Elsevier

Clinical Neurophysiology

Volume 115, Issue 10, October 2004, Pages 2195-2222
Clinical Neurophysiology

Invited review
EEG source imaging

https://doi.org/10.1016/j.clinph.2004.06.001Get rights and content

Abstract

Objective: Electroencephalography (EEG) is an important tool for studying the temporal dynamics of the human brain's large-scale neuronal circuits. However, most EEG applications fail to capitalize on all of the data's available information, particularly that concerning the location of active sources in the brain. Localizing the sources of a given scalp measurement is only achieved by solving the so-called inverse problem. By introducing reasonable a priori constraints, the inverse problem can be solved and the most probable sources in the brain at every moment in time can be accurately localized.

Methods and Results: Here, we review the different EEG source localization procedures applied during the last two decades. Additionally, we detail the importance of those procedures preceding and following source estimation that are intimately linked to a successful, reliable result. We discuss (1) the number and positioning of electrodes, (2) the varieties of inverse solution models and algorithms, (3) the integration of EEG source estimations with MRI data, (4) the integration of time and frequency in source imaging, and (5) the statistical analysis of inverse solution results.

Conclusions and Significance: We show that modern EEG source imaging simultaneously details the temporal and spatial dimensions of brain activity, making it an important and affordable tool to study the properties of cerebral, neural networks in cognitive and clinical neurosciences.

Introduction

Distributed neuronal networks assure correct functioning of the human brain (Mesulam, 1998, Lutkenhoner and Grave de Peralta Menendez, 1997, Seeck et al., 1997). Inhibitory and excitatory feedforward and feedback processes are the basic mechanisms of interaction between different modules of these networks (Bullier, 2001). Localizing the different modules of the functional network implicated in a given mental task is the principal aim of functional neuroimaging studies. A large body of research, using positron emission topography (PET) and functional magnetic resonance imaging (fMRI), has been devoted to this aim (Cabeza and Nyberg, 2000). However, these methods are not the most suitable for addressing the question of when during the mental task the different modules become active and hence in what processing step(s) each module is involved. Nor can they readily answer the important questions of sequential versus parallel activation, feedforward versus feedback processes, or how information is ‘bound’ together to form unified percepts.

In order to investigate such temporal properties of brain circuits, methods that directly measure neuronal activity in real time are needed. Electro- and magneto-encephalography (EEG, MEG) offer this possibility by measuring the electrical activity of neuronal cell assemblies on a submillisecond time scale. Unfortunately, these techniques face the problem that the signals measured on the scalp surface do not directly indicate the location of the active neurons in the brain due to the ambiguity of the underlying static electromagnetic inverse problem (Helmholtz, 1853). Many different source configurations can generate the same distribution of potentials and magnetic fields on the scalp (for a review see Fender (1987)). Therefore, maximal activity or maximal differences at certain electrodes do not unequivocally indicate that the generators were located in the area underlying it. However, and as will be discussed in further detail below, the converse holds: different scalp topographies must have been generated by different configurations of brain sources. Capitalizing on this fact, a first step in defining whether and when different neuronal populations were activated over time or between experimental or pathological conditions is to identify differences in scalp topographies. Spatial enhancement algorithms, such as current source density calculations or deblurring (Nunez, 1981, Gevins et al., 1991, Babiloni et al., 1996, He et al., 2001) can help for this purpose.

While the analysis of the scalp potential or magnetic field distribution is the precursor for source localization, it does not provide conclusive information about the location and distribution of the sources. The only way to localize the putative electric sources in the brain is through the solution of the so-called inverse problem, a problem that can only be solved by introducing a priori assumptions on the generation of EEG and MEG signals. The more appropriate these assumptions are the more trustable are the source estimations. During the last two decades different such assumptions have been formulated and implemented in inverse solution algorithms. They range from single equivalent current dipole estimations to the calculation of three-dimensional (3D) current density distributions. Each approach uses different mathematical, biophysical, statistical, anatomical or functional constraints.

Several reviews on EEG/MEG source imaging exist, that explain in detail the formal implementation of the a priori constraints in the different algorithms (Hämäläinen et al., 1993, George et al., 1995, Grave de Peralta Menendez and Gonzalez Andino, 1998, Gonzalez Andino et al., 1999, Michel et al., 1999a, Fuchs et al., 1999, Baillet et al., 2001, He and Lian, 2002). While these rather mathematically oriented reviews are of utmost importance for the specialist in inverse solutions, the practical user might be more interested in a summary of the critical requirements for successful source localization. In fact, electromagnetic source imaging should involve many more analysis steps than applying a given source localization algorithm to the data. Each step should be carefully considered and selected on the basis of the information one would like to obtain from the measurements. The judgment on the validity of the results presented in a given study should be based on all these points, and not only on the choice of the inverse solution algorithm, because they are intimately linked. This review therefore not only discusses inverse solution algorithms, but also critical issues in steps preceding and following source estimation, such as the number and positioning of electrodes (including the reference electrode) and the determination of relevant time points or periods for source localization. We also discuss how the source estimations can be integrated with MRI and how one can go beyond simple pictures of inverse solutions by analyzing source localizations statistically.

This review concentrates on EEG recordings, though most of the aspects discussed here similarly concern MEG. Similarities and differences between EEG and MEG have been discussed elsewhere (e.g. Anogianakis et al., 1992, Wikswo et al., 1993, Malmivuo et al., 1997, Liu et al., 2002; see also discussion in Barkley and Baumgartner (2003)).

EEG source imaging is not only used in cognitive neuroscience research, but has also found important applications in clinical neuroscience such as neurology, psychiatry and psychopharmacology. In cognitive neuroscience, the majority of the studies investigate the temporal aspects of information processing by analyzing event related potentials (ERP). In neurology, the study of sensory or motor evoked potentials is of increasing interest, but the main clinical application concerns the localization of epileptic foci. In psychiatry and psychopharmacology, a major focus of interest is the localization of sources in certain frequency bands. While the issue of source localization is similar for these different applications, the pre-processing of the data is somewhat different.

Section snippets

Number and positioning of the electrodes

This section discusses some of the basic questions regarding the recording of the data for EEG source imaging. It concerns the number and the distribution of the electrodes on the scalp and the spatial normalization of the individual potential maps for group averages. We will show that localization precision of epileptic sources drastically increases from 31 to 63 electrodes and also, though less drastically, from 63 to 123 electrodes. We also give an example that illustrates the importance of

The choice of the inverse model

This section will give an overview on some of the currently available source localization algorithms. Generally speaking, these algorithms try to most optimally explain the scalp potential field by intracranial sources. The fundamental problem of EEG/MEG source reconstruction is the ambiguity of the electromagnetic inverse problem (Helmholtz, 1853). That is, a given electric potential or magnetic field recorded at the scalp can be explained by the activity of infinite different configurations

Integration of EEG source imaging with MRI

The ultimate goal of modern EEG source imaging is the localization of the EEG sources in anatomically defined brain structures so that direct comparison with other imaging methods, with lesion studies, or with intracranial recordings can be made. Most of the commercial and freely available software packages provide this possibility. Most often, it is mainly applied for visualization purposes. More recently, this information has been used to define the coordinates of the sources in terms of

Incorporating time and frequency in source imaging

While the high temporal resolution is considered as the most important advantage of EEG/MEG measures, it carries the side effect of greatly increasing the amount of data collected, and consequently of introducing a temporal dimension to the data analysis procedures. As such, experimenters must devise methods for determining the relevant events within a continuous time series of data to which source analysis will be applied. Some of these methods are heavily influenced by the experimenter,

Post-processing of EEG source images

While clinical EEG source imaging studies typically apply source localization algorithms to the individual patient's data, most experimental studies limit source localization to only the group-averaged data in order to illustrate putative sources at certain time points/periods (using spatiotemporal models) or their activation sequence over time. In this case, no indication about the statistical reliability of these sources or their correspondence to the actual source locations in the individual

Conclusion

EEG source imaging has made tremendous progress in recent years to provide statistically based neurophysiological interpretations of scalp recordings. To achieve such requires that researchers abandon ambiguous waveform analyses and the phenomenological description of grapho-elements. Instead, comprehensive analyses of the electric field at the scalp must be conducted that serve as the basis for estimating the sources underlying these fields. Many recent publications using new source analysis

Acknowledgements

The Swiss National Science Foundation supported the studies reported in this article. We thank Denis Brunet for his excellent software Cartool, Olaf Blanke, Christine Ducommun, Asaid Khateb, and Gregor Thut for technical assistance in data recording and analysis, and Margitta Seeck and Theodor Landis for their important input on clinical issues.

References (254)

  • A.M. Dale et al.

    Dynamic statistical parametric mapping: combining fMRI and MEG for high resolution imaging of cortical activity

    Neuron

    (2000)
  • J.C. De Munck et al.

    A practical method for determining electrode positions on the head

    Electroencephalogr Clin Neurophysiol

    (1991)
  • J.E. Desmedt et al.

    Beware of the average reference in brain mapping (Review)

    Electroencephalogr Clin Neurophysiol Suppl

    (1990)
  • A. Devor et al.

    Coupling of total hemoglobin concentration, oxygenation, and neural activity in rat somatosensory cortex

    Neuron

    (2003)
  • V. Diekmann et al.

    Localisation of epileptic foci with electric, magnetic and combined electronic models

    Electroencephalogr Clin Neurophysiol

    (1998)
  • T. Dierks et al.

    Brain electrical activity in depression described by equivalent dipoles

    J Affect Disord

    (1993)
  • T. Dierks et al.

    Dementia of the Alzheimer type: effects on the spontaneous EEG described by dipole sources

    Psychiatry Res

    (1993)
  • T. Dierks et al.

    Electrical brain activity in schizophrenia described by equivalent dipoles of FFT-data

    Schizophr Res

    (1995)
  • C.Y. Ducommun et al.

    Segregated processing of auditory motion and auditory location: an ERP mapping study

    Neuroimage

    (2002)
  • F.H. Duffy et al.

    Significance probability mapping: an aid in the topographic analysis of brain electrical activity

    Electroencephalogr Clin Neurophysiol

    (1981)
  • A.J. Fallgatter et al.

    Brain electrical dysfunction of the anterior cingulate in schizophrenic patients

    Psychiatry Res

    (2003)
  • J.J. Foxe et al.

    Right hemisphere control of visuospatial attention: line-bisection judgments evaluated with high-density electrical mapping and source analysis

    Neuroimage

    (2003)
  • M. Fuchs et al.

    Boundary element method volume conductor models for EEG source reconstruction

    Clin Neurophysiol

    (2001)
  • M. Fuchs et al.

    A standardized boundary element method volume conductor model

    Clin Neurophysiol

    (2002)
  • N.G. Gencer et al.

    Optimal reference electrode selection for electric source imaging

    Electroencephalogr Clin Neurophysiol

    (1996)
  • I.F. Gorodnitsky et al.

    Neuromagnetic source imaging with FOCUSS: a recursive weighted minimum norm algorithm

    Electroencephalogr Clin Neurophysiol

    (1995)
  • O. Hauk et al.

    Comparison of data transformation procedures to enhance topographical accuracy in time-series analysis of the human EEG

    J Neurosci Methods

    (2002)
  • B. He et al.

    High-resolution EEG a new realistic geometry spline Laplacian estimation technique

    Clin Neurophysiol

    (2001)
  • C. Huang et al.

    Discrimination of Alzheimer's disease and mild cognitive impairment by equivalent EEG sources: a cross-sectional and longitudinal study

    Clin Neurophysiol

    (2000)
  • H.J. Huppertz et al.

    Cortical current density reconstruction of interictal epileptiform activity in temporal lobe epilepsy

    Clin Neurophysiol

    (2001)
  • H.J. Huppertz et al.

    Localization of interictal delta and epileptiform EEG activity associated with focal epileptogenic brain lesions

    Neuroimage

    (2001)
  • S.P. Ahlfors et al.

    Spatiotemporal activity of a cortical network for processing visual motion revealed by MEG and fMRI

    J Neurophysiol

    (1999)
  • J.P. Ary et al.

    Location of sources of evoked scalp potentials: corrections of skull and scalp thickness

    IEEE Trans Biomed Eng

    (1981)
  • G.E. Backus et al.

    Uniqueness in the inversion of gross earth data

    Phil Trans R Soc

    (1970)
  • S. Baillet et al.

    A Bayesian approach to introducing anatomo-functional priors in the EEG/MEG inverse problem

    IEEE Trans Biomed Eng

    (1997)
  • S. Baillet et al.

    Electromagnetic brain mapping

    IEEE Signal Processing Magazine

    (2001)
  • G.L. Barkley et al.

    MEG and EEG in epilepsy (Review)

    J Clin Neurophysiol

    (2003)
  • G.R. Barnes et al.

    Statistical flattening of MEG beamformer images

    Hum Brain Mapp

    (2003)
  • C. Baumgartner

    Clinical applications of magnetoencephalography (Editorial)

    J Clin Neurophysiol

    (2000)
  • C.G. Benar et al.

    Non-uniform spatial sampling in EEG source analysis

    Proc 23rd Conf IEEE-EMBS

    (2001)
  • O. Bertrand et al.

    Finite element method in brain electrical activity studies

  • O. Blanke et al.

    Visual activity in the human frontal eye field

    Neuroreport

    (1999)
  • P. Boon et al.

    Interictal and ictal source localization in neocortical versus medial temporal lobe epilepsy

    Adv Neurol

    (2000)
  • D. Brandeis et al.

    Scopolamine effects on visual information processing, attention, and event-related potential map latencies

    Psychophysiology

    (1992)
  • R. Cabeza et al.

    Imaging cognition. II. An empirical review of 275 PET and fMRI studies

    J Cogn Neurosci (Review)

    (2000)
  • D. Cabrera Fernandez et al.

    Some limitations of spatio-temporal source models

    Brain Topogr

    (1995)
  • N.B. Cuffin

    Effects of head shape on EGG's and MEG's

    IEEE Trans Biomed Eng

    (1990)
  • N.B. Cuffin

    Effects of local variations in skull and scalp thickness on EEG's and MEG's

    IEEE Trans Biomed Eng

    (1993)
  • N.B. Cuffin

    EEG localization accuracy improvements using realistically shaped head models

    IEEE Trans Biomed Eng

    (1996)
  • B.N. Cuffin

    Effects of modeling errors and EEG measurement montage on source localization accuracy

    J Clin Neurophysiol

    (2001)
  • Cited by (1414)

    View all citing articles on Scopus
    1

    Present address: The Functional Electrical Neuroimaging Laboratory, Division Autonome de Neuropsychologie and Service de Radiodiagnostic et Radiologie Interventionnelle, Centre Hospitalier Universitaire Vaudois, Hôpital Nestlé, 5 Av. Pierre-Decker, 1011 Lausanne, Switzerland.

    View full text