Even if a coronavirus patient has no obvious signs, their cough typically sounds slightly different from the cough of a healthy person due to the effect of the virus on the lungs and vocal cords of patients.
In order to design an algorithm that listens to these subtle variations, which are normally inaudible to the human ear, MIT scientists used artificial intelligence. According to a study which looked at thousands of user-submitted forced cough recordings over the spring, their model correctly identifies coronavirus patients 98.5 percent of the time, and it is accurate 100 percent of the time when patients are asymptomatic.