EPD-ECNN: Enhanced PCOS Detection with Preprocessed Ultrasound Images and Ensemble CNN Model

Polycystic Ovary Syndrome (PCOS) is a prevalent hormonal disorder in women of reproductive age, characterized by irregular menstrual cycles and elevated androgen levels, significantly impacting fertility. PCOS is also associated with conditions such as acne, hair loss, excessive hair growth, elevated androgens, and infrequent ovulation. Traditionally, PCOS diagnosis is performed manually by gynecologists through the labor-intensive and subjective process of counting and measuring ovarian follicles. Recently, Machine Learning (ML) approaches for PCOS detection have gained traction due to their potential for more accurate and early diagnosis, minimizing health risks for women. While various ML models for PCOS detection exist, they often overlook the importance of preprocessing, which is essential for developing robust and reliable models. This study introduces the EPD-ECNN model for detecting PCOS from ultrasound images. First, image quality is enhanced using the CLAHE technique. Then, images are segmented using the watershed algorithm. The model combines pre-trained CNNs, specifically InceptionV3 and DenseNet121, for feature extraction. Finally, a fully connected layer is used for classification. The EPD-ECNN model achieved an accuracy of 98.59%, demonstrating its efficiency in medical cases of PCOS detection.