AI Can Accurately Detect Left Ventricular Dysfunction From Electrocardiograms

FRIDAY, July 31, 2026 (HealthDay News) — Artificial intelligence (AI) analysis of electrocardiogram (ECG) data can simultaneously detect subtypes of left ventricular (LV) dysfunction and heart failure with preserved ejection fraction (HFpEF), according to a study published online July 28 in the Journal of the American Heart Association.

Ibrahim Karabayir, Ph.D., from Wake Forest University in Winston‐Salem, North Carolina, and colleagues developed two ECG-AI models and validated them using data from nearly 1.08 million digital ECGs from 165,243 patients, as well as an external pediatric cohort of 72,832 ECGs from 42,880 patients.

The researchers found that for the 12‐lead ECG‐AI model, areas under the curve were 0.90 for reduced LV ejection fraction (rEF; EF<40), 0.81 for midrange EF (mEF; 40≤mEF<50), and 0.80 for HFpEF. In the single-lead model, the corresponding values were 0.89, 0.78, and 0.75. In the validation cohort, areas under the curve were 0.92, 0.76, and 0.73, respectively, for the 12-lead, and 0.90, 0.75, and 0.74, respectively, for the single-lead. Clinical data‐only machine‐learning models lacked generalizability, while clinical + ECG‐AI models did not show any significant improvement versus ECG‐alone models. Based on 8,418 ECGs (142 cases and 8,276 controls) from the pediatric dataset, the 12‐lead model achieved areas under the curve of 0.97, 0.71, and 0.64 for rEF, mEF, and HFpEF, respectively, while the single-lead model had areas under the curve of 0.94, 0.77, and 0.67, respectively.

“ECG‐AI can accurately detect LV dysfunction and HFpEF even from single‐lead ECG, enabling potential low‐cost screening,” the authors write.

Abstract/Full Text