Retinal Diseases Classification using Deep Learning

Optical coherence tomography(OCT) technique is non-intrusive technique and is used for taking cross-sectional pictures of retinal layers in our eyes. It utilizes light waves in order to capture images of our retina and thus permits ophthalmologists to determine eye diseases. Thus, it has become a crucial methodology to identify and estimate retinal abnormalities and diseases, if present. With OCT images ophthalmologists can see distinct layers of retina across different images obtained, but due to more number of pictures to be visualized it becomes more time consuming for them to go through them and diagnose the pathology in the patient’s eye. In this work, we built various models with the help of Deep Learning technology that could classify various diseases including Drusen, Diabetic Macular Edema(DME), Choroidal Neovascularization(CNV), and Normal. The proposed models are LeNet5, AlexNet, GoogLeNet, VGG19 and MobileNet v2. The highest accuracy among all of the above 98.62 % is obtained with MobileNetV2.