ESTRO 2025 - Abstract Book

S2698

Physics - Dose calculation algorithms

ESTRO 2025

3035

Digital Poster Evaluation of a GPU Monte Carlo code against the XVMC code implemented in Monaco TPS Matthias Kowatsch 1 , Eva Partoll 1,2 , Christian Attenberger 1 , Philipp Szeverinski 2 , Patrick Clemens 3 , Thomas Künzler 2 , Peter Tschann 4 1 Faculty of Medical Sciences, Private University in the Principality of Liechtenstein (UFL), Triesen, Liechtenstein. 2 Institute for Medical Physics, Academic Teaching Hospital Feldkirch, Feldkirch, Austria. 3 Department for Radiotherapy, Academic Teaching Hospital Feldkirch, Feldkirch, Austria. 4 Department of General and Thoracic Surgery, Academic Teaching Hospital Feldkirch, Feldkirch, Austria Purpose/Objective: Achieving high accuracy (via Monte Carlo (MC)) and speed (GPU computing) is critical for radiotherapy dose calculation. GPU-based MC codes have been developed to address these requirements. This study benchmarks two algorithms – XVMC and GPUMC – implemented in the Monaco treatment planning system (TPS) to evaluate the accuracy of GPUMC. Material/Methods: An alpha version of the Monaco TPS (Elekta, Crawley) was used to compare XVMC and GPUMC. Beam models for both algorithms were based on identical beam datasets generated by Elekta. Fifty radiotherapy cases were selected randomly, ensuring comprehensive regional coverage. Original plans were recalculated using GPUMC with identical grid size and statistical uncertainty. Plan quality was assessed by evaluating the PTVMean dose. Additionally, 100 plans (3 mm grid size, 1% statistical uncertainty) were recalculated using an IBA Body Phantom (IBA Dosimetry, Schwarzenbruck) and validated with measurements from a 0.125 cc ionization chamber (PTW, Freiburg). Results: The statistical analysis showed that the results for the comparison of XVMC to GPUMC and GPUMC to measurement are normally distributed, whereas those for XVMC to measurement are not normally distributed. The median of the difference between XVMC and GPUMC for the PTVMean dose was 0.3% with an interquartile range of 0.7%. Comparing GPUMC and XVMC to the measurements, the median differences were -0.2% with an interquartile range of 1.5% and -1.3% with an interquartile range of 1.1% respectively.

Conclusion: The study included 50 diverse patient cases, ensuring broad applicability. Results demonstrate strong agreement between XVMC and GPUMC, though only the high-dose region (PTVMean) was analyzed. Future studies should investigate low-dose regions. Measurements were taken in a homogenous area, with deviations within the 1.4% ionization chamber tolerance (2σ). Gradient region accuracy could not be confirmed, necessitating further analysis.

Keywords: GPUMC, XVMC, Comparison