A clustering approach for exudates detection in screening of diabetic retinopathy

Diabetic Retinopathy (DR) is an extensively spread retina disease which is the outcome of long term or uncontrolled diabetes on the retina. Exudates are prominent sign of DR which is the crucial cause of loss of sight in patients suffering with diabetes. Early diagnosis of the disease through automated screening and regular treatment has proven helpful in preventing the spread of disease and irreparable visual impairment. This paper proposes a method using K-means clustering and morphological image processing for detection of exudates on low-contrast retinal images. The publicly available retinal images of DIARETDB1 database are used as the input samples for testing the algorithm. The exudates obtained using proposed algorithm are verified by comparing with hand-drawn ground truths images available along with DIARETDB1 database. The sensitivity and specificity of the algorithm obtained for the database is 88.34% and 99.27% respectively.