Review of advanced techniques for the estimation of brain connectivity measured with EEG/MEG

Abstract

Brain connectivity can be modeled and quantified with a large number of techniques. The main objective of this paper is to present the most modern and widely established mathematical methods for calculating connectivity that is commonly applied to functional high resolution multichannel neurophysiological signals, including electroencephalographic (EEG) and magnetoencephalographic (MEG) signals. A historical timeline of each technique is outlined along with some illustrative applications. The most crucial underlying assumptions of the presented methodologies are discussed in order to help the reader understand where each technique fits into the bigger picture of measuring brain connectivity. In this endeavor, linear, nonlinear, causality-assessing and information-based techniques are summarized in the framework of measuring functional and effective connectivity. Model based vs. data-driven techniques and bivariate vs. multivariate methods are also discussed. Finally, certain important caveats (i.e. stationarity assumption) pertaining to the applicability of the methods are also illustrated along with some examples of clinical applications.

Introduction

There has been a growing interest in studying both normal and pathological brain function with respect to identifying variations in activation within and interactions between brain areas. Understanding and modeling brain function is based not only on the correct identification of the active brain regions, but also on the functional interactions among the neural assemblies distributed across different brain regions. The aforementioned concepts are addressed in theoretical neuroscience, as the functional segregation (activation of specialized brain regions/neural assemblies) and integration (coordinated activation of very large numbers of neural assemblies distributed across different cortical areas that constitute large-scale distributed systems of the cerebral cortex) principles [1].

Integration of cerebral areas can be measured by assessing brain connectivity. Brain connectivity can be subdivided into neuroanatomical (or structural), functional and effective connectivity. Neuroanatomical connectivity is inherently difficult to define given the fact that at the microscopic scale of neurons, new synaptic connections or elimination of existing ones are formed dynamically and are largely dependent on the function executed [2]. But for the sake of simplicity structural connectivity may be considered as fiber pathways tracking over extended regions of the brain, which are in accordance with general anatomical knowledge [3]. Magnetic Resonance Imaging (MRI) and especially Diffusion Tensor Imaging (DTI) can be used to examine structural connectivity and convey information concerning the white matter fiber tracts. Techniques for measuring neuroanatomical connectivity are discussed in other articles within this special issue.

Functional connectivity is defined as the temporal correlation (in terms of statistically significant dependence between distant brain regions) among the activity of different neural assemblies [4]. Many neurophysiologic signals can be assessed with functional connectivity techniques, including signals derived from single unit and local field potential (LFP) recordings, Electroenchaphalography (EEG), Magnetoencephalography (MEG), Positron Emission Tomography (PET) and Functional Magnetic Resonance Imaging (fMRI).

Effective connectivity is a relatively new concept defined as the direct or indirect influence that one neural system exerts over another [5]. It describes the dynamic directional interactions among brain regions. Effective connectivity can be estimated from the signals directly (i.e. data-driven) or can be based on a model specifying the causal links (i.e. model-based combination of both structural and functional connectivity).

Several different modalities can be used to assess brain connectivity. fMRI is widely used mostly due to the large availability of MRI scanners. fMRI provides a high spatial resolution (1–10 mm), while EEG/MEG has more limited spatial resolution (1–10 cm). On the other hand, fMRI has a limited temporal precision (∼1 s), primarily due to the limitations of the hemodynamic response, while EEG/MEG has high temporal precision of the EEG and MEG techniques (<1 ms). Because functional and effective connectivity techniques are largely dependent on calculating the correspondence of neural signals over time, techniques such as EEG and MEG, which have excellent temporal resolution, are optimal for calculating such connectivity.

This review focuses on the most promising methodologies for assessing functional and effective connectivity from EEG or MEG signals. The introductory section provides an overview of brain connectivity, whereas Section 2 provides a historical and methodological perspective of different families of functional and effective connectivity techniques. Section 3 discusses the merits and the limitations of these techniques. The underlying assumptions of each technique are also discussed along with some illustrative clinical paradigms. Finally, the fourth section concludes this review and points out future research directions.