pGPUMCD: an efficient GPU-based Monte Carlo code for accurate proton dose calculations.

Area of Science:

• Medical Physics
• Computational Physics

Background:

• Accurate dose prediction is crucial in proton therapy.
• Monte Carlo simulations offer high precision but can be computationally intensive.

Purpose of the Study:

• Introduce and benchmark pGPUMCD, a novel GPU-accelerated Monte Carlo code.
• Evaluate its accuracy and efficiency for proton therapy applications.

Main Methods:

• Proton transport in voxelized geometry using a class II condensed history scheme.
• Incorporation of L_eq formalism, Fermi-Eyges scattering, and discrete interactions.
• Validation against Geant4, assessing dose differences and computation times.

Main Results:

• Dose differences with Geant4 were <1% in the Bragg peak and up to 3% in the distal fall-off.
• 99.5% of positions showed voxelwise dose differences <1%.
• One million protons were transported in <0.5 s for 1 mm³ voxels.

Conclusions:

• pGPUMCD provides accurate dose prediction comparable to Geant4.
• Its computational efficiency makes it a strong candidate for clinical integration.
• The L_eq formalism variance reduction technique enhances its performance.