Filtering of neurophysiologic signals.

Clinical neurophysiologic signals cover a broad range of frequencies. Filters help to emphasize waveforms that are of clinical or research interest and to mold their frequency characteristics to suit the purpose of the investigation. Some frequency content is obvious and well known, such as the alpha rhythm (8-11Hz) or spindles (12-14Hz) in the EEG. Other frequencies are not initially discriminable from background activity and require filtering in order to examine them, such as high-frequency oscillations (80-500Hz) in EEG and brainstem auditory evoked potentials (100-3000Hz). Often used to mitigate the effects of background noise or artifact, filters can be used specifically to attenuate unwanted frequencies, such as mains interference (50 or 60Hz) and electrode offset potential (<0.1Hz). For digital instrumentation, an antialiasing filter (below Nyquist) is always needed prior to sampling by the analog-to-digital converter. Once the signals are in the digital realm, sophisticated filtering operations can be carried out post hoc; but in order not to be misled, the neurophysiologist must always bear in mind the effect of filtering on the physiological waveform.