Automatic Discrimination for Screening of the Eyes in a Deep Learning-Based Ensemble Model Using Optical Coherence Tomography Images

Abstract

The average life expectancy has increased globally, and the risk of visual impairment is expected to increase as well. Therefore, eye checkups have become increasingly important to maintain good vision and quality of life. As the population requiring eye checkups increases, so does the clinical work burden of physicians. Hence, an automatic discrimination algorithm to reduce the clinicians' workload is necessary. The convolutional neural network (CNN), a deep learning algorithm, has recently become one of the chief techniques for automated image recognition and is a helpful tool for identifying ocular diseases. However, the accuracy of a single CNN model ranges within 70%–95%, and it may fail to identify some diseases. In this study, we aimed to compare the diagnostic performances of four CNN models trained with optical coherence tomography (OCT) images, the machine-learning (ML) model trained with data on the retinal and choroidal area, and the ensemble model that integrated the CNNs and ML models using OCT images obtained during eye checkups. Our results show that the ensemble model had a superior diagnostic performance over the CNN and ML models. The ML model, which evaluated diseases using data regarding the temporal peripheral retinal area, improved on the CNN model, which misrecognized the temporal peripheral retinal structures. Our study indicates the strong potential of the ensemble model combining the CNN and ML models in accurately predicting abnormalities during eye checkups.