Detection and classification of hard exudates in retinal images.

Diabetic retinopathy (DR) is a chronic disease of the retinal microvasculature which leads to loss of central visual acuity. Early detection of hard exudates in retinal images using computer aided tool helps the ophthalmologist to diagnose the blindness problem. This article presents a novel method to detect and classify the hard exudates in retinal images. For detection, the optic disc (OD) of the retinal image is masked and then the bright patches that contribute to hard exudates are segmented based on thresholding and morphological reconstruction techniques. Here OD is identified using brightness and variance features of the OD followed by Circular Hough Transformation. For classification, features such as color, size, and texture are extracted from each segmented candidate regions based on these features and the regions are classified by using multilayered perceptron neural network (MLP). The proposed method is experimented on the DIARETDB1 retinal dataset and also compared with the existing methods.