The American College of Radiology® (ACR®) Data Science Institute® (DSI) has published a new COVID-19 use case for chest radiography that offers clinical expertise to researchers for developing an artificial intelligence (AI) solution for determining the likelihood of an imaging diagnosis for COVID-19-compatible lung disease.
The chest X-ray use case developed by the ACR DSI Thoracic Panel includes inputs, outputs and possible corollary features for developing an AI solution to detect COVID-19. “Since X-ray testing is more widely accessible than CT, AI research using chest radiographs to detect COVID-19 is worth exploring,” says Eric Stern, MD, chair of the ACR DSI Thoracic Panel.
Published studies describe how the appearance of COVID-19 abnormalities with X-ray imaging are similar to those observed with computed tomography (CT). Both demonstrate bilateral peripheral consolidation, though the sensitivity and specificity rates of detection with chest radiography are not as high as with a real-time, reverse transcription polymerase chain reaction (rRT-PCR) test.
Machine learning research to detect COVID-19 on chest radiographs, rather than CT, could lead to several potential clinical benefits. These include easier and simpler access to testing, lower patient risk and reduced burden on clinical staff. Radiography is also typically less expensive and exposes patients to less radiation compared to CT. Workflows may benefit because radiographic equipment is typically easier to disinfect.