Diabetic Retinopathy Detection using Deep Learning Techniques

Abstract

Diabetic retinopathy is referred as diabetic eye disease. It causes damage to the retina of the light sensitive tissues at the rear portion of the eye. It mainly affects the working age population in the developing country. Right now, recognizing DR is a tedious and manual interaction that requires a prepared clinician to analyze and assess advanced shading fundus photos of the retina. The rate of diabetes is more in local populations and the detection of diabetic retinopathy is needed but there is a shortage of equipment because there are expensive. With persistent advancement of deep learning models we hope to increase the accuracy of the technique and extend it to glaucoma diagnostics. In Early days convolutional neural network were used it takes more time and gives the low accuracy rate. In this paper Regional convolutional neural network and resnet were used to increase the accuracy rate and reduce the time consumption. Convolutional neural network takes image as an input and process it in different ways and assigns important to that images and produces the output by the images. In convolutional neural network there are many layers mainly input layer, hidden layer and output layer. If this technique is implemented. Diabetic retinopathy can be detected at the early stage and we can reduce the number of blindness and increase the accuracy rate and time consumption. Keyword Diabetic retinopathy, Fundus photography, Deep learning and Data set I.INTRODUCTION Diabetic retinopathy (DR) is a common complication of diabetes associated with ret inal vascular damage caused by long standing diabetes. Furthermore, the diagnosis of DR mostly depends on the observation and evaluation to fundus photographs of which procedure can be time consuming even for experienced experts. Therefore computer aided automated diagnosis approaches have great potential in clinical to accurately detect DR in a short time which can further help to improve the screening rate of DR and reduce the number of blindness. For a deep learning model, the most important parts that should be focused on are data set, network architecture and training method. Before being used to train our model, fundus images data set obtained from public resources is pre processed and augmented. The model accepts two fundus images corresponding to the left eye and right eye as inputs and then transmits them into the Siamese like blocks. The information from two eyes is gathered into the fully connected layer and finally the model will output the diagnosis result of each eye respectively.