A first realization of reinforcement learning-based closed-loop EEG-TMS.

A closed-loop system learned when each participant's motor network was in relatively higher or lower excitability states, supporting a more individualized approach to stimulation timing research.

Humaidan D, Xu J, Chen J, Zrenner C, Vetter DE, Marzetti L, Belardinelli P, Roine T, Ilmoniemi RJ, Romani GL, Ziemann U.
NeuroImage · Publication date: 2026-08-18
Added to site: 2026-10-08
DOI: 10.1016/j.neuroimage.2026.122176

First published online August 18, 2026; the October issue date is later. Added October 8, 2026 as a historical addition.

Modalities

EEG, TMS/rTMS, Closed-loop stimulation

Clinical focus

Adaptive stimulation timing, Sensorimotor networks

Population

Adult research participants

Sample

25 participants

Intervention / method

Real-time EEG-TMS targeting the supplementary motor area-primary motor cortex network; a reinforcement-learning algorithm identified each participant's mu-rhythm phases associated with higher and lower corticospinal excitability before repeated phase-specific stimulation

Comparator

Individually learned higher-excitability versus lower-excitability mu-rhythm phases

Clinician summary

Reinforcement learning identified participant-specific mu-rhythm phases associated with higher and lower motor-evoked-potential amplitudes. Repeated stimulation at those phases produced directionally different longer-term EEG coherence changes in the stimulated sensorimotor network.

How to interpret the findings

This was an experimental physiology and feasibility study, not a clinical treatment trial. It did not test symptoms, functional recovery or patient benefit. The checked abstract does not report adverse events, exact effect estimates, attrition or independent replication, and EEG coherence changes should not be equated with clinical improvement.

Equipment and disclosures

Real-time EEG-TMS setup; equipment manufacturers and specific models were not verified from the checked abstract.

European Research Council grant 810377 is indexed. One author reported patents and patent applications related to TMS technology; all other authors reported no conflicts of interest.

Related evidence

Compare findings from the same clinical area, including mixed results.