AI Finds Hidden Heart Signal That May Predict Sudden Cardiac Death

Cardiovascular disease remains the world's biggest killer, claiming an estimated 19.8 million lives in 2022, according to the World Health Organization (WHO). While heart attacks and strokes account for most of these deaths, sudden cardiac death often strikes without warning, sometimes in people who were unaware they had an underlying heart condition.
A sudden cardiac arrest is a condition where the heart's electrical system malfunctions without warning. Even young athletes or people with no history of heart issues could perish from this, not just high-risk older adults.

Using artificial intelligence (AI), researchers at University of California (UC) Berkeley have discovered a previously unknown warning sign of potential sudden cardiac death hidden inside a routine ECG also known as an EKG, that doctors were unaware of. An ECG (electrocardiogram) records the electrical signals pulsing through your heart. It translates these natural electrical impulses into a visible wave pattern on a screen or paper. Doctors use this information to measure your heart rate, check its rhythm, and look for signs of damage or disease.

A heart can pump normally yet still have dangerous electrical problems that lead to sudden cardiac arrest. Existing screening methods miss many patients. Doctors currently rely heavily on Left Ventricular Ejection Fraction (LVEF), which measures how well the heart pumps blood. However, many people who later die from sudden cardiac arrest have normal pumping function.

A Decade Of Research

It took researchers at UC Berkeley led by Associate Professor Ziad Obermeyer nearly a decade to compile the data for the study. Researchers linked over 440,000 ECGs from Sweden with death certificate data to train AI to recognize tiny waveform patterns associated with future sudden cardiac death. Then they validated it using patient data from the United States and Taiwan.

But the AI did more than simply predict who was at risk.

It identified a previously unknown pattern in ECG readings that appears to signal an increased likelihood of sudden cardiac death, one that cardiologists had not recognised before. In other words, the system didn't just analyse existing medical knowledge; it helped uncover a new biological clue hidden in one of medicine's most commonly used diagnostic tests.

"One thing that makes the problem very tragic, but also very well suited for AI, is that we have the cure for this problem," Obermeyer said. "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."

AI identified a group of patients with an estimated 7% annual risk of sudden cardiac death, compared with 4.6% under the current standard. Over 86% of patients flagged by AI would not have been identified using today's standard screening.

We already have implantable defibrillators that can prevent many sudden cardiac deaths. The challenge has been knowing who actually needs one and AI may help solve exactly that problem.
"I wouldn't suggest going out and getting a defibrillator implanted just because we say your ECG is high risk," Obermeyer said "What's nice about this is you don't have to believe the AI at all. You can just use it to target additional testing like doing traditional risk markers."