Source Density Driven Multi-Subjects Spatiotemporal Independent Component Analysis and its Applications

Abstract

It is well known that brain reactions to stimuli are not temporally synchronized and spatially comparable among subjects. In multi-subject EEG source analysis, studies using concatenated data across subjects might have overlooked the fact that projections from intracranial sources to the scalp electrodes may vary widely among subjects. In this talk, an alignment scheme that spatially registers current sources among subjects to a well-defined spherical topology shared by individual cortical surfaces, and that temporally anchors event-related dynamics to a standard latency-by-frequency image is demonstrated.? An innovative functional link between normal distributions and a wide class of non-normal distributions is addressed. Along with the notion of functional nonlinearity in independent component analysis (ICA), this link is useful for modeling the probability density functions of individual components. The essential role of the proposed multi-Subjects spatiotemporal ICA is illustrated in an EEG experiment involving the stop-signal paradigm. Finally, the key areas are highlighted on further research and applications pertinent to the proposed schemes.