital Done w/ FPGA
B
Gain Adjust,Anti-Alias
Gain Adjust,Anti-Alias
Gain Adjust,Anti-Alias
Gain Adjust,Anti-Alias
Sum
ADCVin
Vref
A
C
D
Crystal Lookup(X & Y ID)
EnergyValidation
(E & ID 511?)
Time Stamp(T & ID T Stamp)
Event FormattingCFD TDC
+5V
A
T
FPGA & Memory
Custom ASIC
ADCVin
Vref+5V
B
ADCVin
Vref+5V
C
ADCVin
Vref+5V
D
Free-Running(~75 MHz)
Page 23
• Nuclear Medicine Research Community Needs“Industrial-Strength” Electronics
• Needs Can Be Met By Single, Flexible Design!
• Nuclear Medicine Research Community Needs“Industrial-Strength” Electronics
• Needs Can Be Met By Single, Flexible Design!
Page 24
Electronics System RequirementsElectronics System Requirements
Like Open Source Software!Like Open Source Software!
High-Performance•# of Channels, Rate, Energy, Timing, …
Very Flexible•Type of Detector, Camera Configuration,Event Word Definition
User-Modifiable•Schematics, Source Code, Knowledge Base
User-Friendly• Instructions, Documentation, Can Buy Boards
Page 25
All Detector Outputs Look the SameAll Detector Outputs Look the Same
• Tremendous Variation in How Outputs Are CombinedCombine Outputs in Firmware
• Tremendous Variation in How Outputs Are CombinedCombine Outputs in Firmware
Voltage
Time
Extract Timing Signal from Leading Edge
Extract “Energy” from Area Under the Curve
100 mV – 2 V
Page 26
OpenPET Implementation (~2010)OpenPET Implementation (~2010)PMTs
Analog Done w/ Discrete, More Digital Done w/ FPGAAnalog Done w/ Discrete, More Digital Done w/ FPGA
B
Gain Adjust,Anti-Alias
Gain Adjust,Anti-Alias
Gain Adjust,Anti-Alias
Gain Adjust,Anti-Alias
ADCVin
Vref
A
C
D
Crystal Lookup(X & Y ID)
EnergyValidation
(E & ID 511?)
Time Stamp(T & ID T Stamp)
Event FormattingTDC
+5V A
FPGA & Memory
ADCVin
Vref+5V B
ADCVin
Vref+5V C
ADCVin
Vref+5V D
Free-Running(80 MHz)
TDC
TDC
TDC
Discriminator
Discriminator
Discriminator
Discriminator
Page 27
OpenPET System ArchitectureOpenPET System Architecture
DetectorBoard
Data
Control
Detectors
16 AnalogSignals In
SupportBoard
P
8 Boards In
DigitalMultiplexer
8 Boards In
Coincidence
P
8 Boards In
Host PC
• Supports 256 Block Detectors (2048 With Multiplexers)• PSB + 8 DPBs Makes Nice Test Stand (32 Block Detectors)
• Supports 256 Block Detectors (2048 With Multiplexers)• PSB + 8 DPBs Makes Nice Test Stand (32 Block Detectors)
Page 28
Detector 1: Conventional Block DetectorDetector 1: Conventional Block Detector
• Pre-Prototype Circuit Boards• Excellent Flood Map and Energy Resolution
• Pre-Prototype Circuit Boards• Excellent Flood Map and Energy Resolution
12% 11%
10%12%
12x12 array of 4x4x22 mm3 LSO crystals4 Hamamatsu R-9800 PMTs
Page 29
Detector 2: SiPM ArrayDetector 2: SiPM Array
Same Electronics with Very Different Type of DetectorSame Electronics with Very Different Type of Detector
16x16 array3x3 mm2 SiPMs
3x3x20 mm3 &3x3x30 mm3 LSO
Adapter Board
16x16 arrayto
16 Row &16 Column
Analog Sums
Natural LSO Activity
Pixel Intensity Energy x Count Rate
Page 30
Timing ResolutionTiming Resolution
HPTDC(CERN)
FPGA
Test Pulse TOF Module Pair
0
1 105
2 105
3 105
4 105
5 105
6 105
-400 -200 0 200 400
Counts per Bin
Time (ps)
255 ps fwhm
0
1 105
2 105
3 105
4 105
5 105
6 105
-200 -150 -100 -50 0 50 100 150 200
Counts per Bin
Time (ps)
90 ps fwhm
0
500
1000
1500
2000
2500
3000
-400 -200 0 200 400
Counts per Bin
Time (ps)
286 ps fwhm
0
500
1000
1500
2000
2500
-200 -150 -100 -50 0 50 100 150 200
Counts per Bin
Time (ps)
56 ps fwhm
• 16 Channel TDC in Cyclone II FPGA• Performance Good Enough for Time-of-Flight PET
• 16 Channel TDC in Cyclone II FPGA• Performance Good Enough for Time-of-Flight PET
Page 31
Vision
Vision
http://OpenPET.LBL.govhttp://OpenPET.LBL.gov