Researchers in the US have shown how brainwaves detected in unresponsive patients can help predict if and when they will make a full recovery from traumatic brain injury. By analysing the electrical signals using machine learning, a team led by Jan Claassen at Columbia University Irving Medical Centre found that patients who recover faster tended to generate brainwave activity in response to verbal commands, even when their bodies couldn’t respond physically.
For clinically unresponsive patients who have suffered from traumatic brain injury, it can be incredibly difficult for doctors to predict how long it will take for them to fully recover. As patients start to show signs of recovery, rehabilitation is crucial to ensuring that their brains regain their usual function.
To maximize their chances of success, these rehabilitation programmes must be tailored to the unique rate of progress of each patient. Yet for reasons that neurologists don’t yet understand, the timescale of recovery can vary drastically between patients: ranging from just a few months to potentially several years. Ultimately, this makes it far harder for doctors to decide on how rehabilitation should proceed.
Currently, the extent of a patient’s recovery is often assessed by asking them to respond to simple verbal commands to move a certain part of their body. Those who do not respond to these commands are considered unconscious.
Recently, however, more advanced techniques have emerged, based on electroencephalography (EEG). Here, electrodes placed on a patient’s scalp pick up their brainwaves: oscillations in electrical current generated by large clusters of synchronized neurons in their brains. Studies have shown that even if patients with traumatic brain injury can’t respond to verbal commands directly, their brainwaves indicate that they are aware of them to at least some extent. In this case, patients are said to be in a state of “covert consciousness”.
In their study, Claassen and his team worked with 193 intensive care patients with traumatic brain injury, all of whom were unresponsive to verbal commands at the start of the study. To identify covert consciousness in the patients, the researchers applied machine learning to their EEG recordings – allowing them to distinguish whether the brainwaves appearing after verbal commands to “keep moving” were different from those triggered by instructions to “stop moving”.
AI predicts coma outcome from EEG trace
In total, the researchers identified brainwaves associated with covert consciousness in 27 of the patients. Out of this group, 41% had made a full recovery after just one year; while nearly all of them showed visible signs of improvement after just three months. In contrast, just 10% of patients without covert consciousness had made a full recovery over the same period.
The result is an important step forward in neurologists’ understanding of how the timescale of recovery in unresponsive patients can be predicted from their brain activity. Based on these insights, Claassen’s team hopes that doctors could develop smarter rehabilitation programmes for their patients; while also helping their families to make more informed decisions about their care.
The study is reported in The Lancet Neurology
