time-of-flight positron emission tomography using cherenkov

Time-of-Flight Positron Emission Tomography Using Cherenkov Radiation

Rok Dolenec

Supervisor: Prof. Dr. Samo Korpar

Time-of-Flight Positron Emission TomographyUsing Cherenkov Radiation

Time-of-Flight Positron Emission TomographyUsing Cherenkov Radiation

Doctoral Thesis

2/24Time-of-Flight Positron Emission Tomography Using Cherenkov Radiation

Overview

● Introduction:– Positron emission tomography (PET)

– Time-of-flight (TOF) PET

– TOF PET with Cherenkov radiation

● Experiments:– Setup

– TOF resolution

– Detector efficiency

● Simulations:– Timing & Efficiency

– Intrinsic suppression of scatter events

– Possible efficiency improvements

● Image reconstruction● Conclusion


Positron Emission Tomography

● Nuclear medicine method, used for in-vivo imaging of live tissue

● Biomolecules, marked with radioactive isotope (eg. 18F), are concentrated by metabolism in tissue of interest (eg. cancerous growth)

● Isotope decays via β+ decay pairs of →collinear 511 keV gammas

● Position sensitive gamma camera surrounding the patient reconstruction of →activity distribution

● PET scanner - multiple rings of detectors:

– scintillating crystal (BGO - Bi4Ge3O12)

– photodetector (photomultiplier tubes - PMT)


Image reconstruction

● In principle:– Detection of two gammas line, on which the decay is presumed to occur →

(line of response — LOR)

– Image formed on 2-D matrix of discrete pixels

– Each pixel intersected by LOR is incremented (Simple Projection)

– Many LORs approximate image of activity distribution→

● In practice:– Filtered Back Projection (Fourier transforms, frequency filtering)

– Iterative reconstruction algorithms

1 event 4 events 64 events


Time-of-flight PET

● Non-TOF PET image reconstruction: all pixels along LOR are incremented for each event noise→

● Measurement of two gamma arrival times (TOF) position of decay →along LOR

● TOF PET: number of pixels incremented limited by TOF resolution →improved S/N ratio (contrast) of images

● TOF resolution mainly limited by– scintillator time response (scintillation decay time, raise time)– photodetector timing resolution

● Limitations due to scintillators can be avoided by using radiators of Cherenkov light instead– charged particles passing trough matter at speed vThr > c0/n → prompt

Cherenkov photons

non-TOF TOF


Cherenkov radiators

● To obtain Cherenkov photons from 511 keV annihilation gammas, they must transfer their energy to an electron in a suitable Cherenkov radiator:– high stopping power for 511 keV γ high → ρ, high Z

– high fraction of γ interactions via photoeffect — electron receives more energy than via Compton scattering high Z→

– high index of refraction – more electrons produced above Cherenkov threshold vThr

– good optical properties – transmission for visible & near UV Cherenkov photons

● Most promising available radiators: PbF2 and PbWO4 (PWO) crystals

PbF2

PbWO4

Refractive index 1.8 2.3

Density (g/cm3) 7.77 8.28

e- Cherenkov threshold (keV) 104 56

Optical transmission λcutoff

(nm) 250 350

Scintillation LY (ph/MeV) - 200

Scintillation decay time (ns) - 6/30

Scintillation peak (nm) - 440/530


Cherenkov photon production and detection

● Simulation results for PbF2 and PbWO4 radiators

– 25х25х15 mm3 large, black painted

– coupled to photodetector with realistic PDE

PbF2

PbWO4

Gammas interacting 79.7% 80.1%

Electrons produced 1.53 1.57

Ch. photons produced * 15.1 22.2

Ch.photons reaching photodetector 2.11 1.27

Detected Ch. photons 0.14 0.07

Detected scint. photons - 0.47

* in 200 - 800 nm wavelength range

PbF2

PbWO4

● More Cherenkov photons produced in PbWO4

● More are detected in PbF2 due to better optical transmission (lower λcutoff)


Cherenkov photon timing

● Cherenkov photons are produced promptly, but still need to reach the photodetector

– Radiator dimensions, refractive index travel time spread due to different →gamma interaction depths

– Different photon emission angles

– Reflections from radiator entry and side surfaces

● total internal reflection (high refractive index)● reflective wrapping

● Black paint reduces total internal reflections and stops many photons

– improved timing

– reduced detection efficiency (but from photons with worse timing)

phot

odet

ecto

r

511keV gamma Cherenkov photon

t = d∙n/c0 = 90 ps

phot

odet

ecto

r

t = d/c0 = 50 ps ∆→ t = 40 psd = 15 mm, n = 1.8:


63.2 psFWHM

Experimental setup

● Cherenkov radiators:

– monolithic: 25x25x5,15 mm3 (PbF2, PbWO4)

– 4x4 segmented: 22.5x22.5x7.5 mm3 (PbF2)

– black painted, Teflon wrapped, bare

● Photodetector:– 3 samples of Hamamatsu 16(=4x4) channel MCP PMT

– 22.5x22.5 mm2 active area, d = 10 μm microchannels

– Peak QE ~ 24%, collection eff. ~ 60% (one sample 30%)

– Excellent timing < 70 ps FWHM (incl. laser and electronics)


Experimental setup (2)

● Data analysis:

– time-walk correction

– maximum charge cut - only channels with maximum charge on their MCP PMT selected (to reduce crosstalk between anodes and supress background noise)

● 2 Cherenkov detectors in back-to-back configuration:

– D = 200 mm

– 22Na β+ point source


TOF resolution

● Best TOF resolution with readout of only two central anodes per MCP PMT– 74 ps FWHM (25x25x5 mm3, black painted

PbF2)

– wider contributions from● delayed events in MCP PMT time response● reflections on radiator surfaces

● With readout of 8 anodes per MCP PMT– 104 ps FWHM (25x25x5 mm3, black painted

PbF2)

– two additional peaks● detection of annihilation gammas in MCP,

now apparent due to lower QE of MCP PMTs used

● Difference between detection times on the two detectors (TOF measurements)


TOF resolution (2)

● TOF resolution for different radiator surfaces (15 mm thick PbF2):

– black painted: 121 ps FWHM, bare: 193 ps FWHM, Teflon wrapped: 284 ps FWHM

● TOF resolution for PbWO4 (black painted):

– 5 mm thick: 1.2 ns FWHM, 15 mm thick: 1.7 ns FWHM

– time distributions dominated by scintillation background


TOF resolution (3)

● Thicker radiator worse time resolution→

● Surface treatment

– black paint: suppresses most reflections mostly direct photons contribute to →timing (narrow peak)

– bare surfaces: more photons internally reflected due to high refractive index →more events with large time delay (wider distribution)

– Teflon wrapping: same as with bare surfaces, but even more delayed events due to reflections

● PbWO4:

– when both sides detect Cherenkov photon narrow peak→

– worse timing than PbF2 due to higher refractive index

– timing dominated by events due to scintillations

– simulations and estimates suggest worse efficiency (Cherenkov photons)

→ much worse performance than PbF2


Detection efficiency

● One of the Cherenkov detectors replaced with reference scintillation detector, which provided energy measurement

● Events detected on Cherenkov detector, when 511 keV on reference / 511 keV events on reference detector efficiency→

– corrected for events due to Compton scattering of 1275 kev gammas from 22Na source

● Results: 4.3% (5 mm thick, black painted PbF2) — 18% (15 mm thick, Teflon wrapped PbF2)


Measurements summary

● Detector efficiency (ε), coincidence detection efficiency (ε2), TOF resolution (FWHM) and Figure of merit (FOM = ε2/FWHM) for PbF2 Cherenkov radiators:Radiator ε [%] ε2 [%] FWHM [ps] FOM [%/ns]

5mm, black paint 4.3 0.18 103 1.8

15mm, black paint 6.1 0.37 132 2.8

5mm, bare 7.0 0.49 125 3.9

15mm, bare 12 1.44 186 7.7

5mm, Teflon 11 1.21 162 7.5

15 mm, Teflon 18 3.24 271 12.0

4x4, black paint 4.3 0.18 116 1.6

4x4, Teflon 8.6 0.74 301 2.5

Scintillator PET 10 300 33.3

● For comparison values for traditional, scintillator based PET system

● Cherenkov PET:

– Excellent timing but low efficiency up to 20x worse FOM→

– Teflon wrapping 3x worse FOM, but effects of long tails in timing distribution →not included in FOM value


Simulations - coincidence time distributions

● Without photodetector time response

– PbF2: 5 mm black painted, 15 mm black painted, 15 mm Teflon wrapped

● With measured MCP PMT time response included

– 15 mm black painted PbF2, 15 mm black painted PbWO4


Simulations - TOF resolution, efficiency

● Simulation results:

● Experimental results (reminder):

Radiator ε [%] ε2 [%] FWHM w/o photod. resp. [ps]

FWHM w/ photod. resp. [ps]

FOM [%/ns]

5mm, black paint 6.2 0.38 37.7 84.4 4.5

15mm, black paint 11 1.21 92.8 148 8.2

15mm, bare 16 2.56 118 215 11.9

15 mm, Teflon 18 3.24 130 241 13.4

4x4, black paint 6.7 0.45 45.7 92.2 4.9

4x4, Teflon 12 1.44 62.5 154 9.4

Radiator ε [%] ε2 [%] FWHM [ps] FOM [%/ns]

5mm, black paint 4.3 0.18 103 1.8

15mm, black paint 6.1 0.37 132 2.8

15mm, bare 12 1.44 186 7.7

15 mm, Teflon 18 3.24 271 12.0

4x4, black paint 4.3 0.18 116 1.6

4x4, Teflon 8.6 0.74 301 2.5


Simulations - Intrinsic suppression of scatter events

● Annihilation gammas scatter in patient or detector unwanted →background when scattered gamma is detected in coincidence

● Traditional PET

– number of scintillation photons proportional to energy deposited

– measurement of gamma energy rejection of scattered (lower energy) events→

● Cherenkov PET

– at most a few photons detected no energy information available→

– detection efficiency drops with gamma energy intrinsic → suppression


Simulations - efficiency improvements● Photodetector:

– improved photon detection efficiency● photocathode with better QE

● window, transmissive to lower λ (quartz 160 nm)→

– example: Hamamatsu 500S photocathode

→ 1.4x efficiency (2x FOM)

– Aging — photocathode of MCP PMT used in efficiency measurements might have been degraded 2x lower experimental efficiency than expected→

● Transport of photons from radiator to photodetector:– optimal coupling (n=1.9), window refractive index (n=2.0) → 1.4x efficiency (2x FOM)

– photonic crystals (periodic structure exit surface — guides photons out instead of total reflection) → 1.5x efficiency (~2x FOM)

● Hypothetical, PbF2-like radiator optimization (using 500S photocathode):

– refractive index, thickness (n=2.0, d~14mm) → 1.5x efficiency (3x FOM)

– optical transmission (λcutoff = 160 nm) → 2.4x efficiency (8x FOM)


Reconstruction

● Cherenkov PET tested experimentally– data equivalent to one PET ring obtained with only two

detectors

– source rotated in discrete steps

– data collected at each step for the same amount of time

– D = 185 mm, H = 22.5 mm

● Full body PET scanner simulated– D = 800 mm, 15 rings (H = 340 mm)

– phantom with d = 270 mm, 4 hot spheres (d: 10 - 22 mm) and 2 cold spheres (d = 28, 37mm)

● Reconstruction algorithms:– Filtered backprojection (FBP): basic non-TOF algorithm

– TOF weighted FBP: pixels along LOR incremented with TOF response defined weight

– Most likely position (MLP): point of decay on LOR calculated from TOF information

– Filtered MLP: MLP image deconvoluted for TOF response


Reconstruction - experiment

● 22Na point source at +10 mm and -10 mm

● 30 min acquisition time (A = 3200 kBq)

● 4x4 segmented, black painted PbF2 radiators

TOF w. FBP MLP Filtered MLP(non-TOF) FBP


Reconstruction - simulation

● Hot spheres activity concentration: 3x phantom background

● Statistics equivalent to 163 s of PET examination

● 4x4 segmented, Teflon wrapped PbF2 radiators

● 20 mm thick axial slices



Reconstruction - simulation (2)● Contrast and contrast-to-noise ratios (CNR) of hot spheres:

black painted Teflon wrapped

● Black painted (better TOF resolution) better contrast, Teflon wrapped (higher →statistics) better CNR (despite the tails in timing distributions)→

● TOF information significantly improves CNR● simple, very fast MLP very good CNR→


Summary

● TOF PET with Cherenkov light– excellent time resolution: <100 ps — 300 ps– low efficiency: 4 — 18% → 3x — 20x worse FOM than scintillator TOF PET– despite no energy measurement → intrinsic supression of scattered events

● Improvements possible:– Available technologies:

● photodetector QE: up to 2x FOM (8x if MCP PMT photocathode was aged)● better light extraction: up to 4x FOM

– More optimal Cherenkov radiator: up to 8x FOM● TOF PET image reconstruction:

– excellent time resolution:● significantly improved images (CNR)● faster, simpler reconstruction algorithms possible

– despite worse time resolution, tails reflective radiator wrapping > black →paint

● Using thicker, Teflon wrapped radiators with some of the described efficiency improvements Cherenkov could surpass traditional scintillator TOF PET→


Backup slides


Experimental setup - electronic readout

● 2 or 8 anodes (out of 16)/MCP PMT charge & time→

● other anodes summed charge→

● readout boards (& HV distribution circuits)

– provided by the producer

– designed in-house (shorter connector lengths)


Reconstruction - simulation

● 4x4 segmented, Teflon wrapped (top) vs. black painted (bottom) PbF2 radiators



Reconstruction - experiment (BW)

● 22Na point source at +10 mm and -10 mm

● 30 min acquisition time (A = 3200 kBq)

● 4x4 segmented, black painted PbF2 radiators



Reconstruction - simulation (BW)

● Hot spheres activity concentration: 3x phantom background

● Statistics equivalent to 163 s of PET examination

● 4x4 segmented, Teflon wrapped PbF2 radiators

● 20 mm thick axial slices



Simulation

● Simulation constructed in GEANT4 framework

● Takes into account all important physical processes:

– gammas: photoeffect, Compton scattering

– e—: multiple scattering, ionization, annihilation (e+)

– optical photons: Cherenkov, scintillation process; absorption, boundary processes

● Incorporates important parameters of detectors used in experiments:

– radiator dimensions, surface treatment

– coupling to the photodetector

– photodetector PDE and optionaly, time response

● Surface treatments:

– black paint: n=1.5, R=0.0

– bare surface: n=1.0, R=0.0

– Teflon wrapping: n=1.0, R=1.0

reflectivity (R)

refractive index (n)

radiator


MCP PMT timing

● Surfaces of MCP PMTs illuminated with very (single photon level) weak red (636 nm) and blue (404 nm) laser light pulses

● Time responses of 3 MCP PMT samples (incl. laser and electronics):

Red

Blue

JY0002 JY0005 JY0008

58.7 psFWHM

63.2 psFWHM

63.4 psFWHM

74.8 psFWHM

83.4 psFWHM


Simulations - efficiency improvements

● Photodetector:

– improved photon detection efficiency

● photocathode with better QE

● window, transmissive to lower λ (quartz 160 nm)→

– example: Hamamatsu 500S photocathode → 1.4x efficiency (2x FOM)

– Aging — photocathode of MCP PMT used in efficiency measurements might have been degraded 2x lower experimental efficiency than expected→

● Transport of photons from radiator to photodetector

– coupling, window refractive index (n=1.9, 2.0) → 1.4x efficiency

– photonic crystals (periodic structure on crystal exit surface — guides photons out instead of total reflection) → 1.5x efficiency

P. Lecoq, Progress on Photonic Crystals, 2010 IEEE Nuclear Science SymposiumConference Record.


Simulations - efficiency improvements (2)

● Hypothetical PbF2-like radiator optimization (using 500S photocathode)

– refractive index, thickness (n=2.0, d~14mm) → 1.5x efficiency (3x FOM)

– optical transmission (λcutoff = 160 nm) → 2.4x efficiency (8x FOM)

time-of-flight positron emission tomography using cherenkov

Documents