A person who is contagious but asymptomatic for COVID-19 coughs differently from healthy individuals.

The difference in coughing may not be discernible to the ordinary person, but a trained AI model can pick it up, according to researchers at MIT. Their system analyzes forced-cough recordings taken on a cellphone or laptop and submitted via the internet. It highlighted COVID-19 positivity with near-total accuracy. (MIT News)

The report by the MIT researchers was published recently in the IEEE Journal of Engineering in Medicine and Biology.

AI analyzes coughs to detect if a person is COVID positive

Persons may be unknowingly carrying COVID and display none of the known symptoms of the virus. Also, they may see no need for undergoing a COVID test. They are, however, still contagious and can spread the virus to others.

The MIT researchers trained an AI model on thousands of audio samples of coughs and spoken words. They then tested the model by feeding it new cough recordings.

Surprisingly, the AI model could accurately identify 98.5% of coughs from confirmed Covid-19 positive patients. The model was 100% accurate in detecting the virus in coughs from asymptomatics — who did not its symptoms but were tested and found positive for the virus.

The model can do this by identifying patterns in the four biomarkers — vocal cord strength, sentiment, lung and respiratory performance, and muscular degradation — that are unique to Covid-19.

App on the cards

The findings present an easy but reliable means to test people for COVID on a preliminary basis. If the AI signals infection, the patients may immediately get themselves formally tested for COVID in a lab.

The MIT team is developing an app that would use the AI model and instantly report positivity.

It would be a free, convenient, noninvasive prescreening tool that a person could use daily and check if they need a COVID-19 test by coughing.

“The effective implementation of this group diagnostic tool could diminish the spread of the pandemic if everyone uses it before going to a classroom, a factory, or a restaurant,” says co-author Brian Subirana, a research scientist at MIT’s Auto-ID Laboratory.

Related Story:   Baricitinib, An AI-flagged Rheumatoid Drug, Could Treat COVID-19