Optimization of LSOfor Time-of-Flight PETOptimization of LSO

for Time-of-Flight PET

W. W. Moses1, M. Janecek1, M. A. Spurrier2, P. Szupryczynski2,3 ,W.-S. Choong1, C. L. Melcher2, and M. Andreaco3

1Lawrence Berkeley National Laboratory2University of Tennessee, Knoxville

3Siemens Medical Solutions

October 21, 2008

• Motivation• Reflector Optimization• LSO Optimization• PMT Optimization

Outline:

This work was supported by the NIH (NIBIB grant No. R01-EB006085).

Time-of-Flight in PETTime-of-Flight in PET

• Can localize source along line of flight.

• Time of flight information reduces noise in images.

• Variance reduction given by 2D/ct.

• 500 ps timing resolution 5x reduction in variance!

c = 30 cm/ns500 ps timing resolution

7.5 cm localization

•Time of Flight Provides a Huge Performance Increase!•Largest Improvement in Large Patients

•Time of Flight Provides a Huge Performance Increase!•Largest Improvement in Large Patients

D

Commercial TOF PET w/ LSOCommercial TOF PET w/ LSO

~550 ps Coincidence Timing Achieved~550 ps Coincidence Timing Achieved

Our Goal:“Demonstration” TOF PET Camera

Our Goal:“Demonstration” TOF PET Camera

Achieve the Best Timing Possible w/ LSOAchieve the Best Timing Possible w/ LSO

• With better timing resolution (t), huge gains predicted(23x variance reduction for 100 ps timing)

• Measure image improvement vs. timing resolution

• Use LSO scintillator

– Don’t change other factors that influence SNR(efficiency, scatter fraction, etc.)

What Limits Timing Resolution?What Limits Timing Resolution?

CrystalGeometry

326 psPMT

PMT

PMT 422 ps

Light Sharing 454 ps

PMT Array 274 ps

Baseline 160 ps

Non-TOF Block Detector Module

•Many Factors•“Optical Geometry” Particularly Important

•Many Factors•“Optical Geometry” Particularly Important

Timing Values will be for a Single Detector*Timing Values will be for a Single Detector*1 cm3

BaF2H5321 R9800

ShapingAmplifier

CFDCanberra 454

F=0.2, D=0.6 ns

TACOrtec 566

Start Stop

ADCNI-7833R

Ge-68LSO

+/- fwhm

Photopeak events

• Only Accept Events in Photopeak Window• Subtract (in Quadrature) 150 ps Trigger Contribution

• Only Accept Events in Photopeak Window• Subtract (in Quadrature) 150 ps Trigger Contribution

*Unless explicitly mentioned otherwise

CFDCanberra 454

F=0.2, D=0.6 ns

Pulse Height

Timing

Photopeakevents

Windowed Timing

TriggerDetector

TestDetector

Proposed Side-Coupled DesignProposed Side-Coupled Design

Proposed Geometry(Side-Coupled Crystal)

ScintillatorCrystal

PMT

PMT

Shorter Optical Path Length & Fewer ReflectionsShorter Optical Path Length & Fewer Reflections

Conventional Geometry(End-Coupled Crystal)

384 ps(543 ps coinc.)

218 ps

Detector Module DesignDetector Module Design

PMT

(HamamatsuR-9800)

Two LSO Crystals(each 6.15 x 6.15 x 25 mm3)

Reflector

(on all five faces of each crystal, including the face between the

two crystals)

Optical Glue

(between lower crystal faces and PMT)

Hole in ReflectorOn Top Face of

Crystals

Two Side-Coupled Scintillator Crystals per PMTTwo Side-Coupled Scintillator Crystals per PMT

Detector Ring GeometryDetector Ring Geometry

Crystals Decoded by Opposing PMTCrystals Decoded by Opposing PMT

Exploded View

• Top face of each crystal (with hole in reflector) is coupled via a small (<1 mm) air gap to the edge of one opposing PMT.

• Light seen by the opposing PMT is used to decode the crystal of interaction.

Crystal ofInteraction

Camera GeometryCamera GeometrySection of Detector Ring

• Detector ring is 825 mm diameter, 6.15 mm axial• 192 detector modules, 384 LSO scintillator crystals• Adjustable gap (6 – 150 mm) between lead shields allows

“scatter-free” and “3-D” shielding geometries

Lead Shielding

Modules

“Real” Single-Ring PET Camera for Humans & Phantoms“Real” Single-Ring PET Camera for Humans & Phantoms

Optimization: Surface Finish & ReflectorOptimization: Surface Finish & Reflector

Both Performance & Manufacturing ImportantBoth Performance & Manufacturing Important

• Best Possible Timing Resolution

• Good Light Collection Efficiency

• Good Energy Resolution

• Manufacturable:

• Easy

• Reliable & Reproducible

• Rugged

• Well-Controlled Light from Top Surface

• Starting Point: Saw Cut LSO w/ Teflon Tape

Requirements

Surface & Reflector Optimization MethodSurface & Reflector Optimization Method

Measure Percentage Change in TimingMeasure Percentage Change in Timing

• Measure Timing of “Raw” Crystal(saw cut finish, Teflon tape reflector)

• Apply Surface Treatment

• Apply Reflector

• Re-Measure Timing

• Compute Percent Change

• Repeat for 5 Crystals & Average Results

• Do for All Surface / Reflector Combinations(>100 crystals, each measured twice)

R-9800

• 6.15 x 6.15 x 25 mm3

• Reflector on 5 Sides• Optical Grease• No Hole on Top

Same PMT forall measurements

Surface & Reflector ResultsSurface & Reflector Results

Reflector Saw Cut Chemically MechanicallyEtched Polished

Air GapTeflon 1.00 ± 0.17 0.94 ± 0.10 1.06 ± 0.09ESR 1.01 ± 0.08 0.96 ± 0.16 1.08 ± 0.08Lumirror 1.03 ± 0.13 0.96 ± 0.06 1.04 ± 0.12

GluedESR 0.99 ± 0.09 0.98 ± 0.03 1.01 ± 0.18Lumirror 1.04 ± 0.10 0.97 ± 0.10 0.98 ± 0.22Melinex 1.01 ± 0.16 0.99 ± 0.06 1.01 ± 0.20Epoxy 1.00 ± 0.16 0.95 ± 0.09 1.00 ± 0.15

Paint 0.96 ± 0.03

Average

1.001.021.01

0.991.001.000.98

Average 1.01 0.96 1.03

Finished CrystalFinished Crystal

Etched Surface, White Primer Spray PaintEtched Surface, White Primer Spray Paint

Pulse Height PerformancePulse Height Performance

Good Energy Resolution & Crystal DecodingGood Energy Resolution & Crystal Decoding

0

1000

2000

3000

4000

5000

6000

7000

8000

0 20 40 60 80 100

Counts per Bin

Pulse Height Bin

Well-Coupled PMT Poorly-Coupled PMT

18% fwhm18% fwhm

0

200

400

600

800

1000

1200

0 10 20 30 40 50

Counts per Bin

Pulse Height Bin

Optimization: LSO CompositionOptimization: LSO Composition

• Predicted Timing Resolution 1/sqrt(I0)• Want High Total Light Output & Short Decay Time

• Possible By Co-Doping LSO With Calcium

• Predicted Timing Resolution 1/sqrt(I0)• Want High Total Light Output & Short Decay Time

• Possible By Co-Doping LSO With Calcium

0

20

40

60

80

100

120

0 50 100 150 200

25 ns decay time50 ns decay time

Intensity

Time (ns)

• Both Scintillators Have Same Light Output (photons/MeV)• Red Decay Time is 2x Longer Than Blue Decay Time

I(t) = I0 exp(-t/)

Light Output = I0

I0

25

30

35

40

45

50

0.95 1 1.05 1.1 1.15 1.2 1.25 1.3

Decay Time (ns)

Relative Light Output

Optimization: LSO CompositionOptimization: LSO Composition

Ca-Doping Gives High Light Output & Short Ca-Doping Gives High Light Output & Short

Normal LSO High Light Out

Short The Good Stuff!

= Ca-doped

0.1%0.2%0.4%

0.3%

LSO Composition Optimization MethodLSO Composition Optimization Method

Measure Time Resolution vs. Initial IntensityMeasure Time Resolution vs. Initial Intensity

• Get Samples of Normal & Co-Doped LSO

• Measure Decay Time

• Measure Relative Light Output

• Compute Initial Intensity I0

• Measure Timing Resolution

• Plot Timing Resolution vs. I0

R-9800

• 5 x 5 x 5 mm3

• Saw Cut Surface• Teflon on 5 Sides• Optical Grease

Same PMT forall measurements

• Ca-Doping Gives Good Timing Resolution• ~15% Improvement Over Normal LSO

• Ca-Doping Gives Good Timing Resolution• ~15% Improvement Over Normal LSO

150

160

170

180

190

200

210

220

230

80 100 120 140 160 180 200Time Resolution (ps fwhm)Initial Intensity (I

0)

Normal LSO

Scaled by1/sqrt(I0)

Measured Results: LSO CompositionMeasured Results: LSO Composition

= Ca-doped

0.1%

0.2%

0.4%0.3%

Optimization: Photomultiplier TubeOptimization: Photomultiplier Tube

• Predicted Timing Resolution 1/sqrt(QE)• Want High Quantum Efficiency Version of PMT

• Predicted Timing Resolution 1/sqrt(QE)• Want High Quantum Efficiency Version of PMT

Peak QE

Blue Sensitivity Index

PMT Optimization MethodPMT Optimization Method

Measure Time Resolution vs. Blue SensitivityMeasure Time Resolution vs. Blue Sensitivity

• Couple “Standard” LSO Crystal to PMT(“normal” LSO, etched surface finish,Lumirror reflector glued to five sides)

• Measure Timing

• Repeat for all PMTs

• Measure twice for each PMTR-9800

• 6.15 x 6.15 x 25 mm3

• Chemically etched• Lumirror reflector• Optical Grease• Same Crystal for

all measurements

190

200

210

220

230

240

250

10 11 12 13 14 15Time Resolution (ps fwhm)Blue Sensitivity Index

• Increased QE Improves Timing Resolution by 7%• Expect 10% Improvement with 35% SBA PMT

• Increased QE Improves Timing Resolution by 7%• Expect 10% Improvement with 35% SBA PMT

Normal (“28% QE”) PMTs

Measured Results: High QE PMTsMeasured Results: High QE PMTs

Scaled by1/sqrt(Blue Index)

= “32% QE” PMTs

• Intrinsic Timing Resolution is 63 ps fwhm• With Detectors, Same Timing as NIM Electronics

• Intrinsic Timing Resolution is 63 ps fwhm• With Detectors, Same Timing as NIM Electronics

Shaper

Shaper

4 ADCs

4 ADCsCFD

Sum

Sum

CFD

CERNTDC

FPGA

Out…

Based on Siemens “Cardinal” electronics.

CFD triggers if any of 4 adjacent modules fire.

CERN HPTDC digitizes arrival time w/ 24 ps LSB.

Pulse height from all 8 modules read out on every trigger.

FPGA uses pulse heights to identify interaction crystal.

FPGA also does calibration, event formatting, etc.

ElectronicsElectronics

SummarySummary

• TOF PET with Significantly Better Timing is Possible• To Achieve, We Must “Think Outside the Block Detector”

• TOF PET with Significantly Better Timing is Possible• To Achieve, We Must “Think Outside the Block Detector”

Hardware Single Coinc. TOF (ps fwhm) (ps fwhm) GainEnd-Coupled Crystal 384 544 4.3

Side-Coupled Crystal 218 309 7.6

Etched, Reflector Paint 227 321 7.3

Co-Doped LSO 182 258 9.1

32% QE PMT 155 219 10.6

35% QE “SBA” PMT 148 209 11.1

Future TOF PET Design?Future TOF PET Design?

•Depth of Interaction & 150 ps Timing Resolution•11x Reduction in Variance in Practical Geometry•Depth of Interaction & 150 ps Timing Resolution•11x Reduction in Variance in Practical Geometry

ScintillatorArray

ThinnedSiPM Array