Classification network of Chest X-ray images based on residual network in the context of COVID-19

Abstract

As COVID-19 spreads across the globe, more cases are being confirmed around the world, making it imperative that we take a better approach to fighting the outbreak. To stop the spread of the disease and better screen for cases, we need a more sensitive and efficient test that can classify images of lung abnormalities in patients. In this paper, residual network is used to classify the collected chest radiographs. Feature extraction and classification were carried out on the original chest X-ray images, which were divided into the following three categories: normal lung, bacterial pneumonia and virus pneumonia. This can quickly rule out normal and routine infections, screen out large numbers of cases, and reduce the burden on health care workers who need to further examine cases. At the same time, our results are also very good, with an accuracy of 94%, which has practical classification significance.