Clinically unresponsive patients with acute brain injury activate a brain-computer interface

Published October 22, 2026. Thirteen of 41 clinically unresponsive patients controlled a simple one-way EEG interface by attempting a hand movement. The result suggests that intentional control can sometimes be detected even without a visible response, but the system did not yet provide open-ended communication and cannot determine prognosis by itself.

Shen Q, Eliseyev A, Kansara V, Neves Briard J, Song YL, Carmona J, Gonzales IJ, Velazquez A, Klein AJ, Agarwal S, Park S, Ghoshal S, Roh D, Connolly ES, Claassen J.
Brain · Publication date: 2026-10-22
Added to site: 2026-10-24
DOI: 10.1093/brain/awag210

Published October 22, 2026. Added to Neurofeedback.com October 24, 2026.

Modalities

EEG, Brain-computer interface, Cognitive motor dissociation

Clinical focus

Acute brain injury, Disorders of consciousness, Bedside communication

Population

Critically ill patients who were clinically unresponsive after acute brain injury, plus healthy and conscious critically ill comparison participants

Sample

41 clinically unresponsive patients, 22 conscious critically ill patients and 8 healthy volunteers

Intervention / method

A self-paced bedside EEG interface changed a tone's frequency and pitch while participants attempted repeated right-hand opening and closing or stopped the movement attempt; a separate machine-learning analysis assessed cognitive motor dissociation during motor commands

Comparator

Within-cohort comparison of clinically unresponsive patients with versus without interface control, plus convenience samples of conscious critically ill participants and healthy volunteers

Clinician summary

BCI control occurred in 13 of 41 clinically unresponsive patients and CMD was detected in 8 of 41. Recovery of clinical consciousness at discharge occurred in 4 of 13 with control versus 1 of 28 without control; functional independence at 12 months occurred in 6 of 13 versus 2 of 28. Reported odds ratios were large but highly imprecise, with 95% confidence intervals of 2.3 to 930.7 and 1.8 to 81.6, respectively.

How to interpret the findings

This was a prospective single-center observational cohort rather than a randomized intervention trial. The clinically unresponsive sample was small, outcome events were few, confidence intervals were very wide and the association between BCI control and recovery does not establish causation. The one-way tone interface had a median signal delay of about 49.8 seconds in clinically unresponsive patients and did not enable open-ended messages. Multicenter replication, false-positive and false-negative characterization and clinically validated decision rules are needed.

Equipment and disclosures

The abstract describes a bedside EEG-based interface that modulated an audio tone in real time but does not identify a commercial EEG system. No equipment is attributed to BrainMaster or another vendor.

The Columbia institutional announcement reports support from NIH awards NS1112760 and NS106014, Paris Brain Institute America, the Canadian Institutes of Health Research, Fondation CHUM and the Royal College of Physicians and Surgeons of Canada. It states that the authors reported no competing interests. No indexed correction, retraction or expression-of-concern relation appeared in the retained PubMed or Crossref records; this is a bounded check. AI-assisted source review.

Related evidence

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

• Comparative Efficacy of Noninvasive Neuromodulation in Patients with Disorders of Consciousness: A Systematic Review and Network Meta-Analysis of Randomized Controlled Trials · Positive with limitations
• Detecting EEG index of arousal in sleeping patients with unresponsive wakefulness syndrome · Positive with limitations
• HD-tACS at 40 Hz Improves Consciousness and Brain Connectivity in Prolonged Disorders of Consciousness: A Randomized Sham-Controlled Clinical Trial. · Positive with limitations