Our method is intended to provide a high resolution spatio-temporal analysis of how activations propagate at different latencies across the networks engaged by the task. We analyze fMRI images according to a data-driven approach by means of ICA, a signal processing method able to separate independent spatio-temporal patterns of brain activity [McKeown et al., 1998]. We also extract with ICA the most prominent features from the ERP data, and we localize them within the areas of the time-locked fMRI networks. Accordingly, the method is not likely to be biased by the use of an incomplete model for the fusion of EEG and fMRI data, although similar spatial distribution of the EEG and fMRI sources is assumed in our method. Nonetheless, inter-individual differences in electrode locations, and in cortex and volume conductor geometry, may possibly result in the blurring of EEG inverse solutions, in turn yielding an overall decreased spatial resolution.
