The PennPET Explorer Scanner for Total Body Applications

Transcription

1 The PennPET Explorer Scanner for Total Body Applications JS Karp, MJ Geagan, G Muehllehner, ME Werner, T McDermott, JP Schmall, V Viswanath, University of Pennsylvania, Philadelphia, PA AE Perkins, C-H Tung Philips Healthcare, Cleveland, OH Physics & Instrumentation Group Department of Radiology, University of Pennsylvania Funding support: R01-CA NIH R01-CA (sub to UC Davis) Philips Healthcare Acknowledgments: U Penn: Michael Parma, Margaret Daube-Witherspoon, Samuel Matej, Suleman Surti, TJ McSorley Philips: Thomas Bulgrin, Ilya Brodskiy, Alan Reed, Patrick Bender, Joseph Futey, Gregory Doughty, Cory Nesbitt, Chuck Carmen, Jim Liu, Joe Molyneux UC Davis: Simon Cherry, Ramsey Badawi U Ghent: Stefaan Vandenberghe IEEE MIC Atlanta, GA, October 26, 2017

2 Why large Axial FOV? Commercial PET/CT (~15-25 cm AFOV) - Total body PET - surveys with sequential beds - Provides best cost-benefit for clinical FDG

3 Why large Axial FOV? Commercial PET/CT (~15-25 cm AFOV) - Total body PET - surveys with sequential beds - Provides best cost-benefit for clinical FDG 3 minutes 1.28 mci/2.6 msv 1 minute 0.42 mci/0.9 msv High Sensitivity per volume - Improve SNR - Lower dose - Reduce scan time 5 yo Rhadomyosarcoma with metastatic disease Courtesy, L. States, CHOP

4 Why large Axial FOV? Commercial PET/CT (~15-25 cm AFOV) - Total body PET - surveys with sequential beds - Provides best cost-benefit for clinical FDG 3 minutes 1.28 mci/2.6 msv 1 minute 0.42 mci/0.9 msv High Sensitivity per volume - Improve SNR - Lower dose - Reduce scan time Simultaneous imaging of large volume - Bio-distribution studies of new tracers - Dynamic imaging of multi-organ systems 5 yo Rhadomyosarcoma with metastatic disease Courtesy, L. States, CHOP Fluorthanatrace (FTT): Dynamic imaging characterizes PARP inhibitor pharmacodynamics Courtesy, A. Pantel, R. Mach, UPenn

5 Why Total Body PET? September 12, 2017 U Penn EXPLORER Consortium: UC Davis funded Develop 2-m Total Body PET scanner September 21, 2017 June 30-July 2, 2018 U Ghent

6 Sensitivity (collected/emitted) (%) Total Body PET Design: Performance considerations Scanner length - Longer axial FOV increases sensitivity increased solid angle in 3D - Gain in point source (per slice) sensitivity limited due to attenuation Point source in 20-cm diam. cylinder Gain 2.3 Gain Diam. = 20 cm diameter 27 cm diameter 10 cm diameter 10 cm 20 cm AFOV 20 cm 70 cm 140 cm cm Axial field of view (cm) Axial field-of-view (cm) 200 cm Uniform activity in 20-cm diam. Gain 9.0 Gain 25 Gain 40

7 Total Body PET Design: Performance considerations Scanner length - Longer axial FOV increases sensitivity increased solid angle in 3D - Gain in point source (per slice) sensitivity limited due to attenuation Crystal - Smaller crystal width improves contrast, but not necessarily SNR - Thicker crystal improves sensitivity, but degrades spatial resolution Surti et al, whole-body PET Scanner with Improved Spatial and Timing Resolution. IEEE TNS 2013 Surti et al, using clinical metrics to optimize the scanner axial FOV and crystal thickness. PMB 2013 Surti et al, performance of a long axial field-of-view, whole-body PET scanner. PMB 2015 Schmall et al, Parallax error in long-axial field-of-view PET scanners. PMB cm AFOV, 600 ps TOF Activity Concentration (mci/cc) TOF - Better TOF improves SNR esp. at low activity Activity Concentration LROC LROC cm AFOV Act. Conc. 0.5mm lesions 3:1 contrast 10 min 100 cm ps Act. Conc. 300ps 450 4x4x20 450psps 600 ps 3x3x20 DOI72 3x3x20 600ps cm AFOV, 300 ps TOF X 10 lower dose

8 Total Body PET Design: Performance considerations Scanner length - Longer axial FOV increases sensitivity increased solid angle in 3D - Gain in point source (per slice) sensitivity limited due to attenuation Crystal - Smaller crystal width improves contrast, but not necessarily SNR - Thicker crystal improves sensitivity, but degrades spatial resolution TOF - Better TOF improves SNR esp. at low activity Axial resolution - Degrades due to parallax, but less than transverse parallax; can be modeled in reconstruction Transverse 23 cm 70 cm XCAT phantom, 3-min scan (GATE), DIRECT reconstruction center - 1 cm 3.96 mm 4.06 mm radial 10 cm radial 20 cm Axial 23 cm 70 cm center - 1cm cm cm ~ 0.5-mm loss

9 Total Body PET Design: Practical considerations Cost - Commercial (state-of-the-art) scanner: typical 20-mm thick crystals PET/CT ~$2-3M (MSRP) cm AFOV -> 8-13 liter scintillator, m 2 sensors - Long axial FOV scanner 70 cm AFOV -> 32 liter scintillator, 1.7 m 2 sensors -> x 3 cost ~$2M (materials) 140 cm AFOV -> 64 liter scintillator, 3.4 m 2 sensors -> x 6 cost ~$4M (materials) 210 cm AFOV -> 96 liter scintillator, 5.1 m 2 sensors -> x 9 cost ~$6M (materials) Reliability and HASM (health and safety monitoring) - Multiple scanners electronics, computers, etc ,000 (or more) SiPM devices requires high reliability and fault tolerance Heat load and detector cooling - Water cooling for SiPM sensors temperature control is critical - Air cooling for electronics

10 PennPET Explorer Design: Multi-ring Multi-ring construction for variable axial FOV 3 rings 70 cm torso, pediatric 6 rings 140 cm full body Detector module design with small gaps between rings 70 cm 70 cm Adult female 165 cm 5 5 Child 115 cm cm

11 counts TOF resolution (ps) PennPET Explorer Design: SiPM technology The detector tiles: 3.86 x 3.86 x 19 mm LYSO - Same as Philips Vereos, 64-channel array Read out by PDPC digital SiPMs - 1-to-1 coupling optimal light collection pixelated scintillator DSiPM array row 1 ave. peak to valley: trigger 2: 310 ps trigger 1: 240 ps 0 o C,T2 5 o C,T2 10 o C,T2 0 o C,T1 5 o C,T1 10 o C,T1 Detector sensitivity vs. Temperature trig 1 trig 2 0 C 95% 97% 5 C 93% 96% 15 C 71% 94% flood pixel Near perfect crystal ID Est. scanner singles rate (MHz) Improved timing resolution - trigger 1 PennPET Explorer Vereos Lower temp - reduced deadtime

12 Detector module design Copper tile plate Enable adjacent rings with minimal gap Temperature Control Coolant to tile differential <4 C. Spatial temperature uniformity across copper tile plate <2 C. Coolant flows through all modules in parallel uniform flow rate. System temperature uniformity (module to module) <1 C. Coolant flow Al plate w/ fins Copper plate tiles Condensation Control Enclose module bays. Infuse dry air (<2% RH) - slight positive pressure. Enclosed module bays TOF < 250 ps for PennPET Explorer can be achieved with 5 C at tile - Chiller must provide ~1 C water to gantry Manifolds to distribute chilled water to each module

13 Data Acquisition Architecture Data process (reconstruct) Storage Consumer (OTS) hardware ACQ Console Event sorter Uses production Philips Vereos ACQ hardware - Parallel ACQ requires clock alignment between rings Easily scaled from 3 rings to 6 rings or more Supports up to 100 MCPS singles events/ring - Off-line coincidence sorting Vereos hardware Detector Rings Predicted rates: Ac = 1 kbq/ml Singles = 7 Mcps Prompts = 0.52 Mcps Trues = 0.36 Mcps Ac = 5 kbq/ml (clinical FDG dose) Singles = 34 Mcps Prompts = 3.2 Mcps Trues = 1.8 Mcps Ac = 30 kbq/ml Singles = 205 Mcps Prompts = 27 Mcps Trues = 11 Mcps

14 PennPET Timeline 10/16 Tile testing & module assembly 2/16 Mechanical design gantry & detector modules 8/17 1 st ring complete 3/17 Gantry assembly started 10/17 2 nd ring complete

15 First phantom Measurements Initial 20 C Uniform Cylinder(s) V bias, Inhibit, TDC Energy, Timing Ring 1 Energy 12.1% Timing 332 ps Ring 2 All modules working Good uniformity Can be used for normalization correction Demonstrates performance with Vereos hardware

16 First phantom Measurements NEMA IEC Phantom with 9:1 fill ratio Good uniformity background (without norm) Good contrast - hot lesions and cold lesion Test data correction and image reconstruction software

17 First phantom Measurements Correct alignment 70-cm Pipe Phantom Activity: 5 mci Ring1 singles rate: 11.5 Mcps Ring2 singles rate: 10.6 Mcps Total prompt rate: 1.20 Mcps 1) Aligned the singles streams based on a spike in the coincidence/singles fraction. 2) Performed periodic checks to determine model of the alignment over time account for drift. 3) Merge the two singles streams using the dynamic offset to get the final coincidence list. 2 rings 5 minutes (without norm) Demonstrates running rings with independent clocks - data-derived time alignment to sort coincidences (within 4 ns window) - may require external master clock for 3 rings to avoid drift

18 Next steps towards PennPET Explorer PET/CT Complete implementation of data acquisition - testing of inter-ring alignment Build Ring 3 Phantom measurements of 70-cm (3-ring) system Integrate with CT and couch in clinical research space Human studies Initiate research protocols 3 to 6 rings Expansion (on-site) from 70-cm to 140-cm

19 Acknowledgments Funding support: R01-CA NIH R01-CA (sub to UC Davis) Philips Healthcare Acknowledgments: U Penn: Michael Parma, Margaret Daube-Witherspoon, Samuel Matej, Suleman Surti, TJ McSorley Philips: Thomas Bulgrin, Ilya Brodskiy, Alan Reed, Patrick Bender, Joseph Futey, Gregory Doughty, Cory Nesbitt, Chuck Carmen, Jim Liu, Joe Molyneux UC Davis: Simon Cherry, Ramsey Badawi U Ghent: Stefaan Vandenberghe