Detection of hard exudates in colored retinal fundus images using the Support Vector Machine classifier

Diabetic Retinopathy (DR) is a disorder which alters the blood vessels in the retina, and is a common cause of blindness and vision defects in the world today. Hard exudates are associated with diabetic retinopathy, and have been found to be one of the most prevalent clinical signs of retinopathy. One of the research areas which is currently drawing intense interest of scientists and physicians is medical image analysis to aid in clinical diagnosis. Special efforts have been made to develop algorithms related to retinal image analysis, providing fundus images enhancement and detecting visual signs of DR. In this paper, we propose a system to analyze the fundus images for the detection of hard exudates using image enhancement, based on Lifting Wavelet Transform (LWT) and an image classifier based on Support Vector Machine (SVM). We prospectively assessed the algorithm performance using a database of 60 retinal images of diabetic subjects with variable color, brightness, and quality.