Prediction of atherosclerosis pathology in retinal fundal images with machine learning approaches.

Atherosclerosis is a common cause of cardiac attack and its early detection prevents further complications. In this paper, a research concept is proposed focusing on a novel method of classification system. This method is carried out with image features derived from fundus photographs. It depends upon the arteries and vein classification process and also by the morphological appearance. Further, the proposed mixed algorithm, by using the retina fundal images, this method achieves an accuracy of detecting Atherosclerosis. In spite of the method being somewhat a hard one, of late, several methods are developed which employ advanced retinal photographic imaging techniques. These techniques involve characterizing, measuring and quantifying any variations and dissimilarities in the retinal structure. The hallmark of these methods, which have both qualitative and quantitative prediction, illustrated the allied symptoms found on cardiovascular diseases. This paper deals with providing the accurate input for the atherosclerosis detection by way of image preprocessing method. The study focuses on the reducing the disease independent variations without damaging any information related to the differences between the images of healthy and atherosclerotic eyes. The propose method enables correction of illumination in the blood vessels, by repainting them. Further, there is a normalization of the focus region for the feature extraction and classification process. Finally, Enhanced Bayesian Arithmetic Classifier (EBAC) is implemented for effective classification of the blood vessels. MATLAB software of 2014b version is employed for deriving the simulation results. [ABSTRACT FROM AUTHOR]