Development and clinical application of a probabilistic robustness evaluation tool for pencil beam scanning proton therapy treatments.

Purpose: to implement a probabilistic-Robustness-Evaluation (pRE) tool for proton therapy treatments and to correlate these results with the worst-case approach (wRE) implemented in commercial TPS for clinical applications.

Materials And Methods: 12 skull base patients were planned with a robust multiple field optimization (MFO) approach. 10 years of machine QA were analysed to derive the uncertainties of our treatment system (beam delivery and patient positioning system). For a large cohort of patients, post-treatment imaging was acquired to determine the intra-fraction uncertainty. The pRE, considered explicitly all these uncertainties, the fractionation and range uncertainty. For each plan a wRE with different combinations of range and setup uncertainties was simulated. wRE results were then compared, in terms of target coverage and OAR dose limits, with pRE results.

Results: 43,400 dose distributions were analysed. pRE simulations lasted 18.6 h (±11.5 h). The results showed that the combination of wRE uncertainty parameters that surrogated the best pRE results with a confidence level of 95 % were (1.0 mm/3.5 %). The median OAR's dose indexes difference (D/D) between pRE and wRE was 1.90 (±1.49) GyRBE, while for target D and D it was -0.66(±0.95) and -0.67 (±0.52) GyRBE, respectively.

Conclusion: A tool able to explicitly simulate the source of treatment uncertainties and the effect of the fractionation was implemented to have a more realistic evaluation of plan robustness. This tool was used to find the best wRE parameters that surrogate the pRE results while maintaining clinically acceptable timing. These results are now used in our clinical workflow.