Retinal Disease Screening through Statistical Texture Analysis and Local Binary Patterns using Machine Vision

This paper scrutinizes the most prominent area of machine vision which is medical imaging application to detect high risk diseases for allowing screening method for referral to eye care practitioner and the capabilities of Local Binary Patterns and Statistical Texture Analysis in the analysis of fundus images to distinguish healthy and diseased retina. The goal is to discern between diabetic retinopathy (DR), glaucoma (G), and healthy fundus images analyzing the texture of the retina background using NI Vision. This paper also studied the efficacy and efficiency of the algorithm by measuring the sensitivity and specificity of the diagnosis developed by the researchers and validated by the Philippine Eye Research Institute, University of the Philippines-Manila. Finally, the images were classified as Healthy, DR or G. The results show that the sensitivity, specificity, and precision of Local Binary Patterns and Statistical Texture Analysis are 70.00%, 93.33%, 91.30% respectively. By adding a color matching classifier it will make the screening process more accurate resulting to the following values, 83.33%, 93.33%, and 83.33%.