Variable 3D resolution in complex detector geometries for Monte Carlo simulations

Objective.While current scoring algorithms in GEANT4 and Monte Carlo simulations support complex, mesh-based geometries, they are often limited in their ability to score physical quantities with variable spatial resolution within those volumes. We present a novel Monte Carlo simulation and analysis workflow that overcomes this limitation. The workflow enables customizable three-dimensional scoring of physical quantities and resembles a precise and flexible tool for analyzing this quantity as distribution in irregular formed geometries. Its application is exemplified by ruthenium-106 brachytherapy for intraocular tumors, while also being adaptable to broader clinical and computational scenarios.Approach.In GEANT4, a universal scoring algorithm that combines particle tracking with external post processing was implemented. Spatial coordinates and in our example deposited energy are recorded for each interaction as part of the simulation process. In the postprocessing phase, the complex geometries are binned into virtual volumes, allowing dose resolution to be adjusted as a separate step apart from the Monte Carlo simulation itself. The method was applied to the simulation of a ruthenium-106 eye applicator brachytherapy case using a 3D eye model constructed in computer-aided design software, enabling the generation of dose-volume histograms and precise analysis of three-dimensional dose distributions in structures at risk.Main results.The workflow achieved variable dose resolution across the different structures of the complex eye model, with high accuracy in clinically significant regions such as the tumor volume. Runtime analyzes show the scalability of the approach with multi-threading and its adaptability to diverse detector geometries and radiation modalities. The flexible resolution adjustment allows precise post-simulation analysis, minimizing computational costs while maintaining data accuracy.Significance.Our workflow represents an advancement in Monte Carlo simulation by enabling flexible and high-resolution scoring in complex geometries, surpassing the limitations of conventional algorithms.