Models and methods of cerebrovascular reactivity mapping

Thesis

Cerebrovascular reactivity (CVR) serves as a critical biomarker for assessing the capacity of cerebral blood vessels to adapt to physiological and pathological challenges. Although numerous methods exist to measure CVR, accurately capturing its full dynamic range remains an ongoing challenge. This thesis presents advances in the underlying principles, measurement techniques, and modelling of CVR, integrating data from transcranial Doppler ultrasound (TCD) and blood-oxygen-level-dependent functional magnetic resonance imaging (BOLD-fMRI).

The thesis begins with a review of the pathophysiological bases of vascular smooth muscle cell dysfunction, showing how impaired vascular regulation can occur before clinical symptoms of neurodegeneration emerge and motivating CVR as a promising tool for assessing brain health. The subsequent chapters develop and utilise different methodologies to map CVR dynamics. First, CVR measured using TCD in combination with a fixed carbon dioxide challenge is compared with the pupillary light response (PLR). Although the PLR reflects certain autonomic responses, it only partially aligns with the broader haemodynamic changes measured by TCD. To investigate whether a probability-driven approach would enhance CVR accuracy, a Bayesian modelling framework is introduced for breath-hold-induced CVR in BOLD-fMRI. By treating both data and parameters as probability distributions, this framework demonstrates greater adaptability and computational speed than general linear modelling while maintaining comparable accuracy. The TCD-based research is then extended using a novel ramp protocol that delivers progressive hypercapnia, enabling fine-grained, non-linear (sigmoidal) characterisation of vascular responses, elucidating inflection points and upper/lower limit plateaus that linear models are unable to identify. Incorporating the Bayesian framework into these TCD ramp analyses further enhances the characterisation of the full CVR response. Applying these non-linear methods to BOLD-fMRI then facilitates side-by-side assessment with TCD, revealing both consistent trends and modality-specific differences. Finally, the thesis validates key physiological assumptions by comparing TCD-derived velocity measures with standard models linking altered cerebral blood flow to changes in the BOLD signal.

Taken together, these studies contribute to a more nuanced understanding of cerebrovascular function, emphasizing the importance of non-linear modelling and multimodal approaches. By advancing experimental protocols, computational frameworks, and preliminary non-invasive metrics (such as the PLR), this thesis opens avenues for earlier detection of vascular involvement in neurodegeneration and for broader clinical adoption of CVR as a routine biomarker of cerebrovascular health.

Authors: Hayes, G

• DOI: 10.5287/ora-ozeo67kgo
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: neurodegeneration; transcranial Doppler ultrasound; vascular smooth muscle; sigmoid modelling; hypercapnia; fMRI; cerebral blood flow; Bayesian modelling; cerebrovascular reactivity
• Subjects: Magnetic resonance imaging; Modeling; Cerebral circulation; Medical physics; Biomedical engineering; Transcranial Doppler ultrasonography