Abstract: The robust identification of red lesions in digital color fundus photographs is a vital pace in the development of automated screening systems for diabetic retinopathy. The retinal images are first subjected to preprocessing for color normalization and contrast enhancement. Contextual clustering algorithm is used to segment the retinal image. Before classifying the fragments, it is obligatory to locate and eliminate the optic disc. The detected candidate objects are classified as exudates or non-exudates using the features `Convex Area` and �Solidity` and �Orientation`. The modular neural network is trained using a set of 25 images consisting of 5 normal images and 20 abnormal images. The trained system has been tested with 15 images and is found to acquire satisfactory results with 93.4% sensitivity and 80% specificity.
C. Jayakumari and T.Santhanam , 2007. Detection of Hard Exudates for Diabetic Retinopathy Using Contextual Clustering and Fuzzy Art Neural Network. Asian Journal of Information Technology, 6: 842-846.