DEEP LEARNING APPROACHES FOR MULTI-CLASS DISEASE DETECTION AND CLASSIFICATION IN RETINAL IMAGES

Many techniques are designed to increase accuracy diagnoses of diseases in retinal images such as hypertensive retinopathy. Convolutional neural networks (CNN) have been employed in various techniques to segment blood vessels in retinal images. These techniques sometimes fail to effectively segment the blood vessels of retina of eye and produce extra noise and we have with this drawback, we present a technique to detect blood vessels in retina of eye images using thousands of blocks (patches) from each of the image. This technique is on a CNN where it contains four convolutional layers followed by relu, max pooling four layers, two layers are fully connected and one layer is the softmax for segmenting blood vessels.