Could AI soon let us diagnose covid-19 simply by a cough?

·1-min read

Researchers at the Massachusetts Institute of Technology have developed a technique based on machine learning to detect whether someone who is asymptomatic could be a carrier of covid-19 simply by the sound of their cough. The idea is to be able to very quickly diagnose -- using artificial intelligence -- whether someone is infected with covid-19. The first results are already promising.

If you become worried as soon as you cough or hear the cough of someone close to you, then you may be interested in this new breakthrough. Because while it's difficult to distinguish 50 different nuances of coughs by ear, artificial intelligence can compare thousands of cases.

To create this system, the researchers carried out recordings of forced coughs, nearly 200,000 in all, including 2,500 people who tested positive for covid-19. The automatic learning process makes it possible to compare each new case to this huge sample. The result is that this formidable algorithm is now capable of accurately identifying 98.5% of covid-19-infected individuals.

The team is already working on the integration of this model in an application for the general public, which would still require approval from the US Food and Drug Administration (FDA) before being released. It could then be enough to cough close to your smartphone for it to tell us whether or not we have been infected with covid-19.