GPU-accelerated Monte Carlo simulation of electron and photon interactions for radiotherapy applications

Area of Science:

• Medical Physics
• Computational Science
• Radiotherapy Technology

Background:

• Monte Carlo (MC) simulations are crucial for accurate dose evaluation in radiotherapy treatment planning.
• Current computational demands limit the clinical application of MC methods.
• There is a need for faster MC software to enable real-time dose assessment and plan optimization.

Purpose of the Study:

• To present an innovative GPU-accelerated MC software, FRED (Fast paRticle thErapy Dose evaluator), for dose valuation in electron and photon radiotherapy.
• To reduce the computational time required for MC simulations without compromising accuracy.
• To demonstrate the potential of FRED for clinical applications where speed is critical.

Main Methods:

• Implementation of electromagnetic interaction models (1 MeV-1 GeV) optimized for GPU execution.
• Particle transport simulation through a 3D voxel grid with energy deposition scoring.
• Benchmarking FRED against established MC codes (FLUKA, GEANT4) for specific radiotherapy scenarios.

Main Results:

• FRED achieved dose evaluation accuracy comparable to full MCs (within 2%) for dose-depth profiles and dose maps.
• Processing time was reduced by factors of 200 to 5000 compared to traditional MC methods.
• Electron beam treatment plans could be computed in minutes with high accuracy.

Conclusions:

• The GPU-accelerated FRED software offers a significant speed-up for MC-based dose calculations in radiotherapy.
• FRED's performance makes it suitable for rapid plan re-calculation and potentially adaptive radiotherapy.
• This advancement facilitates wider clinical adoption of accurate MC dose evaluation for photon and electron beams.