Truace
Subscribe

Human-artificial intelligence collaboration for comanaging epilepsy in real time

One-third of the world's 70 million people with epilepsy have seizures uncontrolled by medication, and implantable devices are a promising treatment. These devices can detect impending seizures, missed medication and impaired cognition, but, like most medical devices, they cannot share this information with patients in real time. This is a missed opportunity for devices to learn from clinical events and teach patients how their behaviors affect their health. Here we present a device platform that converses with…

By the Truvace desk2 min read1 sourcePeer-reviewedTRV-2026-1616· cite & verify
Human-artificial intelligence collaboration for comanaging epilepsy in real time
Use case analysis: the ambulatory EEG in navy medicine for traumatic brain injuries by Burner, Tahnyah D.. Public domain

In brief

Researchers built an inpatient prototype that connects scalp or intracranial EEG to a secure large language model, which exchanges messages with patients through a smartphone app. It tracks sleep, medication, cognition and seizure probability, alerts patients to events and trends, and logs their reported symptoms.

The work matters because implanted epilepsy devices usually cannot share what they detect with patients in real time, which limits learning for both device and user. The evidence so far is a small hospital-based validation and a simulation, so real-world benefit outside monitoring units remains unproven.

Main points

  1. About one-third of the 70 million people with epilepsy have seizures that medication does not control.
  2. The platform links EEG to a secure LLM and communicates two ways through a smartphone app.
  3. It processes biomarkers of sleep, medication, cognition and seizure probability.
  4. It was validated in 13 patients undergoing video-EEG monitoring, 11 with scalp and 2 with intracranial recording.
  5. Precision of seizure detectors improved from patient feedback alone in a pseudoprospective simulation.

The gain

In a pseudoprospective simulation, patient feedback alone improved seizure detector precision without expert intervention, reducing false alarms in a prototype that converses with patients.

The rundown

The authors frame the problem as a missed opportunity: implantable devices can sense impending seizures, missed medication and impaired cognition, but cannot tell patients in real time, so patients cannot learn how their behavior affects their health.

The system is designed as a feedback loop. The AI notifies patients of clinical events, records their symptoms and responses, and uses that input to retrain its own seizure detector, reducing false alarms.

The authors propose it as a scalable, safeguarded platform that could adapt to patient and provider needs across epilepsy and other medical conditions, though this is a proposal rather than a demonstrated result.

Sources

  1. Peer-reviewedNature Medicine2026-10-13

The debate