Title:
A Population Neuroscience Approach for Analyzing Regional Structural Brain MRI Data in Cognitive Aging
Author:
Michelle Caunca
Date:
2019
Executive Summary:
As the number of older adults increases in the US over the next few decades, the racial and ethnic diversity of the older adult population is also expected to increase. As such, the burden of age-related cognitive impairment in racial and ethnic minorities is an urgent public health priority. Though the most common form of dementia is Alzheimer’s disease, cerebral small vessel disease resulting from cumulative vascular risk factor burden often co-exists with neurodegenerative pathology and brain atrophy. Both brain atrophy and cerebrovascular damage can be imaged in vivo with structural brain magnetic resonance imaging (MRI). Global MRI-derived markers of brain atrophy and cerebral small vessel disease have extensively been associated with vascular risk factor burden and cognitive impairment. However, evidence suggests that the natural history of cognitive aging-related pathology is region-specific. Regional markers of these pathologies are increasingly available in epidemiologic datasets due to the widespread use of standardized, valid, and reliable brain MRI post-processing software, such as Freesurfer and FSL. However, the statistical and methodological challenges that accompany these data are often not considered in neuroepidemiologic studies to date, including selection bias, high dimensionality and correlation, and multiple comparisons. Analyzing these data using a population neuroscience approach can help address these issues. Population neuroscience encompasses a multidisciplinary, translational approach that involves the application of epidemiologic and biostatistical methods to large-scale neuroscience studies.
In this dissertation proposal, I seek to examine associations between regional brain MRI-derived metrics of gray matter structure and white matter injury with cognitive function in a racially and ethnically diverse, urban sample of community-dwelling participants from the Northern Manhattan Study. Using a population neuroscience approach, I will address these methodological issues while answering the research questions set out in this dissertation research. The specific aims of this proposal are as follows: 1) Examine associations between neurodegeneration and cerebral small vessel disease in regions related to Alzheimer’s disease (AD) using multivariable linear regression and linear mixed-effects models; 2) Construct prediction models for cognitive performance using region-specific neuroimaging biomarkers derived from structural MRI; and 3) Investigate whether neurodegeneration mediates the association between cerebral small vessel disease and cognitive performance using structural equation modeling.
From Aim 1, we found that participants with cholinergic white matter lesion load in the 4th quartile exhibited smaller Alzheimer’s disease-signature cortical thickness, compared to participants with cholinergic WMHV in the 1st quartile. Further, the association of cholinergic white matter lesion load differed by region within the AD signature, such that the strongest association was found in the temporal pole. Results were largely null for hippocampal volume. From Aim 2, we found that MRI markers of gray matter atrophy and white matter injury do not improve the prediction of domain-specific cognitive performance above and beyond demographics and APOE e4 allele. However, analyses in permuted datasets showed that the MRI biomarkers selected by elastic net regression are significantly different from random biomarkers. From Aim 3, we found that greater cholinergic white matter lesion load was most strongly associated with worse baseline processing speed and change in processing speed over time. This association was also partially mediated by AD signature cortical thickness. Taken together, this dissertation work shows that subclinical cerebrovascular disease in cholinergic regions are related to worse cognition as well as gray matter structure in AD signature regions.
