AI can detect asymptomatic COVID-19 infections by listening to a cough

Shane McGlaun - Nov 2, 2020, 5:57am CST
AI can detect asymptomatic COVID-19 infections by listening to a cough

MIT has invented a tool that uses artificial intelligence to identify those with asymptomatic COVID-19 infections. Asymptomatic people are the most difficult to detect and can spread the infection without knowing that they have it. Since they are asymptomatic, they are unlikely to seek testing and continue their daily routine unknowingly infecting others.

MIT’s research found that despite being asymptomatic, some changes go along with COVID-19 infections. Researchers found that asymptomatic people could still differ from healthy individuals in the way they cough. While the changes aren’t detectable using the human ear, they can be discerned by AI by analyzing recorded coughs.

The system takes a recording of a forced cough that people voluntarily submit via web browsers, cell phones, or laptops. The team trained their AI using tens of thousands of samples using both coughs and spoken words. When cough recordings were fed to the AI, it identified 98.5 percent of coughs from people confirmed to have COVID-19 infections.

It was able to identify 100 percent of coughs from asymptomatic individuals reporting no symptoms but having tested positive for the virus. The team is currently incorporating its AI model into an app that is seeking FDA approval. If the app is approved, it would be a free and easy prescreening tool to identify those who are likely to be asymptomatic COVID-19 sufferers.

MIT says that using the app, users can test themselves daily to get information on if they might be infected and if they should confirm that with a formal test. There is no indication of when or if MIT’s system will receive FDA approval. Presumably, the FDA will want to rush the app through the approvals process to help curb the spread of COVID-19.


Must Read Bits & Bytes