AI can detect heart failure from ECG, study finds

Artificial intelligence (AI) can accurately identify different types of left ventricular dysfunction and heart failure with preserved ejection fraction (HFpEF) using electrocardiogram (ECG) data, according to a new study published in the Journal of the American Heart Association.

Researchers at Wake Forest University developed two AI-powered ECG models and evaluated them using a large dataset of nearly 1.08 million digital ECGs from more than 165,000 patients. The algorithms were also validated in a pediatric cohort comprising 42,880 patients with 72,832 ECG recordings.

The AI models accurately detected reduced left ventricular ejection fraction (LVEF), one of the key indicators of heart failure. Using standard 12-lead ECGs, the models achieved 80% to 90% accuracy in identifying different types of left ventricular dysfunction. A model based on single-lead ECG recordings produced comparable results.

The researchers found that conventional machine learning models relying only on patients' clinical data performed less well when applied to different patient populations. Moreover, adding routine clinical information to the AI-based ECG analysis did not significantly improve diagnostic performance compared with using ECG data alone.

According to the study authors, the technology could become a low-cost tool for the early detection of heart disease, particularly because single-lead ECGs can be recorded using portable and wearable devices.

"AI-enabled ECG analysis can accurately identify left ventricular dysfunction and heart failure with preserved ejection fraction, even from a single-lead recording, opening new opportunities for low-cost, large-scale cardiovascular screening," the researchers concluded.

Follow NEWS.am Medicine on Facebook and Twitter