ESTRO 2025 - Abstract Book

S356

Brachytherapy - Physics

ESTRO 2025

brachytherapy. However, in today’s clinical settings, robustness against these uncertainties is not assessed due to unavailable tools and the computational burden. Thus, the goal of this project is to develop a robustness assessment framework for HDR brachytherapy. Material/Methods: An existing in-house developed quality assurance related framework featuring the TG43 dose calculation algorithm was extended for robustness calculation. For a robustness assessment, the framework prepares first a user defined number of uncertainty scenarios. Each uncertainty scenario encompasses several systematic and random uncertainties on the nominal plan. The value of each uncertainty is randomly sampled from user-defined Gaussian distributions. The uncertainties include positioning accuracy, rigid catheter movements, reconstruction uncertainties, dwell time and dose calculation uncertainties. Afterwards, a TG43 dose calculation of each uncertainty scenario is performed. Finally, the robustness is evaluated using statistics by means of box plots of a specific plan quality parameter such as the implant dose non-uniformity index (DNR) = V100%/V150% and dose-volume histogram (DVH) bands. The robustness assessment tool was tested on treatment plans for two breast cases including 14 and 11 catheters, respectively. The prescribed dose was 8 x 4 Gy = 32 Gy for both cases. Robustness assessment included the calculation of 100 uncertainty scenarios per treatment plan with respect to clinical acceptance criteria defined for the nominal scenario. Results: The framework successfully assessed the robustness of the two plans within 50 min per plan using 12 CPU cores. The uncertainties in the dwell time and dose calculation have major impact on the robustness. Especially dose volume quantities of the target are most sensitive on the tested uncertainties followed by the implant DNR. On the other side, the evaluated dose-volume parameters of the organs at risk are substantially more robust. Conclusion: A promising robustness assessment framework is developed which was used to assess the robustness of two breast brachytherapy treatment plans. Further investigations of other treatment sites and improvements in computational performance are suggested in future work. Poster Discussion AI-based treatment planning for breast HDR brachytherapy using BRIGHT Anton Bouter 1 , Daniela Toader 1 , Leah R. Dickhoff 2 , Dorin Todor 3 , Peter A. Bosman 1 , Tanja Alderliesten 2 1 Evolutionary Intelligence Research Group, Centrum Wiskunde & Informatica, Amsterdam, Netherlands. 2 Radiation Oncology, Leiden University Medical Center, Leiden, Netherlands. 3 Radiation Oncology, Virginia Commonwealth University, Richmond, VA, USA Purpose/Objective: ‘BRachytherapy via artificially Intelligent GOMEA-Heuristic based Treatment planning’ (BRIGHT) is a multi-objective treatment planning method based on AI [1], which has been in clinical use for prostate HDR brachytherapy since 2020. BRIGHT's key strengths lie in performing direct multi-objective optimization of dosimetric indices and other arbitrary non-differentiable aims, resulting in a set of high-quality patient-specific treatment plans that each have a different trade-off between objectives of interest (typically: target coverage and organ sparing), providing insight-at a-glance regarding achievability of aims. Here, we present the first results of the retrospective application of BRIGHT to breast HDR brachytherapy, with optimization objectives tailored towards this tumor site. Material/Methods: Keywords: robustness, uncertainties, dose calculation 3940