Follicle Detection in Polycystic Ovary Syndrome Using Ovarian Ultrasound Images

Polycystic Ovary Syndrome (PCOS) is a prevalent hormonal condition, impacting approximately 8–13% of women worldwide, with profound effects on fertility, metabolism, and general health. Detecting follicles accurately using ovarian ultrasound images is significant in diagnosing PCOS. This study presents a conventional image processing method and machine learning (ML) methods with a custom CNN model for automatic identification of follicles in PCOS ultrasound images. The methodology includes image acquisition, pre-processing steps such as enhancement, noise removal, and segmentation using Otsu’s thresholding. The post-processing phase focuses on feature selection to accurately identify follicles. Classification is then performed for follicle detection in two scenarios: with pre-processing applied and without pre-processing. The study achieves an accuracy of 85.29% using a conventional image processing method and 99.14% with a custom CNN-based classification approach that included pre-processing, demonstrating exceptional performance.