Polycystic Ovary Syndrome (PCOS) Detection and Prevalence Report using Computational Intelligence

Polycystic Ovary Syndrome (PCOS) is a common endocrine disorder which impacts 6–12% of women worldwide and can result in infertility and metabolic disorders. Despite how common it is, PCOS is often underdiagnosed simply because it relies on subjective diagnostic measures. This project attempts to navigate this issue by applying DenseNet121, a deep learning model, to automatically detect PCOS from the ultrasound images. DenseNet121 has an efficient architecture that allows for substantial accuracy with often the smallest of datasets. This feature makes it a fantastic application in medical images where data can sometimes become extremely limiting. The model was trained and developed on labeled ultrasound images to classify ultrasound images with and without PCOS lesions as a diagnosis, which were found to be reliable. After detection, the cysts were analyzed manually to extract valuable diagnostic metrics including type, shape, clustering, and severity. These metrics then were combined to construct web-based reports that are user friendly with even more detail to diagnostics and recommendations to action. Combining this technology with clinical expertise, allows for an enhancement in the accuracy of diagnosing PCOS, while providing a practical tool for patients and health care professionals to enhance early detection and management of PCOS.