Scientists from the Massachusetts Institute of Technology (MIT) have developed an easy, quick way to determine whether individuals are healthy or Covid19 positive, but asymptomatic. The solution, still in its early stages, comes in the form of an algorithm taught to detect a Covid cough that is otherwise impossible for human ears to discern. MIT scientists say people who have Covid19 — even if they are asymptomatic — have a cough that is different from healthy individuals, which they have taught an algorithm to detect through “forced-cough recordings.”
People who are unaware they are Covid19 positive because they have no symptoms are thought to be responsible for much of the virus’s spread. The innovation could aid communities in limiting the spread of the virus while reopening gathering places like restaurants, pubs, offices, and educational institutions.
The researchers taught the algorithm by feeding it forced-cough sounds and, in some instances, spoken words from more than 70,000 people who provided 200,000 samples. They say the algorithm accurately identified a Covid cough — as in, a cough coming from a person who was known to be Covid19 positive — 98.5% of the time and identified asymptomatic people 100% of the time.
Related on The Swaddle:
If approved by the U.S. Food and Drug Administration (FDA), MIT scientists plan to launch the algorithm as a free phone app, which could enable people to simply fake-cough into their devices before deciding if they want to step out for a social event or go into their office or educational institution. The app would give the user instant information based on the sound of their cough, either issuing an all-clear or raising a red flag for the person to get a medical test and refrain from exposing others. The app, scientists stress, would not be a diagnostic tool, but merely a way to let a user know their cough sounds different from a healthy cough. Other infections, such as the flu, could also be a reason behind a red-flag cough. The app will at least instill a deeper state of vigilance in the user, scientists hope.
The MIT algorithm is the latest of many that have tried to map health sounds, including coughs and words, for diseases such as pneumonia. For Covid19 especially, Cambridge University’s Covid19 Sounds project found it could identify positive Covid19 cases 80% of the time based on breath and cough sounds. But the project’s source pool at the time of reporting was merely 459 samples; it has now increased to 30,000 people.
“It’s the same principle as feeding a machine a lot of X-rays so it learns to detect cancer,” artificial-intelligence expert Calum Chace told the BBC. “It’s an example of AI being helpful. And, for once, I don’t see a lot of downside in this.”