People with COVID-19 Infection Identified by the Sound of Their Cough

By Mara Anton

Posted on November 5, 2020

Until now, the only way to identify a person that had contracted COVID-19 was by doing a proper swab test. Truthfully, it still is the only sure way, approved by the scientific world, to know if a person has or doesn’t have COVID-19. However, researchers developed an algorithm that can tell a person whether they might have the virus by the sound of their cough.

The Massachusetts Institute of Technology (MIT) lab has built an algorithm that can perceive coughing sounds that are characteristic to COVID-19 even when they are not audible to humans.

“The way you produce sound changes when you have Covid, even if you’re asymptomatic.”, detailed MIT scientist Brian Subirana in the IEEE Journal of Engineering in Medicine and Biology.

To train the AI to make this distinction, MIT collected around 70,000 audio samples, each containing a number of coughs. Among those, there were 2,500 from confirmed coronavirus victims.

What was its success rate? A staggering 98.5% among people who had been officially diagnosed with COVID-19 and 100% in the cases of those who did not display other symptoms.

Naturally, MIT isn’t suggesting replacing medical tests with their algorithm to reach a valid diagnosis. Instead, scientists believe it could be a quick way to screen students or workers daily “or for pool testing to quickly alert of outbreaks in groups”.

The easiest way to provide access to their AI for large groups of people is via app but for that to happen, they would need regulatory approval.

Meanwhile, MIT isn’t the only research center to use technology with this purpose in mind. Cambridge University and Carnegie Mellon University have been doing the same, with Cambridge’s Covid-19 Sounds research having an 80% success rate this past July.