Identifying advanced heart failure usually requires a lot of specialized equipment, like a cardiopulmonary exercise test (CPET), to see how a patient’s heart handles physical stress. It works well, but you usually need to be at a major medical center with highly trained staff to get it done. Because of this hurdle, many of the 200,000 people in the U.S. living with advanced heart failure don’t get the specific care they need.

Researchers from Cornell, Columbia, and New York-Presbyterian are looking to change that by using AI to analyze standard cardiac ultrasounds. Instead of a complex exercise test, their new model looks at ultrasound images, heart valve dynamics, and electronic health records to predict a patient’s risk level.

Efficiently Spotting Heart Disease

The study, recently published in npj Digital Medicine, showed that this AI-powered method is surprisingly accurate. It predicted “peak oxygen consumption,” the main metric used to flag high-risk patients, with about 85% accuracy. This is a big deal because ultrasounds are already a routine part of heart care and are much easier to access than specialized exercise labs.

“This opens up a promising pathway for more efficient assessment of patients with advanced heart failure using data sources that are already embedded in routine care,” said senior author Fei Wang, a professor at Weill Cornell Medicine.

The project started by asking over 40 heart specialists where they thought AI could help. This collaboration led to a model that can process moving images of the heart alongside text-based medical records to find patterns that might be hard for a human to spot quickly.

Tech and Medicine

What makes this project unique is how the two fields influenced each other. Deborah Estrin, a professor at Cornell Tech, noted that the needs of the doctors actually pushed the AI researchers to create brand-new techniques.

“The close interaction between clinicians and AI researchers on this project ended up driving the development of new AI techniques that would not have been explored otherwise,” Estrin said. “So, this was a case of medicine shaping the future of AI – not just AI shaping the future of medicine.”

The team is already planning more clinical studies. If the FDA eventually clears the tool for routine use, it could help doctors catch high-risk cases much earlier.

Dr. Nir Uriel of NewYork-Presbyterian said, “If we can use this approach to identify many advanced heart failure patients who would not be identified otherwise, then this will change our clinical practice and significantly improve patient outcomes and quality of life.”