AI-Powered Headset Predicts Epileptic Seizures, Researchers Say

Scientists in Scotland have developed an AI-powered headset they describe as “unique” for predicting epileptic seizures minutes before they occur. The team from Glasgow Caledonian University said the wearable device could transform how the condition is managed, the Mirror has reported.

Currently, no readily available device offers people with epilepsy advanced warnings of seizures. The headset analyzes brainwaves and heart functions to detect an impending seizure, allowing the wearer to take precautions or alert others.

Professor Hadi Larijani, who leads the research team, told the Press Association the headset is “very unique.” He said they hope to one day release it as a wearable device similar to a cap. “We want it to be as discreet as possible, to be wireless, light – it’s a lot of features that none of the other tools right now have,” he said.

The system was trained using thousands of hours of historical electroencephalography and electrocardiography recordings. Researchers also used a custom-made “phantom head” with attached electrodes to simulate a real user.

By learning to identify subtle electrical and physiological patterns that precede a seizure, the AI can issue a timely alert. It has an accuracy of up to 95%, and the algorithm can display its confidence levels about the chance of an impending seizure.

Professor Larijani said the greatest benefit could be giving people with epilepsy more confidence and helping them avoid secondary injuries. “Providing even a few minutes’ warning could be life changing,” he said. “Our goal is to empower patients and families with time to act to stay safe and live more independently.”

He added: “For families affected by epilepsy, even a brief warning means they can prevent injuries and reduce fear. That is the true measure of success for us.”

The team is still refining the design, which they intend to be suitable for children. Initial prototypes used third-party hardware, but they hope to develop their own models.

Due to the need for regulatory approvals, the product is likely to become widely available in some years. “The most regulated things are medical devices. It takes time to get through the hurdles,” Professor Larijani said. “But I think we’ve done the initial feasibility testing both on the software and the hardware characteristics. Now we’re looking at a more commercial-ready product that hopefully will be going through the hurdles of the initial accreditation for medical devices.”

The research team has secured fresh support from UK Research and Innovation through its £9 million ($12 million) Proof of Concept program.