Date of Award
5-29-2020
Degree Name
Doctor of Philosophy (PhD)
Degree Type
Dissertation
Abstract
Recent functional magnetic resonance imaging (fMRI) studies have examined moment-to-moment variability in the blood oxygen level-dependent (BOLD) signal, demonstrating that BOLD variability is related to age differences, behavioral task performance, and clinical pathology, including symptomatic Alzheimer disease (AD). These relationships suggest that BOLD variability may be a meaningful signal that is sensitive to age-related and clinically-relevant functional decline. However, there are several limitations and open questions in the BOLD variability literature. First, age relationships with BOLD variability might be contaminated by individual differences in head motion, global signal artifacts, and/or cardiovascular health. Thus, careful characterization and control of these influences is necessary to test and validate mechanistic interpretations of BOLD variability. Second, BOLD variability has not been examined in the context of preclinical AD. These relationships are important to consider for evaluating BOLD variability as a potential biomarker of early dysfunction and for interpreting the potential contribution of undiagnosed pathology to previous age relationships, which are otherwise interpreted to reflect “healthy” age differences. Third, resting-state and task-driven BOLD variability might be differentially related to cognition, as they may be sensitive to distinct sources of variance in the BOLD signal. Moreover, like age relationships, behavioral relationships with BOLD variability might be influenced by cardiovascular, rather than neural factors, yet these relationships have not been characterized or controlled in the prior literature. The three studies in this dissertation attempt to address these questions using a large, well-characterized sample of cognitively normal older adults. Importantly, a network-based supervised machine learning approach was used to test the predictive accuracy of BOLD variability at the level of functional networks. The first study evaluated relationships between BOLD variability, age, and global cognition after characterizing and controlling for the influence of motion, global signal, and measures of cardiovascular health (CVH), including pulse, blood pressure, BMI, and white matter hyperintensities (WMH). BOLD variability was negatively related to age and positively related to cognition, replicating prior demonstrations after maximally controlling for head motion and global signal artifacts. Age relationships also survived correction for cardiovascular health, but not correction for WMH alone. These results suggest that network-based machine learning analyses of BOLD variability might yield reliable, sensitive measures to characterize age-related functional decline across a broad range of networks. Age-related differences in BOLD variability may be sensitive to processes related to WMH burden. The second study evaluated relationships with preclinical AD, including established biomarkers of amyloidosis, tauopathy, and neurodegeneration. BOLD variability, particularly in the default mode network, was related to cerebrospinal fluid (CSF) amyloid β 42. Further, BOLD variability, particularly in subcortical areas, was also related to markers of neurodegeneration, including marginal relationships with CSF neurofilament light protein and strong relationships with hippocampal volume and cortical signature estimates in AD-sensitive regions. Notably, relationships with hippocampal volume and cortical signature estimates survived correction for CVH and WMH burden, and also contributed to age-related differences in BOLD variability. These results support the contention that BOLD variability may reflect a functionally relevant signal that is sensitive to preclinical pathology. Specifically, BOLD variability may be sensitive to subtle amyloid pathology in default mode areas, as well as non-specific neurodegeneration in AD-sensitive areas. The third study evaluated relationships between measures of behavioral task performance and BOLD variability during both resting state and task-driven performance of a Stroop and an animacy judgment task. Resting-state BOLD variability was related to composite measures of global cognition and attentional control, but these relationships did not survive correction for age or cardiovascular factors. In contrast, task-driven BOLD variability was related to estimates of attentional control measured both inside and outside the scanner, and importantly, these relationships persisted after correction for age and the cardiovascular measures. Overall, these results support the proposal that BOLD variability (during both rest and task performance) is a functionally relevant measure. However, relationships with resting-state BOLD variability may be sensitive to more general age-related mechanisms, such as CVH. In contrast, task-driven estimates of BOLD variability may offer indices of behaviorally relevant neural processing or network organization above and beyond broader age-related differences.
Language
English (en)
Chair and Committee
David Balota
Recommended Citation
Millar, Peter R., "Evaluating BOLD Variability as a Potential Biomarker of Age- and AD-Related Differences" (2020). Arts & Sciences Electronic Theses and Dissertations. 3260.
https://openscholarship.wustl.edu/art_sci_etds/3260