AI Predicts Sudden Cardiac Arrest Days Before It Strikes

AI can now predict deadly heart arrhythmias up to two weeks in advance, potentially transforming cardiac care.

Artificial intelligence could play a key role in preventing many cases of sudden cardiac death, according to a new study published in the European Heart Journal. Researchers from Inserm, Paris Cité University, and the Paris public hospital group (AP-HP), in collaboration with colleagues in the United States, have developed an artificial neural network modeled after the human brain.

By analyzing data from over 240,000 ambulatory electrocardiograms (ECGs), the AI algorithm was able to identify patients at high risk of experiencing a life-threatening arrhythmia capable of triggering cardiac arrest within two weeks—with an accuracy exceeding 70%.

Sudden cardiac death claims more than 5 million lives globally each year. Often, these fatal events occur without warning, affecting individuals with no prior diagnosis of heart disease.

This study highlights the potential of AI to significantly improve the early detection of dangerous arrhythmias, abnormal heart rhythms that can lead to sudden cardiac arrest, by identifying at-risk patients before symptoms appear.

Developing an AI That Mimics the Brain

As part of this study, a network of artificial neurons was developed by a team of engineers from the company Cardiologs (Philips group) in collaboration with the universities of Paris Cité and Harvard. What this algorithm does is imitate the functions of the human brain in order to improve the prevention of cardiac sudden death.

The researchers analyzed several million hours of heartbeats thanks to data from 240,000 ambulatory electrocardiograms collected in six countries (USA, France, UK, South Africa, India, and Czechia).

Thanks to artificial intelligence, the researchers were able to identify new weak signals that herald the risk of arrhythmia. They were particularly interested in the time needed to electrically stimulate and relax the heart ventricles during a complete cycle of cardiac contraction and relaxation.

New Predictive Signals Identified

“By analyzing their electrical signal for 24 hours, we realized that we could identify the subjects susceptible of developing a serious heart arrhythmia within the next two weeks. If left untreated, this type of arrhythmia can progress towards a fatal cardiac arrest,” explains Dr. Laurent Fiorina, first author of the study, researcher at the Paris Cardiovascular Research Centre (PARCC) (Inserm/Paris Cité University),” cardiologist at Cardiovascular Institute Paris-Sud (ICPS) (Ramsay, Massy), and medical director in charge of artificial intelligence at Philips.

While the artificial neural network is still in the evaluation phase, it showed itself in this study to be capable of detecting at-risk patients in 70% of cases, and no-risk patients in 99.9% of cases.

In the future, this algorithm could be used to monitor at-risk patients in hospitals. If its performance is refined, it could also be used in devices such as ambulatory Holters that measure blood pressure to reveal hypertension risks. It could even be used in smartwatches.

A Paradigm Shift in Cardiac Risk Prediction

“What we’re proposing here is a paradigm change in the prevention of sudden death, “comments Eloi Marijon, Inserm research director at PARCC (Inserm/Paris Cité University), professor of cardiology at Paris Cité University and head of the cardiology department at Georges Pompidou European Hospital AP-HP.

“Until now we’d been trying to identify patients at risk over the medium and long term, but were incapable of predicting what could happen in the minutes, hours, or days that precede a cardiac arrest. Now, thanks to artificial intelligence, we can predict these events in the very short term and potentially take action before it’s too late.”

The researchers now wish to conduct prospective clinical studies to test the efficacy of this model under real-world conditions.

“It’s essential for this technology to be evaluated in clinical trials before being used in medical practice,” insists Dr. Fiorina. “But what we’ve already shown is that AI has the potential to radically transform the prevention of serious arrhythmias.”

Reference: “Near-term prediction of sustained ventricular arrhythmias applying artificial intelligence to single-lead ambulatory electrocardiogram” by Laurent Fiorina, Tanner Carbonati, Kumar Narayanan, Jia Li, Christine Henry, Jagmeet P Singh and Eloi Marijon, 30 March 2025, European Heart Journal.
DOI: 10.1093/eurheartj/ehaf073