A third of 'unresponsive' ICU patients moved a sound with their minds

Imagine lying in an intensive care bed after a brain injury. Doctors shine lights in your eyes, ask you to squeeze a hand, call your name. You hear all of it. You cannot move. On paper you are unresponsive, and the decisions about your care, up to and including whether to continue it, are being made on that basis.

A study from Columbia University, published in the journal Brain, suggests that for some patients this is not a thought experiment. Researchers gave 41 clinically unresponsive ICU patients a simple brain-computer interface. About one in three of them used it.

A beep you can move

The setup is almost aggressively low-tech, which is the point. No implants, no scanner. The team used EEG electrodes, the same stick-on sensors already used every day to monitor brain-injured patients, and hooked them to a bedside computer that read the signals in real time.

Patients wore headphones and heard a beeping tone. A recorded voice asked them to push the pitch up or down by imagining opening and closing their hands. When the computer saw the right pattern in their brain activity, the tone moved. The voice encouraged them as they tried and congratulated them when they hit the target. There was a short training period first.

The patients were tested only a few days after being admitted. About a third managed to control the tone with thought alone, according to Columbia. That group included every patient who had already been flagged, by a separate exam, as having what neurologists call cognitive motor dissociation: a brain that understands and tries to respond, cut off from the muscles that would show it.

"EEG is used in clinical practice every day in the ICU," said Jan Claassen, the neurologist who led the work and runs the neurocritical care unit at NewYork-Presbyterian/Columbia. That is why this matters more than a flashy implant demo. If the hardware is already on the patient's head, the barrier to using it is software and willingness.

The hidden patients

Claassen has been chasing this for years. In 2019 his team used standard EEG and machine learning to find signs of hidden consciousness in 15 percent of unresponsive brain-injured patients. At the time the field had documented the phenomenon in only a handful of cases. Later work linked it to a breakdown in communication between the brain regions that plan an action and the ones that carry it out. The patients were trying. The signal just never reached their hands.

His position, quoted by News-Medical, is that "many patients with acute brain injury who appear to be in a coma actually have some level of preserved consciousness." Patients with cognitive motor dissociation are also more likely to recover consciousness and meaningful day-to-day function, which makes spotting them early a matter of more than academic interest. It is the difference between a family being told there is nothing there and being told someone is listening.

Earlier attempts to talk to these patients mostly ran through fMRI, which has managed to get yes-or-no answers out of a single unresponsive patient. That is a remarkable result and nearly useless in practice: you cannot roll a critically ill person into an MRI tube every time you want to ask if they are in pain.

What it does not do yet

Claassen is careful about the limits. "Our brain-computer interface is only a first step," he said. Moving a tone up and down shows awareness and intent. A conversation is still a long way off. Patients cannot yet use it to say they are cold, or hungry, or hurting.

The sample is small, and the news releases do not give the exact number of patients who succeeded or how big the cognitive motor dissociation group was. Tests for hidden consciousness are still confined to research settings and are not part of routine care anywhere.

The people who used the system had their own feature requests. Patients and family members suggested using a loved one's voice in place of the recording, adding culturally familiar content, and building in ways to detect pain, hunger and awareness. The goal, Claassen says, is a version that lets patients express those needs directly. "I think that's where our work can be fundamentally transformative," he said.

Most brain-computer interface news is about implants, billionaires and cursor speed records. This one is a beep, a pair of headphones and a sticker on someone's scalp. It may end up mattering more, because the people it is built for are already in the bed, wired up, waiting for someone to ask the right question.

Sources: Medical Xpress (Columbia University), News-Medical, Brain (journal), Shen et al. 2026.