Prediction of Cardio Vascular Disease from Retinal Fundus Images Using Machine Learning

Abstract

The publicly available DIARETDB1 datasets contain retinal images that can be used to diagnose and forecast the risk of cardiovascular disease (CVD). Exudates, microaneurysms, and blood vessel segmentation in the retinal fundus images can all be signs of cardiovascular disease. The K-nearest neighbor (KNN) approach is demonstrated for CVD prediction. Anomalies in fundus images are used to train the proposed model using image processing technologies. The model was trained and tested using about 89 pictures from the publicly available DIARTDB1 dataset. This paper describes the feature extraction in blood vessel and aneurysm segmentation, and then, supervised KNN classification model is used for the prediction of the occurrence of heart disease from retinal fundus images with higher rate of accuracy. The CVD prediction accuracy of the trained model is 96.6%.