C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 The 3D Axial PET Concept Christian Joram, CERN, PH-Department Positron Emission Tomography - Principle and intrinsic limitations - State of the art Hybrid Photon Detectors: Principle, performance and fabrication The 3D axial PET camera - concept - strong and critical points - discussion of the key components (scintillator, photo detector, electronics, integration) - Brain PET study: geometry and first performance estimates

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 Positron Emission Tomography Principle and fundamental limits Patient injected with positron (  + ) emitting radiopharmaceutical, e.g. FDG, marked with 18 F.  + annihilates with e - from tissue, forming an annihilation photon pair (back-to-back, 511 keV) 511 keV pairs detected in scintillator crystals via time coincidence  + emission point lies on line defined by detector pair (line of record LOR, chord) Reconstruct 2D image using Computed Tomography Use stacked detector rings to obtain a 3D volumetric image. + many mathematical tricks (filtering, density corrections, …) Standard PET geometry FDG structure

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 PMT2 d/2 Reconstruction Algorithm 1.25 (in-plane) 1.0 (axial) Factor Detector Crystal Width Photon non-collinearity 180º ± 0.25º Shape Resolution (FWHM) multiplicative factor 0.5 mm, 18 F 4.5 mm, 82 Rb 1.8 mm (depends on det. ring radius) 0 (individual coupling) 2.2 mm (Anger logic, empirically) d PMT1 Anger Logic Positron range Courtesy of B. Moses, Mattinata 2002

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 Standard PET geometry The parallax dilemma L Crystals need to have a minimum thickness L Efficiency for pair detection a = photon attenuation length of crystal L = a    ~ 40%, L = 2 a    ~ 75%, … a = 1-2 cm (depending on material, see below) A standard PET does not measure the depth of interaction (DOI) in the crystal.  pp This introduces a parallax error The resolution in the off-center region degrades significantly Solution: reduce L (  bad    or measure DOI or invent a different geometry

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 Resolution is important, but also sensitivity and image quality matter ! Sensitivity = True Event Rate / µCi / cm 3 (measures detector efficiency) Determined by geometry, choice of scintillator, data acquisition… Trues ScattersRandoms But the signal is subject to background from Compton scattering and random coincidences. Noise Equivalent Count Rate

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 With DT T dead   e –(   t) R dead   2 e –(   t) With DT & Coincidence Processor Limit Total throughput const. NEC ~ 1/ρ 2 The NEC rate depends on many factors. More activity and efficiency does not always help ! Courtesy of B. Moses, Mattinata 2002

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 Concrete example: The High Resolution Research Tomograph (HRRT) (CTI, MPI, Karolinska …) Currently the “reference” in brain PET AFOV 25 cm 31 cm 8 panels with 9  13 blocks 2  64 crystals per block Crystal dimensions: 2.1  2.1 x 7.5 mm PMT2PMT1 LSO GSO or LSO Pulse Shape Discrimination = Phoswich approach  = 7 ns 8  8 matrix DOI is known with a precision of 5 mm (FWHM), 10 mm without phoswich. However 15 mm detector length correspond only to about 1 a. Efficiency  2 ~ 40% 7.5 K. Wienhard et al., IEEE Trans. Nucl. Sci. 49 (2002) 104–10

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 HRRT performance spatial resolution (FWHM) transaxial: mm axial: 3-4 mm Energy resolution: 17% Sensitivity: 4.5 cps/kBq NEC: 140 kcps measured by moving a 18 F point source (1 mm Ø) over a cylinder of 20 cm diameter and 20 cm length cylinder of 20 cm diameter and 20 cm length filled with 18 F dissolved in water to an activity of 0.35  Ci/ml.

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 Time-Of-Flight PET – an old idea recently revived (Si-PM ! ) Idea Use very fast scintillators and photodetectors (<3 ns FWHM) Use TOF information of photons to constrain source distribution Benefits Smaller coincidence time  lower background  higher peak NECRR Reduced noise (less ‘false’ intersections) Simultaneous recording of emission / transmission (ext. source) from W.W. Moses IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 50, NO. 5, OCTOBER 2003, p. 1325

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 Hybrid Photon Detectors ( HPD ) Background: Applications in LHC experiments require photon detectors featuring High sensitivity in visible and UV range Ability to see single photons high speed (remember 40 MHz bunch crossing) High filling factor (>70%) Large area coverage (~m 2 )  Development of large area Hybrid Photon Detectors Combination of sensitivity of PMT with excellent spatial and energy resolution of silicon sensor

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 The 5-inch Pad HPD h e-e- Bialkali photocathode 16 front-end chips Ceramic PCB Si Sensor 2048 pads (1 x 1 mm 2 )

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 Single photon imaging with 2048 Channels. 1 p.e. 2 p.e. 3 p.e. signal amplitude (a.u.) counts Electronics noise well separated from signal Signal definition and energy resolution. S/N ~ 10. Same spectrum measured with PMT (schematically)

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 HPD fabrication Facilities and infrastructure for the fabrication of large HPDs (up to 10ӯ) have been developed at CERN. All ingredients for photodetector production are available: Design/simulation Photocathode processing (bialkali, Rb 2 Te, CsI) Glass / ceramic tube manufacturing Indium sealing technique

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 The 3D axial PET camera Conventional concept Rings of block detectors New 3D axial concept Axial arrangement of camera modules based on matrices of long crystals read out on both sides by HPDs Illustration: F. Schönahl, HUG

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 Full 3D reconstruction of  quanta without parallax error x,y from silicon pixel address z from amplitude signal ratio of the 2 HPD’s  Precise Depth of Interaction DOI measurement  No limitation in detector thickness  improved sensitivity. Measurement of light yield on both sides of crystals Negligible statistical fluctuations in HPD  Very good γ energy resolution 3D reconstruction provides possibility to recuperate part of  ’s which underwent Compton scattering in the scintillator crystals  Compton enhanced sensitivity Main advantages of the concept HPD2 HPD1 x y z

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 z-resolution, if obtained by light ratio, scales with crystal length σ z ~ L  Axial field of view limited to about cm Little experience in fabrication of long crystals (cost) Critical aspects of the concept The concept seems to be very promising for brain PET, where both high spatial resolution and high sensitivity are required. PEM (PE mammography), in compressed mode, is another interesting option.

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 Criteria to be taken into account: light yield, absorption length, photo fraction, self absorption, decay time, availability, machinability, price. YAP is OK for proof of principle, however suffers from low Z (high absorption length, low photo fraction) All preliminary performance estimates are based on YAP (availability!) LaBr 3, L(Y)SO and LuAP are the really interesting candidates. YAP:CeLSO:CeLuAP:CeLaBr 3 :Ce Density ρ (g/cm 3 ) Effective atomic charge Z Scintillation light output (photons / MeV) ~10000~61000 Wavelength of max. emission (nm) Refractive index n at wavelength of maximum emission ~1.88 Bulk light absorption length L a (cm) at 370 nm ~14~20 Principal decay time (ns) ±5 Mean γ attenuation length at 511 keV (mm) ~20 Photo fraction at 511 keV (%)4, Energy resolution at 663 keV Scintillation crystals for the 3D axial concept BGO ~ ~ ~

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 Planned geometry for proof of principle 4 camera modules R=160 mm EuroMedIm workplan ~2005

C. Joram - HPD-PET team - Bari (Italy), 17 January mm 52 mm A camera module based on HPD PCR5 208 crystals (YAP) 3.2 × 3.2 × 100 mm 3 HPD PCR5 Ceramic envelope with sapphire entrance window Baseplate (from Pad- HPD) 2 Silicon sensors (208 pads 4×4 mm 2 )

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 First fully operational PET HPD E. Chesi et al., NIM A 564,2006,

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 vacuum pump (turbo) P < mbar DAQ Readout card Pulsed LED (blue) MgF 2 sapphire mirror Si sensor (300  m) 208 pads (4×4 mm 2 ) VME -U PC = kV Experimental set-up for PET-HPD tests collimator

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 Mapping from photo cathode to silicon sensor has very good linearity. Mapping was measured in a previous prototype with 1 mm pixel size. Deviations from linearity only ~ 60  m RMS!

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 Hit Distributions from Light Spot This lego plot shows that a threshold of ~ 20 fC (~ 50 / 20 p.e. at 10 / 20 kV) eliminates easily any dark current hits  Background free images

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 Total Charge Summed from All Pads above Threshold Histograms of the total charge for two different HPD acceleration voltages. Left: U C = -10 kV ; Right: U C = -20 kV. The mean of charge distributions left and right is not a factor 2 apart since there is more energy loss in dead layer of pad sensor at lower electron energy and also more pads with partial energy above threshold at higher electron energy

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 Mean Charge  and  /  of charge distributions as function of Cathode Voltage Mean charge  (left axis) and ratio of Gaussian width to mean charge σ/μ (right axis) versus cathode voltage UC (kV). Note: mean charge is linear but intercepts at ~6 keV due to energy loss in dead layer. This can be improved in next sensor production run. Expect intercept at ~0.5 keV, which will considerably improve the charge gain in the HPD  reaches almost a plateau around 17 keV since energy straggling becomes small. One can estimate an energy loss of ~ 1.6 keV at 20 kV with nearly negligible straggling.  Gain at 20 kV is 5090 From  (ENF/N) 1/2  N = 507 photo electrons ( ~ expected N p.e. per HPD from a 511 keV  in LYSO) The absolute gain of the chain can now be calculated: 0.94 fC/(ADC count) for chip1

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 Full ring scanner Possible configuration for a Brain PET R = 170 mm 34 cm inner diameter 10 cm axial length 2496 crystals 24 HPDs total detection volume 2556 cm 3  coverage 66%  coverage 18% x y z

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 Resolution x,y  ~1.5 mm Resolution x,y  ~2.2 mm Resolution z  ~4.5 mm reconstructed (0,0,0) transaxial (x,y) planeaxial (x,z) plane 100 mm 90 mm 3D axial PET geometry Crystal dimensions: 3.2 x 3.2 x 100 mm Material: YAP:Ce 33 point sources, no background source (0,0,0) 5 mm

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 Resolution x,y  ~1.5 mm FWHM Resolution x,y  ~6 mm FWHM reconstructed (0,0,0) 100 mm 90 mm Standard PET geometry Crystal dimensions: 3.2 x 3.2 x 30 mm 33 point sources, no background source transaxial (x-y) planeaxial (x-z) plane

C. Joram - HPD-PET team - Bari (Italy), 17 January 2007 (HRRT) Detected photoelectrons for a γ of 511 keV:~ per HPD (f( b )(?) Energy resolution: ~10 % (FWHM) (f( b ) (17) Spatial resolution in transaxial plane: mm (FWHM) ( ) Spatial resolution in axial direction: ~4-5 mm (FWHM) (f( b ) (3-4) Coincidence interval: ~ 5-10 ns (  2 ns?) Compton gain: ~ 2(1) Sensitivity~ 5.5 cps/kBq(4.5) NEC: 130 kcps (?)(140) HPD based Brain PET should beat HRRT in spatial and energy resolution and should at least be comparable in other disciplines. Performance estimates 3D Axial Brain PET (with LSO/LYSO)