Bridging hormones and imaging: artificial intelligence advances multimodal quantitative and qualitative assessment of ovarian reserve

To review how artificial intelligence (AI) integrates multimodal data—hormonal, imaging, clinical, and genetic—to improve quantitative and qualitative assessment of ovarian reserve, overcoming limitations of traditional single‑marker approaches. A structured literature search of PubMed and Web of Science (2013–2025) was conducted to narratively review recent AI applications in ovarian reserve evaluation From approximately 400 records, 68 articles were included, focusing on machine/deep learning models for biochemical and imaging data analysis. AI enhances ovarian reserve assessment by automating follicle detection on ultrasound, fusing AMH/FSH/E2 with imaging features, and extracting clinical data via natural language processing, with reported sensitivity of 90% and specificity of 80% in representative studies. Multimodal integration facilitates personalized “endocrine age” modeling and may improve ART outcome prediction. Challenges include data standardization, model explainability, and multicenter validation. AI-driven multimodal assessment holds considerable potential to advance ovarian reserve evaluation from isolated biomarkers toward personalized, precise reproductive care, though large-scale multicenter validation, development of clinician-oriented decision-support tools, and clinical integration remain essential.