ABSTRACTAlzheimer's disease (AD) is a progressive neurodegenerative disease marked by cognitive decline and memory loss due to the accumulation of Amyloid-Beta plaques and hyperphosphorylated Tau tangles. Recent research has implicated vascular dysfunction as a significant contributor to AD onset and progression, with mid-life Hypertension (HT) being an important modifiable risk factor. One possible mechanism underlying this association is the increased pulsatility of cerebral blood flow (CBF) in HT, which leads to small vessel disease, disruption of the blood-brain barrier, and changes in cerebrospinal fluid (CSF) waste clearance. This hypothesis is supported by strong associations between HT and White Matter Hyperintensities (WMH), considered biomarkers of small vessel disease, and studies demonstrating that arterial pulsations drive CSF glymphatic movement. However, the effects of blood pressure management on cerebrovascular health and CSF flow are not fully understood due to the complex interplay between vascular structural changes and mechanisms of cerebral blood flow (CBF) regulation. An examination of changes in CBF pulsatility, CSF flow, and markers of small vessel disease in hypertensive older adults after one year of blood pressure management can provide insights into the effectiveness of BP treatment in improving CBF and CSF flow dynamics, as well as reducing the risk of vascular disease. The first aim of this dissertation was to evaluate the dynamic relationship between CBF and CSF flow in adults with normal blood pressure and those with HT before and after one year of blood pressure management. Volumetric flow data were acquired using CINE Phase-Contrast Magnetic Resonance Imaging, and a novel semi-automatic algorithm was developed to improve the reliability of vessel segmentation for flow analysis. Mean pressure gradient waveforms were estimated using spatial CSF velocities based on a previously validated method that utilizes a simplified Navier-Stokes equation. Parameters of mean pressure gradient, CBF, and CSF flow waveforms were analyzed both cross- sectionally and longitudinally, alongside additional clinical risk measures. Significantly increased pulsatility of blood flow in Internal Carotid Arteries was found in HT, which was potentially driven by significantly reduced diastolic CBF. This finding did not improve after one year of blood pressure management, suggesting it is due to permanent vascular changes. No significant differences were found in any aqueduct CSF parameters or spinal CSF flow parameters, but significantly larger cranial, caudal, and peak-to-peak pressure gradients were found in spinal CSF of adults with HT. Future work is needed to determine how increased pressure gradients in spinal CSF can affect waste clearance in the distal CSF regions. The second aim of this dissertation was to evaluate the performance and versatility of several novel algorithms for the automatic segmentation of WMH, compare them against currently widely used algorithms, and analyze the results of the top-performing algorithm in relation to the findings of the first aim. All algorithms were evaluated for their total segmentation volumes and their similarity and percent volume difference from manual labels on a publicly available high WMH burden testing dataset with 110 FLAIR images. Three blinded independent reviewers evaluated algorithm performances for overall quality, False Positives, and False Negatives on 100 FLAIR images without Fat Saturation and 20 FLAIR images with Fat Saturation with equal distribution of low, medium, and High WMH burden. Two novel algorithms from the 2017 MICCAI WMH challenge, PGS and FMRIB TrUE-Net 2, were identified for superior performance across three datasets and recommended for automatic WMH segmentation in cross- sectional studies. PGS was used for cross-sectional examination of 90 FLAIR images for the HIPAC clinical trial, and Logarithmic periventricular WMH volumes showed significant correlations with age, systolic blood pressure, pulse pressure, and ventricle volume. No associations were found for diastolic CBF or spinal pulse pressure gradients. Future work is needed for longitudinal comparisons of WMH volumes.
Read
