Publication

Abstract : Optical coherence tomography (OCT) is a crucial imaging modality in ophthalmology, offering detailed retinal images essential for diagnosing various retinal disorders. In this study, we investigate the application of convolutional neural networks (CNNs) and transfer learning to classify retinal OCT images. We compare a custom CNN model with two pretrained models, MobileNet and VGG16, to explore their effectiveness in categorizing retinal diseases. Our study focuses on improving classification accuracy through image enhancement techniques, training methodologies, and model evaluation using confusion matrices. The results highlight the VGG16-based CNN model as the most effective, demonstrating its potential for accurate classification of retinal disorders based on OCT images. Furthermore, we delve into strategies to further enhance the performance of these models, such as data augmentation, finetuning, and hyperparameter tuning. We discuss the importance of these techniques in optimizing model performance and reducing overfitting. Additionally, we emphasize the significance of evaluating models using confusion matrices to gain insights into their performance across different classes. Overall, our findings suggest that CNNs, particularly the VGG16based model, hold promise for improving the automated diagnosis of retinal diseases using OCT imaging, which could lead to better patient outcomes in ophthalmology.