AI-Driven Prediction of Oocyte Retrieval in Older Women with Low Ovarian Reserve, a model based on systematic literature review

Abstract

Objective: To introduce a potential paradigm shift in ovarian stimulation for women aged 40 and above, focusing on the rationale for mild stimulation doses, supported by a novel predictive AI model.Methods: The "EFRE Predict AI model," also termed the "AI Accelerated Knowledge Synthesis Paradigm," was developed using an AI-driven methodology. This involved an autonomous systematic literature review (SLR) guided by the PICO framework to synthesize predictive insights from high-quality scientific literature, including international guidelines and primary research. The model is designed to calculate the number of obtainable oocytes based on patient-specific factors: Age, Anti-Müllerian Hormone (AMH), Antral Follicle Count (AFC), Body Mass Index (BMI), previous stimulation history, and gonadotropin doses used (75, 150, 225, and 300 IU). Initial validation included a survey of over 100 experts and specialists.Results: The development process yielded the EFRE Predict AI model capable of estimating oocyte yield based on key clinical parameters. Initial validation through expert surveys indicated very satisfactory acceptance among specialists.Conclusion: The EFRE Predict AI model offers a novel tool to personalize ovarian stimulation protocols for women over 40. By providing data-driven predictions of oocyte yield at varying gonadotropin doses, including mild doses, this AI-accelerated knowledge synthesis approach supports a potential paradigm shift towards more individualized and potentially milder stimulation strategies in this patient population.