AI Detects Hidden ECG Marker of Sudden Cardiac Death

Sudden cardiac arrest is a lethal emergency caused by abrupt failure of the heart’s electrical system. Identifying who is at high risk remains difficult, leading to missed opportunities for implantable defibrillators and unnecessary procedures. The burden is significant, with more than 300,000 U.S. deaths each year. To help address this challenge, researchers have developed an artificial intelligence approach that detects a previously unrecognized electrocardiogram signal associated with sudden cardiac death.

Developed at the University of California (Berkeley, CA, USA), the deep learning model analyzes standard electrocardiogram (ECG) tracings to uncover waveform patterns linked to sudden cardiac death. The system detects a previously unrecognized ECG biomarker and stratifies risk directly from widely available clinical images. It is designed to support decisions about implantable cardioverter-defibrillators by identifying patients who appear low risk under current criteria.

The investigators trained the model on more than 440,000 ECGs from Sweden that were paired with death certificates. The model was exposed to tracings from healthy individuals, at-risk patients, and people who later suffered cardiac death until it learned patterns associated with subsequent events. Over multiple years, the team then tested the approach on thousands of additional patient files from the U.S. and Taiwan.

Performance exceeded current screening that relies on how much blood the heart ejects with each beat. Standard tests identify a high-risk group with a 4.6% annual rate of sudden cardiac death. The artificial intelligence system isolates a high-risk group with a 7% annual rate, capturing thousands more patients each year, many of whom look low risk by today’s standards. The model’s outputs are derived from ECG images that are already routine in hospitals.

According to the report, the findings were published in Nature on June 24, 2025. The data effort spanned about a decade and drew on programs at UC Berkeley, including the joint UCSF–UC Berkeley Computational Precision Health program. Next steps include deploying the algorithm across hospital ECG databases in Sweden, Taiwan, and the U.S., notifying flagged patients, and offering optional continuous monitoring with a wearable patch that could inform downstream decisions, including consideration of an internal defibrillator.

“One thing that makes the problem very tragic, but also very well suited for AI, is that we have the cure for this problem. If you knew you were one of the people who was going to drop dead, you would go to a cardiologist and you’d get a defibrillator implanted. The problem is that doctors can’t figure out who needs one before it’s too late,” said Ziad Obermeyer, an associate professor at UC Berkeley’s School of Public Health and the study’s lead author.

"In some fraction of those people, we could have prevented those deaths if we had just known it in time. There are a lot of lives being lost from people who are dropping dead of sudden cardiac death that are preventable if we just had better AI tools to find these things," added Obermeyer.

Related Links
UC Berkeley