Boston University
Utilizing functional connectivity to understand reasoning, attention, and spatial navigation
Abstract
dc:description.abstractFunctional Magnetic Resonance Imaging (fMRI) offers a non-invasive measurement of human brain activity, allowing researchers to replicate decades of essential findings from animal models, while also adding to what we know about human higher-order cognition. Early fMRI work was done using univariate methods. More recently, functional connectivity (FC), a multivariate analysis method, has gained traction because of its ability to assess whole-brain dynamics both during cognitive tasks and at rest. FC during resting-state has resulted in the recognition of well-established and replicated intrinsic functional networks. These functional networks suggest an inherent underlying organization of function in the neurotypical brain. Studies examining these intrinsic networks have increased our understanding of how they reconfigure and/or remain stable during task-state, vary across individuals and different cognitive domains, and contribute to neurological and psychiatric disorders. In this dissertation, I provide three examples of how to use varying methods of FC to probe the network-based neural underpinnings of higher-order cognition. The goals of the 3 studies were to 1) employ a ridge regression machine learning model to use resting-state connectivity (rs-FC) to predict abstract reasoning task-activation across individual participants in the lateral Prefrontal Cortex (PFC), 2) assess the organization of functional brain networks and their nodes in facilitating long-term memory (LTM)-guided and stimulus (STIM)-guided attention, and 3) examine hippocampal to whole-brain connectivity during different phases of a spatial navigation task. In study 1, I investigated the effectiveness of connectome fingerprinting (CF) trained with rs-FC data during an abstract reasoning task. While previous studies have successfully employed CF on fMRI data using tasks that span sensory processing, motor control, and working-memory, this study served to test its use during a higher-order cognitive task in a highly variable brain area, the lateral PFC. CF was carried out using a nested leave-one-out cross-validation (LOOCV) scheme, resulting in individual activation predictions that were then compared to the group average. The model demonstrated higher prediction accuracy of lateral PFC activation in individual participants than the group average. Further, a subset of nodes generating smaller functional connectomes of only multi-modal brain networks was as accurate as a model using the whole-brain connectome. These results suggest that CF can be used to predict individual activation patterns during higher-order cognition, and that there are key networks whose connectivity patterns contribute to connectome accuracy. In study 2, I used task-based connectivity to assess the contributions across higher-order brain networks in facilitating LTM-guided attention, a task that requires both internal memory recall and external attentional demands. I employed a general Psychophysiological Interaction (gPPI) model suited for event-related fMRI data to analyze FC of the default mode (DMN), cognitive control (CCN), and dorsal attention (DAN) subnetworks, characterizing large-scale communication between brain regions underlying this process, and a STIM-guided attentional control. The results of this study identified flexible network-based interactions, particularly from the DMN, to facilitate the different components of LTM-guided attention. I found a crucial role for the precuneus and retrosplenial cortex along the posterior midline, between DMN and CCN, in aiding the ability to guide attention based on LTM recall. I identified this posterior midline region as crucial for the integration of exogenous memory and endogenous attention to visual stimuli in the environment. Alternatively, DAN and CCN subnetwork interactions were more important for STIM-guided attention, where DMN played less of a role. In study 3, I conducted a task-residual FC analysis from the hippocampus during the navigation of a virtual spatial environment. The study assessed two phases of spatial navigation, from acquisition of the routes on day 1 to performance at ceiling accuracy criterion on day 2. Connectivity seeds were generated from 3 hippocampal segments, the head, body, and tail, as the hippocampus is a known hub for spatial navigation. There are few FC studies examining the human hippocampus during spatial navigation, and to our knowledge none looking at this across the different segments of the long axis. This study aimed to identify whole-brain subcortical and network-level connectivity architecture that facilitates the different phases of spatial navigation, from novelty to familiarity. The results suggest a flexible navigation network, where acquisition involves greater connectivity to subcortical basal ganglia, mainly in the striatum, and cortical networks involved in goal-directed attention and cognitive control. During performance, we see a shift in connectivity to cortical networks involved in top-down attention, memory-recall, and somatomotor processing. The hippocampal head showed the most widespread connectivity across the 3 seeds. These findings suggest that flexibility from hippocampal connectivity throughout the whole brain facilitates higher-order spatial navigation. Together, these projects suggest the use for integrating FC techniques with task-based fMRI to advance our understanding of human cognition. The results show that functional resting-state networks are stable enough to predict underlying properties of task activations (Aim I) while being flexible enough to change their organization to meet task demands (Aims II & III), as resulting from varying methods of FC.
Degree
thesis:*- Name thesis:degree_name
- Doctor of Philosophy
- Level thesis:degree_level
- doctoral
- Discipline thesis:degree_discipline
- Neuroscience
- Grantor
- Boston University
- Year dc:date.issued
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Isenburg, Kylie
- Advisor dc:contributor.advisor
-
- Stern, Chantal
- ORCID dc:identifier.orcid
Subjects
dc:subject × 6Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/2144/53177