S2601
Radiobiology - Radiobiological modelling
ESTRO 2026
when accelerated via GPU [5]. Here we propose GRWU, GPU-based Restricted Water diffUsion simulator, combining C++ speed with GPU parallelisation for faster diffusion simulations in complex microenvironments. Our aim is to provide a practical and time- efficient tool for tumour microstructural characterisation, compatible with the temporal constraints of standard clinical research. Material/Methods: GRWU implements the classic Monte Carlo simulations of water diffusion that have been presented in previous works [3, 4, 5, 6]; GRWU is cross-platform, requiring only a CUDA-capable GPU. Its functioning is summarised in Figure1. The simulator allows the implementation of user-defined complex substrates and acquisition schemes, including the possibility to define the echo time.To verify GRWU performance and accuracy, a comparison between three other simulators, Camino [3], MCDC Simulator (single core) [4] and Disimpy [5] was carried out. The speed was verified on the substrates defined by [1] and [2], with a growing number of mesh faces, analysing how an increase in substrate complexity impacted computational times.
Conclusion: The presented coupled framework is a versatile tool for
radiobiological research, accurately reproducing experimental spheroid growth and radiation response with cell-by-cell resolution. Crucially, the model incorporates oxygen concentration to modulate the induction of indirect damage, providing a powerful platform to analyze radiobiological dynamics within the tumor microenvironment. This in silico platform enables the exploration of spatial and temporal dynamics inaccessible in standard experiments, offering a valuable method to guide and optimize future radiobiological studies. References: Schuemann, J., McNamara, A. L., Ramos-Méndez, J., Perl, J., Held, K. D., Paganetti, H., Incerti, S., & Faddegon, B. (2018). TOPAS-nBio: An Extension to the TOPAS Simulation Toolkit for Cellular and Sub- cellular Radiobiology. Radiation Research, 191(2), 125. https://doi.org/10.1667/RR15226.1 Stewart, R. D. (2001). Two-Lesion Kinetic Model of Double-Strand Break Rejoining and Cell Killing. Radiation Research, 156(4), 365–378. https://doi.org/10.1667/0033-
7587(2001)156%255B0365:TLKMOD%255D2.0.CO;2 Keywords: DNA repair, Tumor Microenvironment
Results: GRWU results are comparable with the other tools with a mean difference on MR signals of 0.33% with respect to Camino, 0.35% with MCDC and 0.67% with Disimpy, a variability that is observable also when comparing the other tools among each other.Tests show faster simulation times with respect to all other simulators, both CPU- and GPU-based, as visible in Figure2. GRWU demonstrates an average speed-up of 930 × over Camino, 368 × over MCDC (single core) and 2 × over Disimpy.
Digital Poster 2725 GRWU: a GPU-based Restricted Water diffUsion simulator for tumour microstructural characterisation. Chiara Scotti, Fabio Casaccio, Chiara Tinelli, Letizia Morelli, Guido Baroni, Chiara Paganelli DEIB, Politecnico di Milano, Milano, Italy Purpose/Objective: Diffusion Weighted Magnetic Resonance Imaging (DW-MRI) is emerging as a technique to personalise radiation therapy treatments, as it allows tumour microstructural characterisation and outcome prediction. Frameworks such as those by Buizza et al. 2021 [1] and Morelli et al. 2023 [2], have been developed to match cancer patients DW-MR signals, with signals simulated in virtual tumour substrates, allowing to extract the patients’ voxel-wise tumour microenvironment.To perform DW-MRI simulations many tools are available but they require long computational times [3, 4], even
Made with FlippingBook - Share PDF online