Skip to main content
. 2021 Jan 9;17(2):539–548. doi: 10.7150/ijbs.53982

Figure 1.

Figure 1

Development of the classifier for differentiating coronavirus disease 2019 (COVID-19) from influenza A/B and structure of the deep neural network (DNN). (A) A total of 525 patients with COVID-19 and 107 patients with influenza A/B were enrolled and separated into a training set of 290 cases and a test set of 328 cases after exclusion. A DNN was applied for feature extraction, selection, classification. The proposed fusion network, clinical network, chest x-ray (CXR) network and computed tomography (CT) network were established for final diagnosis. (B)The combined network system has two input streams: image data and clinical data. The two kinds of data are processed by two streams of deep neural layers, which are ultimately concatenated. When processing CXR image or clinical data only, the other one data stream is removed.