Researchers develop an AI system with near-perfect seizure prediction

Devindra Hardawar
Senior Editor
mr.suphachai praserdumrongchai via Getty Images

We've seen a smart arm bracelet that can predict nightly seizures, but now a pair of researchers have created something even more promising: an AI system that can predict epileptic seizures with 99.6-percent accuracy. Even better, it can do so up to an hour before they occur. As IEEE Spectrum reports, that gives people enough time to prepare for the attack by taking medication. Around 50 million people around the world currently have epilepsy, according to the World Health Organization, and 70 percent of those patients can control their seizures with medication.

While it's not a complete fix, the new AI system, developed by Hisham Daoud and Magdy Bayoumi of the University of Louisiana at Lafayette, is a major leap forward from existing prediction methods. Currently, other methods analyze brain activity with an EEG (electroencephalogram) test and apply a predictive model afterwards. The new method does both of those things at once, with the help of a deep learning algorithm that maps brain activity and another that can predict the electrical channels lighting up during a seizure.

It'll still be some time before this technique will be available for widespread use -- the team is now working on a custom chip that can help process the necessary algorithms -- but it could be life-changing news for patients with epilepsy.