Spatio-temporal dynamics of human resting state fmri

• Autores: Katharina Glomb
• Directores de la Tesis: Gustavo Deco (dir. tes.)
• Lectura: En la Universitat Pompeu Fabra ( España ) en 2017
• Idioma: español
• Tribunal Calificador de la Tesis: Petra Ritter (presid.), Matthieu Gilson (secret.), Maria Victoria Sanchez Vives (voc.)
• Programa de doctorado: Programa de Doctorado en Tecnologías de la Información y las Comunicaciones por la Universidad Pompeu Fabra
• Texto completo no disponible (Saber más ...)
• Resumen

  • The brain is the basis of behavior, personality, and conscious experience. One of the main challenges it faces is to balance flexibility and stability in order to enable us to survive in unknown situations, develop new solutions, and learn, in other words: to adapt, and at the same time draw from experience and remember. This is illustrated in an important concept of brain function: segregation versus integration. Segregation is necessary in order to be able to optimally perform a certain function in a specialized brain region, something that we have studied extensively over the last decades by identifying countless brain areas that support certain functions tested in experiments, like decision making, perception, problem solving, emotional processing, and many more. This endeavor has been immensely aided by imaging methods such as functional magnetic resonance imaging (fMRI). Nevertheless, we have always been interested in how brain regions communicate, recognizing that flexible integration between functions is necessary to perform complex cognitive functions.

    Computational neuroscience is an approach that casts the idea of a "function" into the concept of computations and asks what kind of computation a brain area is suited for and why. Interplay between different kinds of computations over a staggeringly large range of temporal and spatial scales in a hierarchical fashion is the basis for performing tasks like navigating a car on a crowded highway, learning how to speak, or writing a PhD thesis.

    We have long since learned that what fires together, wires together. A main topic in this thesis is the connectivity between brain areas that allows this teamwork to occur, and which structures a bunch of brain regions into a hierarchy inside of which flexible integration and segregation of functions takes place. On a microscopic level, this entails synaptic plasticity, on a mesoscopic scale, neural assemblies, but the scale that is important here is the macroscopic one, the one that is measured with fMRI.

    About 20 years ago, fMRI was instrumental to a discovery that radically changed the way we think about macroscopic connectivity: Even in the fMRI scans that were taken while subjects were not performing any tasks - until then referred to as "noise scans" - the correlation between fluctuations in the involved brain regions (in this case, motor regions) was very close to that observed during task. On the one hand, this meant that activations that are observed during task could no longer be considered as merely differing from a flat baseline consisting of uncorrelated noise. On the other, it allowed scientists to map distributed brain networks that are jointly involved in similar tasks - i.e. they are functionally connected - without having to design specific experiments. The emerging term "resting state" stands as much for an "idling" brain state as for an experimental paradigm. Immediately the question arises of what supports these functional relationships, and computational models have been invaluable in linking temporal and spatial scales and thus findings from different fields of neuroscience ranging from single cell recordings to fMRI. But it becomes more complicated: About a decade ago, scientists started realizing that this spatially very complex baseline also had a temporal structure, meaning that functional relationships between brain regions change all the time. Chapter 1 of this thesis will introduce "Resting state and its dynamics".

    Chapter 2, "Robust extraction of spatio-temporal patterns from resting state fMRI", contains the first study that was conducted within this PhD. Its goal is to show that a certain dynamic model of resting state, the dynamic mean field model, does not only reproduce the average functional connectivity, as was already shown elsewhere, but also exhibits specific functional patterns over time. It shows how a decomposition technique that is routinely in use in other fields, but so far not in neuroscience, namely tensor factorization, can be used to obtain common sets of regions that tend to be functionally connected from empirical as well as simulated data; these "sets of regions", or communities, are well-known in the literature as resting state networks. A discussion of which consequences our findings have for our understanding of the dynamics present in resting state is also included.

    Chapter 3, "Temporal dynamics of human resting state fMRI", continues where chapter 2 leaves off and characterizes the temporal changes in functional connectivity. We establish suitable measures to track these changes and show that they are not explained by random fluctuations in global connectivity, but are a specific signature of resting state dynamics. This is tied to the temporal evolution of the resting state networks mentioned above. Furthermore, we discuss which properties a model that reproduces these dynamics should have.

    Chapter 4 contains the "General discussion" and ties together all the findings presented in this thesis, including a discussion of future steps.

    The work presented in chapters 2 and 3 could not have been done without collaboration, and two manuscripts resulting from it, carrying the same titles as the chapters of this dissertation, list the authors as follows: Katharina Glomb, Adrián Ponce-Alvarez, Matthieu Gilson, Petra Ritter, and Gustavo Deco.