Slides (.pdf) - EPFL

Demonstration of a
novel PET concept
Chiara Casella, ETH Zurich
EPFL Lausanne, March 12 th 2012
. .

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Outline
AX-PET : AXial Positron Emission Tomography (PET)
•Brief introduction about PET (Positron Emission Tomography)
•Axial concept
What is it ? Why?
•AX-PET detector
choice of the detector elements
AX-PET modules
“demonstrator” for a PET scanner
•AX-PET detector performance
from characterization measurements with 22-Na sources
•Simulations
•Tomographic image reconstruction
description of the reconstruction methods
measurements campaigns with phantoms and radiotracers
reconstructed images
•Perspectives
preliminary results with Digital Si-PM as alternative photodetectors
•Conclusions
Chiara Casella, 12/3/2012 1

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
AX-PET in the context of HEP
‣ Long tradition of using technologies from High Energy Physics into other fields,
and particularly medical applications.
‣ AX-PET : small size calorimeter, using scintillating crystals, WLS, photodetectors
“borrowed” from HEP
Selection of most relevant AX-PET papers :
Novel Geometrical Concept of a High Performance Brain PET Scanner
Principle, Design and Performance Estimates
J. Seguinot et al, Nuovo Cimento C Vol 29, No 4, pp 429-463 (2006)
High precision axial coordinate readout for an axial 3-D PET detector
module using a wave length shifter strip matrix
A. Braem et al, NIM A 580 (2007), 1513-1521
Wavelength shifter strips and G-APD arrays for the read-out of the
z-coordinate in axial PET modules
A. Braem et al, NIM A 586 (2008), 300-308
The AX-PETdemonstrator—Design, construction and characterization
P. Beltrame et al, NIM A 654 (2011) 546-559
The AX–PET Concept: New Developments And Tomographic Imaging
P.Beltrame et al, 2011 IEEE NSS Conference Record MIC 22-5
axial concept,
HPD readout
WLS strips
G-APD as photodetectors
Characterization
& performance
Tomographic images
for more details : https://twiki.cern.ch/twiki/bin/view/AXIALPET/WebHome
2

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Positron emission / annihilation
‣ basis of the PET system Positron Emission :
‣ β + decay of different radionuclides
Radionuclide Half life Emax_e+(MeV)
11 C 20.4 min 0,96
15 O 122 sec 1,73
18 F 109,8 min 0,63
22 Na 2.6 years 0,55
‣ Positron Annihilation :
2 photons emitted
‣“back - to - back”
‣ Eγ = 511 keV
Unstable
parent
nucleus
p → n + e + + νe
e + e − → γγ
(Eγ = 511 keV )
p → n + e + + νe
e + e − → γγ
(Eγ = 511 keV )
positron
annihilation
positron emission
Physics of the positron annihilation => fundamental limits to the spatial resolution of PET
‣ Finite positron range (ρ)
annihilation position ≠ emission point
ρ depends on the energy of the positron (i.e. on the radioisotope)
‣ Non-collinearity of the 2 photons
residual momentum of the e+e- at the annihilation
the 2 photons are emitted with a small deviation from 180˚ (Δθ ~ 0.5˚)
blurring of the spatial resolution R_FWHM ~ 0.0022 x D [mm]
3

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Positron Emission Tomography : General principles
PET detector principle :
coincidence of 2 photons of defined energy (511 keV) and emitted on the same line (“back to back”)
(3)
(2)
(1)
(5)
(4)
(1) Inject the radiotracer into the body
(2) Wait for uptaking period
(3) Start the acquisition (i.e. detection of coinc. events)
(4) Feed the data into the reconstruction algorithms
(5) image of the activity concentration
‣ PET : “in-vivo” functional imaging technique
‣ get a (quantitative) image of the radio-tracer concentration
A.Del Guerra, CERN Academic Training 2009
4

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Requirements for an ideal PET
number of counts does matter !!!
50000 counts 100000 counts 200000 counts 500000 counts 1M counts 2M counts
‣ how to improve counting statistics ?
‣ increase radio-tracer concentration
‣ extend the scan duration time
‣ OPTIMIZED DETECTOR SENSITIVITY
( geometry / materials )
‣ HIGH SPATIAL RESOLUTION (=> be able to resolve small structures)
‣ GOOD TIMING RESOLUTION (small coincidence window => reduce random coincidences)
‣ GOOD ENERGY RESOLUTION (=> reject scattering events in the body)
A.Del Guerra, CERN Academic Training 2009
5

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Requirements for an ideal PET
number of counts does matter !!!
2009
Training Academic CERN Guerra,
50000 counts 100000 counts 200000 counts 500000 counts 1M counts 2M counts A.Del
‣ how to improve counting statistics ?
‣ increase radio-tracer concentration
‣ extend the scan duration time
‣ OPTIMIZED DETECTOR SENSITIVITY
( geometry / materials )
‣ HIGH SPATIAL RESOLUTION (=> be able to resolve small structures)
‣ GOOD TIMING RESOLUTION (small coincidence window => reduce random coincidences)
‣ GOOD ENERGY RESOLUTION (=> reject scattering events in the body)
5

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
δp
High resolution vs high sensitivity
in “conventional”
PET scanners :
θ
‣ always a compromise between
!"
δp = L · sinθ
‣ solution : add DOI (Depth Of Interaction) information
‣ several attempts / different strategies
scintillator based
radial arrangement
ɛ =1− e −µ·L max interaction efficiency,
long L
min parallax error
- deterioration of the spat. resol.
- non uniformity in the field of view
short L
‣ good spatial resolution (small L, small δp)
‣ good detection efficiency (long L)
6

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
DOI attempts :
(a) multiple
crystals/
photodetector
layers
Conventional PET schemes
(b) dual-ended
photodetectors
DOI from the ratio of
the 2 PD responses
only partial DOI information !
(c) Phoswich design
different types of
crystals with diff decay
times
=> DOI from pulse
shape discrimination
(e) dual layer offset
position.
DOI from different light output
profiles
need a precise 3D identification of the photon interaction point
(d) monolithic. Iterative stat.
models with DOI from light output
intensity and/or light spread profiles
Peng, Levin - Current Pharmaceutical
Biotechnology, 2010, Vol 11, No. 6
7

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
The AXIAL concept
AX-PET approach to the DOI problem : change the geometry !
from radial ... ... to axial !
‣ long and axially oriented crystals
‣ DOI information = position of the hit crystal
8

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
axial direction
The detector concept
(1) TRANSAXIAL COORDINATE (x,y)
scintillator crystals
- Transaxial coordinate: from position of the hit crystal
- Transaxial resolution = d/√12 FWHM
Y
- To increase spatial resolution => Reduce crystals size (d)
- To increase sensitivity => Add additional layers
Z
X
d
9

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
axial direction
The detector concept
(1) TRANSAXIAL COORDINATE (x,y)
scintillator crystals
- Transaxial coordinate: from position of the hit crystal
- Transaxial resolution = d/√12 FWHM
Y
- To increase spatial resolution => Reduce crystals size (d)
- To increase sensitivity => Add additional layers
Z
X
d
WLS (Wave Length Shifter) array
w
(2) AXIAL COORDINATE (z)
scintillation
process
wls
photodetector
ʻescaping coneʼ
(outside total
internal reflection)
crystal
reflective Al
coating
- Axial coordinate : center of gravity method
- Axial resolution < w
Z
9

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
axial direction
The detector concept
(1) TRANSAXIAL COORDINATE (x,y)
scintillator crystals
- Transaxial coordinate: from position of the hit crystal
- Transaxial resolution = d/√12 FWHM
Y
- To increase spatial resolution => Reduce crystals size (d)
- To increase sensitivity => Add additional layers
Z
X
3D localization of the photon interaction point
without compromising between spatial resolution and sensitivity
d
WLS (Wave Length Shifter) array
w
(2) AXIAL COORDINATE (z)
scintillation
process
wls
photodetector
ʻescaping coneʼ
(outside total
internal reflection)
Z
crystal
reflective Al
coating
- Axial coordinate : center of gravity method
- Axial resolution < w
9

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
AX-PET Collaboration
10

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
AX-PET GOAL : PET demonstrator
Goal of the AX-PET collaboration:
Build and fully characterize a “demonstrator” for a PET
scanner based on the axial concept. Assess its performances.
Demonstrator =>
Two identical AX-PET modules, used in coincidence
Characterization / Performance =>
- test each individual module in a dedicated setup
- characterization in the coincidence setup
- reconstruction of the images of extended objects
- simulations
11

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
detect photons, Eγ = 511 keV =>
bare scintillating crystals with escaping light =>
Detector choice: Scintillator
• inorganic scintillator
• high Z, high ρ
• non hygroscopic
• nude crystals (i.e. unwrapped, uncoated)
• perfectly polished surfaces ; sharp edges
LYSO crystals (Lu1.8Y0.2SiO5: Ce) Prelude 420 from Saint Gobain :
ZLu = 71
ZLu = 71
176-Lu is a naturally radioactive β emitter:
• lyso bars : 3 x 3 x 100 mm (A ~ 39 Bq/g; β-decay followed by a γ-cascade)
useful for the calibration
3 each
• Al coated on the non-readout face (R ~ 80-85%)
• measured : λ_opt = (412 ± 31) mm ; (ΔE/E)_intr = (8.3 ± 0.5) % FWHM , @511 keV
12

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
• requirement : high absorption of the crystal
scintillation photons (blue)
• EJ-280-10x from Eljen Technology
• blue to green WLS
• highly doped (x10 higher dye concentration
than standard) to maximize the absorption
(absorption length _ blue light ~ 0.4 mm)
• each WLS strip : 3 x 0.9 x 40 mm 3
• decay time = 8.5 ns
Detector choice: WLS strips
• measurements : λ_opt = (188 ± 36) mm 4 cm
photodetector side
mirror coated side
13

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
‣ newest frontier in photo-detection:
‣ excellent photon counting capabilities
Detector choice: Photodetectors
‣ array of commonly biased APD used in Geiger mode
‣ output : analogue sum of all the firing APD cells
‣ high gain (10 5 to 10 6 ) at low bias V (~ 70 V)
‣ advantages of a Si sensor:
‣ high QE √
‣ compactness √
‣ insensitive to magnetic field (MRI comp.)
‣ temperature dependent √
‣ dark rate (~ 1 MHz @ thr = 0.5 pe)
SiPM (Si photomultipliers)
also called G-APD (Geiger-mode APD)
also called MPPC (Multi Pixel Photon Counter) - from Hamamatsu
√
√
√
√
for crystals for WLS
~ 1000 pe ~ 10 - 50 pe
MPPC S10362‐33‐050C :
• 3x3 mm 2 ac(ve area
• 50 μm x 50 μm pixel
• 3600 pixels
MPPC 3.22×1.19 Octagon‐SMD :
• 1.2 x 3.2 mm 2 ac(ve area
• 70 μm x 70 μm pixel
• 782 pixels
• custom made units
AX-PET operational
parameters
14

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
LYSO crystals
3 mm
10 cm
3.5 cm
2.8 cm
WLS
The AX-PET module
MODULE
• 6 layers
• 8 crystals / layer
• 26 WLS / layer
• 48 crystals + 156 WLS = 204 channels
• staggering in the crystals layout to optimize photon interaction probability
• optical separation between layers
• each crystal and WLS strip individually coupled to its photodetector
• crystals / WLS strips readout on alternate sides (to optimize packing density)
15

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
The AXPET module
16

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Individual analogue readout of MPPC output
Custom designed DAQ system
‣ fully analogue readout chain
‣ not optimized at all for this specific
application
‣ Amplifiers: OPA486 (Lyso) / OPA487
(WLS)
‣ Fast energy sum for all the crystals in
the module
‣ VATA GP5 chip
- 128 ch charge sensitive integrating
- fast (~ 50ns shaper + discriminator) /
slow (~ 250ns shaper) branches
- sparse readout mode: only the
channels above thr are multiplexed into
the output
‣ analogue info processed by custom made
VME ADC
Readout & DAQ
17

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Individual analogue readout of MPPC output
Custom designed DAQ system
‣ fully analogue readout chain
‣ not optimized at all for this specific
application
‣ Amplifiers: OPA486 (Lyso) / OPA487
(WLS)
‣ Fast energy sum for all the crystals in
the module
‣ VATA GP5 chip
- 128 ch charge sensitive integrating
- fast (~ 50ns shaper + discriminator) /
slow (~ 250ns shaper) branches
- sparse readout mode: only the
channels above thr are multiplexed into
the output
‣ analogue info processed by custom made
VME ADC
Readout & DAQ
17

PPC
PPC
intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
‣ analogue sum of the whole module
(i.e. total energy over 48 crystals)
Fast energy sum & Trigger
‣ with a proper threshold choice (LL x HL X notHHL)
=> select only events with 511 keV total
energy deposition
TRIGGER
OPA486
OPA486
test
! E (crystals only)
test
ΣE_Mod1
! E (crystals only)
MODULE1
fast
ΣE_Mod2
LYSO SUM
MODULE2
LYSO SUM
fast
‘slow’
VATAGP5 (128 ch.)
‘slow’
VATAGP5 (128 ch.)
First estimate of the time resolution:
•! measure delay of coincidence wrt Module2
•! measurement from the scope [Lecroy Waverunner LT584 L 1GHz]
S&H
S&H
Channels
above threshold
Analog Channels output:
above Light threshold per
crystal / WLS
Analog output:
Light per
crystal / WLS
LL
HL
HHL
Mod 1, energy sum (scope measurement), 22 Na source
TRIGGER = 2 modules
- each one discriminated @ 511 keV energy sum
- used in coincidence
=> Selection of the good events
18

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Compton scattering in the detector
‣ fast energy sum & trigger
‣ high granularity in the module
‣ good energy resolution
Compton scattering
E1 + E2 = 511 keV
if the event is :
• correctly identifiable (e.g. Compt. kinematics)
=> the event can be included in the reconstruction => improved sensitivity
• ambiguous (1-2 OR 2-1 ?)
=> the event is rejected => improved resolution
}
Compton scattering in the body
Events killed by the trigger !
Possibility to tag Compton scattered events
in the detector
“Inter-Crystal Scattering” events (ICS)
E1 + E2 = 511 keV
win - win situation !
Esum = 511 keV Esum < 511 keV
19

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
AX-PET inspired other developments
COMPET :
University of Oslo, Norway - E. Bolle et al.
research project for a pre-clinical PET scanner with high sensitivity,
high resolution. MRI compatible
- no axial geometry
- 3D reco of photon interaction point with LYSO + WLS + G-APD
E. Bolle et al, NIM A 648(2011) S93-S95
Tampere University (Finland) :
build a small specific scanner based on AX-PET (toward possible commercialization...)
Low cost planar detector for PET
Triumph, Canada - F. Retiere et al.
F. Retiere, NIM A (2011) doi:10.1016/j.nima.2011.12.084
20

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
single module
characterization
Characterization measurements
with point-like 22 Na source (diam = 0.25 mm, A~900 kBq) , @ CERN
2-D
moving
station
tagger
PMT
22Na source
module
- two different taggers
- different distances tagger / source / module
=> both uniform illumination of the module and precisely collimated beam spot
module
22Na source
2-D
moving
station
module
characterization
of modules
in coincidence
AX-PET performance:
NIM A 654 (2011) 546-559
1. Energy measurement / energy calibration => LYSO response
2. Position measurement / spatial resolution => WLS response
21

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
LYSO No. 21 - 22Na coinc. trigger
450
400
350
300
250
200
150
100
50
0
Lu X-ray
photopeak
Compton
0 100 200 300 400 500 600 700
ADC counts
typical energy spectrum of one LYSO inside the
module :
‣ photopeak (511 keV)
‣ Compton continuum (0 - 340 keV)
‣ Lu X-ray peak (~ 55 keV)
LYSO energy response
511 keV
all events
N_lyso = 1
fit: gaussian
55 keV
511
keV 511
keV
Light yield at 511 keV ~ 1000 pe
(from independent calibration measurements)
< 511 keV
R_FWHM (@511 keV) [%]
Energy resolution
14
13.5
13
12.5
12
11.5
11
10.5
Energy resolution
‣ from gaussian fit of the photopeak
‣ AFTER ENERGY CALIBRATION
Module 1
Module 2
10
0 20 40 60 80 100
LYSO_ID
< R_FWHM > ~ 11.8% @511 keV
(averaged on 96 LYSO crystals)
22

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
LYSO energy calibration
Photopeak + Intrinsic Lu radioactivity: very good tool for the energy calibration
‣ LYSO contains Lu-176
‣ A ~ 39 cps/g
=> ~ 250 Bq / bar
=> ~ 12 kHz / module
LYSO No. 21 - 22Na coinc. trigger
450
400
350
300
250
200
150
100
50
0
0 100 200 300 400 500 600 700
LYSO No. 21 - intrinsic radioactivity
[511 keV]
ADC counts
600
500
400
300
200
100
0
with source,
coincidence
0 100 200 300 400 500 600 700
[303 keV]
ADC counts
[55 keV]
[202 keV]
all events
no source,
internal trg.
N_lyso = 1
fit: gaussian
Peak position [ADC counts]
Energy Calibration ( LYSO No. 21)
500
450
400
350
300
250
200
150
100
50
same procedure applied
identically for every channel
deviation from linearity
(~ 5% effect)
En(ADC) =E0 − a × ln 1 − ADC
b
0
0 100 200 300 400 500
Energy [ keV ]
MPPC saturation. Due to:
- limited nr of cells in the MPPC (3200)
- important light yield in the scintillator (~1000).
23

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
WLS response
typical integrated raw spectra of few WLS strips
• beam spot collimated at the center of the module (WLS 13)
• 511 keV energy deposition in the LYSO
Collimated beam spot, WLS response
1800 WLS Nr. 13 (central)
1600
1400
1200
1000
800
600
400
200
WLS Nr. 14 (side)
WLS Nr. 24 (peripheral)
0
0 100 200 300 400 500 600 700 800
ADC counts
• more than1 WLS participate to the event (typically 2-4)
• noise should not be included
WLS cluster: Summed ADC counts
1200
1000
800
600
400
200
0
0 100 200 300 400 500 600 700 800
ADC counts
Light yield in WLS cluster ~ 100 pe
@511 keV LYSO energy deposition
(from independent calibration measurements: 1pe ~ 4 ADC)
axial coordinate :
derived from center of gravity method
from all the WLS participating to the cluster
cluster
noise
24

layer1
clustPPLayCutsum_Zmm_lay1
Z c.grav. in layer 1 (PPeak samelay 1!cutsum)
Entries 4032
400
350
300
250
200
150
100
50
0
1000
800
600
400
200
0
intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Mean 48.4
RMS 3.356
2
! / ndf 253.2 / 10
Prob 0
p0 304.8 ± 7.5
p1 48.13 ± 0.06
p2 2.213 ± 0.066
35 40 45 50 55 60
z [mm]
clustPPLayCutsum_Zmm_lay4
Z c.grav. in layer 4 (PPeak samelay 1!cutsum)
Entries 9623
Mean 48.08
RMS 2.868
2
! / ndf 10.71 / 7
Prob 0.152
p0 1056 ± 16.9
p1 47.91 ± 0.02
p2 1.501 ± 0.026
35 40 45 50 55 60
z [mm]
layer2
layer4 layer5
z coordinate (COG) :
clustPPLayCutsum_Zmm_lay2
Z c.grav. in layer 2 (PPeak samelay 1!cutsum)
Entries 5716
500
400
300
200
100
0
1200
1000
800
600
400
200
0
Mean 48.47
RMS 3.017
2
! / ndf 17.16 / 8
Prob 0.02848
p0 561.3 ± 11.8
p1 48.4 ± 0.0
p2 1.744 ± 0.043
35 40 45 50 55 60
z [mm]
clustPPLayCutsum_Zmm_lay5
Z c.grav. in layer 5 (PPeak samelay 1!cutsum)
Entries 10762
Mean 48.32
RMS 3.37
2
! / ndf 30.27 / 6
Prob 3.489e-05
p0 1218 ± 19.0
p1 48.24 ± 0.02
p2 -1.391 ± 0.024
35 40 45 50 55 60
z [mm]
layer3
500
400
300
200
100
0
Axial spatial resolution
clustPPLayCutsum_Zmm_lay3
Z c.grav. in layer 3 (PPeak samelay 1!cutsum)
900
Entries
Mean
7442
48.36
800
RMS
2
! / ndf
3.099
73.49 / 8
700
Prob
p0
9.906e-13
765.8 ± 13.9
600
p1 48.17 ± 0.03
p2 1.587 ± 0.029
3000
2500
2000
1500
1000
500
0
35 40 45 50 55 60
z [mm]
layer6
clustPPLayCutsum_Zmm_lay6
Z c.grav. in layer 6 (PPeak samelay 1!cutsum)
Entries 17215
each measured σz includes :
• intrinsic spatial resolution
• positron range (ρ ~ 0.54 mm)
• non collinearity (÷d)
• beam divergency (÷d)
• (source dimensions; ø=250μm)
Mean 48.08
RMS 3.356
2
! / ndf 39.68 / 3
Prob 1.248e-08
p0 2854 ± 36.7
p1 48.07 ± 0.01
p2 -0.876 ± 0.012
35 40 45 50 55 60
z [mm]
Intrinsic axial spatial resolution
Mod 1 : 1.75 mm FWHM
Mod 2 : 1.83 mm FWHM
Trans-axial spatial resolution
• = d/√12 = 0.87 mm (digital) ; FWHM ~ 2 mm
• = d/√12 = 0.87 mm (digital) ; FWHM ~ 2 mm
z
x
y
layer6
layer1
25

Y[mm]
intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
TOP View - d(Mod1,Mod2) = 150 mm
100
50
0
-50
-100
N_coinc_evts=1
-100 -50 0 50 100
X[mm]
Two modules coincidences
Z[mm]
SIDE View - d(Mod1,Mod2) = 150 mm
100
50
0
-50
-100
AX-PET very first coincidence event !
N_coinc_evts=1
-100 -50 0 50 100
X[mm]
26

Y[mm]
intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
TOP View - d(Mod1,Mod2) = 150 mm
100
50
0
-50
-100
-100 -50 0 50 100
X[mm]
Two modules coincidences
N_coinc_evts=10
Z[mm]
SIDE View - d(Mod1,Mod2) = 150 mm
100
50
0
-50
-100
N_coinc_evts=10
-100 -50 0 50 100
X[mm]
26

Y[mm]
intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
TOP View - d(Mod1,Mod2) = 150 mm
100
50
0
-50
-100
-100 -50 0 50 100
X[mm]
Two modules coincidences
N_coinc_evts=100
Z[mm]
SIDE View - d(Mod1,Mod2) = 150 mm
100
50
0
-50
-100
N_coinc_evts=100
-100 -50 0 50 100
X[mm]
26

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Z[mm]
SIDE View - d(Mod1,Mod2) = 150 mm
100
50
0
-50
-100
-100 -50 0 50 100
X[mm]
Rintr =
N_coinc_evts=100
Z_projection with X= 0.00
1200
1000
800
600
400
200
R 2 meas − R 2 ρ − R 2 180
(0.54 mm) 2
Axial resolution
Intersection of LOR with central plane
no tomographic reconstruction Z_proj Z_proj !!!
Entries Entries 20000 20000
Mean Mean 0.1968 0.1968
RMS RMS 1.104 1.104
22
! ! / / ndf ndf 53.9 53.9 / / 21 21
Prob Prob 0.0001019 0.0001019
Constant Constant 1125 1125 ± ± 11.7 11.7
Mean Mean 0.1977 0.1977 ± ± 0.0059 0.0059
Sigma Sigma 0.6485 0.6485 ± ± 0.0064 0.0064
0
-10 -5 0 5 10
Z[mm]
≈ 1.35 mm
[0.0022 x Diam = 0.33 mm] 2
2200
2000
1800
1600
1400
(R_FWHM)z 1200 ~ 1.5 mm
Y_projection with X= 0.0
1000
including :
• intrinsic resolution 800
• positron range
600
• non collinearity
• (source dimensions 400 ; ø=250μm)
∼
200
0
-10 -5 0
(R F W MH)z SingleMod
(2)
27

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
δp
F2F
OBL
θ
!"
Parallax free demonstration
parallax error is more
and more important
outside the center of
the FOV
intersection of LORs with the plane containing the source
14000
12000
10000
8000
6000
4000
2000
0
160
140
120
100
80
60
40
20
0
3
" 10
Z_projection with X'= 0.00
axial trans-axial
Z_proj
Entries 226472
Mean 0.07255
RMS 0.9407
2
! / ndf
438 / 22
Prob 0
Constant 1.264e+04 ± 3.845e+01
Mean 0.07968 ± 0.00172
Sigma 0.6546 ± 0.0018
-4 -2 0 2 4
Z[mm]
50000
40000
30000
20000
10000
Y'_projection with X'= 0.00 and expected Y'= 0.00
0
Y_proj
Entries 226472
Mean -0.079
RMS 0.8928
-4 -2 0 2 4
Y'[mm]
Z_projection with X'= 0.00 Z_proj
Y'_projection 3 with X'= 0.00 and expected Y'=-45.00
" 10
Y_proj
Entries 2875919
Mean 0.1839
RMS 0.9481
2
! / ndf
1540 / 21
Prob 0
Constant 1.62e+05 ± 1.39e+02
Mean 0.1946 ± 0.0005
Sigma 0.6431 ± 0.0005
-4 -2 0 2 4
Z[mm]
350
300
250
200
150
100
50
Entries 2875919
Mean -44.65
RMS 0.9086
-49
0
-48 -47 -46 -45 -44 -43 -42 -41 -40
Y'[mm]
-50
-50 -40 -30 -20 -10 0 10 20 30 40 50
Y'[mm]
Intrinsic resolution is not degraded by parallax effects, even in very oblique configuration !
Z[mm]
Z[mm]
50
40
30
20
10
0
-10
-20
-30
-40
-30
-40
ZY'_projection with X'= 0.00 and expected Y'= 0.00
X_proj
Entries 226472
Mean x
Mean y
-0.08777
8000
0.07623
RMS x
RMS y
1.103
1.818 7000
axial transaxial
F2F σ=0.655 RMS=0.893
ZY'_projection with X'= 0.00 and expected Y'=-45.00
X_proj
50
40
30
Entries
Mean x
Mean y
RMS x
RMS y
2875919
-44.61 45000
0.1884
1.115 40000
1.835
35000
20
10
OBL 0 σ=0.643
30000
25000
RMS=0.909
-10
20000
-20
15000
-50
-50 -40 -30 -20 -10 0 10
Y'[mm]
6000
5000
4000
3000
2000
1000
0
10000
5000
0
28

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
AX-PET is a fully simulated system
Simulations
Dedicated Monte Carlo simulations for AX-PET .
Why? (1) get a better understanding of the detector
(2) train Compton reconstruction algorithms
(3) support image reconstruction
(4) simulate an hypothetical full ring scanner
and estimate its potential performance
GATE simulation package (Geant4 application for tomographic emission)
Non conventional nature of AX-PET
=> several challenges for simulations (standard templates could not be applied)
- geometry (axial; staggering; layered structure)
- WLS modeling
- dedicated sorter for the coincidences
(module sum; trigger logic... up to DAQ rate modeling)
29

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Simulations: some results
Excellent agreement data / simulations :
One AXPET Module
illuminated by a collimated
511 keV gamma beam :
Data and Simulations
Two AXPET Modules in coincidences
!"#$%&'()%*%+,)%-($,(&$.'+-(/).0+)%-($$
LYSO energy spectrum LYSO multiplicity
1%**'.'()$%&'()%*%+,)%-($,23-.%)45/$,.'$)'/)'&$,(&$)4'%.$'**%+%'(+6$%($$!"#$
.'+-(/).0+)%-($%/$'/)%5,)'&$-($/%502,)%-(/7
! 82'%(9:%/4%(, ;,/'& -( 3'-5').6 -. '('.36<
! =,>%505 ?('.36<
! "-5@)-( 8%('5,)%+/ A"8B<
! :'0.,2 :')C-.D7
#%502,)%-($%/$@'.*-.5'&$;6$0/%(3$EFP$'('.36$.'/-20)%-($,)$QEE$D'ST$C%)4$@-%()92%D'$/-
%($)4'$+'().'$-*$)4'$UVST$;,+D9)-9;,+D$3,55,$'5%//%-(T$F$5-&02'/$,)$WQ$55$&%/),(+
• identify ICS events before image reconstruction.
• several identification algorithms tested
axial resolution from WLS strips Inter-Crystal Scattering (ICS)
!"#$%& '()*+(,%-$ -./0,120340," 2/56". 2/+7(683
OEP OQP9OOP OEP9OIP RQP
• identification rate for ICS ~ 60%
very incomplete collection of results here !
simulations paper is in preparation !
EFGHIGHJ #%502,)%-($,(&$%5,3'$.'+-(/).0+)%-($*-.$KL9M?N
30

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Towards a tomographic reconstruction...
How to mimic a full scanner with 2 modules only available?
Central FOV :
rotating the phantom...
θ = 0˚, 20˚, 40˚... 180˚ (9 steps)
1 tomographic acquisition
= 27 steps acquisition
φ = 0˚
Extended FOV :
...and rotating also the module
φ = 20˚
θ = 0˚, 20˚, 40˚... 360˚ (18 steps)
mimics a 18-modules
ring, with coincidences
between face-to-face ±
one adjacent modules
31

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
FE
G-APD bias
Setup for tomographic reconstruction
Module1
trigger NIM crate
source
rotating
table
setup @ CERN, Blg 304
Module2
power supply box
ADC VME crate
FE
The two modules are mounted on top of a portable
platform, which houses also the electronics, power
supply, etc...
• One rotating motor for
the source / phantom
• One module fixed (Mod1);
the other rotating (Mod2)
32

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Measurements with phantoms
Three different measurement campaigns :
@ ETH Zurich (CH), Radio-Pharmaceutical Institute - Apr 2010
@ AAA (Advanced Acceleration Applications) St. Genis-Pouilly (FR) - July 2010
@ AAA (Advanced Acceleration Applications) St. Genis-Pouilly (FR) - July 2011
in-situ cyclotron production of the radiotracer.
- Phantoms filled with F-18 based tracers (F-18 in water solution)
- t1/2 ~ 110 mins
- Concentration diluted in water; A0: few MBq up to ~ 100 MBq
- Analysis restricted to photoelectric absorption events only (no ICS events for the moment)
33

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Measurements with phantoms
Three different measurement campaigns :
@ ETH Zurich (CH), Radio-Pharmaceutical Institute - Apr 2010
@ AAA (Advanced Acceleration Applications) St. Genis-Pouilly (FR) - July 2010
@ AAA (Advanced Acceleration Applications) St. Genis-Pouilly (FR) - July 2011
in-situ cyclotron production of the radiotracer.
- Phantoms filled with F-18 based tracers (F-18 in water solution)
- t1/2 ~ 110 mins
- Concentration diluted in water; A0: few MBq up to ~ 100 MBq
- very first measurements with high activity and extended objects
- limited to the central FOV (i.e. fixed modules, rotating source)
- extended FOV coverage (one module rotating, rotating source)
- larger phantoms / placed off-centered phantoms
- extended FOV as before
- improved DAQ performance
- improved acquisition methods
33

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
(1) capillaries
L = 3cm ;
Diam = 1.4 mm ;
Pitch = 5 mm
(2) micro Derenzo
H = 1.5 cm ;
Diam = 2 cm;
Rods_Diam = 0.8÷2 mm
(3) mini DeLuxe
H = 5 cm ;
Diam = 7.5 cm;
Rods_Diam = 1.2÷4 mm
Phantoms for reconstruction
(4) homogeneous cylinder
H = 9 cm ;
Diam = 6 cm;
(5) NEMA phantom
H = 6.3 cm ;
Diam = 3.3 cm;
pictures not on scale!
34

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Image reconstruction
Goal :
recover the activity distribution, starting from the acquired data
Data :
LOR of the various coincidence events i.e. projections
(typically organized in “sinograms”)
Angle θ
Two different approaches :
projections
g(s,θ)
Displacement s
ANALYTICAL METHOD ITERATIVE METHOD
• Filtered Back Projection (FBP) :
1) Fourier analysis of the projection data
2) Different weight to different frequencies (“filtering”)
3) “Back-Project”
• simple and fast
☺ ☹
• not extremely accurate
• assumption : measured data are perfectly consistent with the
source object (never true: e.g. noise, gaps between detector...)
• slow and CPU consuming
• accurate model of the emission and
detection processes
• accurate reconstruction
• optimization procedure until the best
estimate is found
• several optimization strategies exist
activity
f(x,y)
☹
☺
35

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Iterative reconstruction method
36

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Iterative reconstruction method
(2)
obtain the projection that would
derive from such an object
(1)
initial estimate of the
object count distribution
(typically: uniform)
36

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Iterative reconstruction method
(2)
obtain the projection that would
derive from such an object
(1)
initial estimate of the
object count distribution
(typically: uniform)
36

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Iterative reconstruction method
(2)
obtain the projection that would
derive from such an object
build up correction coefficients
(1)
initial estimate of the
object count distribution
(typically: uniform)
36

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Iterative reconstruction method
36

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Iterative reconstruction method
requires a description of the
physical model of emission
and detection processes :
SYSTEM MATRIX
Mij
probability of the activity
in the j-voxel to be
detected by the i-LOR
includes :
- geometry description
- physics (e.g. : attenuation)
- variable fraction of
the voxel contributes to
the counts
- ...
36

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
out of the many possible iterative methods
AX-PET uses ML-EM
Max Likelihood
Iterative reconstruction method
Expectation
Maximization
two steps per iteration :
(1) Expectation step : form the
likelihood of any reconstructed image
given the measured data
(2) Maximization step : find the
image with the greatest likelihood
requires a description of the
physical model of emission
and detection processes :
SYSTEM MATRIX
Mij
probability of the activity
in the j-voxel to be
detected by the i-LOR
includes :
- geometry description
- physics (e.g. : attenuation)
- variable fraction of
the voxel contributes to
the counts
- ... 36

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives
conclusion
y
y
z
z
x
Simple reconstructed images : capillaries
x
counts [a.u.]
capillaries :
L = 3 cm
∅ = 1.4 mm
pitch = 5 mm
y
y
x
x
z
z
z
z
y
y
x
x
z
z
central FOV
no modules rotation
ETH 2010
Profiling of the reconstructed capillaries (3 different measurements) and resolutions (FWHM) of
reconstructed sources. The resolution still includes the capillary finite size (1.4 mm inner diameter).
Profiling
FWHM [mm]
z
z
Resolution
x
x
y
y

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
extended FOV
2nd module rotation
NEMA phantom
Three regions in the same phantom to
address three different aspects
Hot & Cold rods for contrast
Homogeneous cylinder for assessing the
ability to reconstruct
homogeneous distributions
Series of small rods for resolution
{
{
{
34 mm
63 mm
38

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
extended FOV
2nd module rotation
NEMA phantom uniformly filled - AAA 2010
NEMA phantom hot / cold / warm - AAA 2011
NEMA phantom
Three regions in the same phantom to
address three different aspects
Hot & Cold rods for contrast
Homogeneous cylinder for assessing the
ability to reconstruct
homogeneous distributions
Series of small rods for resolution
{
{
{
34 mm
2 mm
63 mm
38

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
extended FOV
2nd module rotation
NEMA phantom uniformly filled - AAA 2010
NEMA phantom hot / cold / warm - AAA 2011
NEMA phantom
Three regions in the same phantom to
address three different aspects
Hot & Cold rods for contrast
Homogeneous cylinder for assessing the
ability to reconstruct
homogeneous distributions
Series of small rods for resolution
{
{
{
34 mm
2 mm
63 mm
38

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
extended FOV
2nd module rotation
NEMA phantom uniformly filled - AAA 2010
NEMA phantom hot / cold / warm - AAA 2011
NEMA phantom
Three regions in the same phantom to
address three different aspects
Hot & Cold rods for contrast
Homogeneous cylinder for assessing the
ability to reconstruct
homogeneous distributions
Series of small rods for resolution
{
{
{
34 mm
different color scale !
1 mm (i.e. < Resolution)
5 mm
4 mm
3 mm
2 mm
Reconstructed 1 mm rod
=> FWHM ~ 1.6 mm
63 mm
38

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
extended FOV
2nd module rotation
NEMA phantom uniformly filled - AAA 2010
NEMA phantom hot / cold / warm - AAA 2011
NEMA phantom
Three regions in the same phantom to
address three different aspects
Hot & Cold rods for contrast
Homogeneous cylinder for assessing the
ability to reconstruct
homogeneous distributions
Series of small rods for resolution
{
{
{
34 mm
different color scale !
1 mm (i.e. < Resolution)
5 mm
4 mm
3 mm
2 mm
Reconstructed 1 mm rod
=> FWHM ~ 1.6 mm
63 mm
38

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Mini Deluxe phantom
(A)
Rods oriented parallel to Z axis
Resolution phantom : Mini Deluxe
extended FOV
2nd module rotation
AAA 2011
(A) (B)
39

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Mini Deluxe phantom
(A)
Rods oriented parallel to Z axis
Resolution phantom : Mini Deluxe
extended FOV
2nd module rotation
AAA 2011
Rods oriented perpendicular to Z axis
(B)
(A) (B)
Results presented in Valencia, IEEE 2011
39

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
What’s next (1/3) ?
40

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
What’s next (1/3) ?
• improve the reconstruction algorithms
• include ICS (Inter Crystal Scattering) events in the reconstruction
preliminar!
• ICS included:
• only “triple” coinc. (1+2)
• simplest weighting (50% ; 50%)
• Simulated One-Pass List-mode
reconstruction (SOPL)
• No image deterioration when ICS
events are included!
40

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
What’s next (2/3) ?
• full ring simulations:
• performance of (hypothetical) full ring (e.g. resolution, sensitivity, NEC...)
• two different approaches
41

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
What’s next (3/3) ?
• images of (probably dead) small animal
• new measurement campaign
• @ETH Zurich, Radiopharmaceutical Institute
• Spring 2012 ?
• F-18 (bone scan) or FDG ?
Mouse bone scan,
NanoSPECT (Bioscan)
42

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives
D-SiPM
conclusion
What’s next (3/3) ?
• images of (probably dead) small animal
• new measurement campaign
• @ETH Zurich, Radiopharmaceutical Institute
• Spring 2012 ?
• F-18 (bone scan) or FDG ?
• test digital Silicon Photomultipliers (from Philips)
as alternative photodetectors
• extremely good timing performance
• GOAL : demonstrate the possibility
a TOF-PET with the axial concept
Mouse bone scan,
NanoSPECT (Bioscan)
42

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives
D-SiPM
conclusion
Digital Silicon Photomultiplier (D-SiPM)
D-SiPM :
Single Photon Avalanche
Photodiode (SPAD)
integrated with
conventional CMOS
circuits on the same
substrate.
All SPADs (including
their corresponding
electronics) connected
together
- to photon counter
- to TDC
Thomas Frach,
CERN Detector seminar, Oct 2011
43

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives
D-SiPM
conclusion
Digital Silicon Photomultiplier (D-SiPM)
Advantages of D-SiPM for AX-PET:
Intrinsically very fast photodetector (~ 50ps).
Great potential for TOF-PET.
Small size. High level of integration.
Compactness.
Interesting concept of simplifying the readout chain.
Bias supply included.
Early digitization of the cell output; integrated
electronics on chip => low noise.
Digital => Temperature and gain stability is less
critical than in analogue SiPM.
Possibility to disable cells with high Dark Count.
MRI compatible.
1 sensor
full tile,
64 sensors
Technology Evaluation Kit (TEK) :
MPPC :
3200 cells
3x3 mm 2
- 4 tiles :
DLS_6400 (x2 tiles) ; 6400 cells/sensor
DLS_3200 (x2 tiles) ; 3200 cells/sensor
- each tile : 8x8 sensors
- 1 sensor : 3.9 x 3.3 mm 2
received at CERN on Jan 12 th , 2012
44

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives
D-SiPM
conclusion
22 Na
source
• 2 LYSO scintillator crystals
• non AX-PET standard
• (2x2x12) mm 3 and (2x2x15) mm 3
• wrapped with teflon
• coupling done with optical grease
• LYSO crystal placed in the center of the pixel
(no precise mechanical alignment)
D-SiPM: first preliminary results
time difference between the two tiles for coincident events
(coincidence window = 2 ns)
#Events
1000
800
600
400
200
0
PRELIMINARY !!!
DLS 3200
T ° C
=5
Peltier
Entries 10964
Constant 930.5 ± 13.1
Mean 0.2449 ± 0.0011
Sigma 0.08534 ± 0.00123
Time resolution = 200 ps FWHM
-0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Δt
[ns]
FWHM ~ 200 ps
(cutting on events at the photopeak)
45

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives
D-SiPM
conclusion
D-SiPM: first preliminary results
Minimal AX-PET like set-up
WLS strip
(3 x 0.9 x 40 mm 3 )
DLS_3200
(one pixel enabled)
22-Na source
(~ 3 MBq)
LYSO crystal
(3 x 3 x100 mm 3 )
DLS_3200
(1/2 pixel enabled)
46

intro axial concept AXPET simulations 800 image reconstructions
Raw LYSO spectrum, Na-22 source
1400
1200
1000
800
600
400
200
Raw spectrum
600
D-SiPM: first preliminary results
400
200
511 keV
0
0 500 1000 1500 2000 2500 3000
Nr photons
Energy spectrum
(calibrated, corrected
for saturation)
300
250
200
150
100
PRELIMINARY !!!
50
Nr photons
1000
AXPET performance perspectives
D-SiPM
conclusion
LYSO light yield
1.27 MeV
LYSO energy spectrum, Na-22 source
2500
2000
1500
1000
500
0
0 500 1000 1500 2000 2500 3000
Energy Calibration [0]*(1-TMath::Exp(-x*[1]/[0]))
! / ndf
5.868e-11 / 1
p0 2767 ± 2.05e-05
p1 4.453 ± 4.72e-08
0
0 200 400 600 800 1000 1200 1400
300
250
200
150
100
PRELIMINARY !!!
0
0 200 400 600 800 1000 1200 1400
50
ENERGY CALIBRATED SPECTRUM
2
Energy [keV]
Entries
Mean
RMS 2
! / ndf
Constant
Mean
Sigma
SpectTile4
0
0 200 400 600 800 1000 1200 1400
Energy [keV]
Energy [keV]
R_FWHM ~ 12.3%
47

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives
D-SiPM
conclusion
• 2 tiles operated in coincidence
Correlation
LYSO / WLS response
Still to do :
- add more WLS strips
- full WLS cluster
- extract axial coordinate
D-SiPM: first preliminary results
LYSO Energy [keV]
Minimal AX-PET Setup
2000
1800
1600
1400
1200
1000
800
600
400
200
PRELIMINARY !!!
0
0 20 40 60 80 100 120 140 160 180 200
WLS [Nr photons]
48

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives
D-SiPM
conclusion
two needed arrangements
Questions to be answered :
- light yield and ΔE/E from LYSO crystals ?
- axial coordinate reconstruction: does it work ?
- axial resolution through WLS readout ?
- time resolution (long crystals) ?
- cooling for DC reduction ?
D-SiPM: Outlook
AX-PET performance demonstration
timing performance demonstration
49

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Conclusions
Axial concept for a PET scanner :
i.e. long and axially oriented scintillation crystals
Intrinsically parallax free system (DOI information directly from the axial geometry)
Spatial resolution and sensitivity could both be optimized
AX-PET implementation :
3D spatial information of the photon interaction point with :
matrix of LYSO crystals and WLS strips
individual readout of each channel (Si-PM)
Two modules built (i.e. AX-PET demonstrator)
Energy resolution ~ 12% FWHM,@ 511 keV
Spatial resolution ~ 1.35 mm FWHM
(competitive with state of the art PET)
Fully simulated device
Simulations - fully validated on the demonstrator - will assess the final performance of an hypothetical full
ring scanner. Flexible design: scalable in size / dimensions / nr. layers....
=> flexibility in the final target of AX-PET (small animal PET / brain PET)
AX-PET demonstrator :
Extensively tested with sources and successfully used with phantoms !
‣ calorimeter with tracking
capabilities (granularity)
‣ novelty as a PET detector :
- geometry
- WLS implementation
- module sum / trigger
- Compton scattering reconstruction
Currently under test :
Digital SiPM as alternative for photodetector for AX-PET (TOF capabilities)
50

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Conclusions
50

intro axial concept AXPET AXPET performance simulations image reconstructions perspectives conclusion
Conclusions
50

han s!
51

AX-PET collaboration
A. Braem, M. Heller, C. Joram, T. Schneider and J. Séguinot
CERN, PH Department, CH-1211 Geneva, Switzerland
V. Fanti
Università e Sezione INFN di Cagliari, Italy.
C. Casella, G. Dissertori, L. Djambazov, W. Lustermann, F. Nessi-Tedaldi, F. Pauss, D. Renker 1 , D. Schinzel 2
ETH Zurich, CH-8092 Zurich, Switzerland
1 Currently with Technical University München , D-80333 München, Germany
2 Currently with Massachussetts Institute of Technology, Cambridge 02139-4307, USA.
J.E. Gillam, J. F. Oliver, M. Rafecas, P. Solevi
IFIC (CSIC / Universidad de Valencia), E-46071 Valencia, Spain
R. De Leo, E. Nappi
INFN, Sezione di Bari, I-70122 Bari, Italy
E. Chesi, A. Rudge, P. Weilhammer
Ohio State University, Columbus, Ohio 43210, USA
E. Bolle, S. Stapnes
University of Oslo, NO-0317 Oslo, Norway
U. Ruotsalainen, U. Tuna
Tampere University of Technology, FI-33100 Tampere, Finland
Many thanks for the appreciated technical support from :
Volker Commichau, Michael Droge, Hanspeter Von Gunten, Christian Haller,
Urs Horisberger, Bozidar Panev (ETH Zurich)
Jerome Bendotti, Bernard Cantin, Claude David, Niel Dixon, Didier Gabriele,
Markus Joos, Miranda Van Stenis (CERN)
52