ErpICASSO: a tool for reliability estimates of independent components in EEG event-related analysis

Independent component analysis and blind source separation methods are steadily gaining popularity for separating individual brain and non-brain source signals mixed by volume conduction in electroencephalographic data. Despite the advancements on these techniques, determining the number of embedded sources and their reliability are still open issues. In particular to date no method takes into account trial-to-trial variability in order to provide a reliability measure of independent components extracted in Event Related Potentials (ERPs) studies. In this work we present ErpICASSO, a new method which modifies a data-driven approach named ICASSO for the analysis of trials (epochs). In addition to ICASSO the method enables the user to estimate the number of embedded sources, and provides a quality index of each extracted ERP component by combining trial-to-trial bootstrapping and CCA projection. We applied ErpICASSO on ERPs recorded from 14 subjects presented with unpleasant and neutral pictures. We separated potentials putatively related to different systems and identified the four primary ERP independent sources. Standing on the confidence interval estimated by ErpICASSO, we were able to compare the components between neutral and unpleasant conditions. ErpICASSO yielded encouraging results, thus providing the scientific community with a useful tool for ICA signal processing whenever dealing with trials recorded in different conditions.