Polycystic ovary syndrome detection using optimized SVM and DenseNet

Abstract

Polycystic ovary syndrome (PCOS) is a complicated endocrine disease that significantly impacts the health of women, affecting fertility and leading to various critical conditions. Unfortunately, around 70% of PCOS cases remain undiagnosed, emphasizing the importance of early detection. Ultrasound imaging has emerged as a valuable tool for detecting polycystic ovaries, providing crucial details such as follicle count, size, and position. However, manual diagnosis through ultrasound imaging is laborious and prone to errors, highlighting the need for more objective diagnostic methods. In this study, we propose two distinct predictive models for PCOS detection, utilizing both text and image based datasets. Firstly, an Optimized Support Vector Machine based PCOS detection model is developed using text-based datasets. Secondly, we introduce an image dataset based PCOS detection model using DenseNet. Experimental results demonstrated the suggested models’ effectiveness in accuracy, recall, F-score, and precision for both developed methods. The results showed that the present approaches offer superior performance compared to other methods.