GPU-based Monte Carlo source modeling and simulation for radiation therapy involving Varian TrueBeam LINAC

Abstract

Monte Carlo particle transport methods are well recognized as the most accurate approach to dose calculations for radiation therapy treatment planning. However, the lengthy computation time due to the inherent statistical nature of the method has hindered the acceptance of Monte Carlo dose calculation method by clinicians for a long time. In recent years, the emerging parallel computing hardware accelerators, such as general-purpose Graphics Processing Unit (GPU), have significantly boosted the software performance in many fields and also cast lights on enhancing the simulation efficiency of Monte Carlo dose calculation methods. Under the circumstance of external beam radiation therapy dose calculations, an accurate dose distribution calculation depends on two major factors: an accurate modeling of the clinical LINear ACcelerator (LINAC) that serves as the source term and an accurate methodology to transport particles. Therefore, the objective of this study is to develop an accurate and fast Monte Carlo dose engine for Intensity-Modulated Radiation Therapy (IMRT) dose calculations during the treatment planning process.

To achieve this goal, a source model of the Varian TrueBeam LINAC and an efficient, First-Compton-based source particle sampling method are developed and are integrated with a fast coupled electron-photon transport module of the recently developed high-performance Monte Carlo code, ARCHER. The simulation results of different parts of the source model are verified with established dose calculation codes and validated with experimental measurements. Isodose distributions, dose difference maps and three-dimensional Gamma index tests are performed to demonstrate and quantify the benchmark results. To demonstrate the capability of the platform in clinical applications, three clinical treatment plans (the lung, breast and prostate) are simulated and analyzed. The hardware devices involved in this study include one NVIDIA TitanX GPU, one NVIDIA K40 GPU, one NVIDIA K20 GPU and one 6-cores Intel X5650 CPU. From the perspective of performance, the entire simulation of each clinical plan, starting from the source modeling to the end of particle transport within the patient phantom, can be finished in less than one minute using a single Titan X GPU card with 1% statistical error. In comparison, the same cases take the EGSnrc code hundreds of CPU hours to finish. The presented results indicate that this new version of ARCHER code was successfully developed and tested as a clinically deployable dose engine for external photon beam IMRT plans in the clinical setting.