Diagnosing Cardio Vascular Disease (CVD) using Generative Adversarial Network (GAN) in Retinal Fundus Images

Abstract

In most developing countries cardiovascular diseases (CVDs) is one of the primary causes of mortality. In India, it is estimated that the disease prevalence is very high and one in 4 deaths happens due to CVD. There is no proper specific programmes have been addresses to control CVDs as like existing communicable diseases. The proposed work uses Retinal fundus images in identifying the potential risk factors causes CVD in terms micro structural changes happens in its blood vessels. Hypertensive Retinopathy and Cholesterol-Embolization Syndrome (CES) are the most adverse factors affect the heart function which leads to CVD. These two factors are exactly identified in optic-disc of retinal vasculature. Generative Adversarial Network (GAN) is uses as a deep learning model to synthesize the images with high resolution, which is used in predicting the severity level of the CVD. This work also uses an existing retraining ImageNet model for solving custom image classification tasks. It observes that the experimental results show that the prediction accuracy rate will be high when compared with other existing deep learning approaches. The proposed work diagnoses the prevalence rate of CVD effectively and gives solution to the medical practitioners to adhere proper treatment in avoiding adverse risk.