Click here to


Are you sure ?

Yes, do it No, cancel

GPET: A GPU-Based Accurate and Efficient Monte Carlo Simulation Tool for PET

Y Lai1*, Y Zhong2 , M Jin1 , X Jia2 , Y Chi1, (1) University of Texas at Arlington, Arlington, TX, (2) UT Southwestern Medical Center, Dallas, TX


(Sunday, 7/14/2019) 4:00 PM - 6:00 PM

Room: 304ABC

Purpose: To fully address the time consuming issue in current CPU-based Monte Carlo (MC) packages for positron emission tomography (PET), a GPU based efficient and accurate MC tool—gPET, was developed, further optimized and validated.

Methods: gPET was built on the NVidia CUDA platform. The simulation process was divided into three functional parts: 1) gamma pairs generation, including positron decay, transport and annihilation, 2) gamma transport inside the voxelized phantom, and 3) signal detection and processing inside the parametrized detector, with 1) and 3) the new functions added compared to the initial development in 2018. The detector was formed in three levels: panel, module and crystal scintillator. A user can selectively tally quantities as outputs, like intermediate photon phase space file (PSF), hits, singles and coincidence. The performance of gPET was compared with GATE8.0/Geant4 in a comprehensive simulation study.

Results: 1) 1e7 positrons from C-11 point source centered in an 8cm3 cubic water phantom were simulated in both gPET and GATE/Geant4. Positron stopping points were tracked and gamma PSF was recorded on a spherical surface 3cm away from the phantom center. The mean of the positron ranges was 0.99mm and 1.14mm for gPET and GATE/Geant4, respectively. 0.5% difference in angular distribution was found for the PSF. 2) The PSFs were further transported in an eight-panel-detector 6.65cm away from the center. For hits (singles) recorded on discretized crystals, the energy and spatial differences were 2.2% (2.1%) and 1.6% (1.5%), respectively. The mean±standard deviation of the reconstructed images based on back-retrace of the coincidences were 0±3.49mm and 0±3.13mm for gPET and GATE/Geant4, respectively. Corresponding execution time for gPET on single Titan Xp GPU (1.58GHz) and GATE8.0 on single Intel i7-6850K CPU (3.6GHz) were 20s and 4260s.

Conclusion: gPET is an efficient and accurate MC tool for PET with open-source gPET-v1.0 available at


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email