EEG & MEG: spatial localization

Lecture: PSY394 Foundations of Neuroimaging

Logan Trujillo, Ph.D.Department of Psychology

UT Austin

Non-Invasive Brain Imaging

–

Good spatial resolution, poor temporalresolution

–

Good temporal resolution, poor spatial resolution

Magnetic Resonance

Imaging (MRI)

FunctionalMRI

Electro-encephalography

(EEG)

Magneto-encephalography

(MEG)

&

&

Figure adapted from Cohen and Bookheimer

(1994)

Small neural source at

cortex

Bioelectric signals are spatially blurred as they pass through the csf, brain, skull, & scalp.

Large spatial blurring at scalp

To understand the reason for the blurring,need to understand biophysics of EEG/MEG

9.67

-9.67

0 μV

EEG measures the electric potential differences across scalp arising from current flow within the head & brain.

V = IR

V = electric potential (“voltage”) analogous towater pressure

I = electric current (analogous to rate of water flow)

R = resistance to flow of electricity (size of pipes)

σ

~ 1/R = conductance of flow of electricity

What are EEG and MEG Signals?

VFigure from Halliday, Resnick, & Krane, 1992)

MEG measures the magnetic fields produced by current flow within the head & brain.

An electric current (flow of electrons) produces a magnetic field around it.

Images from Wikipedia

EEG and MEG signals arise primarily from synaptic activity

Figures from Speckman

and Elger

(1999)

EEG/MEG signals represent the summed activity of large populations of synchronously firing neurons

Figure from Speckmannand Elger(1999)

EEG/MEG sources arewell approximated ascurrent dipoles/dipole layers

Neural dipoles => electromagnetic field changes throughout the head

=> current flow through head and scalp.

-

+

Figure adaptedfrom Spehlman, 1981

Image Adapted from Wikipedia

FigurefromNunez (1995)

2) EEG ~ Gyrus

sourcesMEG ~ Sulci

sources

Figures from Lorente

de No. (1947) & Hubbard et al. (1969).

Dipolar nature of neural sourcesImplies:

1)primarily cortical EEG & MEG signals observable at scalp

EEG signal spatial distortion arises from differencesin head tissue electrical resistance (R) /conductance (σ).

Brain

σ1

CSF

σ2

Skull

σ3

Scalp

σ4

R1 R2 R3R4

Source Point(R0

,θ0

)

Field Point(R4

,θ)

High R (low sigma):=> More collisions of electrons

with tissue molecules & thusgreater deviation from trajectories determined by electric fields of neural sources.

Skull & scalp have lower σ

(high R) than brain and CSF => most spatial blurring of electric current due to skull

& scalp.

Blurring of scalp potentials is even greater.

MEG signals are less affected by conductance/resistance (signal falls off rapidly from source ~ 1/r2

& magnetic fields of impressed currents are very small), & thus have better spatial resolution than EEG.

Figure from Halliday, Resnick, & Krane, 1992)

Outer Scalp

RInner Scalp

Scalp Level Spatial Enhancement of EEG SignalsIt is possible to improve the resolution of scalp EEG

signals by transforming the signals into measures of electric current. (Nunez & Srinivasan, 2005)

R = radius of head

Laplacian

Current Source Density (CSD) Transform

02

2

2

2

2

22 =

∇=

∂Φ∂

+∂Φ∂

+∂Φ∂

=Φ∇σJ

zyxLaplace’s

Eq.

z

x

y

Jz-

=> no net current through spherical scalp “shell”.0=

∇σJ

J ~ I/m2

Jz+

Jx-

Jx+

Jy-

Jy+

ScalpLaplacian ⎟⎟

⎠

⎞⎜⎜⎝

⎛∂Φ∂

+∂Φ∂

−=∂Φ∂

=yxz

yxL 2

2

2

2

2

2

),(

z

x

y

ΔΦx

ΔΦ y

ΔΦz

radial electric currentthrough scalp

~

02

2

2

2

2

2

=∂Φ∂

+∂Φ∂

+∂Φ∂

zyx

ModelSource

Distribution

IsopotentialMap

LaplacianMap

Law, Nunez, and Wijesinghe, R.S. (1993)

EEG/MEG Source Reconstruction-

estimating cortical source distribution of scalp

recorded electromagnetic signals

Source reconstruction principles are the same for EEG & MEG; only differ in the mathematical details.

Usually performed on ERPs

in order to reduce effects of noise on source solutions.

The Inverse Problem:

Any measured scalp electromagnetic distribution may be described by an infinite number of possible source distributions

Must use functional–anatomical criteria to constrain the number of possible source distributions

?=

1) 2)

or 3)

Two steps to EEG/MEG Source Reconstruction:

1)Create a model of brain, skull, and scalp.

2)Use this model to solve for “best fitting” source distributions, where “best fit” determined by

i) minimization of error btw model prediction & actual measurements.

ii) Neuroanatomical

constraints (e.g. expectedregions of activation based on nature of stimulus & task).

Head (Volume Conductor) ModelsTypical models in EEG & MEG: spherical, ellipsoidal

Brain

σ1

CSF

σ2

Skull

σ3

Scalp

σ4

R1 R2 R3R4

Source Point(R0

,θ0

)

Field Point(R4

,θ)

Structural MRI-based

models

MRI informs creation of head model

Individual subject:Volumetric boundary element method

=>

Group Analyses:Use MNI or Tailarach

brain 3D templates

Requires spatial co-registration of EEG-fMRI coordinate systems

LORETA-KEY figures

Electrode coordinates may be default 10-20 International System

Electrode coordinates can be scanned in from actual scalp locations via fMRI

scanner or

through use of 3D scanning technology

Dipole modeling

(Scherg, 1989)

Dipole position and moment is what gives best “fit” to scalp EEG/ERP data.

Assume a given brain response arises from brainelectrical activity equivalent to a few EM dipoles.

Dipole Current Source

P100 (rv

= 1.6) N170(rv

= 2.17)

-100 -50 0 50 100 150 200 Time (ms)

ERPs

may reflect activation of more than one brain region.

UP

Independent Components Analysis (ICA)decomposes signals into maximally temporally independent sub-signals.

P1 & N1 responses composed of early/late ICA components with different cortical locations.

Deblurring

Method (Le and Gevin, 1993)

02 ≠∇

=Φ∇σJ Poisson’s Eq.

=> net current throughvolume defined by scalp and cortical surfaces

Cortical surface expresses current distribution from active neural sources

Current flowsthrough scalpsurface, but does not leave it

0≠∇σJ

Cortex

Lead Field Method –

can estimate current sources in 3D space

Forward solutionΦ

= KJ is scalp potential

K represents the “lead field”, i.e. the biophysical volumetric model of scalp, skull, & brain.

J is vector of current elements, with 1-3 elements (depending on the model) per “voxel”.

Inverse Solution is thenJ = TΦ

where T = f(K)

Inverse Problem => must choose T that gives the “best” solution.

Several different methods to choose T.

Minimum Norm

Solution (Hämäläinen

and Ilmoniemi, 1994)

Notice small degree of localization error. Could be due to poor head model and/or limitations of Min Norm algorithm.

1

-1

0 A/m2

Limitations of Minimum Norm:Can sometimes exhibit large spatial error when localizing maxima of activations, especially with deep dipole sources.

Best source solutions obtained when recording EEG & MEG simultaneously & when temporal information taken into account (Dale & Sereno, 1993)

RadialDipole

EEG& MEG

EEGonly

MEGonly

Dale & Sereno(1993)

TangentialDipole

EEG& MEG

EEGonly

MEGonly

Dale & Sereno(1993)

EEG& MEG

Dale & Sereno(1993)

Deep Radial Dipole

EEG & MEG withInformation from other timepoints

taken into account.

Low Resolution Electromagnetic Tomography (LORETA) (Pascual-Marqui

et al., 1994)

LORETA solutions have the lowest activity maxima localization error of “lead field” methods (Pascual-Marqui,

1994), but solutions are more spread out spatially, hence the term “Low Resolution”.

0 A/m2

1

1) Only as good as head model2) Strong effect of data variance on the reconstruction

algorithms3) Inverse problem restrictions4) Intrinsic neuroanatomical

limitations of EEG (gyri)

& MEG (sulci). Best results when using EEG & MEG simultaneously (Dale & Sereno, 1993)

In the end: ~ same order of temporal resolutionOne -

two orders of magnitude better spatial

resolution then straight scalp topography

METHODLOGICAL LIMITATIONS

This problem may be reduced by new technology allowing simultaneously recording of EEG & fMRI

with task

paradigms involving irreversible cognitive/perceptual changes.

fMRI

& EEG/MEG Source LocalizationMRI technology can also play a role besides head model

creation.

fMRI

activations can be used to constrain dipole solutions, reduce source solution space created by inverse problem, & provide convergent evidence for a particular source solution.

Caveat: This role is limited in non-simultaneous combination of EEG & fMRI

with task paradigms

involving irreversible cognitive/perceptual changes.

Simultaneous method: fMRI

signal is modeled and subtracted from raw EEG signal

EX: Neuroscan

MAGLINK systemUncorrected/Corrected 64 Channel Spontaneous EEG

REFERENCESCohen, M.S. and Bookheimer, S.Y. (1994). Localization of brain functions using magnetic

resonance imaging. Trends Neurolog. Sci., 17, 268-77.

Dale, A.M. and M.I. Sereno

(1993) Improved localization of cortical activity by combining EEG and MEG with MRI cortical surface reconstruction: A linear approach. J. Cogn. Neurosci. 5, 162-176

Halliday, D., Resnick, R., and Krane, K.S. (1992). Physics, 4th

Edition. New York: John Wiley & Sons.

Hamalainen, M.S. and Ilmoniemi, R.J. (1994). Interpreting magnetic fields of the brain: minimum norm estimates. Med. Biol. Eng. Comp. 32, 35-42.

Hubbard, J.I., Llinas, R., and Quastel, D.M.J. (1969). Electrophysiological Analysis of Synaptic Transmission. London: Edward Arnold, Ltd.

Law, S.K., Nunez, P.L., and Wijesinghe, R.S. (1993). High resolution EEG using spline

generated surface laplacians

on spherical and ellipsoidal surfaces. IEEE Trans. Biomed, Eng. 40, 145-153.

Le, J. and Gevins, A.S. (1993). Method to reduce blur distortion from EEG’s using

a realistic head model. IEEE Trans. Biomed. Eng. 40, 517-528.

Lorete

do No., R. (1947). Action potential of the motoneurons

of the hypoglossus

nucleus. J. Cell. Comp. Physiol. 29, 207-287.

Nunez, P.L. (1995). Neocortical Dynamics and Human EEG Rhythms. New York: Oxford University Press.

Nunez, P.L. and Srinivasan, R.R. (2005). Electric Fields of the Brain: The Neurophysics of EEG, 2nd Edition. New York: Oxford University Press.

Pascual-Marqui

RD. (1999) Review of methods for solving the EEG inverse problem. International Journal of Bioelectromagnetism 1, 75-86.

Pascual-Marqui, R.D., M. Esslen, K. Kochi, D. Lehmann. (2002). Functional imaging with low resolution brain electromagnetic tomography (LORETA): a review. Methods & Findings in Experimental & Clinical Pharmacology 24C, 91-95.

Pascual-Marqui

RD, Michel CM, Lehmann

D. Low resolution electromagnetic tomography: a new method for localizing electrical activity in the brain. International Journal of Psychophysiology. 1994, 18:49-65.

Scherg, M. (1989). Fundamentals of dipole source potential analysis. In: Auditory Evoked Magnetic Fields and Potentials. Advances in Neurology. Vol. 66 (Hoke, M., Grandori, F., Romani, G.L. eds.), Basel: Karger.

Speckmann, E-J and Elger, C.E. (1999). Introduction to the neurophsyiological

basis of the EEG and DC potentials. In: Electroencephalography: Basic principles, Clinical Applications, and Related Fields, 4th Edition. (Niedermeyer, E. and Lopes Da

Silva, F., eds.), Baltimore: Williams & Wilkins.

Spehlmann, R. (1981). EEG Primer. New York: Elsevier.