Advanced Monte Carlo scoring options for Proton Therapy treatment planning

Proton therapy offers highly conformal dose delivery through the Bragg peak phenomenon, making it a powerful tool for treating deep-seated tumors while minimizing damage to surrounding healthy tissues. Accurate dose modeling and biological optimization in this context demand robust simulation tools. This thesis investigates advanced Monte Carlo (MC) scoring strategies for proton therapy treatment planning by comparing two state-of-the-art platforms: GATE, a Geant4-based CPU simulation toolkit known for its high-fidelity physics modeling, and FRED, a GPU-accelerated engine optimized for clinical speed and efficiency. After detailing the physical principles underpinning proton therapy, the thesis introduces the simulation setup replicating PSI’s Gantry 2 beamline. A harmonized geometry and beam model were implemented in both platforms to ensure comparability. Dose distributions were first evaluated in homogeneous and heterogeneous phantoms to validate the consistency between GATE and FRED. Particular attention was then devoted to the novel scoring of proton Track Ends, a quantity increasingly recognized for its correlation with biological effects such as LET and RBE. Track-end distributions were analyzed in water, tissue-equivalent phantoms, and patient CT data, highlighting the role of secondary particle production and the importance of consistent filtering between platforms. The results demonstrate proper agreement in dose calculations and good consistency in Track Ends scoring, especially when appropriate filters and production cuts are applied in GATE to match FRED’s simpler scoring logic. The study concludes that FRED, despite its simplified transport model, can serve as a fast and reliable alternative to GATE for scoring Track Ends and supporting biologically informed treatment planning. These findings support the integration of GPU-based simulations into clinical workflows for proton therapy.