COVID-19 MIT cough test

If anything can help us tackle the spread of a novel virus head-on, it is technology — a medium that truly levels the playing field.

In the latest example, Massachusetts Institute of Technology (MIT) researchers have developed an AI algorithm that diagnoses you for COVID-19 by merely listening to your coughing sound.

Upon receiving approval from the Food and Drug Administration (FDA), MIT researchers would soon integrate the algorithm in a mobile app, allowing users to test for COVID-19 by only coughing close to their phone.

For their research, MIT scientists leveraged the power of data by analyzing 70K+ cough audio samples, out of which about 2500 were from confirmed COVID-19 patients. Using this collection of audio, they developed an AI algorithm that can tell if you have contracted the Coronavirus based on your cough’s sound.

During the testing stage, the researchers recorded a success rate of 98.5% when tested on COVID-19 patients. The same test recorded an accuracy of 100% in identifying COVID-19 positive people among an asymptomatic group.

Although the data pool is not significant, the highly accurate results make a strong case for the test’s reliability. Once the FDA gives its thumbs-up, we could see this test transform into an app, making it extremely easy for us to regularly carry out a self-test.

The “cough test” is especially promising for detecting asymptomatic Coronavirus carriers who are hard to track. Quarantining such patients quickly is essential to prevent the virus from spreading.

As the test gains popularity, offices, restaurants, and other public places might even adopt it as part of a daily check to further curb Coronavirus’s spread.

Priye is a tech writer at Fossbytes, who mainly covers games but also writes about anything remotely related to tech, including apps, phones, CPUs, and GPUs. He prefers to be called a "video game journalist" and grimaces when he doesn't get to be "Player 1." If you want to talk about games or send any feedback, drop him a mail at [email protected]