MOQUI: an open-source GPU-based Monte Carlo code for proton dose calculation with efficient data structure.

Area of Science:

• Medical Physics
• Computational Science
• Radiotherapy

Background:

• Monte Carlo (MC) codes are crucial for accurate radiotherapy dose calculations, especially in proton therapy.
• MC simulations require significant computational resources for statistical accuracy.
• Graphic Processing Units (GPUs) offer computational speedups but have limited memory, posing challenges for large-scale simulations.

Purpose of the Study:

• To develop an open-source GPU-based MC package capable of scoring large quantities with limited GPU memory.
• To improve the computational efficiency of MC simulations in radiotherapy.

Main Methods:

• Developed a novel GPU-based MC code named moqui.
• Integrated a hash-table data structure to efficiently manage memory on the GPU.
• Validated moqui against the TOPAS MC code using homogeneous and heterogeneous phantoms and clinical treatment plans.

Main Results:

• Moqui demonstrated agreement with TOPAS within 2% for dose calculations, with >99% gamma pass rates (2 mm/2% criteria).
• The hash-table approach enabled scoring of dose-influence matrix and dose-rate on a GPU using only 10 GB of memory.
• This is a significant reduction compared to the >100 GB required by conventional array structures.

Conclusions:

• Moqui offers an efficient and accurate solution for GPU-based MC dose calculations in proton therapy.
• The hash-table implementation effectively addresses GPU memory limitations for large-scale simulations.
• This advancement can accelerate radiotherapy treatment planning and research.