COHERENT MEG/EEG SOURCE LOCALIZATION IN TRANSFORMED DATA SPACE

In some cases, different brain regions give rise to strongly-coherent electrical neural activities. For example, pure tone evoked activations of the bilateral auditory cortices exhibit strong coherence. Conventional 2nd order statistics-based spatio-temporal algorithms, such as MUSIC (MUltiple SIgnal Classification) and beamforming encounter difficulties in localizing such activities. In this paper, we proposed a novel solution for this case. The key idea is to map the measurement data into a new data space through a transformation prior to the localization. The orthogonal complement of the lead field matrix for the region to be suppressed is generated as the transformation matrix. Using a priori knowledge or another independent imaging method, such as sLORETA (standard LOw REsolution brain electromagnetic TomogrAphy), the coherent source regions can be primarily identified. And then, in the transformed data space a conventional spatio-temporal method, such as MUSIC, can be used to accomplish the localization of the remaining coherent sources. Repeatedly applying the method will achieve localization of all the coherent sources. The algorithm was validated by simulation experiments as well as by the reconstructions of real bilateral auditory cortical coherent activities.