Dosimetric study of proton boron capture therapy based on Monte Carlo simulation and machine learning algorithms

Proton boron capture therapy offers a promising enhancement to conventional proton therapy by leveraging the 11B(p,a)3a nuclear reaction for localized dose amplification. This study systematically investigates the Bragg curve characteristics of proton beams using the Geant4 Monte Carlo toolkit, with a particular focus on the dosimetric impact of ??B. The introduction of a pure ??B target induced a consistent forward shift in the Bragg peak position, attributed to increased stopping power and nuclear reactions. Furthermore, Proton boron capture therapy demonstrated enhanced local energy deposition along the central axis due to the generation of high linear energy transfer alpha particles, with minimal lateral broadening. To facilitate precise treatment planning, Bragg curve data for 800 distinct proton energies (0-80 MeV) in water were generated. Various machine learning algorithms were subsequently employed to develop predictive models for the Bragg peak position. Comparative analysis identified gaussian process regression as the optimal model, achieving an R? of 0.999997 and a root mean squared error of approximately 0.0273 mm for predicting Bragg peak positions in water. Crucially, this research pioneers a novel pathway for proton boron capture therapy treatment planning by combining high-accuracy machine learning-based prediction of the initial Bragg peak (in water) with a characterized correction for the ??B-induced forward shift, enabling more precise determination of the actual treatment depth. This work provides critical dosimetric characterization, quantifies key ??B-induced phenomena, and offers a validated predictive framework, thereby establishing a theoretical foundation and technical support for dose optimization in this emerging therapeutic modality.