Time varying multivariate system identification of cerebral autoregulation
Journal
Journal of Cerebral Blood Flow and Metabolism
Journal Volume
27
Journal Issue
SUPPL. 1
Pages
BP04-04W
Date Issued
2007
Author(s)
Abstract
BACKGROUND: Blood flow in the brain is controlled by systemic variability (heart rate, blood pressure, and CO2) as well as mental activation. Numerous studies attempt to assess the dynamical properties of the various relationships using low frequency oscillations that can be observed in all of the variables, sometimes at very similar frequencies. Recent studies attempting to assess control of cerebral blood flow by systemic blood pressure have demonstrated significant changes in the dynamical relationship during mental activation paradigms. It is also reasonable to expect that attempts to assess changes in blood flow due to mental activation may be confounded by systemic variability. We have shown that under some circumstances, measurement of additional systemic variability (for example CO2 measurements) can improve the reliability of estimating a dynamical relationship between blood pressure and cerebral blood flow. It could also be expected that measuring variability related to mental activation would further improve the reliability of this process. Difficulties arise because the factors leading to a change in cerebral blood flow during mental activation are still a topic of active research. When it is not possible to measure all variability leading to a dynamic response of a system it could be expected that any system relationship that is identified would vary due to the unmeasured variability. This artefact in the system identification process is termed 'non-stationarity' and might be expected to lead to the low values of coherence found during studies which attempt to perform linear system identification to understand cerebral blood flow control. METHODS: We have developed a technique for performing non-parametric system identification of cerebral autoregulation using the maximal overlap wavelet packet transform to carry out time varying estimation of cross and power spectral densities. This allows assessment of a multivariate linear relationship between inputs: blood pressure and CO2 variability; and outputs: cerebral blood flow velocity measured using transcranial Doppler ultrasonography (TCD) or near infrared spectroscopy (NIRS) measurements. This technique is applied to combined measurements of arterial blood pressure, CO2 variability and TCD measured cerebral blood flow velocity in the resting state, and combined measurements of arterial blood pressure and Near Infrared Spectroscopy during mental activation paradigms. RESULTS AND CONCLUSIONS: In the resting state it appears that a significant proportion of low frequency cerebral blood flow variability can be explained by a linear system acting upon low frequency variability in CO2 and ABP. This relationship appears constant with time and multiple coherence function estimation using traditional stationary spectral techniques appears to work well. It appears that during mental activation this assumption of stationarity breaks down, however, in using time frequency analysis to estimate the properties of the underlying system the extent to which this occurs can be reliably quantified, providing a more general technique of discovering the behaviour of the system from experimental data.
Subjects
carbon dioxide
arterial pressure
autoregulation
blood flow velocity
blood pressure
brain blood flow
brain function
conference paper
Doppler echography
near infrared spectroscopy
nonhuman
priority journal
time
Type
conference paper