Scientists say they can distinguish the sound of a COVID-19 cough.
Researchers with the Embedded Systems Laboratory at the Swiss Federal Institute of Technology Lausanne are developing a diagnostic test on a smartphone app that works by "listening" to the sound of the user's cough.
The World Health Organization reports that two-thirds of COVID-19 patients exhibit a dry cough as opposed to the typical wet cough people often experience with the common cold or allergies.
Scientists have used cough sounds to diagnose whooping cough, asthma and pneumonia in the past. This new diagnostic test uses machine learning algorithms “to determine what kind of frequencies the cough is made up of,” says Lara Orlandic, a research assistant on the project.
The mobile app has not been developed yet, but volunteers can upload their cough recordings to the project’s website. Users are kept anonymous, but they are asked for permission to save their location, Orlandic says. The end goal is to share that information with local governments so they can understand where the virus is spreading.
Many experts say testing for the coronavirus is key to containing its spread and reopening the country. But there have been concerns about nasal swab testing delivering false negatives, and Orlandic says the acoustic testing method her team is developing is no different.
“Our system can never be 100% perfect because … there's that one-third of patients who are asymptomatic, so they don't show any cough symptoms at all,” she says. “What this app can do is it can serve as a sort of preliminary screening tool to first maybe put people at ease in case they're at home, wondering whether or not they have coronavirus.”
Right now, Orlandic says the team needs more data in order to make the method more accurate. This study is extra challenging because researchers can’t record people in the same room with the same device.
“You have people with all sorts of different devices from around the world in different — perhaps some noisy, some quiet — environments sending in their recordings,” she says. “We need lots of data to be able to train the algorithms to distinguish between the types of coughs despite all the additional noise.”
This segment aired on April 15, 2020.