Comparison of DL model for retinal diseases classification using OCT images

Age-related macular degeneration (AMD) and Diabetic macular edema (DME) affect the macula of the retina for the person above 50 years. Accurate diagnosis of eye diseases in the initial stages prevents blindness. Manual detection needs the use of trained human experts. Hence, automated models are necessary. In this research paper, a novel approach for detecting retinal conditions is proposed. This involves the progress of three deep convolutional neural network (CNNs) models viz., AlexNet, ResNet50, and InceptionV3. These models are trained to classify Optical Coherence Tomography (OCT) retina images into three macular diseases viz., Choroidal neovascularization (CNV), DME, DRUSEN, and Normal. The publicly available Kermany dataset is apply to test the accuracy of each model. The outcomes show that the AlexNet model achieves an accuracy of 97.2% followed by the InceptionV3 model (95%) and the ResNet50 model (79.8%). This finding can help the Ophthalmologists for easy and fast diagnosis to ensure quality eye care for individuals who are at the risk of vision loss.