(Sunday, 7/12/2020) [Eastern Time (GMT-4)]
Room: AAPM ePoster Library
Purpose: importance of imaging dose management in image-guided radiotherapy has been highlighted by several reports. Monte Carlo (MC) simulation is considered the most accurate method for 3D dose calculations in dose management. However, the main disadvantage of MC simulation is the long calculation time. The purpose of this study was to accelerate Monte Carlo dose calculation for kilovoltage cone-beam computed tomography (kV-CBCT) through graphics processing unit (GPU)-based parallel computing.
Methods: Monte Carlo-based dose calculation code (GPU-MC code) was implemented on a GPU using NVIDIA’s CUDA framework. The implementation was specifically designed for kV-CBCT and does not model electron interactions accurately. Photon transport (including Rayleigh scattering, photoelectric absorption and Compton scattering) was simulated in an analogous manner to the EGSnrc code. The average electron range for the kV-source is only a few millimeters in water-like tissues. Electron transport was modeled based on the simple continuous energy loss approximation. An analytic photon source model for the kV-CBCT based on the complete MC simulations of the X-ray tube was implemented in the GPU-MC code. The accuracy of the GPU-MC code was validated by comparing its calculated dose distributions in water with those from the EGSnrc code.
Results: GPU-based dose calculation engine was 100–300 times faster than a single-threaded CPU implementation. The dose distributions of open, full, and half bowtie filter fields showed good agreements. The dose differences between the calculated and measured PDD and OAR values were under 1%, except for the penumbra region where the maximum deviation was found to be approximately 3%.
Conclusion: accurate MC dose calculation engine for kV-CBCT was successfully implemented on the GPU.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
