A new concept of vertex ERA and EEG analysis applying inverse filtering.

Evidence for a common anatomical and physiological substrate for the generation of the background EEG and the vertex-evoked response led to the hypothesis that the EEG can be interpreted as the output from a time varying filter driven by (a) a noise generator responsible for the background activity, and (b) a pulse generator creating the evoked response. The characteristics of the filter can be estimated by autoregression. The application of such a model results in considerable theoretical and practical improvements in the detection of evoked responses. Selected averaging on EEGs classified according to the filter characteristics shows improved S/N ratios and considerable diversity between classes. Investigations of an estimate of the input signal to the filter, derived by deconvolution, reveal a sound-evoked potential which is less variable than the response in the EEG. The study concludes that the proposed model and methods have a sufficient applicability and physiological relevance to support further developments.