PATHOLOGY DETECTION OF PCOS AND ITS SEVERITY GRADING USING MACHINE LEARNING ALGORITHM

Computer Engineering and research and development

Polycystic ovarian syndrome (PCOS) stands as a significant threat among gynaecological conditions due to its tendency to manifest with subtle symptoms, often leading to diagnosis at advanced stages. Distinguishing between various types of ovarian cysts poses a formidable challenge for medical professionals. Among imaging techniques, ultrasound (US) imaging emerges as the preferred choice owing to its convenience, non-invasiveness, and real-time capabilities. However, the current screening methods for ovarian cysts via imaging still suffer from limitations, contributing to the poor prognosis associated with ovarian cysts. In recent years, the integration of deep learning models with US images has shown promising results in enhancing diagnostic efficiency, reducing mortality rates, and minimizing diagnostic delays. This project introduces an innovative approach for diagnosing ovarian cysts using US images, employing a Deep Convolutional Neural Network (DCNN) model enhanced with a Generative Adversarial Network (GAN) to address overfitting issues by augmenting training samples. Through this augmentation process, a more robust dataset is created, enabling the DCNN model to effectively classify different types of ovarian cysts. The proposed system not only aids in accurate diagnosis but also serves as a valuable tool for physicians in medical decision-making. By analysing the outcomes generated by the fused DCNN model, medical professionals can gain valuable insights into the nature of ovarian cysts and make informed treatment decisions. The results obtained from this study demonstrate the superior precision and accuracy of the proposed model in diagnosing ovarian cancer and other types of cysts. This advancement marks a significant stride towards improving the management and prognosis of ovarian cysts, potentially saving lives through early detection and intervention.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.