Intelligent estimation of evoked potentials by event analysis

A new method is described for averaging evoked potentials (EP) which can be applied in its most general form to any evoked transient data which is obtained under conditions of poor signal to noise ratio. Event Analysis is a non-linear signal estimation procedure which first reduces the individual trials of EP data into sequences of events marking the times of occurrence of significant amplitude deflections in them. This is followed by selective averaging of only those trials which contained significant events lying within latency windows which cover the expected latencies of major waveform features in the EP under study. The latency windows are selected automatically by an expert system or can be assigned manually. As well as producing enhanced estimates of the EP average, the analysis also provides information on the latency variability associated with each of the major waveform features. As an example the authors use the Brainstem Auditory Evoked Potential