An automated exudates detection in diabetic retinopathy fundus images using multi kernel spatial fuzzy c means clustering method

Abstract

Microvasculature change associated with tenacious Hyperglycemia are the hostile effects accompanying to diabetes mellitus. Diabetic Retinopathy (DR) is a progressive complication, which leads to retinal permeability, ischemia, neovascularization and macular edema. The pathology is characterized by variation in capillary diameter, size of microaneurysm, hemorrhage exudates. Thus it stimulates the growth of new abnormal blood vessels so as to nourish the eye muscles. But these newly grown blood vessels are subtle, and may get burst. Therefore it leads to leakage of blood, protein based particles named as exudates. Early determination of the DR signs will help the diabetic patient to eradicate austere vision damage. Medical image processing methods helps the ophthalmologists in easy diagnosis, and to estimate the severity of the pathology. Fuzzy based clustering methods are simple and effective methods that will classify the pathos. This work furnish an improved fuzzy clustering method with induced multi kernel and spatial constraint. Statistical evaluation is done to evaluate the performance of the proposed method.