Machine learning added to electroencephalography (EEG) may offer a more precise method of determining consciousness in clinically unresponsive patients in the days following severe brain injury, new research shows.
Investigators used EEG recordings coupled with analysis by machine learning to identify brain activity in intensive care unit (ICU) patients who were unresponsive to spoken commands.
"We were able to detect this at the bedside using standard EEG recordings — applying a computer analysis to the EEG — in 15% of patients," study investigator Jan Claassen, MD, PhD, head of neurocritical care and medical director of the neurological intensive care unit at Columbia University Irving Medical Center in New York City, told Medscape Medical News.
Detecting brain activation using this method aligned with patients' 1-year outcomes.
The findings, the investigators note, may carry longer-term implications — including prognostic significance — and potentially inform withdrawal-of-care decisions.
The study was published online June 26 in the New England Journal of Medicine.
Predicting Recovery Challenging
Unconsciousness early after brain injury is frequent, and predicting recovery can be challenging.
"Many patients in the ICU do not show movements on commands, and typically this is interpreted as unconsciousness," Claassen said.
However, a subset of patients have brain activation in response to verbal commands that can be detected by specific patterns on digitally processed EEG, raising the possibility of some preservation of consciousness, he added.
This dissociation between behavior and brain activation is known as cognitive–motor dissociation.
A previous meta-analysis estimated that 14% of chronically unresponsive patients may have cognitive–motor dissociation months or years after injury.
However, Claassen noted the prognostic importance of a dissociation between commanded motor behavior and brain activation in the first few days after brain injury is not well understood.
Investigators conducted a prospective single-center study of 104 consecutive patients admitted to the ICU with acute brain injury and no motor response to verbal commands.
Their EEG recordings were analyzed by machine learning to detect brain activation while the command "Keep opening and closing your right hand " was alternated with the command "Stop opening and closing your right hand."
Daily neurologic examinations during morning rounds included assessment of patients' ability to follow spoken commands. The researchers repeated this protocol for eight consecutive trials over six blocks and collected 240 recordings from the study population.
The investigators also evaluated 10 healthy volunteers using the same protocol. All members of this group showed brain activation on EEG in response to motor commands.
Functional outcome at 12 months was determined using the Glasgow Outcome Scale-Extended (GOS-E).
Sixteen (15%) of 104 unresponsive patients had brain activation detected 4 days after injury. Upon discharge, 50% of those patients improved to the extent of being able to follow verbal commands vs 26% of those who had not shown brain activation.
At 12-month follow-up, 44% of the patients who had shown brain activation scored 4 or higher on the GOS-E (indicating independent function for at least 8 hours) vs 14% of patients who had not shown brain activation.
"I did expect that some patients would show the activation but was surprised by how often we detected this, and the association with outcomes," Claassen said.
By 12 months, 6 patients in the cognitive–motor dissociation group had died vs 50 patients in the group without cognitive–motor dissociation.
"Our findings support those of previous studies that have shown that EEG or functional MRI can, in some cases, reflect activation of parts of the brain in response to spoken commands in unresponsive patients, but whether the detected signal represents recognition or comprehension of commands is uncertain," the researchers note.
EEG recordings have some limitations, the researchers note. "The use of methods such as functional MRI to detect cognitive–motor dissociation may result in more frequent detection than EEG, but functional MRI is challenging to perform in a critical care setting," they write.
Multicenter studies are underway that aim to replicate the findings, Claassen said. Additional goals include trying to understand mechanisms underlying their observations, and studying cognitive and quality of life outcomes of these patients and their families.
In an accompanying editorial, David K. Menon, MD, PhD, and Srivas Chennu, PhD, University of Cambridge, United Kingdom, describe the findings as "intriguing" and note that EEG is more easily used and more widely available than fMRI in the ICU.
They add that the investigators' "use of clinically relevant commands – 'open and close your hand' and 'stop opening and closing your hand' — is similar to what is used in the bedside clinical examination that gives rise to the determination that a patient is unresponsive."
Many patients received sedative drugs, which may have affected the findings, Menon and Chennu note.
In addition, EEG signatures were inconsistent over multiple assessments undertaken in the same patient and methodologic questions remain, they note.
"Nevertheless, detection of cognitive–motor dissociation by EEG was associated with a greater likelihood of recovery to at least a state of upper severe disability [the ability to be left up to 8 hours during the day without assistance] on the extended Glasgow Outcome Scale at 12 months after injury than that among unresponsive patients who did not have evidence of cognitive–motor dissociation," they write.
If confirmed, the findings "could inform prognostication of acute brain injury and potentially provide a means of communication with patients who seem unresponsive on the basis of a conventional clinical examination," the editorialists say.
The study was funded in part by the Charles A. Dana Foundation and the James S. McDonnell Foundation. Claassen disclosed receiving grants from the NIH, National Library of Medicine, and Bard Pharmaceuticals. Menon and Chennu have disclosed no relevant financial relationships.
NEJM. Published online June 26, 2019. Abstract, Editorial
Follow Damian McNamara on Twitter: @MedReporter. For more Medscape Neurology news, join us on Facebook and Twitter
Medscape Medical News © 2019
Cite this: Fresh Approach May Detect Consciousness in Unresponsive Patients - Medscape - Jul 08, 2019.