Predicting post-treatment sexual function outcomes in prostate cancer patients using machine learning–based models

Prostate cancer is one of the most common malignancies among men, and while contemporary advances in early diagnosis and treatment strategies have prolonged survival, post-treatment functional outcomes—particularly erectile dysfunction—remain a key clinical concern. Although the applied therapeutic modalities primarily aim to achieve oncological control, they also play a decisive role in determining erectile function outcomes. This study aimed to develop a machine learning–based model to predict post-treatment erectile function outcomes in men treated for prostate cancer. Data from 48 patients treated at the Urology Department of Süleyman Demirel University between 2023 and 2025 were retrospectively analyzed. Clinical and biochemical variables including age, body mass index, PSA, total testosterone level, Gleason score, D'Amico risk classification, comorbidities, and treatment type were incorporated into the model. Using the PyCaret framework, several algorithms such as logistic regression, support vector machines, XGBoost, and ExtraTreesClassifier were compared. The ExtraTreesClassifier algorithm achieved the best predictive performance (AUC=0.96, accuracy=0.91). The model's predictive power was mainly driven by age, testosterone levels, and treatment modality. Findings indicate that machine learning methods can accurately predict sexual function outcomes after prostate cancer treatment and may provide a valuable foundation for clinical decision support systems. Although this study is limited by its retrospective and single-center design, further prospective multicenter validation may enhance the model's generalizability and clinical utility. In conclusion, the proposed model has the potential to support patient-centered counseling by offering individualized risk estimates, aligning with the modern urological paradigm that emphasizes both oncologic control and functional well-being.