Every year, sudden cardiac arrest claims more than 300,000 lives in the United States.
Unlike a heart attack, which is usually caused by a blocked blood vessel, sudden cardiac arrest happens when the heart’s electrical system suddenly fails.
The heart can no longer pump blood effectively, causing the person to collapse within seconds. Without immediate treatment, including CPR and a defibrillator shock, most people die within minutes.
This condition can affect older adults with heart disease as well as younger people who seemed completely healthy. Because it strikes so quickly, doctors have long searched for better ways to identify people at greatest risk before disaster happens.
A new study from the University of California, Berkeley offers hope that artificial intelligence could make this possible. The research, published in Nature, used a computer program trained to study electrocardiograms, also called ECGs or EKGs.
These simple heart tests record the electrical signals that control each heartbeat and are already performed in hospitals and clinics around the world.
The researchers, led by Professor Ziad Obermeyer, collected more than 440,000 ECG recordings from Sweden and linked them with official death records. The AI system learned to recognize tiny electrical patterns that appeared in people who later died from sudden cardiac arrest.
After training, the program was tested using thousands of additional patient records from hospital systems in the United States and Taiwan to see whether it could accurately predict future risk.
The results were encouraging. Compared with today’s standard screening method, which mainly measures how much blood the heart pumps with each beat, the AI tool identified a larger group of people who truly faced a high risk of sudden cardiac death.
The traditional test identifies a group with about a 4.6 percent yearly risk, while the AI identified a group with roughly a 7 percent yearly risk. That improvement could represent thousands of people whose lives might one day be saved through earlier treatment.
One reason this matters is that doctors already have an effective treatment for some high-risk patients. Implantable cardioverter defibrillators continuously monitor the heart and automatically deliver a life-saving shock if a dangerous rhythm occurs.
The challenge has always been deciding who actually needs one. Many patients receive these devices but never need them, while many others who appear low risk die without warning.
Building the AI system was not easy. It took about ten years to collect, organize, and carefully study the medical data needed to train and test the model. The researchers believe that access to large, high-quality health databases is essential for developing trustworthy medical AI.
The team is now working with hospitals in Sweden, Taiwan, and the United States to test the technology in real clinical practice. Patients identified as high risk may first wear heart-monitoring patches before doctors decide whether an implanted defibrillator is needed.
Although the results are exciting, the AI system is not ready to replace doctors. More clinical testing is needed before it becomes part of routine healthcare. The study also cannot prove that using the AI will reduce deaths until future trials confirm that patients identified by the program truly benefit from earlier treatment.
Even so, the findings represent an important step toward more personalized heart care. If future studies confirm these results, this technology could help doctors detect hidden danger earlier, reduce unnecessary procedures, and save many lives by identifying patients who would otherwise appear healthy.
If you care about heart health, please read studies about how eating eggs can help reduce heart disease risk, and Vitamin K2 could help reduce heart disease risk.
For more information about heart health, please see recent studies about how to remove plaques that cause heart attacks, and results showing a new way to prevent heart attacks, strokes.
Source: University of California, Berkeley
