Corona Virus Disease (COVID-19) Detection in CT Images Using Synergic Deep Learning.

During the global COVID-19 outbreak, it is very important to automatically screen for COVID-19 from chest computed tomography (CT). However, due to slight differences between COVID-19 and other viral pneumonia in chest CT, accurate screening checking COVID-19 remains a huge challenge. We present a deep learning scheme to automatically diagnose COVID-19 from chest CT. In this effort, a Synergic Deep Learning (SDL) model is proposed, which has two main modules. One is a module with 2 Resnet-50, which is used to learn and extract features from the image; the other is a synergic network, which is used to judge whether the input of the two networks mentioned above are of the same type and perform modulation. The proposed method achieves an accuracy of 95.11% in distinguishing COVID-19 from Non-COVID-19 with AUC = 0.9919, which is better than using a single Resnet-50. In addition, when it works in the multi-classification problem of the community acquired pneumonia (CAP), pneumonia, lung infection and COVID-19, it also has outstanding performance with accuracy of 88.56%, AUC = 0.9128.