Retinal Disease Detection and Classification using Improved LBP Technique

Abstract

According to World Health Organization (WHO) in 2012, Diabetic retinopathy (DR) and age related macular degeneration (AMD) are the two reasons for blindness, particularly in the country. Early stage treatment could reduce the risk of blindness. So, there is a need for a lot of retinal screening by experts and hence there is a demand for experts in the increasing scenario of the number of patients. Due to the lack of experts in the specified domain, the scope for automatic screening using image processing techniques is evolved around the world. In this paper, an automatic retinal screening using Coye filter and local binary pattern techniques is proposed. The main agenda of the paper is to discriminate the fundus among healthy (normal), DR, AMD and combination of both DR & AMD with specific LBP features. This paper mainly focuses on deviations of normal, DR, AMD and combination of both DR & AMD images by minimizing the problems of existing retinal segmentation techniques. Coye filter and mask techniques are utilized to filter the blood vessels from the fundus images and LBP is meant for fundus feature extraction and segmenting the query image based on LBP score values. Existing researches have not given the individual score values.