AI can tell the presence of COVID-19 through a cough

COVID-19 is still a huge problem as of late with major cities around the world in lockdown due to the resurgence of the infection. One factor that help spread the virus is that it is undetected until it’s too late.

There are literally a dozen signs and symptoms of COVID-19 and cough is most likely a common from those infected, but how can we distinguish them from a typical Flu?

A team of MIT researchers have developed an AI that may help us out with that as it can potentially recognize force coughing from COVID-19 even from those who are asymptomatic. The system uses a neural networks that can detect subtle changes and patterns that’s distinguishable of COVID-19.

One AI neural network detects sounds through vocal strengths and another listens to the emotional states that shows neurological decline such as increased frustrations or what is known as “flat effect”. There’s also a third neural network that measures changes in respiratory performance. The system uses an algorithm that checks for muscular degradation and will shows someone’s state of health based on the data it receives.

From an early tests, the AI showed a lot of promise at the team trained its model after thousands of coughs and dialog samples. The new technique returned a 98.5 percent accuracy in determining cough from people with COVID-19 and a hundred people from those who were asymptomatic.

While the system were accurate, there were limits though as there might be other conditions that produces similar results. However, as a pre-screening tool for the virus, it’s one that might just help a lot.

The team is making a smartphone app that uses the system that we can have it installed on smartphone devices and let users conveniently check themselves at home. The researchers even suggest that it could end the pandemic if the tools is set to always listen in the background but that becomes a privacy concern for everybody.