AI Based PCOS Prediction with Feature Explainability for Clinical Decision Support

Abstract

Introduction: Polycystic Ovary Syndrome (PCOS) is a common endocrine disorder affecting women of reproductive age, often leading to metabolic, reproductive, and psychological complications. Early detection is critical to prevent long-term consequences, yet many cases remain undiagnosed.

Methods: This study developed a web-based AI and Machine Learning (ML) system for preliminary PCOS risk prediction using real-world survey data. Fifty participants provided physiological and lifestyle-related information, including age, BMI, menstrual cycle regularity, body hair growth, acne, diet, and exercise habits. Data were pre-processed, scaled, and used to train supervised ML models for risk classification.

Results: The system successfully classified participants into high- and low-risk categories. Among respondents, 29.5% reported irregular menstrual cycles, with co-occurring symptoms such as weight gain and acne, supporting the model’s multi-feature approach. The predictive model demonstrated consistent alignment with clinical guidelines, highlighting its potential as an accessible early-screening tool.

Conclusion: A simple web-based AI system can facilitate early identification of PCOS risk, enabling timely lifestyle intervention and medical consultation. Although the study was limited by a small dataset and self-reported data, the approach demonstrates feasibility for scalable preventive healthcare solutions. Future work should expand datasets, integrate biochemical markers, and validate predictions clinically.