Automated Detection of Polycystic Ovary Syndrome Using Machine Learning on Ultrasound Images

Polycystic Ovary Syndrome (PCOS) is a prevalent endocrine disorder impacting millions of women worldwide, often remaining undiagnosed due to symptom variability and limitations in traditional diagnostic methods like ultrasound imaging. This study investigates the application of machine learning (ML) models to improve PCOS diagnosis using ultrasound images. A robust dataset, consisting of 23,088 images, was processed using feature extraction (Histogram of Oriented Gradients) and dimensionality reduction (K-Means clustering and Principal Component Analysis) to optimize classification. Four ML algorithms—Support Vector Machine (SVM), Random Forest, K-Nearest Neighbors (KNN), and Gradient Boosting Machine (GBM)—were trained and fine-tuned through cross-validation and hyperparameter optimization, with SVM achieving the highest accuracy of 94.2%. Visual outputs demonstrated the models’ effectiveness in identifying infected regions, underscoring the potential of ML in standardizing PCOS diagnosis. This approach could greatly enhance diagnostic reliability, especially in under-resourced settings, by reducing dependence on operator expertise and enabling scalable diagnostic solutions. Future work will explore deep learning integration and hybrid clinical data approaches to further refine diagnostic accuracy and support personalized healthcare.