Severity assessment of COVID-19 using imaging descriptors: a deep-learning transfer learning approach from non-COVID-19 pneumonia

Artificial intelligence (AI) has the potential to play a significant role in patient care during a pandemic such as COVID-19. In this preliminary study, we demonstrated the feasibility that deep learning-based techniques may be used to quantify the severity of the disease in chest x-rays. The global descriptors derived from the deep-learning-generated severity map in the lung regions showed substantial correlation with radiologist’s severity ratings of abnormality. The global descriptors could also be used to classify normal and abnormal CXRs. If fully developed, the AI-based tool could potentially be used for decision support to assist in patient management or treatment response monitoring.