Ground-Breaking Method Combines fMRI with ML to Predict Mortality Risk in Severely Brain-Injured ICU Patients

Severe brain injuries, whether stemming from a stroke, cardiac arrest, or a traumatic event, can have life-altering consequences for patients and their families. In the case of patients admitted to the intensive care unit (ICU) for brain injury, uncertainty looms large for their families and healthcare providers regarding the chances of recovery, be it partial or complete. Now, researchers have developed a ground-breaking method for predicting which ICU patients can survive a severe brain injury.

Researchers at Western University (Ontario, Canada) combined functional magnetic resonance imaging (fMRI) with advanced machine learning algorithms to address one of the most pressing challenges in critical care: predicting recovery outcomes following significant brain injuries. Working alongside neurologists, the researchers monitored brain activity in 25 ICU patients during the initial days after their brain injuries. They aimed to find out if these readings could indicate which patients would ultimately survive. Earlier work by the team had shown that potential recovery signs could be captured by how different regions of the brain interacted with each other. Maintaining these inter-regional connections is crucial for the restoration of consciousness.

Image: The new technique can predict which patients will recover from a serious brain injury with 80% accuracy (Photo courtesy of Freepik)

The researchers achieved the breakthrough when they figured out they could combine the fMRI data with machine learning technology. This innovative integration allowed them to predict with 80% accuracy which patients were likely to recover, a rate that surpasses the current standard of care. Despite this promising development, the team emphasizes that their predictive method isn't flawless and warrants additional investigation and validation.

“Modern artificial intelligence has shown incredible predictive capabilities. Combining this with our existing imaging techniques was enough to better predict who will recover from their injuries,” said Matthew Kolisnyk, a graduate student from Western University.

Related Links:
Western University