SU-E-T-493: Accelerated Monte Carlo Methods for Photon Dosimetry Using a Dual-GPU System and CUDA.

Area of Science:

• Medical Physics
• Computational Science
• Radiotherapy

Background:

• Accurate radiation dose calculation is crucial for effective cancer treatment planning.
• Traditional Monte Carlo (MC) methods for dose calculation are computationally intensive.
• Accelerating MC simulations can improve treatment planning efficiency.

Purpose of the Study:

• To develop a Graphics Processing Unit (GPU) based Monte Carlo (MC) code for accelerated dose calculations.
• To optimize the code for a dual-GPU system for enhanced performance.
• To ensure both efficiency and accuracy in the developed MC code.

Main Methods:

• A voxelized abdomen phantom from CT data and a GE scanner were modeled.
• A C++ CPU-based MC code was developed and then translated to CUDA C for GPU execution.
• The code featured automatic task assignment to multiple GPUs and accurate cross-section calculations.

Main Results:

• The GPU-based MC code achieved a 19x speedup over the CPU code on a single GPU, and 42x over MCNPX.
• Speedup factors doubled on a dual-GPU system, demonstrating scalability.
• Dose calculations showed a maximum difference of 1% compared to MCNPX with a relative error below 0.1%.

Conclusions:

• A GPU-based MC code was successfully developed for dose calculations using detailed patient and CT scanner models.
• The developed code demonstrates significant efficiency gains and maintains high accuracy.
• The study confirms the scalability and effectiveness of the GPU-based MC code on a dual-GPU system.