AI drives new prevention efforts

Ever since Mass General Brigham’s Paul Dudley White, MD — considered by many to be the father of cardiology — first introduced the concept of preventative care in the late 1940s, our physician scientists have been finding innovative ways to predict risk and prevent heart disease.

Recently, faculty and staff in the Mass General Brigham Heart and Vascular Institute have been applying artificial intelligence (AI), or machine learning, to unlock new methods of identifying people at risk.

Location matters

Excess fat is known to drive cardiometabolic diseases like heart disease and stroke. Traditional measures like body mass index can’t discern fat from muscle or detect where the fat is located, making it difficult to assess a person’s risk. A team of imaging specialists, co-led by Vineet K. Raghu, PhD, in the Heart and Vascular Institute, has developed an AI tool to accurately measure body composition using MRI data. Using the tool, the team has shown that individuals with fat surrounding their internal organs and with fat deposits in muscle tissue have a higher risk of developing cardiovascular disease.

Routine power

Emily Lau, MD, MPH, co-director of the Women’s Heart Health Program; J. Sawalla Guseh, MD, director of the Cardiovascular Performance Program; and Shaan Khurshid, MD, MPH, from the Telemachus and Irene Demoulas Family Foundation Center for Cardiac Arrhythmias, have leveraged machine learning to identify women who are at high risk for cardiac complications during pregnancy. The team was able to accurately assess cardiovascular fitness in individuals by applying an AI model to the data from a routine electrocardiogram. Those with lower fitness levels showed a higher association with pregnancy-related cardiac complications. This method may help doctors to identify patients who would benefit from preventative screening and increased screening and, ultimately, to save lives.

Residual savings

Patients who have suffered one heart attack, stroke, or other cardiovascular event are at increased risk of having a second — but some people are at higher risk than others. Drawing on data from tens of thousands of patients from two independent databases, Min Seo Kim, MD, and a team of investigators from the Heart and Vascular Institute and elsewhere used machine learning to develop a new scoring system designed to gauge a person’s residual risk of experiencing a subsequent episode in the following 10-year period. This personalized model will improve physicians’ abilities to provide guidelines for follow-up care.