EEG/MEG Source Localisation

36
EEG/MEG Source Localisation SPM Course – Wellcome Trust Centre for Neuroimaging – Oct. 2008 ? Jérémie Mattout, Christophe Phillips Jean Daunizeau Guillaume Flandin Karl Friston Rik Henson Stefan Kiebel Vladimir Litvak

description

EEG/MEG Source Localisation. SPM Course – Wellcome Trust Centre for Neuroimaging – Oct. 2008. ?. Jérémie Mattout, Christophe Phillips. Jean Daunizeau Guillaume Flandin Karl Friston Rik Henson Stefan Kiebel Vladimir Litvak. EEG/MEG Source localisation. Outline. Introduction - PowerPoint PPT Presentation

Transcript of EEG/MEG Source Localisation

Page 1: EEG/MEG Source Localisation

EEG/MEG Source LocalisationSPM Course – Wellcome Trust Centre for Neuroimaging – Oct. 2008

??

Jérémie Mattout, Christophe Phillips

Jean DaunizeauGuillaume FlandinKarl FristonRik HensonStefan KiebelVladimir Litvak

Page 2: EEG/MEG Source Localisation

OutlineEEG/MEGEEG/MEG

Source localisationSource localisation

1. Introduction

2. Forward model

3. Inverse problem

4. Bayesian inference applied to the EEG/MEG inverse problem

5. Conclusion

Page 3: EEG/MEG Source Localisation

OutlineEEG/MEGEEG/MEG

Source localisationSource localisation

1. Introduction

2. Forward model

3. Inverse problem

4. Bayesian inference applied to the EEG/MEG inverse problem

5. Conclusion

Page 4: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation Introduction: overview

Page 5: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation

EEG/MEG source reconstruction process

Forwardmodel

Inverseproblem

Introduction: overview

Page 6: EEG/MEG Source Localisation

OutlineEEG/MEGEEG/MEG

Source localisationSource localisation

1. Introduction

2. Forward model

3. Inverse problem

4. Bayesian inference applied to the EEG/MEG inverse problem

5. Conclusion

Page 7: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation Forward model: formulation

EJfY

Forwardmodel

data sourceparameters

noiseforwardoperator

Page 8: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation

source biophysical model: current dipole

EEG/MEG source models

EquivalentCurrent

Dipoles (ECD)

Imaging orDistributed

Forward model: source space

- few dipoles withfree location and orientation

- many dipoles withfixed location and orientation

Page 9: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation Forward model: imaging/distributed model

EKJY

data dipoleamplitudes

noisegain matrix

Page 10: EEG/MEG Source Localisation

OutlineEEG/MEGEEG/MEG

Source localisationSource localisation

1. Introduction

2. Forward model

3. Inverse problem

4. Bayesian inference applied to the EEG/MEG inverse problem

5. Conclusion

Page 11: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation

« Will it ever happen that mathematicians will know enough about the physiology of the brain, and neurophysiologists enough of mathematical discovery, for efficient cooperation to be possible ? »

Jacques Hadamard (1865-1963)

Inverse problem: an ill-posed problem

Inverseproblem

1. Existence2. Unicity3. Stability

Page 12: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation Inverse problem: an ill-posed problem

« Will it ever happen that mathematicians will know enough about the physiology of the brain, and neurophysiologists enough of mathematical discovery, for efficient cooperation to be possible ? »

Jacques Hadamard (1865-1963)

1. Existence2. Unicity3. Stability

Inverseproblem

Page 13: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation Inverse problem: an ill-posed problem

« Will it ever happen that mathematicians will know enough about the physiology of the brain, and neurophysiologists enough of mathematical discovery, for efficient cooperation to be possible ? »

Jacques Hadamard (1865-1963)

1. Existence2. Unicity3. Stability

Inverseproblem

Introduction of prior knowledge (regularization) is needed

Page 14: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation Inverse problem: regularization

Data fit

Adequacywith other

modalities

Spatial and temporal priors

W = I : minimum norm

W = Δ : maximum smoothness (LORETA)

data fit prior(regularization term)

Page 15: EEG/MEG Source Localisation

OutlineEEG/MEGEEG/MEG

Source localisationSource localisation

1. Introduction

2. Forward model

3. Inverse problem

4. Bayesian inference applied to the EEG/MEG inverse problem

5. Conclusion

Page 16: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation Bayesian inference: probabilistic formulation

likelihood prior

posteriorevidence

Forwardmodel

Inverseproblem posterior

likelihood

Page 17: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation Bayesian inference: hierarchical linear model

sensor (1st) level

source (2nd) level

Q : (known) variance components

(λ,μ) : (unknown) hyperparameters

likelihood

prior

qeqee QQC 1

1

kpkpp QQC 1

1

Page 18: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation Bayesian inference: variance components

Multiple Sparse Priors(MSP)

# dipoles

# d

ipo

les

Minimum Norm(IID)

Maximum Smoothness(LORETA)

kpkpp QQC 1

1

Page 19: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation Bayesian inference: iterative estimation scheme

M-step estimate while keeping constants

E-step estimate while keeping constants,J

Expectation-Maximization (EM) algorithm

J,

Page 20: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation Bayesian inference: model comparison

)()()|(log McomplexityMaccuracyMYpF

model Mi

Fi

1 2 3

At convergence

Page 21: EEG/MEG Source Localisation

OutlineEEG/MEGEEG/MEG

Source localisationSource localisation

1. Introduction

2. Forward model

3. Inverse problem

4. Bayesian inference applied to the EEG/MEG inverse problem

5. Conclusion

Page 22: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation Conclusion: At the end of the day...

Somesthesic data Bilateral auditory tone

Page 23: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation Conclusion: At the group level...

RL

Individual reconstructions in MRI template space

Group resultsp < 0.01 uncorrectedR L

Page 24: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation Conclusion: Summary

• Prior information is mandatory

• EEG/MEG source reconstruction:1. forward model2. inverse problem (ill-posed)

• Bayesian inference is used to:1. incorpoate such prior information…2. … and estimating their weight w.r.t the data3. provide a quantitative feedback on model adequacy

Forwardmodel

Inverseproblem

Page 25: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation

source biophysical model: current dipole

EEG/MEG source models

EquivalentCurrent

Dipoles (ECD)

Imaging orDistributed

Equivalent Current Dipole (ECD) solution

few dipoles with free

location and orientation

many dipoles with fixed location and orientation

Page 26: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation ECD approach: principle

EJfY

Forwardmodel

data dipoleparameters

noiseforwardoperator

but a priori fixed number of sources considered iterative fitting of the 6 parameters of each dipole

Page 27: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation

The locations s and moments w are drawn from normal distributions with precisions γs and γw.

E is white observation noise with precision γy.

w s

Y

w s

yThese are drawn from a prior gamma distribution.

EwsgEJfY )()(

Dipole J with location s and moment w

generated data Y using

ECD solution: variational Bayes (VB) approach

Page 28: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation ECD solution: “classical” vs. VB approaches

“Classical” VB

Hard constraints Yes Yes

Soft constraints No Yes

Noise accommodation

No (in general)

Yes

Model comparison

No YES

Page 29: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation

• can be applied to single time-slice data or average over time (MEG and EEG)

• useful for comparing several few-dipole solutions for selected time points (N100, N170, etc.)

• although not dynamic, can be used for building up intuition about underlying generators, or using as a motivation for DCM source models

• implemented in Matlab and available in SPM8b

ECD solution: when and how to apply VB-ECD?

Page 30: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation Example 1: somestesic stimulation

Scalp distribution, 21ms post-stimulus

ERP data over 64 channels VB-ECD solution

Page 31: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation Example 2: auditory oddball

Oddball stimuli Standard stimuli

Scalp potential for auditory stimulations

Page 32: EEG/MEG Source Localisation

Main referencesEEG/MEGEEG/MEG

Source localisationSource localisation

Litvak and Friston (2008) Electromagnetic source reconstruction for group studies

Friston et al. (2008) Multiple sparse priors for the M/EEG inverse problem

Kiebel et al. (2008) Variational Bayesian inversion of the equivalent current dipole model in EEG/MEG

Mattout et al. (2007) Canonical Source Reconstruction for MEG

Daunizeau and Friston (2007) A mesostate-space model for EEG and MEG

Henson et al. (2007) Population-level inferences for distributed MEG source localization under multiple constraints: application to face-evoked fields

Friston et al. (2007) Variational free energy and the Laplace approximation

Mattout et al. (2006) MEG source localization under multiple constraints

Friston et al. (2006) Bayesian estimation of evoked and induced responses

Phillips et al. (2005) An empirical Bayesian solution to the source reconstruction problem in EEG

Page 33: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation

Page 34: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation

),0(~),( CNMJp

)exp(kk

),(~ Nk

- Log-normal hyperpriors- Enforces the non-negativity of the hyperparameters- Enables Automatic Relevance Determination (ARD)

Bayesian inference: multiple sparse priors

Page 35: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation

SubjectsMRI Anatomical warping

Corticalmesh

Canonicalmesh

[Un]-normalisingspatial transformation

MNI Space

Forward model: canonical mesh

Page 36: EEG/MEG Source Localisation

EEG/MEGEEG/MEGSource localisationSource localisation

From Sensor to MRI space

MRI derived meshes

MEG

Full setup

EEG

RigidTransformation

HeadShape

SurfaceMatching

+

HeadShape

Forward model: coregistration