He was 32. Fit. No family history. No warning signs.
Then one morning, his heart just… stopped.
Doctors call it sudden cardiac death.
And until now, they had almost no way to see it coming.
Every year, roughly 350,000 people in the US alone drop dead from sudden cardiac arrest.
Nearly 90% don't survive it.
Many had normal ECGs. Normal check-ups. Normal lives.
So how do you screen for something that looks like nothing?
At UC Berkeley, Associate Professor Ziad Obermeyer spent nearly a decade chasing that question.
His team fed an AI 440,000 ECGs from Sweden, then cross-checked them against death certificates.
Then validated the model on patients in the US and Taiwan.
What the AI found stopped cardiologists cold.
The AI didn't just get better at spotting known warning signs.
It uncovered a brand-new waveform pattern — a hidden electrical fingerprint of impending sudden cardiac death.
Something doctors have been staring at for a century… and missing.
The numbers are wild:
Today's standard leans on one metric: Left Ventricular Ejection Fraction — basically, how well your heart pumps.
Problem is, most people who die suddenly had normal pumping function.
Their hearts were fine.
Their wiring wasn't.
"One thing that makes this problem very tragic, but also very well suited for AI, is that we have the cure," he said.
Implantable defibrillators already exist.
They're proven. They save lives.
The only missing piece was knowing who needs one.
And he's careful not to oversell it.
👉 "You don't have to believe the AI at all. You can just use it to target additional testing."
No drama. No hype. Just a smarter first filter.
For years we've asked what AI in medicine actually means.
Here's your answer.
It's not a chatbot writing prescriptions.
It's a machine noticing something humans have missed for 100 years of cardiology.
A new biological clue, pulled out of the world's most common heart test.
Cardiovascular disease killed 19.8 million people globally in 2022.
A slice of them collapsed with zero warning.
That silence may finally have a signal.
That's all for now!
