A hybrid phase-space and histogram source model for GPU-based Monte Carlo radiotherapy dose calculation.

Area of Science:

• Medical Physics
• Computational Biology
• Radiotherapy Physics

Background:

• GPU-based Monte Carlo simulations are crucial for radiotherapy dose calculation.
• Source modeling from phase-space data presents an efficiency challenge.
• Phase-space-let (PSL) sources have previously improved simulation efficiency.

Purpose of the Study:

• To enhance the efficiency of GPU-based Monte Carlo simulations for radiotherapy.
• To introduce and validate a hybrid primary photon point source and secondary PSL source model.
• To accelerate dose calculation without compromising accuracy.

Main Methods:

• Developed a novel phase-space derived, histogram-based hybrid source model.
• Integrated the model into gDPM v3.0 software.
• Demonstrated a method for deriving source characteristics from LATCH-less phase-space data.
• Validated the model using open fields and intensity-modulated radiation therapy (IMRT) plans on Varian and TrueBeam machines.

Main Results:

• The hybrid model achieved >97% chi-test agreement (mean 99%) above the 2% isodose with 1%/1mm criteria for open fields.
• Root mean square deviations (RMSDs) were <1% (mean 0.5%) for open fields.
• Source generation time was 4-5 times faster for open fields.
• For IMRT, 95% chi-test agreement was achieved above 10% isodose with 1%/1mm criteria.
• IMRT RMSD was 0.8% with a 2.5x source generation speed-up factor.

Conclusions:

• The hybrid source model significantly accelerates radiotherapy dose calculations.
• The model maintains high accuracy, comparable to existing methods.
• This approach offers a practical solution for improving the efficiency of Monte Carlo simulations in radiotherapy.