ORCID

http://orcid.org/0000-0003-0051-336X

Date of Award

Winter 1-15-2021

Author's School

Graduate School of Arts and Sciences

Author's Department

Biology & Biomedical Sciences (Neurosciences)

Degree Name

Doctor of Philosophy (PhD)

Degree Type

Dissertation

Abstract

Stimulus-free brain dynamics form the basis of current knowledge concerning functional integration and segregation within the human brain. These relationships are typically described in terms of resting-state brain networks—regions which spontaneously coactivate. However, despite the interest in the anatomical mechanisms and biobehavioral correlates of stimulus-free brain dynamics, little is known regarding the relation between spontaneous brain dynamics and task-evoked activity. In particular, no computational framework has been previously proposed to unite spontaneous and task dynamics under a single, data-driven model. Model development in this domain will provide new insight regarding the mechanisms by which exogeneous stimuli and intrinsic neural circuitry interact to shape human cognition. The current work bridges this gap by deriving and validating a new technique, termed Mesoscale Individualized NeuroDynamic (MINDy) modeling, to estimate large-scale neural population models for individual human subjects using resting-state fMRI. A combination of ground-truth simulations and test-retest data are used to demonstrate that the approach is robust to various forms of noise, motion, and data processing choices. The MINDy formalism is then extended to simultaneously estimating neural population models and the neurovascular coupling which gives rise to BOLD fMRI. In doing so, I develop and validate a new optimization framework for simultaneously estimating system states and parameters. Lastly, MINDy models derived from resting-state data are used to predict task-based activity and remove the effects of intrinsic dynamics. Removing the MINDy model predictions from task fMRI, enables separation of exogenously-driven components of activity from their indirect consequences (the model predictions). Results demonstrate that removing the predicted intrinsic dynamics improves detection of event-triggered and sustained responses across four cognitive tasks. Together, these findings validate the MINDy framework and demonstrate that MINDy models predict brain dynamics across contexts. These dynamics contribute to the variance of task-evoked brain activity between subjects. Removing the influence of intrinsic dynamics improves the estimation of task effects.

Language

English (en)

Chair and Committee

Todd S. Braver ShiNung Ching

Committee Members

Camillo Padoa-Schioppa, Nico Dosenbach, Jr-Shin Li, Jeffrey Zacks,

Share

COinS