ESTRO 2023 - Abstract Book

S1238

Digital Posters

ESTRO 2023

Results Median of the DICE coefficient were 0.94 (0.31-0.99) for bladder, 0.97 (0.78-0.99) for rectum, 0.97 (0.89-0.99) for prostate, 0.89 (0.6-0.97) for seminal vesicle, 0.96 (0.89-0.98) for CTV, and 0.96 (0.91-0.97) for PTV, respectively. D max of bladder wall were met optimal in 98% and failed in 2%. D 1cc and D 53% of bladder wall were met optimal in all sessions. D max of rectum wall were met optimal in 98% and failed in 2%. D 1cc and D 53% of rectum wall were met optimal in all sessions. D 95% of prescription PTV were met optimal in 84% and met mandatory in 16%. D 98% of prescription PTV were met optimal in 98% and met mandatory in 2%. D 1cc of prescription PTV were met optimal in all sessions. The case of failed dose constrains in D max of bladder wall was exceeded by 0.012 Gy and the case of failed dose constrains in D max of rectum wall was exceeded by 0.012 Gy. Conclusion In the setting of CTV encompassed prostate and proximal 1cm of the seminal vesicle with hydrogel spacer and sigmoid or intestine was not involved organ at risk, the application of treatment plan using contours propagated by DIR was deemed acceptable. H. Hasannejadasl 1 , H. van de Poel 2 , B. Vanneste 3 , J. van Roermund 4 , K. Aben 5,6 , Z. Zheng 1 , B. Osong 1 , L. Kiemeney 5 , I. Van Oort 7 , R. Verwey 8 , L. Hochstenbach 8 , E. Bloemen 9 , A. Dekker 1 , R. Fijten 1 1 Department of Radiation Oncology (Maastro), Maastricht University Medical Centre+, Maastricht, The Netherlands; 2 Amsterdam University Medical Centers, Department of Urology, Amsterdam, The Netherlands; 3 Department of Radiation Oncology (Maastro), Maastricht University Medical Centre+, Maastricht, The Netherlands; 4 Maastricht University Medical Center+, Department of Urology , Maastricht, The Netherlands; 5 Netherlands Comprehensive Cancer Organization, Department of Research & Development, Utrecht, The Netherlands; 6 Radboud university medical centre, Radboud Institute for Health Sciences, Nijmegen, The Netherlands; 7 Department of Urology, Radboud university Medical Center, Nijmegen, The Netherlands; 8 Zuyd University , Departmenet of Applied Sciences, Heerlen, The Netherlands; 9 Zuyd University , Department of Applied Sciences, Heerlen, The Netherlands Purpose or Objective Urinary incontinence (UI) is one of the most common side effects of prostate cancer treatment, but it is currently difficult to predict in clinical practice without artificial intelligence models. Finding a balance between explainability and predictability of a clinical predictive model is a prerequisite for its adoption, but some black box models are considered to perform better in terms of predictability despite not being as explainable. To determine which algorithm has the highest accuracy and is also easily explainable, we used three machine learning (ML) algorithms: logistic regression (LR), random forests (RF), and support vector machines (SVM). To identify the best algorithm to predict UI following localized prostate cancer treatment, we compared the performance of the generated models. Materials and Methods For our analyses, we used the ProZIB dataset for this study, which included demographics, clinical data, and patient reported outcomes (PROMs) from 69 Dutch hospitals collected by the Netherlands Comprehensive Cancer Organization. This dataset contained information of 964 men with localized prostate cancer for the purpose of training and external validation. In order to perform an external validation in accordance with the TRIPOD Type 3 guidelines, data were split by location so that one hospital's data could be used either for training or validation. Six models were generated for 2 time points; 3 models for UI 1 year after treatment and 3 for UI 2 years after treatment. Results Analyses were conducted on 847 and 670 localized prostate cancers for 1- and 2-year models, respectively. The performance of LR in external validation was superior to other models with an accuracy of 0.76, a sensitivity of 0.82, and an AUC of 0.79 for the 1-year outcome. Training and validation sets of all 2-year models, however, showed markedly different performances. The 2-year models’ accuracy varies from 0.60 (for LR and SVM) to 0.65 (for RF), and both sensitivity and specificity were considerably different for all models. Figure 1 shows the performance results of generated models. The importance of features in each ML model for predicting UI is shown in Figure 2. The importance of features varied among different ML models where 4 variables were selected by all models for 1-year and two for the 2-year outcome. Substantial overlap was observed between variables selected by RF and SVM algorithms. PO-1523 Comparison of machine learning methods to predict urinary incontinence in localized prostate cancer

Figure 1. Performance results of 1-year and 2-year models