ESTRO 2025 - Abstract Book

S2121

Clinical - Urology

ESTRO 2025

Initial feature extraction resulted in 333 features from which only 29 were finally selected. Those included prostate CTV volume, three DVH metrics — bladder Dmax (Gy), urethra Dmax (Gy), and urethra Dmean (Gy) — and age. The final model was developed using the k-Nearest Neighbors (kNN) algorithm [2]. A 5-fold cross-validation technique led to an AUC of 0.71 (SD 0.07) and an accuracy of 0.67 (SD 0.07).

Conclusion: A promising ML model based on dosiomics was developed to predict late GU toxicities in prostate cancer patients treated with SBRT. Optimizing and applying this model could facilitate targeted interventions for high-risk patients and assist in treatment planning to reduce adverse events.

Keywords: Prostate cancer, toxicity, dosiomics

References: [1] van Griethuysen JJM, Fedorov A, Parmar C, Hosny A, Aucoin N, Narayan V, Beets-Tan RGH, Fillion-Robin JC, Pieper S, Aerts HJWL. Computational Radiomics System to Decode the Radiographic Phenotype. Cancer Res. 2017 Nov 1;77(21):e104-e107. doi: 10.1158/0008-5472.CAN-17-0339. PMID: 29092951; PMCID: PMC5672828. [2] Fabian Pedregosa, Gaël Varoquaux, Alexandre Gramfort, Vincent Michel, Bertrand Thirion, et al. Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research, 2011. 〈 hal-00650905v2 〉

4309

Poster Discussion Pattern of care and survival outcomes in older men with localised prostate cancer in an Australian population-based cohort Therese Kang 1,2 , Jeremy Millar 1 , Wee Loon Ong 1,2 1 Department of Radiation Oncology, Alfred Health, Melbourne, Australia. 2 Department of Radiation Oncology, Barwon Health, Geelong, Australia Purpose/Objective: Active treatment for prostate cancer in older men needs to be balanced against their comorbidities and life expectancy. In this study, we evaluate the patterns of care and outcomes in older men with prostate cancer in an Australian population-based cohort.