Stroke Prognosis through Retinal Image Analysis

Abstract

Many eye diseases as well as systemic diseases usually used to manifest in the retina. The innovations in the field of retinal imaging have paved the way to the development of tools for assisting physicians in stroke prognosis. Stroke is one of the leading causes of adult disability in most of the developing countries. Diagnosis of stroke during the initial stage is crucial for timely prevention and cure. Retinal imaging provides a non invasive technique of predicting the possibility of stroke. This research work focuses on the prediction of retinal ischemia from retinal fundus images and thereby predicting the occurance of stroke. Preprocessing of retinal images is done by retinex processing and morphological operations are done to remove noisy background. Branching points are detected and various features like major axis length, mean diameter, orientation, eccentricity, fractal dimension and tortuosity for the branching blood vessels are computed. This has been compared for various diseases like diabetic retinopathy, hypertensive retinopathy and retinal ischemia against a set of healthy retinal images. Classification has been implemented by Artificial Neural Networks which gives an accuracy of 89 % and the results proved to be promising.