ERP Source Analysis Guided by fMRI During Familiar Face Processing.

ABSTRACT

Event related potentials (ERPs) provide precise temporal information about cognitive processing, but with poor spatial resolution, while functional magnetic resonance imaging (fMRI) reliably identifies brain areas involved, but with poor temporal resolution. Here we use fMRI to guide source localization of the ERPs at different times for studying the temporal dynamics of the neural system for recognizing familiar faces. fMRI activation areas were defined in a previous experiment applying the same paradigm used for ERPs. The Bayesian model averaging (BMA) method was used to estimate the generators of the ERPs to unfamiliar, visually familiar, and personally-familiar faces constraining the model by fMRI activation results. For this, higher prior probabilities in the solution space were assigned to the fMRI-defined regions, which included face-selective areas and other areas related to "person knowledge" retrieval. Source analysis was carried out in three-time windows early (150-210 ms), middle (300-380 ms) and late (460-580 ms). The early and middle responses were generated in fMRI-defined areas for all face categories, while these areas do not contribute to the late response. Different areas contributed to the generation of the early and middle ERPs elicited by unfamiliar faces fusiform (Fus), inferior occipital, superior temporal sulcus and the posterior cingulate (PC) cortices. For familiar faces, the contributing areas were Fus, PC and anterior temporal areas for visually familiar faces, with the addition of the medial orbitofrontal areas and other frontal structures for personally-significant faces. For both unfamiliar and familiar faces, more extended and reliable involvement of contributing areas were obtained for the middle compare with early time window. Our fMRI guide ERP source analysis suggested the recruitment of person-knowledge processing areas as early as 150-210 ms after stimulus onset during recognition of personally-familiar faces. We concluded that fMRI-constrained BMA source analysis provide information regarding the temporal-dynamics in the neural system for cognitive processsing.