General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

Users may download and print one copy of any publication from the public portal for the purpose of private study or research.

You may not further distribute the material or use it for any profit-making activity or commercial gain

You may freely distribute the URL identifying the publication in the public portal If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from orbit.dtu.dk on: Dec 08, 2021

Forward Models can be Inferred from EEG Data

Hansen, Sofie Therese; Hauberg, Søren; Hansen, Lars Kai

Publication date:2016

Document VersionPublisher's PDF, also known as Version of record

Link back to DTU Orbit

Citation (APA):Hansen, S. T., Hauberg, S., & Hansen, L. K. (2016). Forward Models can be Inferred from EEG Data. Postersession presented at 22nd Annual Meeting of the Organization for Human Brain Mapping, Geneva, Switzerland.

Forward Models can be Inferred from EEG DataSOFIE THERESE HANSEN (sofha@dtu.dk), SØREN HAUBERG (sohau@dtu.dk)

AND LARS KAI HANSEN (lkai@dtu.dk)Department of Applied Mathematics and Computer Science,

Technical University of Denmark

MOTIVATIONAccurate 3D EEG imaging is contingent on a suitable forward model [1, 2].The forward model describes the propagation path of the EEG sources tothe EEG sensors [3]. Forward models are estimated based on head geome-try and conductivity assumptions thus requiring subject-specific information.We propose an alternative: Learn a forward model based on the EEG data ofthe subject and a data-driven prior over forward models.

FORWARD MODEL REPRESENTATION

ElectrodesScalp

OutermskullInnermskull

Cortex

Highmconductivity

Lowmconductivity

PCAmrepresentationmofmforwardmmodelmcorpus

MRImscans BEMmheadmmodel1 2

3

InfermamforwardmmodelmformamnewmsubjectmusingmEEGmdata

Generatemforwardmmodelsmwithmdifferentmconductivities

Segment

1. Structural scans of 16 subjects are obtained from the multi-subject mul-timodal neuroimaging dataset [4].

2. The sMRIs are for each subject segmented into cortex, skull and scalp.3. Each head model is combined with different skull:soft tissue conductiv-

ity ratios, generating 100 forward models for each subject. The forwardmodel corpus is decomposed using principal component analysis.

FORWARD MODEL INFERENCE

Candidate sources

Candidate sources

Am

plitu

de [V

] TheGfreeGenergyGisGanGapproximationGofGtheGmodelGevidenceG[5].GItGquantifiesGtheGcomplexityGandGdataGfitGofGtheGmodel.WeGobtainGtheGfreeGenergyGfromGtheGVariationalGGarroteG[6,G7]GandGuseGitGtoGinferGaGsuitableGforwardGmodel.

SIMULATION RECOVER A FORWARD MODEL FOR A NEW SUBJECT

Localization.error150

100

50

0[mm]

3200

3100

3000

Free.energy

forward.modelPredicted

modelTest.forward

F.-measure

0

1

0.8

0.6

0.4

0.2

1

Anterior Posterior Sou

rce.

activ

ity.[a

.u.]

True1

0

Time.samples5 15 2510 20

Estimated1

0

Time.samples5 15 2510 20

Source.distribution,.true.and.estimated

Tes

t.sub

ject

Tra

inin

g.su

bje

cts

Forward model PCA prior, Template head, Subject sMRI, Subject sMRI,build using: EEG data template σ template σ true σFree energy 2994 3192 3057 2956MSE 0.63 0.55 0.93 0.61F1-measure 1 0 0.44 0.5Localization error 0 19.6 mm 5.3 mm 2.9 mm

REAL EEG DATA - FACE PERCEPTION TASK

Free5energy

Sum5of5m

PCA5forward5model5with5minimum5free5energy

Free5energy

Time5after5stimulus5[s]

Sou

rce5

ampl

itude

5[a.u

.] 1

0.5

0

-0.5

0.1 0.15 0.2

Time5after5stimulus5[s]

Sou

rce5

ampl

itude

5[a.u

.] 1

0.5

0

-0.5

0.1 0.15 0.2

-1200

-1400

-1600

-1800

-2000

Forward5model5build5using5sMRI5and5template5conductivity

CONCLUSIONThe proposed framework provides simultaneous estimation of a for-ward model and the EEG sources.Forward models can be inferred for new subjects without subject-specific geometry and conductivity. Instead inference is based on aforward model representation and the recorded EEG of the new sub-jects.

Inferred forward models provides similar source distributions aswhen using forward models having structural information.

ACKNOWLEDGEMENT

The work was supported in part by the Novo Nordisk Foundation Interdisciplinary Synergy Pro-gram 2014 [‘Biophysically adjusted state-informed cortex stimulation (BASICS)’] and the Otto Møn-sted Foundation (STH), the Danish Research Council for Natural Sciences (SH), and the DanishInnovation Foundation (LKH).

REFERENCES

[1] Akalin Acar, Z., Makeig, S. (2013). Effects of forward model errors on EEG source localization. Brain Topography, 26(3), 378-396.[2] Stahlhut, C., Mørup, M., Winther, O., Hansen, L.K. (2011) Simultaneous EEG source and forward model reconstruction (sofomore)

using a hierarchical bayesian approach. Journal of Signal Processing Systems, 65(3).[3] Hallez, H. et al. (2007). Review on solving the forward problem in EEG source analysis. J. Neuroeng. and Rehabilitation, 4(46), 1-29.[4] Wakeman, D.G., Henson, R.N. (2015). A multi-subject, multi-modal human neuroimaging dataset. Scientific Data, 2, 150001.[5] Friston, K., Mattout, J., Trujillo-Barreto, N., Ashburner, J., Penny, W. (2007). Variational free energy and the Laplace approximation.

NeuroImage, 34(1), 220-234.[6] Kappen, H.J., Gomez, V. (2014). The variational garrote. Machine Learning, 96(3), 269-294.[7] Hansen, S.T., Stahlhut, C., Hansen, L.K. (2013) Sparse Source EEG Imaging with the Variational Garrote, 3rd Int. Workshop on Pattern

Recognition in Neuroimaging.

For more information check out: Hansen, S.T., Hauberg, S., Hansen, L.K. (2016).Data-driven forward model inference for EEG brain imaging. NeuroImage, inpress. doi:10.1016/j.neuroimage.2016.06.017