Machine learning based model to predict the recurrence after fertility sparing surgery in patients with borderline ovarian tumor

Borderline ovarian tumors (BOTs) are epithelial ovarian tumors with a favorable prognosis, particularly in younger patients. Fertility-sparing surgery (FSS) is often preferred for patients desiring to preserve fertility. This study aimed to develop a machine learning (ML)-based prediction model to identify patients at high risk of recurrence after FSS. A multi-center retrospective study was conducted, including 589 BOT patients who underwent FSS between 2010 and 2023. Clinical, pathological, and laboratory data were collected, and ten ML algorithms were evaluated for their predictive performance. The dataset was preprocessed, and feature selection was conducted using univariate analyses and feature importance scores derived from preliminary ML models. The dataset was randomly split into training and test sets, and hyperparameter tuning was performed using grid search and cross-validation. Model performance was assessed using metrics such as AUC, accuracy, sensitivity, and specificity. The Gradient Boosting model demonstrated the highest accuracy, with an AUC of 0.94, sensitivity of 91%, and specificity of 88%. Key predictors of recurrence included histological subtype, age at diagnosis, tumor size, and surgical approach. The model’s interpretability was enhanced using SHAP values, which highlighted the importance of these features. Additionally, a web-based tool was developed to facilitate clinical application of the model, allowing physicians to input patient data and obtain recurrence risk predictions. This study highlights the potential of ML-based models in improving personalized post-surgical management and optimizing oncological and reproductive outcomes for BOT patients undergoing FSS.