HMN 2025: How Using AI and wearable sensors to detect inflammation before symptoms appear

Modern medicine is largely reactive—treating illness only after symptoms emerge. But a new study from the Research Institute of the McGill University Health Center (The Institute) and McGill University points to a more proactive future: one where silent signs of infection are detected before we even feel sick.

In a world first, this study has led to the development of an artificial intelligence (AI) platform that can accurately predict acute systemic inflammation—an early immune response to viral respiratory tract infections (VRTIs)—by analyzing biometric data from a smart ring, a smart watch or a smart shirt.

By detecting immune signals before symptoms appear, the system opens the door to earlier intervention, potentially saving lives and reducing health care costs by preventing complications and hospitalizations. This multidisciplinary work is published in The Lancet Digital Health.

“By the time an infection is detected based on clinical symptoms or PCR testing, it is generally already well underway,” said Dennis Jensen, Ph.D., senior author of the study, Scientist in The Institute’s Translational Research in Respiratory Diseases Program and Associate Professor in McGill’s Department of Kinesiology and Physical Education.

“By enabling rapid, personalized and objective early warning of systemic inflammatory events due to viral respiratory infections, our predictive tool gives patients and health care providers the chance to intervene early before critical health events occur.”

A proactive approach to infection detection

Acute systemic inflammation is a rapid and intense inflammatory reaction that occurs throughout the body in response to an infection or injury. Although it often resolves on its own, this natural defense mechanism can sometimes lead to serious complications such as organ damage or failure, and even death. This is especially true for vulnerable populations, such as people with chronic obstructive pulmonary disease (COPD).

To simulate real-world infection, the team administered a live attenuated influenza vaccine to 55 healthy adults aged 18–59 who were followed from seven days before inoculation to five days after inoculation.

For the period of the study, participants wore three commercially available wearable devices simultaneously—a ring, a watch and a shirt—allowing for the continuous monitoring of several physiological and activity measures, including heart rate, heart rate variability, body temperature, respiratory rate, blood pressure, physical activity and sleep quality.

The researchers also measured systemic inflammatory biomarkers via repeated blood tests, performed PCR testing for respiratory pathogens and used a smartphone app to collect self-reported symptoms.

In total, over 2 billion data points were collected to train machine-learning algorithms. These algorithms were then used to build different AI models: nine used subtle physiological changes to predict surges in systemic inflammation, while one relied solely on reported symptoms.

The model using the fewest features was chosen for further development because it was considered more practical for everyday monitoring. It was still effective, with close to 90% sensitivity, meaning it correctly predicted nearly 90% of actual positive cases.

“Previous studies have suggested a link between physiological signals measured with wearable sensors and subtle immune activity,” says Amir Hadid, Ph.D., the first author of the study, who was a postdoctoral research fellow at McGill at the time of the study. “Our study marks a significant step forward by using AI to translate these invisible signals into a real-time, accurate early warning system for acute inflammation.”

Outperforming traditional symptom reporting

All wearable-based models outperformed the symptom-based model. The authors of the study explain this by noting that some participants with systemic inflammation did not develop noticeable VRTI-related symptoms (false negatives), and some participants without systemic inflammation reported symptoms (false positives)—a phenomenon known as the nocebo effect.

Remarkably, the algorithms also successfully detected systemic inflammation in four participants who were infected with SARS-CoV-2 during the study. In each case, the algorithms flagged the immune response before symptoms appeared or PCR testing confirmed the infection.

“In future clinical validation studies, our system aims to detect systemic inflammation caused by other common viruses, like rhinovirus, respiratory syncytial virus (RSV) or SARS-CoV-2, using only wearable devices—no blood tests, no specialized hardware, no need for a health care visit,” says Dr. Emily McDonald, Scientist in The institute’s Infectious Diseases and Immunity in Global Health Program and Associate Professor in General Internal Medicine at McGill, whose research team helped run the study.