Researchers aim to build first AI prediction models capable of analyzing coronary angiograms

Coronary artery disease (CAD) is the leading cause of death globally, affecting millions. In 2015, CAD was behind nearly nine million deaths in Canada, and it remains a leading cause of death and hospitalization in the country. 

Coronary angiography, an X-ray technology that involves using a catheter and special dye to view blockages and narrowing of the vessels in the heart, has been the gold standard used to assess CAD since the 1960s. However, data from test results is entered manually, which can be time consuming and may lead to mistakes. It also makes it difficult for researchers to include these results in artificial intelligence (AI) models, which are increasingly being used to help predict patient outcomes and in treatment decisions.

Health AI researcher Dr. Joon Lee, PhD, a professor in the departments of Cardiac Sciences and Community Health Sciences at the Cumming School of Medicine, is leading a team that wants to change that.   

The team, made up of researchers from the University of Calgary, the University of Ottawa Heart Institute and the Montreal Heart Institute, was recently awarded a Michael and Terry Wilson Grant to develop the first AI-powered multi-modal patient outcome prediction models capable of analyzing coronary angiograms and electronic health record data. 

Lee is excited about the project and grateful for the funding. 

“This is an important project because our models will help support treatment decisions, which aren’t always straightforward,” says Lee. “We will be able to do a lot with this grant.” 

To develop the models, this project will expand an existing data set—previously built by Lee and his team—containing information from more than 60,000 catheterization procedures done in Alberta between 2009 and 2019. The models will be designed to predict major adverse cardiovascular events, such as heart attacks, stroke, and heart failure, at 90 days and one, three and five years after receiving a CAD treatment.

According to Lee, once updated to include angiogram images and other patient data, a similar data set will be extracted at the University of Ottawa Heart Institute to further refine and validate the models. 

“This partnership with Ottawa is important, because in machine learning, the bigger your data set is, the more generalizable and useful your prediction models will be,” says Lee. “It’s one thing to build something that works for your own site, but we want to create something that will work at as many sites as possible.” 

Following this work, Lee’s team will apply for grants to conduct multi-center trials to evaluate the clinical usefulness of their models in CAD treatment decision-making, integrate with clinical information systems and commercialize the technology. 

Dr. Joon Lee, PhD, is a professor in the departments of Cardiac Sciences and Community Health Sciences at the Cumming School of Medicine at the University of Calgary. He directs the Data Intelligence for Health Lab. He is a member of the Libin Cardiovascular Institute and the O’Brien Institute for Public Health.