EEG source localization based on a structured sparsity prior and a partially collapsed Gibbs sampler

Facundo Costa; Hadj Batatia; Thomas Oberlin; Jean-Yves Tourneret

doi:10.1109/CAMSAP.2015.7383786

EEG source localization based on a structured sparsity prior and a partially collapsed Gibbs sampler

Facundo Costa, Hadj Batatia, Thomas Oberlin, Jean-Yves Tourneret

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

4 Citations (Scopus)

Abstract

In this paper, we propose a hierarchical Bayesian model approximating the ℓ20 mixed-norm regularization by a multivariate Bernoulli Laplace prior to solve the EEG inverse problem by promoting spatial structured sparsity. The posterior distribution of this model is too complex to derive closed-form expressions of the standard Bayesian estimators. An MCMC method is proposed to sample this posterior and estimate the model parameters from the generated samples. The algorithm is based on a partially collapsed Gibbs sampler and a dual dipole random shift proposal for the non-zero positions. The brain activity and all other model parameters are jointly estimated in a completely unsupervised framework. The results obtained on synthetic data with controlled ground truth show the good performance of the proposed method when compared to the ℓ21 approach in different scenarios, and its capacity to estimate point-like source activity.

Original language	English
Title of host publication	2015 IEEE 6th International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP)
Publisher	IEEE
Pages	261-264
Number of pages	4
ISBN (Electronic)	9781479919635
DOIs	https://doi.org/10.1109/CAMSAP.2015.7383786
Publication status	Published - 21 Jan 2016
Event	6th IEEE International Workshop on Computational Advances in Multi-Sensor Adaptive Processing 2015 - Cancun, Mexico Duration: 13 Dec 2015 → 16 Dec 2015

Conference

Conference	6th IEEE International Workshop on Computational Advances in Multi-Sensor Adaptive Processing 2015
Abbreviated title	CAMSAP 2015
Country/Territory	Mexico
City	Cancun
Period	13/12/15 → 16/12/15

Keywords

EEG
hierarchical Bayesian model
inverse problem
MCMC
source localization
structured-sparsity
ℓ20-norm regularization

ASJC Scopus subject areas

Signal Processing
Computational Mathematics

Access to Document

10.1109/CAMSAP.2015.7383786

Cite this

@inproceedings{f3745284cc82482599ab4c61f7686c76,

title = "EEG source localization based on a structured sparsity prior and a partially collapsed Gibbs sampler",

abstract = "In this paper, we propose a hierarchical Bayesian model approximating the ℓ20 mixed-norm regularization by a multivariate Bernoulli Laplace prior to solve the EEG inverse problem by promoting spatial structured sparsity. The posterior distribution of this model is too complex to derive closed-form expressions of the standard Bayesian estimators. An MCMC method is proposed to sample this posterior and estimate the model parameters from the generated samples. The algorithm is based on a partially collapsed Gibbs sampler and a dual dipole random shift proposal for the non-zero positions. The brain activity and all other model parameters are jointly estimated in a completely unsupervised framework. The results obtained on synthetic data with controlled ground truth show the good performance of the proposed method when compared to the ℓ21 approach in different scenarios, and its capacity to estimate point-like source activity.",

keywords = "EEG, hierarchical Bayesian model, inverse problem, MCMC, source localization, structured-sparsity, ℓ20-norm regularization",

author = "Facundo Costa and Hadj Batatia and Thomas Oberlin and Jean-Yves Tourneret",

year = "2016",

month = jan,

day = "21",

doi = "10.1109/CAMSAP.2015.7383786",

language = "English",

pages = "261--264",

booktitle = "2015 IEEE 6th International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP)",

publisher = "IEEE",

address = "United States",

note = "6th IEEE International Workshop on Computational Advances in Multi-Sensor Adaptive Processing 2015, CAMSAP 2015 ; Conference date: 13-12-2015 Through 16-12-2015",

}

Costa, F, Batatia, H, Oberlin, T & Tourneret, J-Y 2016, EEG source localization based on a structured sparsity prior and a partially collapsed Gibbs sampler. in 2015 IEEE 6th International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP). IEEE, pp. 261-264, 6th IEEE International Workshop on Computational Advances in Multi-Sensor Adaptive Processing 2015, Cancun, Mexico, 13/12/15. https://doi.org/10.1109/CAMSAP.2015.7383786

EEG source localization based on a structured sparsity prior and a partially collapsed Gibbs sampler. / Costa, Facundo; Batatia, Hadj; Oberlin, Thomas et al.
2015 IEEE 6th International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP). IEEE, 2016. p. 261-264.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - EEG source localization based on a structured sparsity prior and a partially collapsed Gibbs sampler

AU - Costa, Facundo

AU - Batatia, Hadj

AU - Oberlin, Thomas

AU - Tourneret, Jean-Yves

PY - 2016/1/21

Y1 - 2016/1/21

N2 - In this paper, we propose a hierarchical Bayesian model approximating the ℓ20 mixed-norm regularization by a multivariate Bernoulli Laplace prior to solve the EEG inverse problem by promoting spatial structured sparsity. The posterior distribution of this model is too complex to derive closed-form expressions of the standard Bayesian estimators. An MCMC method is proposed to sample this posterior and estimate the model parameters from the generated samples. The algorithm is based on a partially collapsed Gibbs sampler and a dual dipole random shift proposal for the non-zero positions. The brain activity and all other model parameters are jointly estimated in a completely unsupervised framework. The results obtained on synthetic data with controlled ground truth show the good performance of the proposed method when compared to the ℓ21 approach in different scenarios, and its capacity to estimate point-like source activity.

AB - In this paper, we propose a hierarchical Bayesian model approximating the ℓ20 mixed-norm regularization by a multivariate Bernoulli Laplace prior to solve the EEG inverse problem by promoting spatial structured sparsity. The posterior distribution of this model is too complex to derive closed-form expressions of the standard Bayesian estimators. An MCMC method is proposed to sample this posterior and estimate the model parameters from the generated samples. The algorithm is based on a partially collapsed Gibbs sampler and a dual dipole random shift proposal for the non-zero positions. The brain activity and all other model parameters are jointly estimated in a completely unsupervised framework. The results obtained on synthetic data with controlled ground truth show the good performance of the proposed method when compared to the ℓ21 approach in different scenarios, and its capacity to estimate point-like source activity.

KW - EEG

KW - hierarchical Bayesian model

KW - inverse problem

KW - MCMC

KW - source localization

KW - structured-sparsity

KW - ℓ20-norm regularization

UR - http://www.scopus.com/inward/record.url?scp=84963813562&partnerID=8YFLogxK

U2 - 10.1109/CAMSAP.2015.7383786

DO - 10.1109/CAMSAP.2015.7383786

M3 - Conference contribution

AN - SCOPUS:84963813562

SP - 261

EP - 264

BT - 2015 IEEE 6th International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP)

PB - IEEE

T2 - 6th IEEE International Workshop on Computational Advances in Multi-Sensor Adaptive Processing 2015

Y2 - 13 December 2015 through 16 December 2015

ER -

EEG source localization based on a structured sparsity prior and a partially collapsed Gibbs sampler

Abstract

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this