Author: Veng Jean Heng, Jan P. Seuntjens, Timothy C. Zhu, Yifeng Zhu 👨🔬
Affiliation: Department of Radiation Oncology, Stanford University, Princess Margaret Cancer Centre & University of Toronto, University of Pennsylvania 🌍
Purpose: We have developed a GPU-accelerated Monte Carlo (MC) simulation to determine the cumulative dose distribution of patients undergoing Total Skin Electron Therapy (TSET) using the Stanford technique.
Methods: The GPU MC (PMC) for the TSET electron beam utilizes the phase space file generated right after the spoiler at TSE stand using TOPAS MC for a TrueBeam 6MeV. PMC relies on EGSnrc pre-calculated electron tracks to significantly speed up electron transport at runtime. The TOPAS MC was commissioned previously for TSET. Both depth dose and profiles for a dual TSET electron beam directed at a flat phantom surface at 500 cm are compared between PMC and TOPAS. PMC calculations are performed for a patient for all 6 postures and cumulative dose profiles, obtained by projecting dose for each posture onto the corresponding fem model which maintains the fem node numbering for the same anatomic site and then summed, are compared with direct TOPAS dose calculation for the same patient. Dose surface histogram (DSH) for the cumulative dose are compared between PMC and TOPAS on the same patient.
Results: Excellent agreement are obtained for both PDD and profiles at 500 cm between PMC and TOPAS on a flat phantom. PMC can account for the reduced depth dose and broadened dose profiles. The dose profiles on patient geometries agree between PMC and TOPAS. DSH for cumulative dose distributions agrees between PMC and TOPAS to within 3%. Calculation speed for TOPAS (on 12 Intel x5670) was 1 day while PMC (on a TITAN RTX) was <1 hour for all 6 patient postures. The dose uncertainty is much improved using PMC (2%) vs. TOPAS MC (10%).
Conclusion: GPU-accelerated MC is feasible for TSET dose calculation that resulted in accurate dose distribution with much improved uncertainty (2%) and greatly reduced calculation time (<1 hour).