McGill University team develops AI that can detect infection before symptoms appear

CTV News

2 days ago

Participants wore a smart ring, a smart watch, and a smart T-shirt that monitored multiple physiological parameters and activities. In the photo, an Apple Watch in 2015 in Tokyo. (Koji Sasahara, File/AP Photo)
MONTRÉAL — Researchers at McGill University say they developed an artificial intelligence platform that can predict when someone is about to come down with a respiratory tract infection before they start to feel sick.
In what researchers are calling a 'world first,' the study involved participants who wore a ring, a watch and a T-shirt, all of which were equipped with censors that recorded their biometric data. By analyzing the data, researchers were able to accurately predict acute systemic inflammation — an early sign of a respiratory infection such as COVID-19.
Published in The Lancet Digital Health, the study says the AI platform can one day help doctors address health problems much earlier than they normally would, particularly in patients who are fragile and for whom a new infection could have serious consequences. It could also potentially reduce costs for the health-care system by preventing complications and hospitalizations.
'We were very interested to see if physiological data measured using wearable sensors … could be used to train an artificial intelligence system capable of detecting an infection or disease resulting from inflammation,' explained the study's lead author, Prof. Dennis Jensen of McGill University's department of kinesiology and physical education.
'We wondered if we could detect early changes in physiology and, from there, predict that someone is about to get sick.'
Jense says the AI model his team created is the first in the world to use physiological measures — including heart rate, heart rate variability, body temperature, respiratory rate, blood pressure — rather than symptoms, to detect a problem.
Acute systemic inflammation is a natural defence mechanism of the body that usually resolves on its own, but it can cause serious health problems, especially in populations with pre-existing conditions.
'The whole idea is kind of like an iceberg,' Jensen said. 'Kind of when the ice cracks the surface, that's like when you're symptomatic, and then it's too late to really do much to treat it.'
During the study, McGill researchers administered a weakened flu vaccine to 55 healthy adults to simulate infection in their bodies. The subjects were monitored seven days before inoculation and five days after.
Participants wore a smart ring, smart watch, and a smart T-shirt simultaneously throughout the study. As well, researchers collected biomarkers of systemic inflammation using blood samples, PCR tests to detect the presence of respiratory pathogens, and a mobile app to collect symptoms reported by participants.
In total, more than two billion data points were collected to train machine learning algorithms. Ten different AI models were developed, but the researchers chose the model that used the least amount of data for the remainder of the project. The chosen model correctly detected nearly 90 per cent of actual positive cases and was deemed more practical for daily monitoring.
On their own, Jensen said, none of the data collected from the ring, watch, or T-shirt alone is sensitive enough to detect how the body is responding.
'An increase in heart rate alone may only correspond to two beats per minute, which is not really clinically relevant,' he explained. 'The decrease in heart rate variability can be very modest. The increase in temperature can be very modest. So the idea was that by looking at … several different measurements, we would be able to identify subtle changes in physiology.'
The algorithms also successfully detected systemic inflammation in four participants infected with COVID-19 during the study. In each case, the algorithms flagged the immune response up to 72 hours before symptoms appeared or infection was confirmed by PCR testing.
Ultimately, the researchers hope to develop a system that will inform patients of possible inflammation so they can contact their health-care provider. 'In medicine, we say that you have to give the right treatment to the right person at the right time,' Jensen said.
By expanding the therapeutic window in which doctors can intervene, he added, they could save lives and achieve significant savings by avoiding hospitalizations and enabling home management of chronic conditions or even aging.
'In a way, we hope to revolutionize personalized medicine.'
This report by The Canadian Press was first published July 30, 2025.
By Jean-Benoit Legault, The Canadian Press