It looks like whenever you cough these days, everybody around you seems to get a mild heart attack.
Thankfully, the good news is here – Researchers from the MIT discovered a way to determine whether a person is infected with COVID-19 just by listening to the sound of their cough.
The tool is based on neural networks that can detect subtle changes in people’s cough, signs that they are infected, even if they don’t manifest other symptoms. Asymptomatic patients are a crucial element in spreading the disease, mostly because they are trickier to manage, mainly because they are less likely to get tested because they don’t feel abnormal.
Therefore, carriers may infect otherwise healthy individuals without realizing it.
It turned out that even asymptomatic carriers have one “symptom” that is proof they are infected, MIT researchers say, and it’s all in the cough.
Infected patients have a slightly, nearly undetectable altered cough that is imperceptible to the human ear.
However, AI and microphones can do what humans can’t! The service returned an accuracy of 98.5% from people with confirmed covid-19 cases, and 100% of coughs were from asymptomatic people.
One neural network analyzes sounds associated with vocal cord strength, while another one detects traces related to a patient’s emotional state, like frustration, which typically produces a “flat affect.”
A third network listens for subtle changes in respiratory performance. The three systems work in tandem to detect the traces of the deadly virus.
The algorithm isn’t necessarily a new idea, as researchers made similar attempts to detect other diseases like pneumonia and asthma just by analyzing the sound produced by coughing patients.