Artificial Intelligence Reveals Heart Secrets Before Sudden Cardiac Arrest
Sudden cardiac arrest is among the most dangerous medical emergencies, given its sudden onset and lack of preceding symptoms, making prediction akin to an intractable challenge. In the United States alone, more than 300,000 people lose their lives annually due to this lethal electrical malfunction, which can affect even healthy young individuals with no clear risk indicators. Although defibrillators capable of saving lives exist, the decision of who qualifies for implantation still relies on imprecise medical assessments.
In a promising scientific step, researchers from the University of California, Berkeley, have developed an artificial intelligence model capable of analyzing electrocardiograms (ECGs) with remarkable precision and detecting hidden electrical patterns invisible to the human eye. The team trained the model using more than 440,000 electrocardiograms from Sweden, linked to mortality data, enabling it to subsequently identify at-risk individuals with exceptional accuracy.
The results demonstrated clear superiority over traditional methods that rely on measuring blood-pumping efficiency. The AI identified a risk group with an annual mortality rate of 7%, compared to just 4.6% using standard methods, meaning thousands of additional cases previously considered low-risk were detected.
Researchers describe this technology as a “black box” for the heart, helping to uncover hidden patterns that precede sudden cardiac arrest, potentially revolutionizing the understanding of the mechanisms underlying this phenomenon.
Currently, the model is being tested in hospitals in the United States, Sweden, and Taiwan, paving the way for integration into medical systems. If successful, it will open the door to a new era of preventive medicine, enabling the identification of patients most in need of monitoring or device implantation. It may also allow some individuals to preliminarily assess their risk level, thereby enhancing AI-supported personal prevention.
