Prediction of Polycystic Ovary and Its Stages Using Deep Neural Networks
Sathish R, Prathvi P Rao, Kavyaharini V, Chandana M R, Nishchita Manjunath Moger
Abstract
Polycystic Ovary Syndrome (PCOS) represents a complex endocrine condition affecting numerous women during their reproductive years, leading to hormonal disruptions, metabolic issues, and fertility challenges that substantially diminish life quality. This disorder stands among the primary causes of female infertility while generating extensive health implications encompassing metabolic, mental health, and heart-related complications. Although widely prevalent, achieving prompt and precise diagnosis continues to pose ongoing difficulties because of diverse symptom presentations and varying diagnostic standards.
Conventional diagnostic methods including hormone testing and ultrasound imaging, though medically dependable, involve invasive procedures, high costs, and require specialist expertise. These constraints limit availability in areas with limited resources and frequently result in varied diagnostic results. This study introduces an advanced deep learning diagnostic system making use of ANNs to identify and categorize PCOS severity levels through organized clinical and lifestyle information.
The framework functions through two separate stages. The initial stage determines PCOS presence or absence, while the subsequent phase categorizes the disorder as either Stage 1 (Polycystic Ovarian Disease – PCOD) or Stage 2 (Polycystic Ovary Syndrome – PCOS). Data preparation methods including median replacement, feature normalization, and anomaly management improve system dependability. The ANN framework utilizes Adam optimization and binary cross-entropy loss calculations, reaching 97% accuracy for PCOS identification and 95% for severity classification.
Additionally, a Streamlit-powered web platform enables immediate implementation, providing accessible, non- invasive, and economical diagnostic capabilities. The framework integrates seamlessly within healthcare systems, especially in emerging regions where imaging-based diagnostics remain scarce. The suggested approach surpasses conventional machine learning methods in accuracy and sensitivity while advancing AI- powered, data-focused women's healthcare solutions that encourage early detection, preventive treatment, and evidence-based medical decisions.
Keywords
Polycystic Ovary Syndrome (PCOS), Deep Learning, Artificial Neural Network (ANN), Streamlit, Clinical Diagnostics, Women's Health, Early Detection, Biomedical Data Analysis
