Combined ICA-Bayesian MCMC Reconstruction of EEG Sources
-
1
Boğaziçi University, Türkiye
-
2
Istanbul University, Türkiye
Reconstructing the source of a given EEG/ERP distribution is a mathematically underdetermined and ill-conditioned inverse problem, which does not, in general, provide a unique physically correct answer. In this study, we combined ICA as a pre-processing step with MCMC methods to reconstruct epileptic EEG activity as well as cognitive activity-related waveforms in ERP data. We use the equivalent current dipole (ECD) model, assuming that the scalp EEG is generated by a modest number of focal sources. Each dipole assumed to generate statistically independent brain processes to satisfy the mathematical assumptions of ICA method. A comparative study with realistically simulated epileptic EEG data is made to decide the most effective ICA approach between FASTICA with symmetric orthogonalization, Joint approximate diagonalization of eigenmatrices (JADE), Infomax and Topographic ICA. In high noise levels, D4 or S8 wavelet denoising is also used with 4 level of decomposition to improve the ICA results. The localization of the decomposed waveforms, namely the inverse problem, are done on a realistic head model constructed by solving the well-posed EEG forward problem with Boundary Elements Method. The inverse problem is modeled with a Bayesian approach: The posterior distributions of dipole parameters are sampled by the Reversible Jump MCMC method combined by Parallel Tempering, a heuristic to escape from local optima. Experiments with simulated and empirical data show that a two step study, localization by MCMC subsequent to ICA decomposition seems very efficient for EEG source reconstruction as it alleviates the ambiguity of the EEG inverse problem. The method offers a flexible framework for the reconstruction of both epileptic and ERP waveforms, focal and widespread activity, which is an advantage not offered by conventional methods.
Conference:
10th International Conference on Cognitive Neuroscience, Bodrum, Türkiye, 1 Sep - 5 Sep, 2008.
Presentation Type:
Poster Presentation
Topic:
Neuroinformatics of Cognition
Citation:
Yildiz
G,
Duru
A,
Ademoglu
A and
Demiralp
T
(2008). Combined ICA-Bayesian MCMC Reconstruction of EEG Sources.
Conference Abstract:
10th International Conference on Cognitive Neuroscience.
doi: 10.3389/conf.neuro.09.2009.01.373
Copyright:
The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers.
They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.
The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.
Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.
For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.
Received:
15 Dec 2008;
Published Online:
15 Dec 2008.
*
Correspondence:
Gökcen Yildiz, Boğaziçi University, Istanbul, Türkiye, Gokcen.Yildiz@esat.kuleuven.be