A novel method for automatic Hard Exudates detection in color retinal images

Diabetic Retinopathy (DR) is one of the major causes of blindness, and Hard Exudates (HEs) which are common and early clinical signs of DR. This paper presented a novel method to automatically detect HEs in color retinal images. We first extract HEs candidate regions by combining histogram segmentation with morphological reconstruction. Next, we define 44 significant features for each candidate region. A supervised support vector machine (SVM) is finally trained based on these features to classify the candidate regions for HEs. We evaluate the proposed method on the public DIARETDB1 database and achieve an sensitivity of 94.7% and an positive predictive value of 90.0%. Experimental results show that our method can produce reliable detection of HEs.