US Edition
Researchers have developed a new artificial intelligence (AI) system which they claim can detect coronavirus by analysing the sound of people coughing.
Scientists at Massachusetts Institute of Technology (MIT) said the AI works because the virus causes temporary muscular impairment which can cause small differences to people’s speech or the sound of their cough – even if they have no other symptoms.
Tests performed on 4,000 recordings of people forcing themselves to cough showed that the system was able to reliably detect coronavirus, the researchers wrote in paper published the IEEE Journal of Engineering in Medicine and Biology.
When given new cough recordings, the AI was able to identify coughs from people who were confirmed to have Covid-19 98.5 per cent of the time. It was also able to identify coronavirus in those who were assymptomatic 100 per cemt of the time.
“We think this shows that the way you produce sound, changes when you have Covid, even if you’re asymptomatic,” Brian Subirana, a co-author of the paper and a research scientist in MIT’s Auto-ID Laboratory told MIT News.
The technology was adapted from the team’s original project of using AI to examine coughs and speech to detect whether people have Alzheimer’s, which causes weakened vocal cords.
The team now hopes to gain regulatory approval to release a free app that could be used to analyse cough samples around the world to detect asymptomatic coronavirus.
It is also collaborating with a number of hospitals to build a more diverse data set.
“The effective implementation of this group diagnostic tool could diminish the spread of the pandemic if everyone uses it before going to a classroom, a factory, or a restaurant,” Mr Subirana said.
UiPath, a software business, has also used AI in an attempt to reliably detect coronavirus. The business adapted its software earlier this year to analyse lung scans, potentially reducing diagnosis times by up to 50 per cent, the business said.