|
Abstract:
|
Resting-state functional magnetic resonance imaging (fMRI), a technique often used to evaluate regional interactions in the brain that occur when a subject is not performing a task, is considered a promising biomarker for Alzheimer's Disease. Dynamic networks inferred from fMRI data are now frequently used to represent the inherent inter-relational structure among the brain regions. We model a subject's resting-state fMRI network over time as longitudinal observations of a continuous-time Markov chain on network space. Network dynamics are represented in these models as being driven by various factors, both endogenous (i.e., network effects) and exogenous, where the latter include in particular certain potential mechanisms and relationships conjectured in the literature. These models are fitted to networks from a subset of healthy and diseased subjects from the Alzheimer's Disease Neuroimaging Initiative (ADNI). A meta-analysis is then conducted on the model output, allowing us formally to test a number of current hypotheses regarding relationships between various phenotypes and disease onset.
|
