Modelling the mechanisms of dynamic cerebral autoregulation and cerebrovascular reactivity in the human brain

Abstract

Healthy cerebral blood flow (CBF) control is vital to maintain sufficient oxygen perfusion to brain tissue and prevent brain damage. CBF is therefore controlled by, amongst others, two main global mechanisms: dynamic cerebral autoregulation (dCA), that responds to changes in arterial blood pressure (ABP) to keep steady state CBF near constant, and dynamic cerebrovascular reactivity (dCVR) that responds to vasoactive substances such as CO2. However, CBF control is impaired in a range of cerebrovascular diseases, including stroke and dementia, and treatments are limited by the lack of understanding of the underlying mechanisms of dCA and dCVR.

This thesis investigated the mechanisms that govern dCA and dCVR by combining physiological modelling with time series analysis under different physiological conditions. Simple physiological models of the CBF mechanisms were developed and model parameters were optimised with experimentally-derived impulse responses (IRs). dCA was investigated by representing its myogenic and metabolic responses each by a gain and time constant in a physiological model, to disentangle their contributions. Parameters were optimised with a dataset where CBF was represented by cerebral blood velocity (CBv) measured with transcranial Doppler ultrasonography (TCD) under different physiological conditions, where normocapnia and thigh cuff conditions represented intact dCA and hypercapnia represented impaired dCA. Both the myogenic and metabolic responses were found to be impaired (p-values < 0.001) and the metabolic response to be specifically slowed down (p-values < 0.001) in hypercapnia.

dCVR was investigated by representing the vasculature’s response to changes in CO2 by a gain and two time constants in a physiological model. The model produced the CO2-flow IR shape most commonly reported in previous data-driven literature, an improvement on models that include the effects of CO2 with only one time constant. Model parameters were optimised with the TCD dataset under normocapnia and hypercapnia, and one dCVR time constant was found to be significantly affected by physiological condition (p-values < 0.001, except one analysis where p-value < 0.05), becoming smaller in hypercapnia compared to normocapnia. Potential sex-specific effects on the gain (p-value < 0.05) and time constants (p-values < 0.001) were also found. The dCVR model was expanded for use with a dataset where CBF was represented by blood-oxygen-level-data (BOLD) measured with functional magnetic resonance imagining (fMRI). A group comparison was performed between postpartum women and controls but significant differences were not found in the brain regions investigated in this thesis. Brain region was, however, found to significantly affect all model parameters (p-values < 0.001), highlighting the spatial heterogeneity of dCVR and motivating application of the dCVR physiological model with BOLD-fMRI data.

The results of this thesis contribute to the understanding of the complexities of dCA and dCVR and how they are affected under different physiological conditions. For dCA, specific mechanisms could be investigated, and for dCVR, a first link between the growing body of data-driven dCVR studies and its underlying physiology was presented, using both TCD data, as is typical for the dCA community, and BOLD-fMRI data, which is typical for the dCVR community.