RetinaFoldNet: An Intelligent Deep Learning Framework for Automated OCT-Based Retinal Disease Diagnosis

Abstract

Optical Coherence Tomography (OCT) is an essential imaging modality for retinal disease diagnosis due to its capacity for high-resolution cross-sectional visualization of retinal structures. Manual interpretation of large-scale OCT scans is time-consuming and subject to inter-observer variability. This study introduces a deep learning framework for automated OCT-based retinal disease diagnosis, which classifies retinal images into four categories: Normal, Drusen, Diabetic Macular Edema (DME), and Choroidal Neovascularization (CNV). The framework combines region-focused preprocessing with multi-scale feature learning to improve the discriminative representation of retinal abnormalities. Experimental evaluation using a publicly available OCT dataset demonstrates superior diagnostic performance, achieving 99% accuracy, sensitivity, and specificity. This framework offers an effective and reliable computer-aided decision-support system for automated retinal disease diagnosis.