and averaging with respect to QRS peaks. The BCG template was then obtained by replicating the same averaged waveform across heart beats, shifted by a fixed delay of 150 ms in order to take into account the typical difference in timing between EKG and BCG signals. Volume MRI and slice MRI artifact templates were created by means of signal peaks, positioned in correspondence of volume and slice onset times, respectively (Fig. 1). Independent component analysis (ICA) was used for the for the decomposition of the EEG data into a set of independent spatio-temporal patterns (independent components, ICs) (Hyvärinen and Oja, 2000), and the subsequent removal of BCG, imaging and ocular artifacts (Fig. 1). The FastICA algorithm (Hyvärinen et al., 1999) was run on the 29 EEG signals for the decomposition into 29 ICs (Makeig et al., 1997), which were then automatically classified into brain signals and artifacts, on the basis of the correlation coefficient rt with the set of EEG artifact templates, including EKG and EOG recordings, simulated BCG, volume MRI and slice MRI artifacts. A specific IC was considered artifactual in case of rt > 0.2 for at least an EEG artifact template. With this procedure, from 8 to
