Traumatic brain injury, or TBI for short, is a serious problem. Despite the most serious of TBI’s being treated in hospitals’ intensive care units, the final outcome for almost one in every three patients with the condition is death. One of the big problems in treating patients with TBI’s is that they’re often unconscious. This makes it much more difficult to monitor the true condition of the patient. 

When a patient’s in an intensive care unit several things are being monitored at once, with just one single variable gathering potentially thousands of data points per day. So, for one human being to correlate and get their head around all this data would be a mammoth task. And, that’s why researchers at Helsinki University Hospital (HUS) took to developing artificial intelligence (AI) algorithm – to help treat TBI patients by predicting the outcome for each individual and adjusting their treatment plan accordingly. 

An artist’s illustration of modern neurosurgical intensive care and the use of machine learning algorithms.
CREDIT Rahul Raj, University of Helsinki

This type of model is the first of its kind and still in the early stages of development. But already it has an accuracy of 80-85% when it comes to predicting the mortality probability of the patient. “We have developed two algorithms,” explains Eetu Pursianen, one of the authors of the paper and a Data Scientist from the Analytics and AI Development Department at HUS. “The first algorithm is simpler and is based only upon objective monitor data. The second algorithm is more complex and includes data regarding the level of consciousness, measured by the widely used Glasgow Coma Scale score.”

As expected, the second, more complex algorithm is more accurate. But the first is also acceptable considering its based on only three main variables opposed to five. However, both algorithms still need to be validated in national and international datasets before they’ll be available for future use.