Prediction of Intracranial Pressure Elevation Events in Traumatic Brain Injury Patients Using Machine Learning
Open Access DepositedThe timely provision of treatment to patients with traumatic brain injury (TBI) requires accurate prediction of intracranial pressure (ICP). Albeit, as machine learning approaches have become more commonly used in the ICU, most of the current research concentrates on the identification of individual events as opposed to the continuous monitoring via repeated classification over time of ICP, which prevents us from monitoring the dynamic processes of physiological alterations of a patient.This praxis investigates how one can make predictions of ICP elevation risk, employing the multivariate ICU time-series of the unit to forecast ICP event prediction, using machine learning models. Models depending only on ICP measurements were compared to models that included a larger regiment of physiological data including the hemodynamic variables as well as respiratory variables. Several machine learning and deep learning architectures were compared to embrace temporal patterns and predict ICP elevation risk within short and long period durations. The results of the study show that the model should be improved with the inclusion of more vital signs and increase in performance as the prediction horizon is longer. This benefit was further reflected at the longer horizons, where the multi-vital models provided more continuous monitoring via repeated classification over time and closer representation of the way in which patient physiology changes. Additional approaches in temporal modeling yielded greater accuracy and explainability analyses revealed that further physiological signals added value to ICP dynamics.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.