| Monte Carlo method is the gold standard in radiation transport calculations. How-ever, the prolonged simulation time impedes its wide applications in the field of medical physics including imaging and radiation treatment. Recently emerged parallel computing hardware accelerators, such as Graphics Processing Unit (GPU), brought an opportunity to improve the speed of Monte Carlo calculations. In this study, an extremely fast Monte Carlo code, ARCHER for RadioTherapy (RT), was developed as a testbed for the GPU environment to perform radiation dose calculations specific to external beam radiation therapy. This dissertation describes the detailed physics models and software develop-ment of the code, and its applications to three clinical TomoTherapyRTM cases. Coupled electron-photon transport, for up to 20 MeV, was modeled in the design of this code. Photon interactions including the photoelectric effect, Compton scattering, and pair pro-duction were considered. For electron, inelastic collision, elastic collision and brems-strahlung emission were included. Furthermore, two different Woodcock tracking algo-rithms were implemented and their relative performance was compared. ARCHER for RT was benchmarked against EGSnrc/DOSXYZnrc using different radiation sources. For clinical cases, phase space files (PSFs) were used as the radiation source input. Pa-tient-specific phantoms were constructed from patient CT images. Batch simulations were employed to facilitate the time-consuming task of loading large PSFs, and to im-prove the estimation of statistical uncertainty. For the clinical treatment cases, dose vol-ume histograms (DVHs) and isodose maps were produced from ARCHER and the gen-eral-purpose Monte Carlo code, GEANT4. Gamma index analysis was performed to evaluate the similarity of voxel doses obtained from these two codes. The hardware ac-celerators used in this study were one NVIDIA K20 GPU, one NVIDIA K40 GPU and six NVIDIA M2090 GPUs. In addition, to perform a fair comparison of the CPU and GPU performance, a multi-threaded CPU code was developed using OpenMP and tested on an Intel E5-2620 CPU. The benchmark results showed good agreement between ARCHER and EGSnrc. For clinical cases, with the 2%/2mm criteria, the gamma test passing rates for the prostate, lung case and head & neck cases are 99.7%, 98.5% and 97.2%, respectively. Meanwhile, ARCHER achieved a fast speed for PSF-based dose calculations. With a single M2090 card, the simulations took about 60, 50, 80 seconds for three cases, respectively, with 1% statistical error in the PTV. Using the latest K40 GPU card, the simulation times were further reduced by 1.7-1.8. When six M2090 cards were used, the total computing time was 8.9-13.4 seconds. For comparison, the same simulations on Intel E5-2620 (12 hyper-threads) cost about 500~800 seconds. In conclu-sion, ARCHER for RT was developed successfully to perform fast and accurate MC dose calculation for radiotherapy using PSFs and patient CT phantoms. |
