After computing all 22 decompositions for each of the 13 EEG datasets, we localized a best-fitting single equivalent dipole corresponding to each returned component using a single equivalent dipole in a best-fitting spherical four-shell model head (radius: 71, 72, 79, 85 mm; shell conductances: 0.33, 0.0042, 1, 0.33 µS) using the DIPFIT plug-in (version 1.02) in the EEGLAB toolbox (version 4.515) [48] (link). To avoid errors based on the simplistic head model we used in the computations, scalp map values for two electrodes sited near the eyes were excluded from dipole fitting. Note that modeling each component map with a single dipole is somewhat idealistic, since in particular some ICA components represent apparently bilateral synchronous source activities (e.g., the component maps in the third column of Figure 1). However, brain components clearly warranting a dual-dipole model appeared to be rare (approximately one per decomposition), as we typically find in other decompositions of more than 32 data channels. Components accounting for most electro-oculographic (EOG) artifacts should also be modeled using two (peri-ocular) dipoles, but as the eyes are relatively close together and accurate forward modeling the front of the skull is difficult, the additional errors introduced by using single dipole models for EOG components is not large.
Free full text: Click here