Signal processing of cerebral blood flow regulation in humans

Abstract

The brain is energetically expensive and maintenance of normal cerebral metabolism is critically dependent on the regulation of the brain's blood supply. Under normal physiological conditions, this regulation is thought to involve many vascular processes, such as the sympathetic neuromodulation, which are vital for cerebral vasomotor control and underpin the functional basis of cerebral autoregulation (CA). In this thesis we determined that the cerebral pressure-flow dynamics are nonlinear, nonstationary and frequency-dependent, intervened by spontaneous CO₂ reactivity, and are mediated through pathways of sympathetic modulation in a closed-loop fashion.

This thesis presents four studies examining the intrinsic nonlinear and nonstationary dynamics of CA by analyzing the cerebral blood flow (CBF) regulation in response to beat-to-beat blood pressure (BP) and end-tidal CO₂ (PETCO₂) fluctuations under different experimental conditions.

The first study sought to determine the nature and consistency of dynamic CA (dCA) across different subjects during the application of oscillatory lower body negative pressure. dCA was characterized using projection pursuit regression and locally weighted scatter plot smoothing. Additionally, we proposed a piecewise regression method to determine the statistical consistency of a dCA curve. We observed heterogeneous patterns of dynamic BP-CBF relations and dCA did not manifest as any single characteristic curve, which implies that the absence of a clear dCA curve does not necessarily represent an abnormal state of CBF regulation.

The second study sought to identify the role of sympathetic neurovascular control using multivariate wavelet decomposition that explicitly incorporate the confounding effects of dynamic PETCO₂. Analyses showed that placebo administration did not alter phase synchronization index (PSI) values while sympathetic blockade increased PSI for frequencies ≤ 0.03 Hz. Additionally, we found that the sympathetic treatment response varied as a function of frequency and varied depending on whether PSI values were PETCO₂-corrected, which implies that some between-subject variability of autoregulatory capacity may be attributed to the cerebral sympathetic modulation.

The third study aimed to characterize the dynamics of sympathetic control by employing Laguerre-Volterra kernels and global principal dynamic modes (PDMs). We observed that very low frequency (< 0.03 Hz) linear components (first-order kernels) of BP variability and PETCO₂ reactivity are mutually coupled to CBF dynamics and can reliably distinguish between status of normal and impaired cerebrovascular control. Moreover, gains of the associated nonlinear functions of global PDMs having low-pass and ~0.03 Hz resonant peak characteristics also have potential as autoregulatory indices to cerebrovascular function.

The fourth study used vector autoregressive models to examine the causality of the dynamic BP-CBF relations, in terms of Granger's causality in the time domain. The analyses showed that contrary to standard models that assume pressure energy drives brain blood flow in a unidirectional manner, a directional feedback in the reverse direction from CBF to BP also exists. We also found that cerebral sympathetic modulation is an integral pathway that enables bidirectional interactions between BP and CBF, which implies the presence of a dynamic BP stabilizing system that is sensitive to CBF changes.

To summarize, the frequency characteristics of BP-CBF dynamics within the very low frequency range (< 0.03 Hz) are highly sensitive to spontaneous PETCO₂ reactivity. Moreover, CBF regulation is intervened by the sympathetic control of cerebrovasculature in a closed-loop fashion. Collectively, these observations show that multivariate signal analysis techniques can quantify and characterize the time-varying closed-loop dynamics of cerebral circulation.