Retinal Disease Detection: Using Deep Learning for Accurate Diagnosis with Optical Coherence Tomography Scans

Retinal diseases such as Choroidal Neovascularization (CNV), Diabetic Macular Edema (DME), and Drusen are leading causes of vision impairment globally. Early and accurate detection is crucial for effective treatment and prevention of blindness. This study presents a deep learning model for the classification of retinal diseases using convolutional neural networks (CNNs). A dataset comprising 968 retinal images evenly distributed across four classes—CNV, DME, Drusen, and Normal—was utilized. The model was trained over 10 epochs, with training and validation loss decreasing significantly, indicating effective learning without overfitting. The training accuracy increased from 65% to 90%, while validation accuracy rapidly reached 90% and stabilized between 95% and 100%, demonstrating strong performance on unseen data. Evaluation metrics revealed an overall accuracy of 96.9%. The model exhibited high precision and recall for the CNV and Normal classes, with F1-scores of 99.17% and 98.97%, respectively. However, the DME class had a lower recall of 90.91%, and the Drusen class had a lower precision of 91.51%, suggesting misclassifications due to visual similarities between these conditions. The findings indicate that while the model performs exceptionally well overall, there is room for improvement in distinguishing between DME and Drusen. Future work will focus on enhancing the model's ability to differentiate these conditions by incorporating more discriminative features and expanding the dataset. The proposed model shows promise for clinical applications in retinal disease detection, potentially aiding in early diagnosis and treatment to prevent vision loss.