Computer-Aided Diagnostic System for Diabetic Retinopathy Using Convolutional Neural Network

An ongoing advancement in the condition of craftsmanship innovation assumes an imperative job in the picture handling applications, like biomedical, satellite picture preparing, artificial intelligence, object recognizable proof, diabetic retinopathy (DR), etc. DR is an eye disease found in people having high blood sugar. It can lead to loss of vision, if it is not treated properly. There is an increase in number of patients in comparison with ophthalmologists. The seriousness of the DR depends upon nearness of microaneurysms, hemorrhages, exudates and neovascularization. Specialists arrange diabetic retinopathy into five stages, namely ordinary, gentle, moderate, non-proliferative DR (NPDR) or proliferative DR (PDR). Convolutional neural network (CNN) results in high accuracy in classifying these diseases by spatial analysis. A CNN is progressively mind-boggling engineering construed more from the human visual perspective. A previous study done on DR suggests the use of CNN but with a different approach. Among other managed calculations involved, the proposed arrangement is to locate a superior and advanced way to classify the fundus picture with little pre-preparing techniques. Different fundus image databases available have been discussed. In this paper, different parameters used for the evaluation of developed systems have been presented.