Meta-cognitive neural network method for classification of diabetic retinal images

Abstract

An eye disease which is assorted in person with diabetes that is occurred by change in blood vessels of the retina is called Diabetic Retinopathy. Retinopathy can occur with all types of diabetes and can cause vision loss if it's not treated on time. Detection of exudate by ophthalmologist normally takes time and energy. In this paper, classification and detection of exudate in color retinal image using automated technique have been proposed. This method reduces work of ophthalmologist. A series of steps or actions need to be taken for exudate detection. Firstly in pre-processing step, green channel extraction is used and optic disk is eliminated to prevent optic disk from interfacing with exudates detection. Robust Spatial Kernel FCM (RSKFCM) segmentation method is used for optic disk elimination which gives good result compared to other FCM based method. The significant features are extracted from the segmented images and are used for classification purpose. Meta-cognitive neural network method is used as classifier. The experiments were conducted on standard diabetic retinal image dataset. Experimental results shows that the proposed method gives promising results.