Localisation de l'activité cérébrale synchrone en neuroimagerie électromagnétique et connectivité fonctionnelle

Abstract

The brain is the physiological interface through which adaptive behavior can be generated in a complex environment. It integrates sensory inputs from different modalities such as vision and audition, and generates internal representations of its environment, encoded in neuronal activity. Modern theories in neuroscience suggest that neural encoding is a selective process aiming at enhancing functionally relevant informations. The mechanisms through which various neural populations, each encoding particular features of sensory inputs, generate an integrated neural representation are the major question behind this thesis. This question is here tackled from the perspective of functional connectivity, i.e. dynamical networks of distributed neural populations sharing information. The goal of functional connectivity is thus to identify neural correlates of brain functions by studying the interactions among the time course of neural activity. This approach is currently a major research field in neuroscience. The pioneering work of Gray et al. (1989) led to the idea that neural synchronization, instead of discharge rate, is a reliable physiological marker of cognitive processes. They showed that a pair of neurons, recorded intracellularly, displays synchronous discharge patterns when encoding functionally related visual information. At a larger scale, recordings of brain hemodynamic activity also showed patterns of neural interactions associated with specific brain functions, thus called functional connectivity. However, neural hemodynamics have much larger time scale than neural interactions which limits their use for understanding neural functional connectivity. In turn, electromagnetic brain activity offers a sufficient temporal resolution for exploring the dynamics of functional connectivity. However, imaging modalities that record electromagnetic activity lack a good spatial resolution because they require solving an underdetermined inverse problem. The main objective of this work is to demonstrate the usefulness of electromagnetic imaging for studying neural functional connectivity. The methodology proposed here aims at pushing the spatial limits of electromagnetic imaging through two complementary approaches : the design of an adaptive filter that reveals synchronous neural activity and tuning the inverse problem solving technique towards adaptively filtered signals. The proposed filtering is based on time-frequency representation of brain signals using complex analytic wavelets. These representations are then used to extract dominant oscillatory modes through the "ridges" techniques. These oscillatory modes, expressed as a complex analytic signal, characterize the neural networks involved in functional connectivity. Subsequently, the developed methodology aims at recovering the synchronous neural networks from the analytical brain signals yielded by electromagnetic imaging. These signals readily express the instantaneous phase and amplitude of neural activity, which allows to characterize neural interactions from the perspective of phase synchrony. This methodology was used to reveal the functional connectivity associated with sleep spindles in humans.

Date	2 Jul 2014
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Jean-Marc Lina (Supervisor) & Boutheina Jemel (Co-supervisor)