ESTRO 2025 - Abstract Book

S2735

Physics - Dose prediction, optimisation and applications of photon and electron planning

ESTRO 2025

Results:

Figure 2 demonstrates several predictive DVH metrics plotted versus ground truth of a sample HA-SRS plan. The average RMSE of three surrogate models for three typical DVH metrics of different structures is listed below: For the right eye D mean , the average R 2 are 0.98, 0.97, and 0.99; For the optic chiasma D 0.03cc , average R 2 0.76, 0.58, and 0.95; For the GTV1 D 98% , average R 2 0.53, 0.89, and 0.95, for the PCE, GP and PCK surrogates respectively. On average, PCK outperforms GP and PCE in terms of predictive accuracies, given identical training sets. We also observed that the predictive accuracies of the models for D mean are generally superior to that for D max , while the model exhibits relatively poor performance for DVH metrics in small structures, e.g. the optic chiasma D 0.03cc . Conclusion: A robustness evaluation tool was implemented through interaction with the TPS for photon radiotherapy. The predictive accuracies of DVH metrics for three surrogate models were validated and compared on HA-SRS plans. The demonstrated accuracies of the surrogate models provide confidence for the following accelerated sensitivity analysis of treatment plans, complementing the drawbacks of the current worst-case-evaluation approach.

Keywords: Uncertainty Quantification, Robustness Evaluation

References: [1] Lomax, A. J. (2008). Intensity modulated proton therapy and its sensitivity to treatment uncertainties 1: the potential effects of calculational uncertainties. Physics in Medicine & Biology , 53 (4). [2] Perkó, Z., Van Der Voort, S.R., Van De Water, et.al. 2016. Fast and accurate sensitivity analysis of IMPT treatment plans using Polynomial Chaos Expansion. Physics in Medicine & Biology , 61 (12).