Classification of Retinal Diseases from OCT images using 2D-Convolutional Neural Networks

Abstract

Purpose: Diabetic macular edema (DME) is the primary cause of vision loss among individuals with diabetes mellitus (DM). An approach for detection of Diabetic Macular Edema (DME) retinal diseases from Optical Coherence Tomography (OCT) images in diabetic patients using 2D-convolutional neural networks (CNN), a commonly used deep learning network model has been presented. Classification of optical coherence tomography (OCT) images can be achieved with high accuracy using convolution neural networks (CNN), a commonly used deep learning network for computer- aided diagnosis. We developed, validated, and tested a deep learning (DL) system for classifying DME using OCT images using optical coherence tomography (OCT) devices.

Methods: In this study, we used 300 B-Scan OCT images (DME and Normal) to evaluate the model. There were 200 sample data in the training and for testing a total of 100 retinal OCT images datasets containing (50 normal and 50 DME affected) are used for detection and classification. All the data are OCT images of retina. The proposed model was constructed based on a deep neural network to diagnose the images accurately and for better accuracy of classification. It was trained using the training dataset and the test dataset was used to evaluate the trained model.

Result: Our study shows that the classification of OCT images using deep learning method achieved accuracy score of 93% and 24% of score of loss respectively. In this study, as per the comparison it can be seen that our 2D-CNN model exhibited higher accuracy than those of the other traditional SVM and Random forest machine learning method.

Conclusion: The proposed deep learning model helps us in better classification of DME from the OCT images. This will help the eye doctor to initiate early and proper treatment. Future research is needed to validate DL algorithms to get faster better accuracy of big data of OCT images and get more details of disease changes in all retina layers.