Computational Methods to Advance Individual Precision in Brain Mapping

Abstract

Neuroimaging research has enabled scientists to observe the human brain at work without invasive procedures, unveiling fascinating insights into the spatial and temporal activity underlying a wide range of brain functions. Despite these advances, our understanding of the mechanisms of brain function remains incomplete. One notable challenge involves inter-individual variability. To achieve a better understanding of brain function in health and disease and to learn what makes each person unique, it is crucial to interpret neuroimaging data at the individual level. In this dissertation, I developed computational methods to address some of the current challenges in individualized mapping of the human brain. First, I investigated how peripheral physiology uniquely influences brain function and functional magnetic resonance imaging (fMRI) measures. I developed learning frameworks to infer heart rate and respiration variation directly from fMRI data itself. I examined the spatial patterns that these two key peripheral physiological processes impose on brain activity measures. Second, I developed parcellation methods for individual-specific functional brain mapping. One approach is a pragmatic, interactive individualization tool to project group parcellations to subject-specific parcellations. The other approach is designed to leverage commonly acquired structural magnetic resonance imaging (sMRI) data (T1/T2) in the absence of less commonly acquired fMRI scans to map human cortex at the level of individuals. Lastly, I contributed towards developing a manual tractography protocol to reproducibly identify white matter pathways across subjects. Overall, these studies present novel computational methods to address the challenges of individualized mapping of the human brain and provide new insights into the inter-individual variability of brain function.