Deep Learning-Based Automated Classification of Multi-Categorical Abnormalities From Optical Coherence Tomography Images

• DOI: 10.1167/tvst.7.6.41
• Publication Date: 2018-12-28T18:30:24Z
• Authors: Wei Lu; Yan Tong; Yue Yu; Yiqiao Xing; Changzheng Chen; Yin Shen
• Keywords: Epiretinal Membrane; Kappa

Purpose: To develop a new intelligent system based on deep learning for automatically optical coherence tomography (OCT) images categorization. Methods: A total of 60,407 OCT were labeled by 17 licensed retinal experts and 25,134 included. One hundred one-layer convolutional neural networks (ResNet) trained the We applied 10-fold cross-validation method to train optimize our algorithms. The area under receiver operating characteristic curve (AUC), accuracy kappa value calculated evaluate performance in categorizing images. also compared with results obtained two experts. Results: achieved an AUC 0.984 0.959 detecting macular hole, cystoid edema, epiretinal membrane, serous detachment. Specifically, accuracies discriminating normal images, detachment, hole 0.973, 0.848, 0.947, 0.957, 0.978, respectively. had 0.929, while physicians' values 0.882 0.889 independently. Conclusions: This learning-based is able detect differentiate various excellent accuracy. Moreover, at level comparable or better than that human study promising step revolutionizing current disease diagnostic pattern has potential generate significant clinical impact. Translational Relevance: great increasing diseases' efficiency circumstances.