Atlas-Free Connectivity Analysis Driven By White Matter Structure

Diffusion tensor imaging allows to infer brain connectivity from white matter, which can then be investigated aiming at finding possible biomarkers of disease. The usual initial step in graph construction is to identify the nodes in the brain using a predefined atlas. However, atlases are usually not considering the white matter structure. As a result, atlas-based brain parcellation and, hence, brain graphs are not fully considering the white matter organization. In this work, we are proposing an atlas-free scheme to map the structural brain networks. The idea is to identify the nodes in the brain exploiting the white matter structure inferred from the data. We first retrieve the white matter pathways from DTI, grouping fiber tracts into bundles. We then use these pathways in a clustering pipeline to identify the brain regions to map into the graph nodes, which are used to define the brain connectivity. We empirically tested the goodness of the proposed approach on a known case-control study obtaining results confirming findings in related literature.

The topology of structural brain connectivity in diseases and spatio-temporal connectomics

Alessandra Griffa

The brain is a complex system, composed of multiple neural units interconnected at different spatial and temporal scales. Diffusion MRI allows probing in vivo the anatomical connectivity between different cortical areas through white matter tracts. In parallel, functional MRI records neural-related signals of brain activity. Particularly, during rest (in absence of specific external task) reproducible dynamical patterns of functional synchronization have been shown across different brain areas. This rich information can be conveniently represented in the form of a graph, a mathematical object where nodes correspond to cortical regions and are connected by edges representing anatomical connections. On the top of this structural network, or brain connectome, individual nodes are associated to functional signals representing neural activity over observation periods. Network science has fundamentally contributed to the characterization of the human connectome. The brain is a small-world network, able to combine segregation and integration aspects. These properties allow functional specialization on the one side, and efficient communication between distant brain areas on the other side, supporting complex cognitive and executive functions. Graph theoretical methods quantify brain topological properties, and allow their comparison between different populations and conditions. In fact, brain connectivity patterns and interdependences between anatomical substrate and functional synchronization have been proved to be impaired in a variety of brain disorders, and to change across human development and aging. Despite these important advancements in the understanding of the brain structure and functioning, many questions are currently unanswered. It is not clear for instance how structural connectivity features are related to individual cognitive capabilities and deficits, and if they have the concrete potential to distinguish pathological subgroups for early diagnosis of brain diseases. Most importantly, it is not yet understood how the connectome topology relates to specific brain functions, and how the transmission of information happens on the top of the structural connectivity infrastructure in order to generate observed functional dynamics. This thesis was motivated by these interdisciplinary inputs, and is the result of a strong interaction between biological and clinical questions on the one hand, and methodological development needs on the other hand. First, we have contributed to the characterization of the human connectome in health and pathologies by adapting and developing network measures for the description of the brain architecture at different scales. Particularly, we have focused on the topological characterization of subnetworks role within the overall brain network. Importantly, we have shown that the topological alteration of distinct brain subsystems may be a biomarker for different brain disorders. Second, we have proposed an original network model for the joint representation of brain structural and functional connectivity properties. This flexible spatio-temporal framework allows the investigation of functional dynamics at multiple temporal scales. Importantly, the investigation of spatio-temporal graphs in healthy subjects have allowed to disclose temporal relationships between local brain activations in resting state recordings, and has highlighted functional communication principles across the brain structural network.

EPFL2015

Exploring dynamic functional connectivity by incorporating prior knowledge of brain structure

Anjali Bagunu Tarun

The synchronized firing of distant neuronal populations gives rise to a wide array of functional brain networks that underlie human brain function. Given the enormous perception, learning, and cognition potential of the human brain, it is not surprising that network-level representations of brain function reveal intrinsically rich organizational structure. What remains puzzling is how the human brain maintains its vast repertoire of functional connectivity (FC) states despite being constrained by the underlying fixed anatomical substrate. Advances in modern neuroimaging technologies, such as diffusion-weighted magnetic resonance imaging (DW-MRI), have made it possible to map the brain's anatomical scaffold, while functional MRI (fMRI) provides complementary information on neural activity. In this thesis, we develop and apply methods for combining DW-MRI and fMRI data into an integrated framework to analyze the interplay between brain structure and function. We first explored the dynamics of brain function during wakefulness and across the different non-rapid eye movement (NREM) sleep stages. We applied the innovation-driven co-activation pattern (iCAP) analysis to uncover the spatial and temporal organization of overlapping large-scale brain networks. Our results reveal new spatial patterns covering regions that support the physiological organization of sleep and arousal. Contrary to the previously observed decreasing FC that accompanies increasing sleep depth, we instead observe a surge of network activity and cross-network interactions during NREM stage 2, followed by an abrupt decrease in NREM stage 3. In the next step of the thesis, we propose a new method that models all brain voxels as nodes of a high-resolution voxel-level brain graph. We provide two ways to construct the graph and we characterize their properties by performing a spectral analysis of the graph Laplacian operator. Our findings show that despite the huge dimensionality of the proposed brain graphs, the majority of the structural information is captured by the Laplacian's lowest frequency eigenmodes. The lower end of the Laplacian spectra also captures about 85% of the energy content of functional MRI, suggesting that functional patterns are overall smooth over the structure, thus providing, for the first time, a direct and quantitative measure of how much brain function is shaped by the anatomy. Going beyond a scalar measure of the SC-FC link, we introduce a new framework that interpolates gray matter signals onto the white matter using the structure embedded in the voxel-level brain grid to guide the process. This enables visualization of key white matter structures that link temporally coherent gray matter areas. We found whole-brain structure-function networks that extend currently known spatial patterns that are limited within the gray matter only. Finally, we assessed the collective mediation of white matter pathways by giving a quantitative measure of the overall anatomical range between temporally coherent gray matter areas. We utilized a canonical model of graph diffusion to extract the anatomical range of functional network interactions. We find that this measure meaningfully differentiates brain regions according to a behaviorally relevant macroscale gradient that divides the cortex between low-level primary sensory areas and high-level cognitive functions.

EPFL2020

Exploring dynamic functional connectivity by incorporating prior knowledge of brain structure

Anjali Bagunu Tarun

EPFL2020

The topology of structural brain connectivity in diseases and spatio-temporal connectomics

Alessandra Griffa

EPFL2015