Russian Scientists Teach an AI System to Detect Epilepsy During Sleep

Researchers at Lobachevsky University in Nizhny Novgorod have developed an AI-driven system that analyzes sleep patterns in patients with epilepsy. According to the university’s press office, the team trained machine-learning models to automatically detect and analyze sleep spindles in EEG recordings from both healthy individuals and patients diagnosed with epilepsy.

Tatiana Levanova, senior researcher at the Laboratory for Artificial Intelligence in Cardio- and Neuroscience at the university’s AI Research Center, explained that treatment decisions often rely on threshold values of brain rhythms that clinicians typically evaluate visually. To automate the review of many hours of EEG data, the researchers turned to AI.

A major part of the project focused on analyzing neural signals recorded during the second phase of sleep. In healthy individuals, the rhythmic waves characteristic of this sleep stage are known as sleep spindles, but in epilepsy, their properties may be altered. The method developed by the Nizhny Novgorod team helps describe and analyze these deviations with higher precision.

All EEG datasets used to train the machine-learning models — as well as the conclusions produced by the prototype software — were verified by neurologists in Moscow and Nizhny Novgorod.

The new program is expected to help clinicians forecast disease progression and design individualized treatment plans. The approach is especially crucial for detecting early and hard‑to‑spot signs of epilepsy in children.