Elimination of brain noise from MEG data using ICA with robust pre-whitening technique

Proc. of the International Conference on Biomagnetism (BIOMAG),pp. 161-162, Boston, 2004

Elimination of brain noise from MEG data using ICA with robust pre-whitening technique

Y. Konno, J. Cao, T. Takeda, H. Endo, M. Tanaka and T. Arai

Abstract: In this paper, single-trial and averaged multiple-trials data are analysed applying proposed robust pre-whitening technique with independent component analysis (ICA), in order to study the performance of source decomposition in each case. To evaluate the performance of source decomposition, we use a synthesized MEG data set. The main advantage of our data set is that dipole location of evoked responses and its dynamics are known in advance, which facilitates the evaluation of the decomposed components. Moreover, some existing lCA algorithms such as JADE, Fast-ICA, and Natural gradient-based ICA with robust pre-whitening technique are used to eliminate brain noise. Our results show the performance of source decomposition applying proposed approach and the effectiveness of JADE algorithm for our MEG data analysis.

Keywords: Magnetoencephalography (MEG), robust pre-whitening technique, independent component analysis (ICA), JADE algorithm, fast-ICA, and natural gradient-based ICA, source localization.

[PDF (243 kB)]