A Comprehensive Review of Machine Learning and Deep Learning Approaches for Early Diagnosis of Polycystic Ovary Syndrome (PCOS)

Polycystic Ovary Syndrome (PCOS) is a common hormonal disorder in women of repro- ductive ages. It is frequently associated with infertility, metabolic abnormalities, and long-term health complications.Due to heterogenous manifestations and the dependence on multiple clinincal factors of this disorder, achieving early and precise diagnosis is difficult. This review provides an overview of machine learning (ML) and deep learning (DL) methods developed for the early detection of PCOS. We examined studies that adopted a range of algorithms including Random Forest, Support Vector Machine, XGBoost, Convolutional Neural Networks, and ensemble methods applied to clinical, biochemical, and ultrasound imaging data. This review describes the important facts about the most frequently utilized datasets, puts an emphasis on key diagnostic markers such as AMH, FSH, LH, BMI, and follicle count, and compares model performance indicators. Particularly, ensemble and stacking approaches showed accuracies above 97%, while explainable AI methods such as SHAP and LIME have improved the trans- parency and clinical interpretability of models. Also, limited diversity of available datasets, inadequate multimodal data fusion, and a lack of extensive validation in real-world clinical environments are some of the long-lasting issues. This paper combines current progress, highlights the existing research gaps, and suggests future pathways for creating robust, interpretable, and clinically applicable PCOS diagnostic tools.