Ensemble Learning based on Convolutional Neural Networks for the Classification of Retinal Diseases from Optical Coherence Tomography Images

Age-related macular degeneration (AMD) and Diabetic macular edema (DME) are retinal disease that can cause permanent vision loss. AMD is the leading cause of irreversible vision loss in individuals aged 65 and above and DME is the largest caused of visual loss of patients with diabetes in the world. Early detection and treatment is important to treat or delay the progress. Ophthalmologists use Optical Coherence Tomography (OCT) as one of key modalities to diagnose the diseases and decide whether to perform anti-VEGF therapy since it provides cross-section of patients' retina layers. Unfortunately, it is a tedious and time consuming work for ophthalmologist to analyze the images since OCT provides several images for each patient. Therefore, this paper propose automated methods to categorize the images into four categories (Choroidal neovascularization (CNV), Diabetic macular edema (DME), Drusen, and Normal) using deep learning and ensemble learning methods. Several Convolutional Neural Networks (CNNs) are adapted for the classification. To standardize training and test images, Fully Convolutional Networks (FCN) is applied to remove noise and a projection method is used to adjust tilted retina layers in the images. We train several CNNs and implement an ensemble learning model based on CNNs to further improve the performance. Among the CNNs, ResNet152 shows the best results with 0.9810 accuracy, 0.9810 sensitivity, and 0.9937 specificity, and the ensemble learning based on three ResNet152 shows 0.989 accuracy, 0.989 sensitivity, and 0.996 specificity.