Automatic Classification of Retinal Eye Diseases from Optical Coherence Tomography using Transfer Learning
Aya Adel, Mona Soliman, Nour Eldeen M. Khalifa, Khaled Mostafa
Abstract
This paper focuses on a four-class classification problem to automatically detect Choroidal Neovascularization (CNV), Diabetic Macular Edema (DME), DRUSEN, and NORMAL class in Optical Coherence Tomography (OCT) images. The proposed deep learning classification model adopted two transfer-learning architecture and a categorical hinge loss as Support Vector Machine (SVM) instances to correctly classify retinal diseases in the retinal OCT dataset. The experiments follow a patient-level 10-fold cross-validation process using a retinal OCT image dataset. The proposed model is tested on the benchmark data with exactly 84495 grey-scale images labelled into 4 classes. The experimental evaluation shows that the proposed model outperforms the conventional deep learning state of the art reported on the OCT dataset with an overall accuracy of 98% for the Xception model and 93% for The inception V3 model.