Assessment and classification of Retinal images using ant colony Optimization based hybrid methods And support vector machines

Abstract

In this work digital retinal images in health and diseases have been newlineanalysed using Optimization based algorithm and hybrid techniques The newlineacquired fundus images N 300 were subjected to various techniques to newlineidentify objects in retinal images such as optic disc macula and blood vessels newlineusing Ant Colony Optimization method For comparison Morphological newlinedilation residue Otsu Matched filter local thresholds and modified newlinewatershed methods were also implemented Further their significant features newlinewere extracted selected and used for classification of normal and abnormal newlineimages using Naive Bayes classifier and Support Vector Machines newlineResults demonstrate the ability of the Ant Colony Optimization newlinemethod to identify optic disc and blood vessels with and without newlinepreprocessing The results provide high visual quality output with better newlineoptic disc and blood vessel identification It provides better delineation newlineextraction of blood vessels and also distinctly differentiates central veins and newlinesmall blood vessels compared to other methods The sensitivity and newlinespecificity obtained for the detection of blood vessels were 87 and 95 newlinerespectively The ratio of vessel to vessel free area using ACO method is newlinedifferent for normal and abnormal images p 0005 and the area under the newlineReceiver Operating Characteristics value is 0 95 The algorithm also detects newlinethe presence of exudates and red lesions in Diabetic Retinopathy images The newlinevalue of sensitivity specificity and Positive Predictive Value newline newline