A major advance in catching heart failure early
A team of researchers from Wake Forest University School of Medicine has developed an artificial intelligence model. It shows promise in detecting heart failure. This includes a form often overlooked during standard examinations. The AI can analyze data from a single ECG lead. This is similar to input from wearable devices. It opens new possibilities for accessible heart health assessments.
Identifying the condition early is crucial, but traditional evaluations require specialized imaging tests that are not always available. With the help of AI, doctors could soon screen patients more efficiently using a standard electrocardiogram (ECG).
How the AI model identifies heart issues
The study introduces an innovative AI tool. It uses data from a standard ECG to detect three types of heart dysfunction. These include reduced ejection fraction (rEF). In this case, the heart’s main chamber fails to pump blood effectively. It also includes mildly reduced ejection fraction (mEF). And heart failure with preserved ejection fraction (HFpEF). This form is especially hard to detect. It is often missed during routine evaluations.
Ejection fraction is a measure of how much blood the heart pushes out with each beat. The AI model was trained on more than 1 million ECGs from Atrium Health Wake Forest Baptist. It was then tested using over 72,000 ECGs from the University of Tennessee Health Science Center to see how well it performed in different patient groups.
What sets this AI apart is its ability to work with just a single ECG lead. This is the same kind of data captured by many smartwatches. It is also captured by wearable ECG devices. The researchers believe this could mean the model may one day be adapted for use in wearable technology. This could make heart health monitoring more accessible to the general public.
Testing the AI in real-world scenarios
The researchers tested two versions of the AI model—one using a 12-lead ECG and the other using a single lead. The 12-lead version was particularly effective at identifying reduced ejection fraction. While it performed slightly less well on the other two types of heart dysfunction, its results were still promising.
However, the team noted that the pediatric group was relatively small.
One of the key findings from the study is that the AI model generalized well across different demographic groups. This means it could be a valuable tool in diverse populations.
Oguz Akbilgic, Ph.D., is the corresponding author and professor of artificial intelligence. He works at the Department of Cardiovascular Medicine at Wake Forest University School of Medicine. He explained that the AI helps doctors detect subtle electrical patterns in the heart. These patterns are often invisible to the human eye. These insights can guide clinicians. They can decide whether a patient needs further heart evaluation.
Akbilgic also emphasized that many heart conditions can progress without obvious symptoms and may go unnoticed until they become more serious. By identifying these issues early, the AI could help doctors intervene before complications arise.
Looking ahead for AI in heart care
The team is currently testing the AI tool in a real-world healthcare environment to assess its impact on clinical decision-making and resource use. If successful, this technology could lead to more efficient heart health screening and earlier identification of at-risk patients.
The findings have been published in the Journal of the American Heart Association, highlighting the growing role of AI in medical diagnostics and treatment.