Automatic Cotton Wool Spots Extraction in Retinal Images Using Texture Segmentation and Gabor Wavelet

Abstract

Hypertensive Retinopathy is a very serious disease that harms the retina of eye and is mainly caused by severe hypertension. A major sign of Hypertensive Retinopathy is Cotton Wool Spots which can lead to visual loss. Hypertensive patient can be saved from ocular complications, stroke and heart attacks by early detection of these spots. An efficient automated method for cotton wool spots detection is presented in this paper. Three major stages are involved in the proposed algorithm namely; pre-processing, segmentation and feature extraction. There are many techniques of pre-processing noisy fundus images that can be applied for noise removal and features enhancement in order to equalize regions having uneven contrast. Optic disc in retinal image has cotton wool spots characteristics like contrast, colour and intensity. The automatic detection of cotton wool spots gets confused during automated evaluation. Because of this reason, in image segmentation stage, optic disc is eliminated earlier by using Texture Segmentation and Gabor Wavelet. Finally in feature extraction, cotton wool spots are detected using Otsu thresholding method. According to the experiments which are conducted on the basis of pixels, the proposed algorithm achieves better results.