Expeditious Prognosis of PCOS with Ultrasonography Images - A Convolutional Neural Network Approach

Polycystic Ovary Syndrome (PCOS) is a complex condition that affects women during their reproductive years. It is caused by a combination of genetic and environmental factors, and it leads to a hormonal imbalance and the formation of cysts on the ovaries. Hyperandrogenism, a clinical feature of PCOS, can lead to inhibition of follicle development, ovarian microcysts, anovulation, and menstrual changes. Symptoms of PCOS include Weight gain, Fatigue, Depression, Acne, Hyperthyroidism, Hypothyroidism infertility etc., PCOS affects 5% to 10% of women age 18 to 44 and early diagnosis and treatment is crucial for managing the condition and reducing the risk of related health problems. It’s important for women to be aware of the symptoms of PCOS and to seek medical advice if they suspect they may have the condition. With proper treatment and management, women with PCOS can lead healthy and fulfilling lives. Machine learning and deep learning algorithms have the potential to revolutionize medical diagnosis. In this paper, convolutional neural networks - ResNets, VGGNet and Inception V3 have been implemented to diagnose PCOS from ultrasound ovary images. VGG 19 produced the highest accuracy of 96% than other models. Generative Adversarial Network (GAN) approach is used to address the overfitting issue and to increase the accuracy of the model by generating new images.