AI can identify patients at risk of dying—of any cause—within the next year or of developing an irregular heartbeat, according to two new studies to be presented at the American Heart Association’s Scientific Sessions 2019.
The same group of researchers carried out both studies, using electrocardiogram (ECG) results of more than 2 million patients to train deep neural networks.
“This is exciting and provides more evidence that we are on the verge of a revolution in medicine where computers will be working alongside physicians to improve patient care,” lead author Brandon Fornwalt, MD, PhD, of Geisinger Medical Center in Danville, Pennsylvania, said in a prepared statement.
For one study, the authors found that their neural network could accurately predict a patient’s risk of death—of any cause—within the next year. A team of three cardiologists examined the ECG data separately, but were “generally unable to recognize the risk patterns that the neural network detected.”
“This is the most important finding of this study,” Fornwalt said in the same statement. “This could completely alter the way we interpret ECGs in the future.”
For the second study, the researchers successfully trained a deep neural network to predict when patients were at an escalated risk of atrial fibrillation (AF) before it even developed. The network was able to analyze 30,000 different data points for each ECG.
“Currently, there are limited methods to identify which patients will develop AF within the next year, which is why, many times, the first sign of AF is a stroke,” senior author Christopher Haggerty, PhD, also from Geisinger Medical Center, said in the prepared statement. “We hope that this model can be used to identify patients with AFvery early so they can be treated to prevent stroke.”
The American Heart Association’s Scientific Sessions 2019 are scheduled for Nov. 16-18 in Philadelphia. More information is available here.