Fig. 1

The flow chart of the study design. First, dataset collection was done at the National Taiwan University Hospital, Hsinchu Branch. After glaucoma specialists reviewed all the collected images as well as the participants’ demographic and OCT extracted numerical data, each participants’ eye was precisely graded as N, PPG, or G. OCT extracted numerical data were not included in the training process. Data from the last two weeks was kept as a test dataset, and the rest of the data were used for training the models with tenfold cross-validation. The model was built on a webpage for telemedicine and the labeled color fundus images will be published as open data. N, normal; OCTA, optical coherence tomography angiography; PPG, pre-perimetric glaucoma; G, glaucoma