A GPU-accelerated radiation dose engine for small animal proton radiotherapy

Area of Science:

• Medical Physics
• Radiation Oncology
• Computational Biology

Background:

• Preclinical small animal experiments are crucial for proton therapy research.
• Accurate dose calculation in small animals is challenging due to low beam energy and submillimeter resolution needs.
• The Monte Carlo method provides precision but is computationally intensive.

Purpose of the Study:

• Develop a GPU-accelerated Monte Carlo engine (pGARDEN) for fast and accurate proton dose calculations in small animal irradiation.
• Optimize dose engine for small animal proton radiotherapy research.

Main Methods:

• Optimized particle transport algorithm for GPU architecture.
• Implemented acceleration techniques for computational efficiency.
• Tuned physical parameters (e.g., energy cutoffs) for small animal conditions.
• Validated against Geant4 simulations and experimental measurements.

Main Results:

• Achieved >1000-fold speedup compared to Geant4 with >97% 3D gamma passing rate (1%/0.15 mm).
• Demonstrated good agreement with measurements for depth dose and profiles.
• In vivo validation showed high 2D gamma passing rates (e.g., 95.52% ± 0.74% in abdomen).
• Calculated a treatment plan in 4.3s with <1% uncertainty and 100% 3D gamma passing rate (2%/0.3 mm).

Conclusions:

• pGARDEN enables rapid and accurate proton dose distribution calculation at submillimeter resolution for small animals.
• It serves as a valuable tool for preclinical proton radiation experiments.
• Supports research on relative biological effectiveness (RBE) and novel therapeutic strategies.