Reliable measurement of cortical flow patterns using complex independent component analysis of electroencephalographic signals

Abstract

Complex independent component analysis (ICA) of frequency-domain electroencephalographic (EEG) data [1] is a generalization of real time-domain ICA to the frequency-domain. Complex ICA aims to model functionally independent sources as representing patterns of spatio-temporal dynamics. Applied to EEG data, it may allow non-invasive measurement of flow trajectories of cortical potentials. As complex ICA has a higher complexity and number of parameters than time-domain ICA, it is important to determine the extent to which complex ICA applied to brain signals is stable across decompositions. This question is investigated for the complex ICA method applied to the 5-Hz frequency band of data from a selective attention EEG experiment1. © Springer-Verlag 2004.