Machine Learning Predicts Oocyte Yield And OHSS Risk

This prognostic study (internal n=6,401; external n=3,805) develops machine-learning models to predict number of oocytes retrieved (NOR) and early-onset moderate-to-severe OHSS across varying FSH starting doses. Gradient boosting regressor achieved R2≈0.798 for NOR, and LightGBM achieved AUROC≈0.759 internal (0.729 external) for OHSS; SHAP highlighted FSH-dose/BMI ratio and baseline AFC as key predictors. Models were deployed in a web prototype, InOvaSGuide, for individualized dose-response guidance, pending prospective validation.