Using Gaussian processes to study traumatic brain injury in sheep

Abstract

The aim of this study was to investigate the dynamic interrelationship between brain tissue oxygenation (PbtO2), intracranial pressure (ICP), mean arterial blood pressure (MABP), and cerebral perfusion pressure (CPP) in an acute sheep model of traumatic brain injury (TBI). PbtO2, ICP, MABP, and CPP were monitored after TBI. Gaussian Processes for machine learning has been used to do simultaneous 3D analysis of the dynamic interrelationship between these parameters. Two critical thresholds for ICP after TBI were identified during our experiments that correlated to human data and may correlate to different intravascular hydrostatic pressures in two different sections of cerebral vasculature. Simultaneous 3D analysis of the dynamic interrelationship between PbtO2 vs. ICP and MABP or CPP could be a better method as opposed to 2-parameter (PbtO2 vs. ICP or CPP) analysis. Current CPP formula may not always reflect the actual cerebral perfusion, therefore we propose the formula as: CPP = PAP - ICP, where PAP is the post arteriolar pressure.