A novel approach for automated detection of exudates using retinal image processing

Even though a number of anatomical structures contribute to the process of vision, many eye disease that causes blindness occur mainly in the retina. One among the important eye disease is Diabetic Retinopathy. DR is a severe eye disease and is the primary cause of blindness in the case of diabetic patients. We describe a novel method for automatically detecting the exudates with the help of retinal photography. The color retinal images are pre-processed using CLAHE and second order Gaussian filter. The pre-processed retinal color images are partitioned into segments with the help of Soft clustering algorithm. Since OD and exudates are similar in color, a segment with OD along with exudates is chosen. A set of features is extracted from the segmented image by Scale Invariant Feature Transformation (SIFT) algorithm. Then active Support Vector Machine Classifier is trained with these extracted feature vector and evaluated on two publicly available database. We tested our approach on a set of 1000 images and obtained a sensitivity of 99.96% and specificity of 96.6%. Our approach has potential to be used as a clinical tool in automatic detection of Diabetic Retinopathy.