Multiscale Statics and Dynamics of Cerebral Functional Connectivity Open Access

Billings, Jacob Charles Wright (Fall 2017)

Permanent URL: https://etd.library.emory.edu/concern/etds/ff3656181?locale=en
Published

Abstract

The advent of whole-brain functional imaging through Blood-Oxygen Level Dependent (BOLD) functional Magnetic Resonance Imaging (fMRI) invites novel analytical frameworks to understand the brain's intrinsic functional organization. As brains are complex self-assembled systems, a mechanistic view of brain activity is expected to observe emergent structures interacting across multiple spectral, spatial, and temporal scales. Analytical frameworks that incorporate information at multiple scales may therefore provide additional insights into brain physiology. Chapter 1 introduces this line of reasoning in greater detail. Because BOLD fMRI is an indirect measure of neuronal activity, Chapter 2 pursues an optimal preprocessing strategy for increasing the information content of the BOLD signal. A stratigy that normalizes voxel-wise BOLD signals via z-scoring and removes motion noise via motion parameter regression was found to effectively isolated BOLD signal energetics to the brain's gray matter. Enhancing the signal-to-noise ratio of gray matter BOLD signals is expected to most effectively enhance the proportion of spontaneous hemodynamic (BOLD) fluctuations attributable to neuronal signaling. This is because synaptic activity accounts for the majority of energy usage in the brain, and the dendritic arbor of the central nervous system is unmylenated gray matter. In Chapter 3, preprocessed, voxel-level BOLD signals are filtered into multiple spectral domains in order to identify the spectral components that best reveal the brain's intrinsic organization. Graphs of the brain's functional connectivity--its spatial network architecture--most closely resemble known brain networks in several pass-bands within the low-frequency fluctuation range (~0.1 to ~0.01 Hz). To discover just why low-frequency spectra of the BOLD signal are most effective at revealing the brain's architecture, Chapter 4 links hemodynamic connectivity to neuroelectric connectivity through multimodal studies in the rodent brain. Long-term (static) BOLD connectivity is demonstrated to correspond to static local field potential (LFP) connectivity when neuroelectric activity is filtered into either delta (1-4 Hz), alpha (8-12 Hz), or gamma (40-60 Hz) pass-bands. These findings support the theoretical interpretation of neurovascular coupling as a diffusion-mediated process involving small signaling molecules that communicate information about changing neuronal metabolic load to the cardiovascular system. Essentially, low-frequency fluctuations in the BOLD signal are low-pass filtered versions of neuroelectric activity. Whereas Chapters 2 through 4 pursue long-term trends in coordinated brain activity, Chapter 5 pursues the question of how to identify the kinds of time-varying BOLD dynamics expected to relate to ongoing mental activity. To this end, the instintaneous state space of multi-scale BOLD dynamics is embedded onto a two-dimensional sheet, thereby providing a visually tractable map of the brain dynamics. Discrete epochs of experimentally defined tasks are shown to agglomerate into densely populated peaks in the map space. The brain activitions associated with each map region are further investigated in order to better understand how the brain produces a range of experimentally defined states. Taken as a whole, the enclosed dissertation research demonstrates the pervasiveness of the brain's multi-scalar architecture, and the utility that this perspective affords towards the interpretation of various and complex brain functions.

Table of Contents

Table Of Contents

List of Figures

Figure 2.1

Global Signal Correlations - Preprocessing Strategies Set 1

17

Figure 2.2

Global Signal Correlations - Preprocessing Strategies Set 2

18

Figure 2.3

Global Signal Correlations - Preprocessing Strategies Set 3

19

Figure 2.4

Global Signal Correlations Segmented by Tissue Type

21

Figure 2.5

Global Signal Correlations Across Individuals

22

Figure 2.6

Global Signal Correlation Time Lags

23

Figure 2.7

Spatial Deviations from Zero Time Lag

24

Figure 2.8

Global Signal Spectrum

25

Figure 3.1

The Wavelet Packet Transform

37

Figure 3.2

Hierarchical Clustering

43

Figure 3.3

Network Entropy Across Wavelet Packets

46

Figure 3.4

Variation in Information among Spectrally Delimited Functional Networks

48

Figure 3.5

Example Functional Connectivity Dendrograms

49

Figure 3.6

Connectivity Networks - Set 1 - 112 Volunteers

50 - 51

Figure 3.7

Connectivity Networks - Set 2 - 30 and 5 Volunteers

52 - 53

Figure 3.8

Connectivity Networks - Set 3 - Volunteer #027 and #039

54 - 55

Figure 3.9

Connectivity Networks - Set 4 - 0.016 - 0.028 Hz

57

Figure 3.10

Connectivity Networks - Set 5 - 0.028 - 0.052 Hz

58

Figure 3.11

Connectivity Networks - Set 6 - 0.052- 0.100 Hz

59

Figure 3.12

Multispectral Functional Network Variation in Information across Scan Types

60

Figure 4.1

Rodent Default Mode Network

74

Figure 4.1

Rodent Gradient Echo - Echo Planar Image

78

Figure 4.1

Rodent Local Field Potentials

80

Figure 4.1

Inter-node Distances Across Spectra and Between Modalities

81

Figure 4.1

Multispectral Cross-Modal Network Correlations

82

Figure 5.1

Embedded Brain Dynamics during Rest and Task

101

Figure 5.2

Watershed Segmentation of Resting-State Dynamics

102

Figure 5.3

Watershed Segmentation of Task-Active Dynamics

103

Figure 5.4

Fifty Functional Networks from Independen Component Analysis

104

Figure 5.5

Cumulative Dwell Time Histograms

105

Figure 5.6

Embedded Brain Dynamics Segmented across Scan Types

106

Figure 5.7

Statistical Similarities among Scan-Segmented Embeddings

107

Figure 5.8

Structural Similarity Values between Scan-Segmented Embeddings

108

Figure 5.9

Visualization and Statistical Similarities among Event-Segmented Embeddings

109 - 110

Figure 5.10

Temporal Evolution of SOCIAL Task Dynamics

111

Figure 5.11

Close-up View of Temporal Evolution of SOCIAL Task Dynamics

112

Figure 5.12

A Labeled State Space of Brain Dynamics

113

Figure 5.13

Statistical Affinity of Experimental Conditions for each State-Space Regions

115

Figure 5.14

Contrasting Brain States Evoked by Contrasting SOCIAL Stimuli

116

List of Tables

Table 2.1

Preprocessing Strategy Definitions

15

Table 3.1

Cophenetic Coefficients from Multiple Hierarchical Clusterings of Filtered BOLD Data

40

About this Dissertation

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
School
Submission
Language
  • English
Research field
Keyword
Committee Chair / Thesis Advisor
Committee Members
Last modified

Primary PDF

Supplemental Files