ESTRO 2025 - Abstract Book

S3468

Physics - Optimisation, algorithms and applications for ion beam treatment planning

ESTRO 2025

of the quadratic term is determined using biological parameters like α 0 , β 0 , R n and r d . Model performance was evaluated using the experimental dataset, 9 which was also used for parameter fitting in the mMK model. 4 Parameter fitting for the eMK model used a combination of global and local fitting methods, including grid search and trust region reflective fitting. Results: Fitting the eMK model with D 10 values showed a higher mean square error (MSE) than the mMK model across the entire dose range. However, fitting with D 10 , D 1 , and D 0.1 values, the eMK outperformed mMK in the 0.1% and 0.01% survival ranges, with a slightly higher MSE at 10% survival. Furthermore, cell survival curves based on high-LET beams showed a better match to experimental data with the eMK model compared to the mMK model.

Conclusion: The improved predictive accuracy of the eMK model is likely due to its quadratic term, which is determined by biological parameters and diminishes as the LET increases, resulting in a less curved survival curve at high doses. Despite its higher MSE with the D 10 metric, the eMK model is more accurate in high-dose and high-LET regions. Future research should focus on obtaining more appropriate experimental data and developing fitting techniques that reliably find the global minimum, as well as refining the model with Monte Carlo-based dose distribution calculations 10 to enhance prediction accuracy.

Keywords: MK model, Hypofractionated ion beam therapy