The American Heart Association’s new AI Assessment Lab has published its first report, and the verdict is upbeat for Ultromics, the Oxford company behind the EchoGo Heart Failure tool. The FDA-cleared software reads standard echocardiogram videos to flag patients who may have heart failure with preserved ejection fraction, or HFpEF, a common condition that is frequently missed.
Researchers analyzed roughly 90,000 real-world echocardiograms from Dandelion Health and paired that retrospective data with clinical and economic modeling over five years. The report estimates EchoGo could identify HFpEF up to 263 days earlier than standard care among patients whose condition would otherwise slip through. For every 10,000 patients evaluated, earlier treatment could prevent 477 deaths, along with 406 fewer hospital admissions, 501 fewer readmissions and 564 fewer emergency department visits.
The assessment matters beyond the numbers. The AHA lab was built to judge clinical AI on more than sensitivity and specificity, folding in patient outcomes, workflow effects, health equity and economics. HFpEF diagnosis is especially delayed in women and people of color, and the report suggests younger and non-white patients may gain the most from the technology.
“Diagnostic accuracy is what earns the AI a seat at the table,” said Roger Owens, chief commercial officer at Ultromics, adding that accuracy alone should not be the final measure. “AI should help close these gaps, not inadvertently reinforce them.”
The findings come with caveats: performance needs continued monitoring across different populations and care settings. But as a template for evaluating AI in cardiology, the report gives hospitals a model for asking not just whether a tool is accurate, but whether it changes outcomes and who benefits.
