Traumatic brain injury (TBI) is identified as one of the major causes of morbidity and mortality in low-income countries. Severe cases of TBI are treated in the intensive care units, during which many tens of variables are under observation continuously. One of the major challenges faced during the treatment of TBI is the lack of precise monitoring of the patient’s condition, owning to their unconscious state.
In the ICU, the patient’s intracranial pressure, cerebral perfusion and mean arterial pressure is measured which also gives insights into the condition of the patient.
The intracranial pressure helps to yield hundreds of thousands of data points daily. It is therefore not possible for the human brain to understand millions of daily collected data points from all the monitored data.
A team of researchers at the Helsinki University Hospital therefore dedicated efforts to create artificial intelligence (AI) based algorithm which can help doctors in the treatment of patients with severe TBI. The latest algorithm can help understand the outcome of the individual patient while also presenting objective data based on the prognosis and condition of the patient during the stages of treatment.
According to Rahul Raj, professor of Experimental Neurology from HUS, the latest study has created the first dynamic prognostic model. It is a proof-of-concept and will require time before algorithms like these can be implemented into daily clinical practice.
The algorithm predicts the probability of the patient’s death within 30-days with an accuracy of 80-85%. Sources reveal the team has developed two different algorithms. The first is simpler and is based on objective monitor data, while the other is more complex and has data which monitors level of consciousness which is measured by Glasgow Coma Scale Score.
The accuracy of complex algorithm is better than that of simpler algorithm. The researchers however claim the accuracy of both algorithms is unexpectedly good. The simpler model is based on three main variables and the more complex algorithm is based on five main variables.
Sources allege, the algorithms in the future need to be validated in national and international external datasets.