Monte Carlo simulation is often employed to derive precise dose calculation results for radiotherapy due to its ability to simulate complex physics interactions between incident beams and irradiated subjects. However, applying Monte Carlo simulation to clinical practices poses a significant challenge because of a huge amount of computational demands and time. In response to this challenge, an in-house dose calculation engine using the condensed history class-II Monte Carlo method is developed with various CPU and GPU accelerated algorithms. Moreover, we adopt the latest container framework, docker, for the ease to deploy the accelerated in-house dose calculation engine onto different acceleration platforms, such as a PC equipped with an NVIDIA GTX 1080 GPU card or a server equipped with hundreds of CPU cores. At the moment, we focus on proton therapy simulation. The physics simulations include electromagnetic processes, such as continuous ionization, energy straggling, and multiple coulomb scattering by Highland's approximation, while the processes via the nuclear forces for elastic and non-elastic scattering are in progress. A typical test example of 160-MeV proton beam into water is presented for comparing the simulation time and results between single-threaded Geant4-10.5.0 and the developed GPU algorithm of our in-house dose calculation engine executed on the platforms, each of which is equipped with NVIDIA GTX 1060 and GTX 1080 GPU cards, respectively. Comparing with the dose simulation results of Geant4, the passing rates of 1%/1mm gamma index achieve 100% and the speedup factors are 104 and 221 times, respectively.

We are developing a containerized Monte Carlo dose calculation engine integrated with CPU and GPU accelerated algorithms.