Detection of Retinal Degeneration via High-Resolution Fundus Images using Deep Neural Networks

Abstract

Retinal imaging analysis is used to diagnose and classify retinal diseases like age-related macular degeneration (AMD), retinal detachment, diabetic retinopathy (DR), retinitis pigmentosa, and retinoblastoma. The automated detection of retinal disorders is a significant step towards early disease diagnosis and the prevention of disease progression. Historically, a number of cutting-edge techniques have been created to aid in the automatic segmentation and detection of retinal landmarks and diseases. But new advances in deep learning and advanced ophthalmology imaging modalities have given researchers access to a whole new domain. In this paper, two multilayer deep neural networks convolutional neural network (CNN) and AlexNet are presented for the early detection of retinal degeneration. Further analysis is carried out by applying three different types of optimizers to train the classifiers, such as ADAM, RMSProp, and SGDM. The performance analysis is carried out with high-resolution fundus images at three different training rates to determine the superiority of the classifiers.