Retina images classification using histogram of equivalent pattern (HEP) texture descriptors

Abstract

Diabetic retinopathy (DR) is a complication of diabetes and is one of the commonest causes of visual loss worldwide. It occurs as a result of long term accumulated damage to the small blood vessels in the retina. Diabetic Retinopathy is asymptomatic in its early stage when it is most easily amenable to treatment. Thus diabetic patients should ideally have their eyes checked or screen at least annually. For the screening to be available widely, computer assisted detection and evaluation of DR must be developed. In this study, we propose a system for automated classification of normal, and abnormal retinal images using texture analysis method. We performed extensive experiments using 38 texture descriptors belonging to Histogram of Equivalent Patterns (HEP) together with 1kNN classifier. A 2-fold cross-validation process is applied to the DIARETTDB0 database to evaluate the performance of the proposed framework. It is shown that the highest accuracy of 84.64% is achieved when using Gradient-based local binary patterns methods.Diabetic retinopathy (DR) is a complication of diabetes and is one of the commonest causes of visual loss worldwide. It occurs as a result of long term accumulated damage to the small blood vessels in the retina. Diabetic Retinopathy is asymptomatic in its early stage when it is most easily amenable to treatment. Thus diabetic patients should ideally have their eyes checked or screen at least annually. For the screening to be available widely, computer assisted detection and evaluation of DR must be developed. In this study, we propose a system for automated classification of normal, and abnormal retinal images using texture analysis method. We performed extensive experiments using 38 texture descriptors belonging to Histogram of Equivalent Patterns (HEP) together with 1kNN classifier. A 2-fold cross-validation process is applied to the DIARETTDB0 database to evaluate the performance of the proposed framework. It is shown that the highest accuracy of 84.64% is achieved when using Gradient-based local bina...