What your ECG could be hiding 

What your ECG could be hiding

The ER Doc and AI Researcher Who Finds Hidden Signals in Data

Ziad Obermeyer trained and worked as an ER physician, and now focuses on creating algorithms to catch things doctors miss: sudden cardiac death risk a normal ECG reading won't flag, or knee pain a radiologist's scan can't explain. On the latest episode of NextLevel, the UC Berkeley professor walks through how his team built a model that found a group of patients at real risk of sudden cardiac death that no existing medical criteria would have caught, and why he thinks data access is the key to unlocking real change in healthcare.

See what doctors missed →

Five Key Takeaways

1. An algorithm found heart risk that standard tests missed entirely.

Trained on 440,000 ECGs linked to death certificates, the model flagged a group with a 7% annual risk of sudden cardiac death, well above the 4.5% rate doctors currently consider high risk, and 85% of them weren't flagged by any existing medical criteria.

2. Getting the right data took a decade. Building the model took weeks.

Obermeyer's team spent years extracting raw ECG waveforms from a vendor's system and navigating privacy law before the actual machine learning even started.

3. The same approach found knee pain that radiologists couldn't see.