Monte Carlo (MC) simulation is an important tool in radiation therapy. It is commonly regarded as the most accurate dose calculation method due to the precise modeling of transport physics and object geometry. In this project, we developed high-performance MC tools on parallel processing platform, e.g. GPU, for coupled photon/electron transport simulations in MeV energy range. We also build accurate source models for medical linear accelerators and automated commissioning methods to facilitate clinical applications of the MC tool. Our tool gDPM and goMC can compute doses for a typical treatment plan in less than 1 minute on a single GPU card with less than 1 percent uncertainty. We have been employing the developed system in a variety of clinical projects, such as secondary dose calculations for treatment plan quality assurance, dose reconstruction for stereotactic body radiotherapy, and MC-based inverse treatment planning.
MC also plays an important role in X-ray imaging studies, such as cone beam CT (CBCT) reconstruction and dose calculations. We have been developing high-performance MC tools gCTD/gMCDRR on GPU platform for keV photon transport simulations for CBCT imaging and dosimetry. With accurate modeling of kV photon transport, these tools can generate CBCT projections accurately at a high efficiency (~1min/projection) and compute dose from X-ray imaging procedures, e.g. CBCT (~10 sec).
In collaboration with Dr. Harald Paganetti at Massachusetts General Hospital, we have developed a GPU-based MC tool gPMC for particle transport simulations in proton therapy. We aim at developing a tool with appropriate physics models to maintain dose calculation accuracy as well as GPU-friendly parallelization schemes to achieve a high computational efficiency. Beam modeling and treatment geometry are also included in the package to allow clinical applications. We have achieved a high computational efficiency such that it took 10-22 sec to transport 107 source protons on a single GPU platform.
Carbon therapy treatment planning also heavily relies on MC simulations, not only for physical dose calculation, but also for evaluation of biology-related quantities. We have developed a GPU-based MC tool goCMC to facilitate carbon therapy in collaboration with Dr. Katia Parodi’s group at Ludwig Maximilian University of Munich. This tool employs accurate particle transport physics to simulate transports of carbon ions and all the secondary particles. Its accuracy has been benchmarked against Geant4 with different phantoms and beam energies. goCMC supports scoring various quantities including physical dose, particle fluence, spectrum, linear energy transfer, and positron emitting nuclei. Depending on the beam energy and voxel size, it took 20-100 seconds to simulate 107 carbons on a GPU card. The corresponding CPU time for Geant4 with the same voxelized phantom was 60-100 hours.