Integrating reproductive and metabolic factors for uterine fibroid risk assessment: a two-center machine learning study with SHAP interpretability.

Background: Uterine fibroids are common benign gynecological tumors that adversely affect reproductive health. This study aimed to develop a machine learning-based model to predict individualized fibroid risk in women of reproductive age. Methods: Six clinical predictors encompassing reproductive and metabolic factors were analyzed. Feature selection was performed using LASSO regression, and multiple machine learning algorithms with cross-validation were compared. The optimal model was assessed for discrimination, calibration, and clinical benefit, with SHapley Additive exPlanations (SHAP) analysis employed to enhance interpretability. Results: A total of 1,274 women were included, of whom 762 (59.8%) had uterine fibroids. Six predictors were retained, and among the classifiers tested, the Random Forest achieved the best validation performance (AUC = 0.734) with balanced accuracy and F1 score. SHAP interpretation further identified age, BMI, menarche age, parity, triglycerides, and fasting glucose as the most influential risk factors. Conclusion: An interpretable Random Forest model was established to predict uterine fibroid risk, enabling individualized risk stratification and supporting timely preventive interventions in clinical practice.