Integrating Ovarian Reserve Biomarkers with Machine Learning for Gonadotoxicity Risk Prediction in Young Female Cancer Patients: A Scoping Review

Abstract

Cancer treatment-related gonadotoxicity poses substantial threats to reproductive health in young female patients. Current risk stratification frameworks, including ASCO clinical practice guidelines and the Oncofertility Pediatric Initiative Network (O-PIN) classification, rely predominantly on population-level statistics and expert consensus, offering limited capacity for individualized prediction. This scoping review examines the emerging intersection of ovarian reserve biomarkers and machine learning (ML) methodologies for gonadotoxicity risk prediction in young female cancer patients. Through systematic analysis of published literature, publicly available datasets, and clinical decision support tools, this review identifies key biomarkers—including anti-Müllerian hormone (AMH), antral follicle count (AFC), and follicle-stimulating hormone (FSH)—that serve as critical input features for predictive models. ML approaches, particularly random forest classifiers, have demonstrated promising discriminative performance (area under the receiver operating characteristic curve up to 0.87) in predicting primary ovarian insufficiency following chemotherapy. This review further evaluates how artificial intelligence may facilitate personalized fertility preservation pathway selection and clinical workflow integration. Challenges including limited sample sizes, absence of external validation, data heterogeneity, and algorithmic fairness concerns are critically discussed. Recommendations for multicenter prospective registries and standardized evaluation frameworks are proposed to advance clinical translation of AI-assisted oncofertility decision-making.