Automated Diagnosis of Age Related Macular degeneration from fundus image

Abstract

Retinal image analysis paves the way for easy diagnosis of retinal pathologies and acts as a first aid tool for ophthalmologist. In this paper a novel approach has been proposed for the automated diagnosis of age related macular degeneration (AMD) from fundus image. A landmark called Drusen, in fundus image whose detection and its location identification play the crucial to detect and grade AMD. In pre-processing step optic disk and blood vessels are detected and removed. By applying log Gabor filter to the pre-processed image energy has been computed. Gray level co-occurrence matrix has been calculated for the image and after applying fuzzy entropy thresholding technique, two discriminative features auto correlation and contrast features have been chosen. Classification is done by using a total of three feature vector using k-nearest neighbour, Support Vector Machine, Random forest classifier. Highest sensitivity is obtained in the case of Random forest classifier. SVM with RBF kernel also does better classification next to random forest.