Abstract

This paper aims to facilitate the diagnosis procedure of the ophthalmology specialists for the treatment of retina abnormalities due to Diabetic Retinopathy and Age related Macular Degeneration. Detection of the initial signatures of these two diseases in retina are challenging to the doctors due to the negligible size of the lesions. A Computer Aided Diagnosis system for the automatic detection of lesions from retina fundus image and classification of the lesion types using Machine Learning eases the job of the doctors. In this paper, an approach depending on the concept of Case Based Reasoning is implemented using Support Vector Machine classifier for the classification of the lesions. Detection of lesions uses Otsu multilevel thresholding for segmentation and morphological operations for the elimination of optic disc and blood vessel tree. Evaluation of a set of selected features for each lesion forms the feature vector set for machine learning. The Support Vector Machine classifier achieves an accuracy of 94.03% for bright lesion classification and an accuracy of 97.9% for dark lesion classification, which is promising.