A spectral approach to noninvasive model-based estimation of intracranial pressure

Abstract

Intracranial pressure (ICP) is the hydrostatic pressure of the cerebrospinal fluid. Ideally, ICP should be monitored in many neuropathological conditions, as elevated ICP is correlated with poor neurocognitive outcomes after injuries to the brain. Measuring ICP requires the surgical placement of a sensor or catheter into the brain tissue or cerebrospinal fluid spaces of the brain. With the risk of infection and brain damage, ICP is only measured in a small subset of those patients whose treatment could benefit from knowing ICP. We expand on a previously proposed model-based time-domain approach to noninvasive, patient-specific and continuous estimation of ICP using routinely measured waveforms that has been validated on patients with traumatic brain injuries. Here, we present a model-based algorithm to estimate ICP using the functional relationship between the spectral densities of the routinely measured waveforms. We applied this algorithm to both a simulated and clinical dataset. For the simulated dataset, we achieved a mean error (bias) of 1.2 mmHg and a standard deviation of error (SDE) of 2.2 mmHg. For the clinical dataset of patients with traumatic brain injuries, we achieved a bias of 13.7 mmHg and a SDE of 15.0 mmHg. While the clinical results are not favorable, we describe sources of estimation error and future directions of research to improve the ICP estimates.