Human-artificial intelligence collaboration for comanaging epilepsy in real time.

A device platform using AI and patient feedback improves seizure detection precision in 13 epilepsy patients, reducing false alarms without expert intervention.

• Why it matters: Uncontrolled seizures affect one-third of the 70 million people with epilepsy, and current implantable devices cannot share real-time information with patients, missing opportunities for personalized management.
• What they did: Researchers developed an inpatient prototype linking EEG data to a large language model-based AI that communicates via a smartphone app, processing biomarkers and retraining its seizure detector based on patient feedback in a simulation involving 13 patients.
• The result: The system enhanced seizure detection accuracy by leveraging patient input, suggesting a scalable platform that can adapt rapidly to patient and provider needs across epilepsy and other conditions.