Diagnosis and early detection of diabetic retinopathy using spectral classifier with predictive rules

Abstract

A Diabetic Retinopathy DR is an eye disease that infects the blood vessels of the retina and finally leads to blindness if not identified and treated in time The early detection and diagnosis of DR disease is imperative for preserving patients vision Generally retinal fundus images are employed to diagnose different kinds of eye diseases Automatic disease diagnosis has attained greater significance than ever before with the rapid growth of computing technology and enhancement in the service through medical field Besides automatic detection of fundus images and the recognition of blood vessels are imperative The blood vessels provide information about length width tortuosity and branching pattern of retinal images and also help in ranking the severity of diseases However manual identification of blood vessels is very difficult due to fundus images having lower contrast As a result consistent and automatic methods are necessary to extract the blood vessel in fundus images for DR disease diagnosis Although Feature based Macular Edema Detection FMED failed to create a competitive diabetic retinopathy feature selection system to transparently analyze the disease state yet the existing FMED method identifies the occurrence of exudation in fundus images with the help of an extensive segmentations process through experts Further Fourier domain Optical Coherence Tomography OCT is developed to identify subtle retinal injury in gently affected individuals However OCT method is meant only for the left eye newline