Intersubject repeatability of independent components obtained by ICA with EEG records

Independent component analysis is a methodology that can be used to take a multichannel EEG recording and returns maximally temporally independent statistical source signal or components, where some of these components are related to brain sources. The independent components could be the starting point to compare homologous components between subjects of study. First it is necessary to evaluate repeatability between the independent components of a population of study. In this work, we develop a methodology that automatically measures repeatability in independent components of a data set. It includes a spatial information based clustering of the components with respect to a template. The repeatability degree is then evaluated based on a mutual information approach. We use 14 EEG recordings of 71 channels each. This work presents 15 independent components groups with a significant degree of repeatability, which suggests that the methodology proposed is efficient to find repeatable independent components in a data set.