MINNEAPOLIS, Minnesota — COVID-19 infects the lungs, and now a chest X-ray could help tell within seconds if a patient likely has the virus that causes COVID-19.
Researchers at the University of Minnesota (UMN) have developed and validated an artificial intelligence algorithm that can identify possible COVID-19 cases from a patient's chest X-ray.
"Everyone is really worldwide working together collaboratively trying to beat this," said Dr. Christopher Tignanelli, a U of M Medical School assistant professor of surgery and co-lead on the project.
UMN researchers have been working with M Health Fairview and Epic — a healthcare software company — on the project.
To develop the algorithm, researchers used 100,000 X-rays of patients who did not have COVID-19 and 18,000 of patients who did.
The algorithm evaluates the X-ray as soon as the image is taken and if it recognizes patterns associated with COVID-19, medical staff can find out within seconds how likely the patient is to have the virus.
"We're not routinely testing patients every day in the hospital so someone could be admitted to get an appendix surgery or hernia surgery, for example. They get their COVID test; initially, they're negative and then maybe they get exposed by a healthcare provider, another patient or whatever to COVID two, three days into their hospitalization," Dr. Tignanelli said. "We're not just doing it every third day or so, get a COVID test, but people do frequently get X-rays in the hospital. They're somewhat routine. So that makes it a very low-hanging fruit to do a diagnosis and obviously COVID infects the lungs so that's a good place to look for it."
Dr. Tignanelli said the algorithm can help serve as another tool to catch more cases of COVID-19, especially when faced with testing delays and false negatives.
But it won't work for every patient.
"There's definitely some patients who have COVID disease that have rather normal-looking lungs and the algorithm does not pick those people up as having COVID. So it's not perfect. It's just an additional tool," Dr. Tignanelli said.
The algorithm will be available at no cost to more than 450 health care systems worldwide through Epic, including all 12 M Health Fairview hospitals.
According to Dr. Tignanelli, it's not live yet. In a few weeks they'll get into beta and pilot testing and then see if any tweaks are needed.
"We're constantly optimizing the algorithms as we get more and more X-rays," he said.