Pseudo-Bayesian Model-Based Noninvasive Intracranial Pressure Estimation and Tracking.

Abstract

A noninvasive intracranial pressure (ICP) estimation method is proposed that incorporates a model-based approach within a probabilistic framework to mitigate the effects of data and modeling uncertainties. A first-order model of the cerebral vasculature relates measured arterial blood pressure (ABP) and cerebral blood flow velocity (CBFV) to ICP. The model is driven by the ABP waveform and is solved for a range of mean ICP values to predict the CBFV waveform. The resulting errors between measured and predicted CBFV are transformed into likelihoods for each candidate ICP in two steps. First, a baseline ICP estimate is established over five data windows of 20 beats by combining the likelihoods with a prior distribution of the ICP to yield an a posteriori distribution whose median is taken as the baseline ICP estimate. A single-state model of cerebral autoregulatory dynamics is then employed in subsequent data windows to track changes in the baseline by combining ICP estimates obtained with a uniform prior belief and model-predicted ICP. For each data window, the estimated model parameters are also used to determine the ICP pulse pressure. On a dataset of thirteen pediatric patients with a variety of pathological conditions requiring invasive ICP monitoring, the method yielded for mean ICP estimation a bias (mean error) of 0.6mmHg and a root-mean-squared error of 3.7mmHg. These performance characteristics are well within the acceptable range for clinical decision making. The method proposed here constitutes a significant step towards robust, continuous, patient-specific noninvasive ICP determination.