����ɫ��

Imagine if doctors had a tool that could detect disease, predict a person’s risk of developing a disease and determine how someone will respond to a specific treatment. 

It may sound like a futuristic dream, but by harnessing the power of artificial intelligence (AI) techniques, it may be possible to provide tomorrow’s health care, today. 

Dr. Na Li, PhD, a data and statistical science researcher at the Cumming School of Medicine, received a Canadian Institutes for Health Research (CIHR) project grant to build Canada’s first federated learning framework for disease surveillance. 

“Federated learning is a fairly new concept in medical health research. It enables numerous devices to work together to share knowledge without transferring or exposing any personal data that would compromise patient privacy,” says Li. “It will be a significant step in digital health care in Canada, and I am definitely excited about the project.”

There is enormous potential to use mountains of data — captured in routine tests like cardiac MRIs and blood analysis and stored in electronic medical records (EMR) across Canada —  to create powerful detection tools with staggering potential for improving health outcomes. While AI is a fixture of modern life, its potential remains relatively untapped in health care.

Privacy concerns addressed

This is due to privacy laws and data regulations that are critical to protecting the individuals’ confidentiality but make it difficult to share data across jurisdictions; therefore, they create barriers to improving health outcomes. Federated learning solves the privacy concern as none of the data is attached to a person. 

Li will build a system to aggregate and analyze medical data from electronic EMRs in Calgary, Sherbrooke, Que., and Ottawa, Ont. Once the framework is built, Li’s team will evaluate its performance to detect acute heart attacks, hypertension and sepsis at the trio of sites. All of this will occur without sharing patient information. 

“The goal of this project is to develop reliable tools that can detect disease and support decision-making at the clinical level,” she says. “This project has the potential to revolutionize the field of health research by allowing for collaborative EMR machine learning while addressing data privacy concerns. The ability to improve model generalizability and adapt to other conditions is also promising.

Overall, this project could have a significant impact on the health-care industry and ultimately benefit patients.

Dr. James White, MD, who leads the Libin Cardiovascular Institute’s Precision Medicine Initiative, is excited about the potential of this project. 

“Advancements in federated learning, such as those being developed by Dr. Li, allow sites to combine knowledge from health-care data without the historic risks to patient privacy, expanding our ability to solve big health-care problems using data across diverse patient groups,” says White.

“This has the potential to accelerate the development of intelligent health-care systems that detect disease earlier and improve outcomes through personalized care. Precision medicine projects like Dr. Li’s are bringing us one step closer to providing tomorrow’s health care today.”