An efficient deep learning approach to detect retinal disease using optical coherence tomographic images

Abstract

Optical Coherence Tomography (OCT) is an effective approach for diagnosing retinal problems that can be used in combination with traditional diagnostic testing methods. We developed and implemented a deep Convolutional Neural Network (CNN) model, which has the capability to effectively identify and classify Optical Coherence Tomography (OCT) images into the following four categories: Normal, DMD, CNV, and DME. The proposed 21 layered CNN model is built with three basic layers: a convolutional layer, a pooling layer, and a fully connected layer along with dropout and dense layers. Our proposed model is able to detect and differentiate between the OCT images with a high amount of accuracy. The 21 layer proposed CNN model was used for the classification and diagnosis of retinal sickness using OCT images. To justify the efficiency of our custom CNN model, seven pre-trained CNN models (VGG16, VGG19, MobNetV2, Resnet50, DenseNet121, InceptionV3, and InceptionResNetV2) were used and testified with the same amount of dataset. In terms of the accuracy, precision, recall, and f1 score, which are all tested in this paper, the suggested CNN model along with seven other pre-trained CNN architectures perform comparable on the available dataset. The proposed model has an accuracy rate of 98.37 percent, which is greater than any of the experimental results of the CNN models utilized in this research due to the fact that the recommended model was developed. When it comes to the diagnosis of retinal problems, the CNN model that was suggested performs far better than any other model that was previously used.