Identifying key determinants of cumulative live birth in women with ovarian endometrioma undergoing ethanol sclerotherapy followed by in vitro fertilization or intracytoplasmic sperm injection: an interpretable machine learning analysis

Background Ovarian endometriomas impair ovarian reserve and fertility in women of reproductive age. Ethanol sclerotherapy is a fertility-preserving alternative to surgery. Nonetheless, predicting cumulative live birth rates after in vitro fertilization remains challenging. This study aimed to develop and validate a machine learning model for predicting the cumulative live birth rate in women with endometriomas who underwent alcohol sclerotherapy followed by assisted reproduction. Methods This retrospective cohort study included 194 patients with ovarian endometriomas who underwent ultrasound-guided ethanol sclerotherapy before in vitro fertilization or intracytoplasmic sperm injection cycles between January 2020 and December 2024 at our institution. Patients were allocated to the training (135 patients, 70%) and validation (59 patients, 30%) groups. Feature selection used univariate logistic regression (p 0.10) to identify 19 predictors, which were refined using the Boruta, Recursive Feature Elimination, and maximum relevance minimum redundancy algorithms. Features identified by all methods were selected as the final predictors. Four machine learning algorithms (Decision Tree, Random Forest, Extreme Gradient Boosting, Support Vector Machine) were compared using discrimination, calibration, and utility metrics. SHapley Additive exPlanations analysis was used to interpret the model. Results The cumulative live birth rate was 50.0% (97/194). Five predictors were identified: antral follicle count, progesterone level on gonadotropin starting day, downregulation, cyst diameter, and previous live birth history. The Extreme Gradient Boosting model showed optimal performance, with an AUC of 0.830 (95% confidence interval: 0.719–0.941), sensitivity of 0.783, specificity of 0.750, and Brier score of 0.176. SHapley analysis revealed​ that a higher antral follicle count and downregulation positively impacted birth prediction, whereas elevated progesterone levels and larger cyst diameters had negative effects. Conclusion We developed an explainable Extreme Gradient Boosting model for predicting cumulative live birth rates in women with ovarian endometriomas after ethanol sclerotherapy and assisted reproductive technology. SHapley Additive exPlanations analysis identified key predictors and revealed their non-linear contributions to outcomes, providing transparent explanations for predictions. This interpretable machine learning approach offers a clinical decision-support tool for patient counseling and treatment optimization, advancing beyond traditional methods in capturing reproductive outcomes.