An Ensemble Approach For Ultrasound-Based Polycystic Ovary Syndrome (PCOS) Classification

Abstract

Polycystic Ovary Syndrome (PCOS) is a prevalent endocrine disorder characterized by hormonal imbalances, ovulatory dysfunction, and metabolic disturbances in women of reproductive age. A novel approach for the early detection of PCOS using an innovative Ensemble Learning technique is proposed. Extreme Gradient Boosting (XGBoost), K-Nearest Neighbors (KNN), and Support Vector Machine (SVM) classifiers are combined, leveraging their complementary strengths to improve classification accuracy. A comprehensive image preprocessing pipeline, including geometric transformation, contrast enhancement, and noise reduction, is introduced to optimize feature extraction. To address class imbalance, an effective bias mitigation strategy using class weighting is implemented. The PCOSGen Dataset, comprising 3200 healthy and 1468 unhealthy ultrasound images, was used for the training and evaluation of the model. A remarkable test accuracy of 92% was achieved by the proposed ensemble, outperforming individual classifiers. Notably, feature extraction was incorporated to reduce input data dimensionality, enhancing both model interpretability and computational efficiency. This approach is made particularly suitable for real-world clinical applications, especially in resource-constrained environments. Robust performance of the model, demonstrated through comprehensive metrics including precision, recall, and F1-score, offers a promising tool for improving PCOS diagnosis using ultrasound image analysis.