A routine heart test may be hiding a warning sign that doctors have missed for years. That is the big takeaway from new UC Berkeley research published in Nature. Researchers trained an artificial intelligence model to study ECGs, also called EKGs, and look for patterns tied to sudden cardiac death.

This is the scary part. Sudden cardiac arrest can strike people with known heart problems. However, it can also hit younger athletes and people who never knew they were at risk.

Each year, hundreds of thousands of Americans die after cardiac arrest. Once it happens outside a hospital, survival can drop fast. CPR and a defibrillator can save lives, but timing is everything.

Now, AI may help doctors spot some patients earlier, while their hearts still look normal by today's common tests.

How AI found a hidden heart risk

An ECG records the electrical activity of your heart. It creates the familiar spikes and waves doctors review to check rhythm and other heart clues.

For this study, researchers used more than 440,000 ECGs from Sweden. They paired those scans with death certificates and health records. Then they trained the AI model to look for waveform patterns linked to sudden cardiac death.

After that, they tested the model on separate patient data from the U.S. and Taiwan. That step is important because medical AI often looks good in one dataset, then fails in the real world. Here, the model held up across very different health systems.