EMG artifact minimization during clinical EEG recordings by special analog filtering

A multichannel analog filter for minimizing EMG artifact in routine EEG recordings is described. A 20-channel readily portable version has been implemented which can be easily interfaced with standard electroencephalographs at the auxiliary input/output connectors, in such a way that the filter can be inserted or removed for all channels simultaneously, or, selected channels can be filtered individually. That such an on-line filter be explored was suggested by the remarkable results for minimizing EMG artifact from EEG recordings at seizure onset by off-line computer-based digital filtering recently reported by Gotman, Ives and Gloor. A 4-pole Butterworth filter, i.e., having a maximally flat amplitude response in the passband and having a cut-off frequency, i.e., 30% attenuation point, of 12.5 Hz was selected, the same cut-off frequency as that of the digital filter used by Gotman et al. With this cut-off frequency, which represents a compromise, residual EMG artifact is relatively small. With higher cut-off frequencies, EMG artifact can appear as pseudo-beta activity. Minimization of pseudo-beta activity is obtained, however, at the expense of the elimination of true beta activity. On the other hand, activity of alpha frequencies or lower can be readily identified in the filter output, even with severe EMG artifact in the unfiltered EEG. Other problems in EMG filtering are discussed. Since the analog filter has a non-linear phase characteristic, in contrast to the linear phase characteristic of the finite-impulse-response digital filter used by Gotman et al., the possibility existed that distortion of the wave form of EEG spikes could result, as well as attenuation of their amplitudes. However, the wave forms of filtered spikes in the write-out of the analog filter were found to be essentially identical with those from a one-channel on-line finite-impulse-response computer-based filter having a linear phase characteristic. This finding indicates that phase shift in the analog filter is not a problem, at least for clinical EEG.