S1728
Physics - Dose prediction/calculation, optimisation and applications for particle therapy planning
ESTRO 2026
1 Department of Radiotherapy, Peking University Cancer Hospital & Institute, Beijing, China. 2 Department of Technology, CAS Ion Medical Technology Co., Ltd., Beijing, China. 3 School of Physics, Beihang University, Beijing, China. 4 Department of Radiotherapy, Mianyang Third People’s Hospital, Mianyang, China. 5 Department of Radiation Physics, Zhejiang Cancer Hospital, Hangzhou, China. 6 Department of Radiation Oncology, Peking University Third Hospital, Beijing, China. 7 Sino-French Carbon Neutrality Research Center, Beihang University, Beijing, China Purpose/Objective: Accurate dose calculation is essential for securing the therapeutic benefit and safety of carbon ion radiotherapy1,2. However, the uncertainty associated with tissue parameter estimation methods for Monte Carlo–based dose calculation has not been systematically evaluated, potentially limiting the reliability of quality assurance in clinical practice. This study aims to systematically investigate how upstream uncertainties in physical density and elemental composition propagate into downstream Monte Carlo dose calculations. By comparing three elemental decomposition strategies, we quantitatively evaluated their impact on the consistency and reliability of both physical and biological dose distributions. Material/Methods: Uncertainties in physical density and elemental composition were assessed and propagated for three approaches: single-energy CT3 (SECT), parameterized dual-energy CT4 (PA-DECT), and machine-learning– based DECT5 (ML-DECT). For each method, relevant uncertainty sources—originating from imaging, CT- number calculation, model approximation, parameterization, and training stochasticity —were identified based on their methodological workflows and combined following the GUM framework6,7. The resulting uncertainty distributions were then used to generate voxel-wise perturbations through Gaussian sampling, which were applied to the ICRP 110 phantom. Dose distributions were simulated using FLUKA with 10 realizations per method, and both physical and biological dose metrics were evaluated. Voxel-wise uncertainty, gamma passing rate, and range uncertainty were examined to characterize the quantitative influence of tissue-parameter perturbations on the resulting dose distributions. Results: Across all decomposition strategies, ML-DECT consistently yielded the lowest uncertainties in tissue parameters, reducing the uncertainties in C, N and O density by up to 77.4% compared with SECT. In Monte Carlo dose simulations, ML-DECT achieved the lowest average relative uncertainties (~5% physical; ~7–9% biological), together with the highest gamma passing
rate (97.96 ± 0.28%) and the smallest range uncertainty (0.5%). The visualized absolute biological dose uncertainties (Fig. 1) demonstrated markedly reduced uncertainty for ML-DECT across orthogonal planes. The biologically weighted IDD curves (Fig. 2) showed that ML-DECT produced the most stable distal fall-off and minimal inter-sample variability.
Conclusion: More accurate modelling of the relationship between CT numbers and tissue parameters substantially reduces uncertainty in both elemental decomposition and subsequent Monte Carlo dose calculation.Among the evaluated methods, ML-DECT demonstrated the greatest robustness and stability, supporting its
Made with FlippingBook - Share PDF online