Date of Award

Fall 2022

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Biomedical Engineering (ENAS)

First Advisor

Constable, Robert

Abstract

Functional magnetic resonance imaging (fMRI) is a key tool in translational neuroscience. It has been used to investigate the neurobiological phenomena which underpin traits, behaviors and disorders. However, it’s utilization far from perfect. The field struggles with statistical false positives, model overfitting, and poor generalizability. These issues have greatly impeded translation. The causes are multi-factorial, including, but not limited to; the use of small, homogeneous sample sizes, uncertainty about which brain states contribute to successful brain behavior models, and an unclear relationship between the fMRI blood oxygenation level dependent (BOLD) signal and underlying neural activity. Solutions to these issues are emerging. The advent of several large, heterogenous publicly available datasets allow for more robust estimation of model generalizability. Recent methodological advances have increased our understanding of short time, dynamic functional connectivity in the brain. Working at previously uninvestigated time scales, it has been possible to identify behaviorally salient, short time brain states, obscured by using data from across longer time periods. Preclinical multimodal imaging has allowed simultaneous measurement of the BOLD signal and underlying neural activity, which could help in ruling out non-neuronal noise sources in the BOLD signal. Each of these approaches can help bolster translational neuroscience efforts. In this thesis I will describe work touching on all of these areas. In chapter one, I give an overview and motivation for the current state of the art in fMRI research. In chapter two I discuss how to improve the generalizability of predictive models of brain and behavior using a resample aggregating approach, in large publicly available datasets. In chapter three I use the methodology I developed from chapter two to reliably relate short time brain states, represented by dynamic functional connectivity maps, and time varying behavior. In chapter four I present a preprocessing pipeline and analytical framework which can enable functional connectivity analyses, commonly used in human fMRI, in preclinical studies involving mesoscale calcium imaging, a more direct measure of neural activity. The work described herein will hopefully contribute to developing more robust models of brain function and behavior, capable of translation into other fields.

Download

Share

COinS