MIT AI could predict COVID-19 by listening to you cough

Finding new ways to stop the spread of COVID-19 has been harder than it should be. Part of this is due to people refusing to comply with things like social distancing and wearing masks. Another problem, however, is the fact that some people infected with COVID-19 have no symptoms.

These asymptomatic people can then pass the virus on to others – including those who are at risk of serious complications. Researchers at MIT believe they have found an ideal solution to identify asymptomatic cases.

The team developed an artificial intelligence (AI) program that can be used to determine whether a person has COVID-19 or not based on the sound of coughing. The research was published in the IEEE Journal of Engineering in Medicine and Biology.

In an impressive first study conducted with the new AI, 100 percent of asymptomatic patients were correctly identified from a record of their cough.

Surprising analytics

The idea of ​​diagnosing a complex medical condition based on the sound of coughing seems absurd. However, the team’s data supports the fact that the AI ​​approach is effective.

From April, the researchers began collecting audio recordings. They reportedly gathered more than 70,000 submissions. Considering that some of them have multiple coughs, the total number of coughs registered is over 200,000. Additionally, 2,500 of the total admissions were from people with COVID-19, including many who were asymptomatic.

From this data, the team used 2,500 COVID records and 2,500 non-COVID records. They trained the AI ​​on 4,000 recordings and used the remaining 1,000 to test whether it could reliably identify the people with COVID-19.

Amazingly, the model correctly identified 98.5 percent of positive COVID-19 records. In addition, 100 percent of the records sent by asymptomatic people were flagged.

Brian Subirana, MIT researcher, says, “We believe this shows that the way you make sound changes when you have COVID, even if you are asymptomatic.”

Unlikely origin

Oddly enough, the AI ​​program that is making waves right now didn’t emerge as a project related to COVID-19. Subirana and his team tried to develop an algorithm that could detect Alzheimer’s disease.

Most people associate Alzheimer’s with its degenerative neurological effects. However, it also leads to a deterioration in the neuromuscular system – especially the vocal cords. This theoretically makes it possible to hear changes in the vocalizations of someone with the disease.

The team first created a neural network to differentiate sounds based on the strength of the vocal cords. Then they created a second algorithm to identify emotional states from voice recordings. A third neural network was then created to analyze a cough for signs of decreased respiratory performance.

When the researchers put the three algorithms together, they found that combining biomarkers made it possible to diagnose the disease.

This year, after the COVID-19 pandemic, the team reapplied the final algorithm. Says Subirana, “We thought, why not try these Alzheimer’s biomarkers? [to see if they’re relevant] for COVID. ”

According to the results of the new study, the algorithm appears to be effective. Researchers are now working with a third party company to develop an app that will allow people to easily analyze their own cough.

In its article, the team writes: “Pandemics could be a thing of the past if the pre-screening tools are always activated in the background and continuously improved.”