An Automated Function for Identifying EEG Independent Components Representing Bilateral Source Activity

Abstract

Independent component analysis (ICA) decomposition can be used for the identification and localization of brain generators. ICA separates EEG data into a sum of maximally distinct signals (independent components, ICs). Source localization algorithms can be directly applied to the component projections (scalp maps). Usually, brain-generated ICs are well modeled using one equivalent dipole or, in the case of IC scalp maps that appear bilaterally symmetric, with two position-symmetric dipoles. Selection of ICs for bilateral dipole fitting is typically performed by visual inspection, a time-consuming and subjective step. We have developed and tested a routine for automated recommendation of ICs that may be best fit with a position-symmetric dual-dipole model. The algorithm is based on near bilateral symmetry of IC scalp map maxima and minima. Results showed good classification accuracy and specificity. Sensitivity can be optimized by adjusting free parameters as suggested.