Abstract
Diabetic Macular Edema (DME) is an advanced indication of diabetic retinopathy (DR). It starts with blurring in vision and can lead to partial or even complete irreversible visual compromise. The only cure is timely diagnosis, prevention and treatment in the early stages. In this paper we have presented a novel automated system for the diagnosis and stage classification of diabetic macular edema, by following certain steps: first the image is enhanced by contrast limit adaptive histogram equalization (CLAHE) and contrast stretching to improve the overall contrast of the image. After the components become distinct, optic disk (OD) is localized and excluded from the image. All possible exudates are detected from OD less image using dynamic thresholding, where the decision parameters are mean and standard deviation of the image. The different stages of the disease are classified on the given criteria by Early Treatment Diabetic Retinopathy Studies (ETDRS). Stage of the disease is classified as normal, less significant, moderate and severe to assess the severity level. The proposed technique gives overall accuracy of 97.4 %, 98.7%, and 97.2 % on publically available databases MESSIDOR, DIARETDB, and HEI-MED, respectively.