AI & Chest CT in the COVID-19 Crisis
Guidelines for Chest CT in COVID-19
On April 7, the Fleischner Society published a consensus statement on the role of chest X-ray and chest CT in the diagnosis and management of patients with COVID-19. The panel brought together the multidisciplinary perspectives of experts from 10 countries around the world.
The panel concluded that chest imaging is not indicated for screening or in patients with mild symptoms of COVID-19. However, chest imaging is indicated in patients with severe symptoms or worsening respiratory status. The panel also found that CT is appropriate to use in patients with functional impairment and/or hypoxemia—low level of oxygen in the blood—after recovery from COVID-19.
Because chest CT images do not necessarily describe findings unique to COVID-19, chest CT should not be used as a diagnostic tool for COVID-19. Chest CT could instead play a vital role in assessing the severity of disease and evaluating potential disease complications when combined with AI techniques.
AI Applications in the COVID-19 Pandemic
AI scientists around the world have been developing AI algorithms to contribute to the battle against the COVID-19 pandemic. In March, researchers in Europe started a multi-center project, Imaging COVID-19 AI, to develop a deep learning model that will automatically detect and classify COVID-19 on chest CT scans. AI projects such as this could bring value to COVID-19 patients by speeding the time it takes to analyze CT images.
There are more opportunities where AI can be applied to chest CT images to help contribute to managing COVID-19 and inform treatment decisions in a timely fashion.
Assessing Disease Severity & Progress
According to a study published in European Radiology, chest CT could be useful in assessing disease severity in patients with COVID-19 pneumonia because chest CT findings in patients with severe illness vary from those in patients with milder diseases. AI could be applied to chest CT scans to automatically produce a quantitative assessment of lesion volume, proportion, and characteristics. Overtime, clinicians could compare consecutive scans of the same patient to assess the progress of COVID-19 pneumonia. This could help clinicians develop treatment for patients with the disease.
Triage and Notification
Deep learning can be applied to detect and prioritize incidental pneumonia findings associated with COVID-19 from non-contrast chest CT images. CT triage systems can be used to notify the radiology worklist that the image includes findings suggestive of pneumonia associated with COVID-19, such that a radiologist can review the exam sooner and make a definitive report. It will be important to develop AI algorithms that can run in the background to alert radiologists of incidental COVID-19 pneumonia findings as COVID-19 cases become less routine.
Bringing Solutions to Clinical Use
Managing COVID-19 will take global collaborations to accelerate progress.
- COVID-19 Clinical Resources for Radiologists – American College of Radiology
- The Role of Chest Imaging in Patient Management during the COVID-19 Pandemic: A Multinational Consensus Statement from the Fleischner Society – Radiological Society of North America
- COVID-19 + Imaging AI Resources – Stanford University Center for Artificial Intelligence in Medicine & Imaging
- Radiology Department Preparedness for COVID-19: Radiology Scientific Expert Panel – Radiological Society of North America