A Novel Deep Learning Framework for Diagnosing Retinal Diseases using OCT images

Abstract

Ophthalmologists employ optical coherence tomography (OCT), a non-invasive imaging technique that produces cross-sectional images of the retinal layers, to diagnose retinal abnormalities. OCT is a crucial tool for diagnosing and assessing retinal diseases. However, analyzing the many images that OCT generates for each patient often takes a substantial amount of time for ophthalmologists. In this work, OCT images are classified into four groups using deep learning models: normal, drusen, diabetic macular edema (DME), and choroidal neovascularization (CNV). Two distinct models are proposed: one employs segmentation based on curvature-based regions of interest (ROI), while the other uses the Folded RESNET101 method for classification. The recommended method's accuracy, sensitivity, and specificity are all 0.99, producing outstanding outcomes.

Keywords