SARS-CoV-2 virus-specific reverse transcriptase polymerase chain reaction (RT–PCR) tests are routinely used to diagnose COVID-19, but they can take up to 2 days to complete and kits are in short supply. Alternative methods for rapid and accurate diagnosis of patients with COVID-19 are urgently needed.

This study describes the development of a unique algorithm to rapidly detect COVID-19 based on computed tomography (CT scans) chest imagery, in combination with patient information including symptoms, age, bloodwork, and possible contact with someone infected with the virus.

The authors examined CT scans of 905 patients who were admitted to 18 medical centers in 13 Chinese provinces between January 17 and March 3, 2020. They developed their algorithm on data from 626 patients, and then tested it on the remaining 279 patients, who were both COVID positive and COVID negative. The compared the AI findings with those from a trained radiologist.

They found that the AI system achieved an area under the curve of 0.92 with 84% sensitivity compared to 75% for radiologists. The algorithm also improved the detection of patients who were positive for COVID-19 via RT–PCR who presented with normal CT scans, correctly identifying 17 of 25 (68%) patients, whereas radiologists classified all of these patients as COVID-19 negative.

The authors conclude that, when CT scans and associated clinical history are available, this AI system can help to rapidly diagnose COVID-19 patients.