Automated Diagnosis of Retinal Diseases from OCT Images Using ResNet-18

Retinal diseases are a major cause of vision impairment and blindness globally. Automated diagnosis of these diseases from optical coherence tomography (OCT) images can significantly improve patient outcomes by facilitating early detection and intervention. The study presents an automated approach for the diagnosis of retinal disease based on OCT images. Using deep learning, a robust model for accurate detection of retinal diseases was developed, specifically on ResNet-18 architecture in a convolutional neural network (CNN). The main points of the study were to explore the effectiveness of transfer learning in training the model and, secondly, to examine its performance in classifying retinal OCT images. The methodology consisted of fine-tuning a pre-trained ResNet-18 model using a dataset of retinal OCT images and classifying images into four categories: Normal, Diabetic Macular Edema (DME), Choroidal Neovascularization (CNV), and Drusen. The model achieved an accuracy of 96.61%, indicating that it successfully classifies retinal disease. A confusion matrix analysis provided more insights into the model's performance and exposed areas needing improvement. Fundamentally, this study has notable social implications in terms of the early diagnosis and treatment of ocular diseases, which may lighten some load on medical systems.