GGEMS-Brachy: GPU GEant4-based Monte Carlo simulation for brachytherapy applications.

Area of Science:

• Medical Physics
• Radiation Oncology
• Computational Science

Background:

• Current brachytherapy planning relies on AAPM TG43, simplifying patients to water phantoms.
• Accurate modeling of patient heterogeneity with Monte Carlo simulation (MCS) is computationally intensive.
• Existing MCS methods are too slow for routine clinical use in brachytherapy.

Purpose of the Study:

• To implement and evaluate a fast, accurate MCS platform on GPUs for low dose rate (LDR) brachytherapy.
• To enable patient-specific dosimetry considering anatomical and physical complexities.
• To assess the clinical feasibility of GPU-accelerated MCS for brachytherapy.

Main Methods:

• Extended a Geant4-based GPU framework (GGEMS) with a hybrid GPU navigator for voxelized patient images.
• Incorporated analytical models for (125)I seeds and track-length estimator dose scoring with uncertainty.
• Validated the GGEMS-brachy platform against Geant4 simulations and reference datasets.

Main Results:

• Achieved a mean patient dosimetry study run time of 9.35s (1 GPU) and 2.5s (4 GPUs) for 2% uncertainty.
• Demonstrated accurate dose scoring, including uncertainty calculations, for LDR brachytherapy.
• The platform's performance is compatible with clinical workflows.

Conclusions:

• The GGEMS-brachy platform provides a computationally efficient and accurate solution for brachytherapy dosimetry.
• Enables routine use of MCS for patient-specific dose calculations, improving treatment planning.
• Applicable to various LDR brachytherapy sites, with potential for high dose rate applications.