AI-powered electrocardiogram detects early signs of heart failure

 

Interpreting relatively inexpensive electrocardiograms (ECGs) with an artificial intelligence (AI) algorithm accurately screened patients for a key precursor of heart failure in Kenya, a study led by UT Southwestern Medical Center researchers shows. The results, published in JAMA Cardiology, suggest AI-augmented ECG (AI-ECG) analysis could be a potential low-cost strategy for identifying patients who have underlying impairment in heart function.

"These findings support AI-ECG as a practical, scalable screening tool that can effectively identify individuals at risk for heart failure in resource-limited settings where access to echocardiography is constrained, addressing a critical gap in global cardiovascular care," said Ambarish Pandey, M.D., Associate Professor of Internal Medicine in the Division of Cardiology and in the Peter O'Donnell Jr. School of Public Health at UT Southwestern. He has a secondary appointment in Internal Medicine's Division of Geriatric Medicine.

In addition to Dr. Pandey, the study's primary lead author, other investigators included Neil Keshvani, M.D., Adjunct Assistant Professor of Internal Medicine at UT Southwestern, and Bernard Samia, M.B.Ch.B., M.Med., M.P.H., consultant physician and cardiologist at M.P. Shah Hospital in Kenya and President of the Kenya Cardiac Society.

Heart failure, a chronic condition in which the heart is unable to pump enough blood to meet the body's needs, is on the rise globally. The burden is particularly severe in sub-Saharan Africa, where health care resources are limited and patients develop heart failure at younger ages and face worse outcomes despite having fewer complicating conditions compared with patients in developed countries.

Before suffering heart failure, many patients develop precursor conditions such as left ventricular systolic dysfunction (LVSD), in which the heart's left ventricle doesn't pump blood effectively. Echocardiograms, which create images of the heart using ultrasound, are the gold standard for diagnosing LVSD and other heart failure antecedents, Dr. Pandey said. But these tests are extraordinarily expensive, and developing countries typically lack the equipment and expertise to perform them, he added.

To address this disparity, Dr. Pandey and his colleagues evaluated the use of AI-ECG, in which a typical ECG—a test of the heart's electrical function—is enhanced by an AI algorithm that searches for evidence of LVSD and other heart failure precursors. AI-ECG has shown promising results when tested in developed countries; it has rarely been evaluated in a developing country.

The team recruited nearly 6,000 patients seeking routine clinical care from eight health care facilities in Kenya to receive AI-ECG. A subset of this group, totaling 1,444 patients, also received echocardiograms to verify their AI-ECG results.

The AI algorithm identified LVSD in 14.1% of those who also received echocardiograms. AI-ECG had a 99.1% negative predictive value, meaning nearly all patients whose results reflected no evidence of LVSD were confirmed negative by echocardiography.

Positive AI-ECG screening in the study was strongly associated with other markers of adverse cardiac remodeling, including left ventricular hypertrophy and diastolic dysfunction. The algorithm demonstrated a high level of sensitivity, correctly identifying 95.6% of people who had LVSD, while also showing high specificity in accurately identifying 79.4% of people who did not have the condition.

The authors said their findings support the use of AI-ECG as a screening tool for LVSD in resource-limited settings where systematic echocardiographic screening is not feasible.