Aug. 27, 2021

Can cough sounds help pre-screen for COVID-19? UCalgary researchers are developing an app for that

The multi-disciplinary team has a technical proof-of-concept. Now they need to do a clinical experiment and fine-tune the technology
CoughSoundAudioFile
A screen-grab of the audio files from cough sounds being analyzed by the UCalgary team. Dr. Behrouz Far

You’re at the grocery store or another public place, wearing a mask, when you let out an inadvertent sneeze or cough.

Others look at you, and you joke about how “it’s just allergies” and “it’s not COVID.”

That situation has played out for many over the last 18 months. Researchers at the University of Calgary’s Schulich School of Engineering are hoping a technological tool will help confirm the diagnosis.

Dr. Behrouz Far, PhD, is a professor in the Department of Electrical and Software Engineering and part of a team that recently had a paper published by Nature about analyzing cough sounds through a mobile app.

He says a multi-disciplinary team of researchers including Dr. Amir Sanati-Nezhad, PHD; Dr. Mohammad Keyhani, PhD; and Dr. Hossein Hejazi, PhD, helped bring the project to life.

LEARN MORE: Biomedical Engineering at UCalgary

“We wanted to know whether there was a fast way of pre-screening COVID-19 that everyone can use at home,” says Far. “We wanted to join the hands of machine learning with the sensor experts and clinicians to come up with a fast and low-cost solution.”

An app for that

Work on the project began with the collection of cough samples from around the world in the spring of 2020.

People were able to use their cellphone microphone to record their cough and then used the app to input the data.

“We have a technical proof-of-concept that different cough types and sequences of coughs can be attributed to different diseases,” says Far. “The technology will indicate to the patient that there is a high probability of being infected and recommends getting a further swab or other tests.”

He adds more than 70 per cent of swab-based tests could be eliminated by using the app, making it more efficient for lab resources and saving money in the long-run.

“This technology could be used at home for personal use as well as in some high-density, in-and-out facilities like workplaces, airports and sports stadiums,” says Far.

Not only would the app work for people who have some of the symptoms of COVID-19, but it could also test asymptomatic users.

Pivoting with new information

Developing the app came with plenty of challenges, as information about COVID-19 changed on an almost-daily basis.

Masoud Karimi Fatemi is a PhD student who worked on the first version of the app and put together the initial cough-analysis program.

“Of course, any software is always subject to change, but this one in particular changed more than I had expected,” Fatemi admitted. “It was because our understanding of COVID-19 wasn’t perfect. We would gather new information about the pandemic every other day.”

Data security and privacy was also a concern as the team continued its work.

Dr. Emad Mohammed, PhD, worked with the group by creating the artificial intelligence module that processes the cough sound and makes the diagnosis.

“The community spread of COVID-19 has changed people’s perspectives on seeking help when they suspect they have contracted the virus,” says Mohammed. “This AI-powered tool will help people feel safe during the time of uncertainty, especially as they might be afraid to visit healthcare facilities or see loved ones.”

Work in progress

There are still a few steps left before the technology can be made available to the public.

“We need to do a clinical experiment with the help of Alberta Health Services to further tune the technology to real cases,” says Far. “We also need to further develop the screening app so that everyone can use it daily to verify if they have the virus.”

Not only are the researchers excited about being a part of this project, but they believe that it could open more doors.

Far says machine-learning-based technologies are quickly becoming useful tools in diagnosing other diseases and conditions such as Alzheimer’s, concussions, prostate cancer and uveal melanoma.