Classification of Age-Related Macular Degeneration Disease with Deep Learning based on Optical Coherence Tomography images

Abstract

Because Deep Learning can predict events accurately and quickly, it has garnered a lot of interest from the scientific community. In this study, age-related macular degeneration was categorized using deep learning. In persons over 50, age-related macular degeneration (AMD) is the main factor causing blindness. AMD first manifests as a black spot in the center of the field of vision and a loss of central vision. With an ageing population, the prevalence of this pathology is continuing to rise, creating a serious public health issue. AI and deep learning are already proving to be disruptive in industries like medicine. Even while in the current state of affairs, an AI’s recommendations cannot completely replace a doctor’s, such systems could nevertheless lessen the overall labour that the doctor has to do. We employ optical coherence tomography (OCT) images and divide them into four groups based on the already recognized retinal illnesses. These groups are called drusen, normal, diabetic macular edema (DME), and choroidal neovascularization (CNV). We have performed image augmentation to prevent the overfitting of the trained models. We train Xception with additional layer, Inception V3 with additional layer and VGG19 with additional layer, out of which VGG19 with additional layer, produced better testing accuracy of 90.93%, which is best when maintaining the train, validation and test split ratio as 0.7, 0.1 and 0.2.