Small Animal Radionuclide Imaging: Instrumentation, Performance, and Applications Craig S. Levin, Ph.D. Stanford University School of Medicine Department of Radiology Stanford Stanford University University MIPS Molecular Imaging Program at Stanford School of Medicine Department of Radiology Small Animal Radionuclide Imaging Outline of talk: •Small Animal Positron Emission Tomography (PET) Instrumentation requirements and challenges Commercially available systems New approaches •Summary Stanford University MIPS Molecular Imaging Program at Stanford School of Medicine Department of Radiology Nuclear decays of interest for imaging generate high energy photons Gamma decay: Nuclear de-excitation + Example: 99mTc + e+ + ν γ−ray Stable state of nucleus Short-lived excited nucleus Positron decay: Nuclear transmutation --> 99Tc + γ Short-lived unstable nucleus More stable isotope Example: 18F positron --> 18O + e+ + ν Annihilation Photons Gamma Ray MIPS Molecular Imaging Program at Stanford e+ Stanford University School of Medicine Department of Radiology 1 Small Animal PET System Design Wish List •Reconstructed spatial resolution ≤ 1 mm •Uniformity of spatial resolution •Sensitivity (coincidence detection efficiency) > 10% •Energy resolution ≤ 12% FWHM at 511 keV •Coincidence time resolution ≤ 2 ns FWHM •Live time > 95% •Robust image reconstruction algorithm •Accurate data correction and calibration •Reasonable cost Stanford University MIPS Molecular Imaging Program at Stanford School of Medicine Department of Radiology Limitations on PET Spatial Resolution Positron Emitter Variations in positron trajectoryEffects on resolution depend on isotope Molecular Probe e+ Detector Element Variations in annihilation photon non-collinearityEffects on resolution depend on system diameter Variations in photon interaction locationEffects on resolution depend on detector element size Detector Gantry Stanford University MIPS Molecular Imaging Program at Stanford School of Medicine Department of Radiology Spatial Resolution Limit for 18F PET Positron Range Photon Non-collinearity Detector Width { FWHM 0 cm 10 cm 20 cm 80 cm System Diameter SPATIAL RESOLUTION (mm) Blurring Functions 6 FWTM 0 cm 10 cm 20 cm 80 cm 18F 5 4 3 2 1 0 0 1 2 3 4 5 DETECTOR ELEMENT WIDTH (mm) ~500µm spatial resolution (fwhm) is possible in theory MIPS Molecular Imaging Program at Stanford Stanford University School of Medicine Department of Radiology 2 Non-Uniform Resolution Due to Photon Penetration in Crystals Upper Limit on Radial Resolution Blurring (see drawing for definition of symbols): Δd detector gantry Δrupper ≈ Δxupper/2 = (Δd/2)·sinθi (FWHM) θi = (Δd/2) r / [R+(Δd/2)] R Δxupper = r / [(D/Δd)+1] r (an “upper” limit since Δx is calculated assuming two isolated crystals as shown; the presence of other adjacent absorbing detector crystals weights the calculation towards shallower average interaction depth) θi crystal finger Stanford University MIPS Molecular Imaging Program at Stanford School of Medicine Department of Radiology Non-Uniform Resolution Due to Photon Penetration in Crystals Δrupper ≈ r / [(D/Δd)+1] Δd For a mouse positioned at center with diameter of 3 cm, at radial position r=1.5 mm, and system depth resolution Δd=10 mm, the radial blurring contribution Δr<1.5 mm FWHM MIPS Molecular Imaging Program at Stanford Stanford University School of Medicine Department of Radiology Challenges for High Resolution PET Detectors I. Getting light out of long, skinny crystals 511 keV photon Requirements for Crystal Arrays: •Narrow (≤1 mm) for high spatial resolution •Long (20-30 mm) and tightly packed for high coincidence detection efficiency •Need robust light signal for best coincident time resolution, energy resolution, detection efficiency, and crystal identification; These parameters determine contrast resolution and quantitative accuracy. MIPS Molecular Imaging Program at Stanford Stanford University School of Medicine Department of Radiology 3 Photon Count Sensitivity (Coincidence Efficiency) for PET High Geometric Efficiency (Ω/4π): High Intrinsic Detection Efficiency (ε): Photon detectors should be as close to the body, cover as large an axial FOV and be as tightly-packed as possible: 511 keV photon detectors should have high Z, high density, and be thick for high stopping power: ε = (1-e -µx)×f Ω ≈ 4π (A/D)×P µ=attenuation coefficient x=crystal thickness/length f=fraction of events within energy window A=axial field-of-view D=diameter P=crystal packing fraction Coincidence Detection: ε2 = [(1-e -µx )×f]2 Sensitivity (%) ≈ 100×(Ω/4π)×ε 2 = 100×(A/D)×P×[(1-e-µx)×f ]2 Stanford University MIPS Molecular Imaging Program at Stanford School of Medicine Department of Radiology Types of Coincidence Events Front view Side view Random Absorbed single photon True Escaped single photon Scatter Scatter and randoms are reduced with better energy and time resolutions Stanford University MIPS Molecular Imaging Program at Stanford School of Medicine Department of Radiology Inorganic scintillation crystals for PET Scintillator “BGO” Bi4(GeO4)3 “LSO” Lu2(SiO4)O:Ce “GSO” Gd2(SiO4)O:Ce “LYSO” Lu1.8Y0.2(SiO4)O:Ce “Sodium Iodide” NaI(Tl) Effective Z Density (g/cc) 75 7.13 1/e attenuation length at 511 keV (cm) 1.06 66 7.4 59 Relative light yield (%NaI) Refractive index Decay time (ns) Peak emission wavelength (nm) Rugged? 15 2.15 300 480 Yes 1.13 75 1.82 42 420 Yes 6.71 1.4 20 1.85 60 440 Yes 65 7.1 1.2 107 1.81 40 420 Yes 51 3.67 2.94 100 1.85 230 410 No MIPS Molecular Imaging Program at Stanford Stanford University School of Medicine Department of Radiology 4 Commercially Available Small Animal PET Instrumentation •Concorde Microsystems, Knoxville TN “microPET” [Scintillation detectors (LSO), fiber-coupled] •Oxford Positron Systems, Oxfordshire, UK “HIDAC” [MWPC Detectors (lead converters, gas)] •Philips Medical Systems, Philadelphia, PA “Mosaic” [Scintillation detectors (GSO)] •GE-Suinsa Medical Systems, Madrid, Spain “eXplore Vista” [Dual-layer scintillation detectors (LSO-GSO)] •Gamma Medica, Northridge, CA “X-PET” [Scintillation detectors (BGO) + SPECT/CT] •Advanced Molecular Imaging, Quebec Canada “LabPET” [Scintillation detectors-Avalanche Photodiodes] Stanford University MIPS Molecular Imaging Program at Stanford School of Medicine Department of Radiology Concorde Microsystems microPET Focus Detector cassette PSPMT 1.5x1.5x10 mm3 LSO Fiber optics Focus 120 Bore size Axial FOV Resolution Coincidence Efficiency Energy Resolution Peak NEC Focus 511 keV flood response of detector block Focus 220 120 mm 780 mm 1.3 mm >6.5% 18% >580kcps R4 & P4 Energy spectrum in LSO 220 mm 780 mm 1.3 mm >4.0% 18% >700kcps Stanford University Courtesy of Stefan Siegel, Concorde Microsystems MIPS 2.2x2.2x10 mm3 LSO Molecular Imaging Program at Stanford School of Medicine Department of Radiology microPET Focus Performance Reconstructed Image point source resolution Resolution 5 Count rate performance (mouse phantom) •250-750 keV energy window •10 ns timing window FWHM ( mm ) 4 800 •Fourier rebinning, 2D FBP 250-750 keV 10 ns 250-750 keV 6 ns 350-650 keV 10 ns 350-650 keV 6 ns 700 3 600 2 1 Focus Tangential Focus Radial Focus Axial 0 0 10 20 30 R4 Tangential R4 Radial R4 Axial 40 50 60 70 80 Radial Offset (mm) Concorde P4, FBP Concorde Focus, FBP 2.4 mm NEC (kcps) m m 500 400 300 200 100 0 0 5 1.6 mm 4.0 mm NEC = 1.2 mm 10 15 Activity (mCi) 3.2 mm T2 T+S+2R 4.8 mm T= “true” coincident rate S= scatter coincident rate R= random coincident rate Fourier rebinning + 2D FBP /Ramp filter. MIPS Courtesy of Yuan-Chuan Tai, Washington University Molecular Imaging Program at Stanford Stanford University School of Medicine Department of Radiology 5 microPET Focus Images Cardiac Gated Images (Rat) end-systole end-diastole non-gated Neuro-receptor imaging (mouse) 20 g mouse injected with 11C-CFT 362.6 g S.D. Rat 0.966 mCi 18FDG 1.8 hr P.I. 30 min scan ~23 ms frames Courtesy of Douglas Rowland, Washington University MIPS Stanford University Molecular Imaging Program at Stanford School of Medicine Department of Radiology GE-Suinsa eXplore Vista (Argus) Dual-layer LYSO-GSO detectors Spatial resolution in central slice: 1.45 mm radial 1.56 mm tangential 1.74 mm axial 3.9 mm3 Volume resolution Coincidence timing resolution: 1.5 ns FWHM Central point source sensitivity: 4.0% [250-700 keV] 5.7% [100-700 keV] Peak NEC rate with mouse phantom: 185,000 cps [250-700keV] @ 15 µCi/cc · Ring diameter: 11.8 cm · Aperture: 8 cm · Effective transverse field-of-view: 6 cm · Axial FOV: drT: 4.6 cm (srT = 2.0 cm) · Number of depth-of-interaction detector modules: 36 (18) PS- PMTs · Number of dual-scintillator depth-of-interaction elements: 6,084 (3,042) · Depth identification method: pulse shape discrimination · Crystal array pitch: 1.55 mm · Total number of crystals: 12,168 (6,084) · 3D (coincidences and singles) · Total number of coincidence lines: 28.8 M (7.2M) MIPS Detector module 511 keV field flood crystal pitch = 1.55 mm Courtesy of Juan José Vaquero, Universitario Gregorio Marañón Stanford University Molecular Imaging Program at Stanford School of Medicine Department of Radiology GE-Suinsa eXplore Vista (Argus) Imaging performance µDerenzo Phantom 3D FORE/2D FBP Awake Rat (F-18 FDG) transverse 3.2 mm 2.4 mm sagittal coronal cortex cortex cortex 4.0 mm 1.6 mm 4.8 mm 1.2 mm spinal cord ARGUS 3D OSEM reconstruction with resolution recovery Courtesy of Juan José Vaquero, Universitario Gregorio Marañón MIPS Molecular Imaging Program at Stanford Stanford University School of Medicine Department of Radiology 6 Research Institutions Developing Small Animal PET Instrumentation MIPS Brookhaven National Laboratory, USA Clear PET Collaboration, Multi-national Hamamatsu University, Japan Hammersmith Medical Center, UK Harvard University, USA Indiana University, USA King’s College UK Montreal Neurological Institute, Canada National Institutes for Health, USA Stanford University, USA Universitario Gregorio Marañón University of California, Davis, USA University of California, Los Angeles, USA University of Julich, Germany University of Munich, Germany University of Pennsylvania, USA University of Pisa, Italy University of Sherbrooke, Canada University of Texas, USA University of Southwestern Texas, USA University of Washington, USA Washington University, USA Molecular Imaging Program at Stanford Stanford University School of Medicine Department of Radiology New Technologies for Small Animal PET System Design •Improved Scintillation Detectors •Semiconductor Detectors MIPS Molecular Imaging Program at Stanford Stanford University School of Medicine Department of Radiology Is it possible to build a high performance PET system with 1 mm crystal pixels? 511 keV photon Requirements for Crystal Arrays: •Narrow (≤1 mm) for high spatial resolution •Long (20-30 mm) and tightly packed for high coincidence detection efficiency and…don’t forget... •Need high light extraction with low variation for best time resolution, energy resolution, detection efficiency, and crystal identification; these parameters will help to optimize contrast resolution and quantitative accuracy by helping to reject background events. MIPS Molecular Imaging Program at Stanford Stanford University School of Medicine Department of Radiology 7 How can we collect >90% of light from tiny array crystals? A Light Collection Efficiency: f ∝ (A/L)(1 - 1/n2) + + + + Collect the light from the long side n = refractive index + Instead of collecting the light from the small end (a small fraction of light is collected) + + L + + (a high fraction of light is collected) + + + + + + + + + + Stanford University MIPS Molecular Imaging Program at Stanford School of Medicine Department of Radiology Light Collection Improvements (1x1 mm2 pixels) 20 mm long 10 mm long 6 mm long 1x1x20 mm 3 1x1x10 mm 3 LSO BGO LSO GSO 100 100 90 90 80 80 70 70 60 60 50 50 40 40 30 30 20 BGO GSO 20 10 10 0 GROUND SURFACE WHITE REFLECTOR POLISHED SURFACE WHITE REFLECTOR 0 GROUND SURFACE WHITE REFLECTOR POLISHED SURFACE WHITE REFLECTOR For proposed scheme •Nearly all available scintillation light is collected (≥95%) •Light collection efficiency is independent of crystal length, width and surface treatment •Light collection efficiency is independent of the light origin •Results in superior energy and time resolution Stanford University MIPS Molecular Imaging Program at Stanford School of Medicine Department of Radiology How could we collect more light from arrays of miniscule (1 mm) crystals? Thin photodetectors at crystal sides Photodetector at crystal ends PSPMT MIPS Instead of this... Molecular Imaging Program at Stanford Can we do this? Stanford University School of Medicine Department of Radiology 8 New Approach for PET Detector Design Scintillation detector arrays with “edge-on” position sensitive avalanche photodiode (PSAPD) arrays between crystal planes 511 keV photon •~2 cm thick of LSO detectors in two stacked block modules 511 keV photon •Each module comprises 8 layers •Each layer comprises 3x8 arrays of 1x1x3 mm3 LSO crystals (left) or 1 mm thick sheets (right) 2 cm thick •This gives ~ 1-3 mm interaction depth resolution •Thin PSAPDs required 1x1x3 mm3 LSO crystals MIPS 9x9x1 mm3 LSO sheets Need: extremely thin PSAPD Stanford University Molecular Imaging Program at Stanford School of Medicine Department of Radiology Selected Scintillation Light Sensor Position sensitive avalanche photodiode (PSAPD) •Made by RMD, Inc. •8x8 mm2 area •Gain ~1000 •Leakage Current ~1-2 µA •Capacitance ~0.7 pf/mm2 ~45 pf •Noise ~130 e- rms A C Scintillation Light flash Corner contacts X = (A+B)-(C+D) A+B+C+D Y = (A+C)-(B+D) A+B+C+D (X,Y) B D Two sizes: 8 mm or 13 mm Problem: Standard PSAPD is not thin enough for high crystal packing fraction in our proposed detector design Stanford University MIPS Molecular Imaging Program at Stanford School of Medicine Department of Radiology New Light Detector: Position Sensitive Avalanche Photodiode (PSAPD) Standard vs. Thin PSAPD Standard chip and ceramic package Flex circuit accomodates two thin PSAPD chips on a same plane MIPS Molecular Imaging Program at Stanford Thin chip and Kapton package (~230 micron thick) Stanford University School of Medicine Department of Radiology 9 Novel “edge-on” detector configuration using position sensitive avalanche photodiodes Flex circuit Flex circuit for signal readout and HV bias 2.2 cm Electrica l traces PSAPD chip 1.3mm interlayer crystal pitch 511 keV photons 1.0 mm intra-layer crystal pitch Polished crystals Ground crystals PSAPD + flex circuit + reflector (<300 µm total thickness) 10 mm of LSO 1x1x3mm3 LSO crystals Second PSAPD chip location (not mounted) Novel, ultra-thin (<300 µm) position-sensitive avalanche photodiodes (PSAPD) are placed between the crystal layers with incident 511-keV photons entering parallel to the PSAPD plane (“edge-on”) as shown. This design gives direct measurement of photon depth-of-interaction and an effective 2-cm thickness of LSO crystal. One layer of detector module comprising PSAPD coupled to a 3x8 array of 1X1X3-mm3 LSO crystals. Half ground and half polished crystals were used to compare spatial, energy and temporal resolution performance. Stanford University MIPS Molecular Imaging Program at Stanford School of Medicine Department of Radiology Flood irradiation of 8x3 array of 1x1x3 mm3 crystals side coupled to thin PSAPD jin new12 x8 x1x3fld r30.l Flood histogram 3 1 Profile through center 200 Polished surfaces Ground surfaces 50 25 { { 100 150 150 20 200 250 15 100 300 10 350 50 400 5 No intercrystal reflectors450 No intercrystal reflectors 500 100 200 300 400 0 100 500 0 150 200 250 300 350 400 15:1 peak:valley ratio Excellent crystal identification due to high light collection efficiency Stanford University MIPS Molecular Imaging Program at Stanford School of Medicine Department of Radiology Spatial resolution of LSO-PSAPD detector layers FWHM of the point spread functions is measured to be about 1.0 mm. Left figure shows the top view of experimental setup used to measure coincidence time resolution and point spread function by edge-on scanning. 22Na LSO array PMT LSO Scan direction PSAPD 35 1000 900 50 30 100 800 25 150 200 20 700 600 250 15 300 500 1 2 3 4 5 6 7 8 400 350 12 3 4 5 6 7 8 400 10 450 500 100 200 300 400 1.04 mm FWHM 300 5 500 0 200 100 0 0 5 10 15 Position (mm) MIPS Molecular Imaging Program at Stanford Stanford University School of Medicine Department of Radiology 10 Thin PSAPD: energy spectra of individual crystals 3x8 array of 1x1x3 mm3 LSO crystals 250 300 200 200 150 100 100 0 0 5 300 0 400 400 300 300 200 200 200 100 0 0 5 0 0 5 200 5 0 10 800 800 800 800 600 600 600 600 400 400 400 400 200 200 200 200 0 0 5 300 200 100 100 5 10 0 5 0 0 0 5 500 0 0 5 0 0 5 400 300 100 0 0 5 0 0 0 5 500 500 400 400 300 300 200 200 100 100 0 5 0 0 5 0 0 5 0 5 0 5 0 5 10 300 400 200 200 200 0 100 50 0 300 300 200 150 100 100 0 250 200 200 100 0 100 300 300 100 0 400 300 200 0 5 500 200 400 300 100 50 0 400 100 100 0 0 5 400 300 0 0 5 300 300 200 200 200 100 100 0 0 5 Ground crystals 0 0 5 100 0 0 5 Polished crystals Stanford University MIPS Molecular Imaging Program at Stanford School of Medicine Department of Radiology Measured energy and time resolutions of LSO-PSAPD modules 22Na Energy Spectrum Coincidence time spectrum 800 600 TAC Data PSAPD START, 700 500 Gaussian Fit PMT STOP 500 Counts Counts 600 9.91% FWHM at 511 keV 400 300 400 300 ← FWHM: 2.1 ns 200 200 100 100 0 0 1 2 3 4 5 6 7 8 9 10 40 45 50 55 60 65 70 Time difference (ns) Voltage (v) Experimental measurements show that the average energy resolution is 11.0±1%, coincidence time resolution is 2.1±0.1 ns. Stanford University MIPS Molecular Imaging Program at Stanford School of Medicine Department of Radiology Energy resolution comparison conventional vs. proposed high res PET Conventional µPET system: LSO-Fiber-PSPMT 26.1% fwhm at 511 keV Proposed PET system: LSO-PSAPD 10.4% fwhm at 511 keV These proposed energy resolution improvements will yield enhanced sensitivity, image contrast and quantitative accuracy MIPS Molecular Imaging Program at Stanford Stanford University School of Medicine Department of Radiology 11 Summary Great progress has been made in pre-clinical radionuclide imaging instrumentation PET: •There are now six vendors for high resolution PET systems for small animal imaging •Over twenty research groups working on high resolution PET systems world wide •Sensitivity and spatial resolution, and energy resolution all continue to improve •Efforts to fuse this information onto high resolution CT and MR are being made In order for small animal imaging to assist existing drug development and testing protocols, must push for highly accurate image data and means to extract quantitative information that accurately characterize molecular signals MIPS Molecular Imaging Program at Stanford Stanford University School of Medicine Department of Radiology 12