Abstract:Diabetic retinopathy occurs due to damage to the blood vessels in the retina, and it is a major health problem in recent years that progresses slowly without recognizable symptoms. Optical coherence tomography (OCT) is a popular and widely used noninvasive imaging modality for the diagnosis of diabetic retinopathy. Accurate and early diagnosis of this disease using OCT images is crucial for the prevention of blindness. In recent years, several deep learning methods have been very successful in automating the process of detecting retinal diseases from OCT images. However, most methods face reliability and interpretability issues. In this study, we propose a deep residual network for the classification of four classes of retinal diseases, namely diabetic macular edema (DME), choroidal neovascularization (CNV), DRUSEN and NORMAL in OCT images. The proposed model is based on the popular architecture called ResNet50, which eliminates the vanishing gradient problem and is pre-trained on large dataset such as ImageNet and trained end-to-end on the publicly available OCT image dataset. We removed the fully connected layer of ResNet50 and placed our new fully connected block on top to improve the classification accuracy and avoid overfitting in the proposed model. The proposed model was trained and evaluated using different performance metrics, including receiver operating characteristic (ROC) curve on a dataset of 84,452 OCT images with expert disease grading as DRUSEN, CNV, DME and NORMAL. The proposed model provides an improved overall classification accuracy of 99.48% with only 5 misclassifications out of 968 test samples and outperforms existing methods on the same dataset. The results show that the proposed model is well suited for the diagnosis of retinal diseases in ophthalmology clinics.

HDB-Net: hierarchical dual-branch network for retinal layer segmentation in diseased OCT images

Multi-Scale and Multi-Branch Convolutional Neural Network for Retinal Image Segmentation

Fully Automated Segmentation of Hyper-Reflective Foci in OCT Images Using a U-shape Network

Multi-Task Dual Boundary Aware Network for Retinal Layer Segmentation

Automatic Segmentation of Hyperreflective Foci in OCT Images Based on Lightweight DBR Network.

ODDF-Net: Multi-object Segmentation in 3D Retinal OCTA Using Optical Density and Disease Features

Boundary-Repairing Dual-Path Network for Retinal Layer Segmentation in OCT Image with Pigment Epithelial Detachment

Multiclass retinal disease classification and lesion segmentation in OCT B-scan images using cascaded convolutional networks

Boundary Aware U-Net for Retinal Layers Segmentation in Optical Coherence Tomography Images

CLC-Net: Contextual and Local Collaborative Network for Lesion Segmentation in Diabetic Retinopathy Images

HRU-TNet: Hybrid Residual U-Transformer Network for diabetic retinopathy multi-lesion segmentation

Embedded Residual Recurrent Network and Graph Search for the Segmentation of Retinal Layer Boundaries in Optical Coherence Tomography

Topology guaranteed segmentation of the human retina from OCT using convolutional neural networks

HRS-Net: A Hybrid Multi-Scale Network Model Based on Convolution and Transformers for Multi-Class Retinal Disease Classification

HDC-Net: A Hierarchical Dilation Convolutional Network for Retinal Vessel Segmentation.

Automatic Segmentation of Retinal Edema in Optical Coherence Tomography Based on Deep Neural Networks

Segmentation of RBCC Disruption and Myopic Stretch Line in Retinal OCT Images Using an Improved U-shape Network

Attennet: Deep Attention Based Retinal Disease Classification In Oct Images

HACDR-Net: Heterogeneous-Aware Convolutional Network for Diabetic Retinopathy Multi-Lesion Segmentation

Robust Layer Segmentation Against Complex Retinal Abnormalities for en face OCTA Generation

Deep Residual Network for Diagnosis of Retinal Diseases Using Optical Coherence Tomography Images