S147
Brachytherapy - Physics
ESTRO 2026
symmetric surface distance (ASSD), and percent volume difference between predicted and clinical V100% volumes. 95% confidence thresholds were derived from the validation data and applied to the independent test set for QA verification. Results: The model demonstrated excellent spatial agreement between predicted and ground-truth V100% regions, achieving a mean ± SD Dice = 0.95 ± 0.01, HD99 = 0.88 ± 0.39 mm, and ASSD = 0.53 ± 0.17 mm across the test cohort. Predicted versus clinical V100% values differed by 0.70 ± 1.0% on average. Performance remained consistent across a wide range of prostate volumes, underscoring robustness to anatomical variability. The model captured expected dose patterns consistent with planning goals and dwell placement. Applying the predefined QA thresholds identified 4.5% of test-set plans as potential outliers, warranting further clinical review.
Conclusion: The proposed AI-based nomogram accurately predicts achievable dose coverage (V100%) from pre-plan anatomy and dwell geometry in HDR prostate brachytherapy. The framework provides a quantitative, interpretable benchmark for plan QA and inter-physician standardization. Future work will extend the model to other treatment sites and incorporate uncertainty estimation to facilitate clinical deployment. Keywords: HDR Prostate, Nomogram, AI The potential of using dose-response models in multi-objective treatment plan optimization Anton Bouter 1 , Thijs J Penning 1 , Danique LJ Barten 2 , Nicole Eder-Nesvacil 3 , Laura A Velema 4 , Bradley R Pieters 2 , Christian Kirisits 5 , Peter AN Bosman 1 , Tanja Alderliesten 4 1 Evolutionary Intelligence Group, Centrum Wiskunde & Informatica, Amsterdam, Netherlands. 2 Department of Radiation Oncology, Amsterdam University Medical Center, University of Amsterdam, Amsterdam, Netherlands. 3 Department of Radiation Oncology, Comprehensive Cancer Center, Medical University of Vienna, Vienna, Austria. 4 Department of Radiation Oncology, Leiden University Medical Center, Leiden, Netherlands. 5 Department of Radiation Oncology, Comprehensive Cancer Center, Medical University of Vienna, Vienna, Netherlands Purpose/Objective: Current clinical decision making in brachytherapy Proffered Paper 4732
Made with FlippingBook - Share PDF online