COVID-19 infections detected through cough by algorithm, new MIT study says. But experts are skeptical

Rachel Grumman Bender
·4 min read

Artificial intelligence may be able to help identify asymptomatic COVID-19 patients just by their “forced” cough, according to a new study published in the journal IEEE Journal of Engineering in Medicine and Biology. But some experts express doubt about the technology.

For the study, MIT researchers set up a website where people could record several different forced coughs from their cell phones. The researchers collected data from more than 5,300 people. According to an MIT press release, approximately 2,500 recordings were submitted by people with a confirmed COVID-19 diagnosis, including those who had tested positive for the virus but were asymptomatic, “along with 2,500 more recordings that they randomly selected from the collection to balance the dataset.”

Young ill woman covering her nose and mouth with arm while sneezing at home. Sick infected girl patient coughing in her elbow. Coronavirus COVID-19 symptoms. Virus protection. Correct cough in hand
Young ill woman covering her nose and mouth with arm while sneezing at home. Sick infected girl patient coughing in her elbow. Coronavirus COVID-19 symptoms. Virus protection. Correct cough in hand

The researchers reported that the AI model had an accuracy rate of more than 98 percent when it came to identifying the coughs of people confirmed to have COVID-19. They also found that the technology correctly distinguished 100 percent of asymptomatic people’s forced coughs from the healthy ones.

“MIT researchers have now found that people who are asymptomatic may differ from healthy individuals in the way that they cough,” according to the press release. (Yahoo Life reached out to the lead study author but did not receive a response.)

In the statement, one of study’s co-authors Brian Subirana, the director of MIT’s Auto-ID Laboratory, said: “We think this shows that the way you produce sound changes when you have COVID, even if you’re asymptomatic.”

While both Dr. Jonathan Chen, an assistant professor of medicine at the Stanford University Medical Center, and Dr. Megan Conroy, an assistant professor of clinical medicine in the division of pulmonary, critical care and sleep medicine at The Ohio State University Wexner Medical Center, tell Yahoo Life that the study results are “interesting,” both are skeptical.

Chen, who specializes in biomedical informatics research, calls the motivation behind the study “fantastic” and “clever,” but questions the plausibility of such high success rates from “self-reported” coughs. “Even a [COVID-19] PCR test isn’t that accurate,” he says. Chen adds that the technology could “still be helpful, but very likely it’s much less accurate than what they're finding.”

Conroy shares that she has several “serious concerns,” saying: “First, this isn't a validated way to screen for any other respiratory disorders or causes of cough; the study didn't include any other respiratory viral infections, which raises some concern for the apparent ability to distinguish between pathology and normal. Further, they state the technology can identify a cough from an asymptomatic person as infected with COVID-19, but it would be important to note that if you have a new cough you wouldn't be considered asymptomatic.”

She also calls the idea of distinguishing diseases by the sound of a cough “questionable.” Adds Chen: “What if I have a cough, but it’s the common cold?”

According to the release, the AI model is not designed to diagnose symptomatic people, “as far as whether their symptoms are due to COVID-19 or other conditions like flu or asthma,” adding: “The tool’s strength lies in its ability to discern asymptomatic coughs from healthy coughs.”

The researchers are now trying to create an app using the technology as a “free, convenient, noninvasive prescreening tool,” according to the press release. “A user could log in daily, cough into their phone, and instantly get information on whether they might be infected and therefore should confirm with a formal test.”

However, Conroy “strongly” cautions against the use of “unproven technology to guide behaviors.” Instead, she encourages the “continued growth of actual testing for the SARS-CoV2 virus, in addition to the many measures that we know work to reduce the spread of the virus: universal masking, staying home when you can, avoiding large gatherings and social distancing whenever you are around those outside of your household.”

For the latest coronavirus news and updates, follow along at https://news.yahoo.com/coronavirus. According to experts, people over 60 and those who are immunocompromised continue to be the most at risk. If you have questions, please reference the CDC’s and WHO’s resource guides.

Read more from Yahoo Life:

Want lifestyle and wellness news delivered to your inbox? Sign up here for Yahoo Life’s newsletter.