Classification of Diabetic Retinopathy Based on Segmentation of Medical Images

Abstract

Diabetic retinopathy (DR) is a prime reason for escapable blindness in the world. As it progresses, the eyesight of the patient starts worsening, which may lead to blindness if not treated in an early stage. Medical image segmentation and analysis techniques are used for this type of detection of abnormality in retina that correlates and defines the harshness of DR. In this chapter, we discuss the method of DR medical image segmentation to automatically detect and classify the condition of DR. Chapter also discusses the feature extraction of blood vessels, optic disk, microaneurysm, exudates, and macula. The texture of features (gray-level co-occurrence matrix features, histogram intensity features, moment invariants, and gray-level run-length matrix features) finally classifies the DR images into four classes—normal, mild, severe, and proliferative. The method was applied by many researchers on medical images for doing accurate classification and testing. Result obtained with the method is accurate and correctly classified medical images. This chapter is not only a collection of information and facts, but it explains methods/procedure for the classification of diabetic retinopathy based on segmentation of medical images and presented information in the form of result.