Wiener Filtering of Auditory Brainstem Responses

Abstract

The effectiveness of Wiener filtering in estimating the auditory brainstem response (ABR) under typical conditions for otoneurological evaluation is examined empirically in this paper. Theoretical and practical difficulties in implementing Wiener filtering are reviewed. Large evoked potential (EP) data matrices were acquired and stored digitally. Wiener-filtered and signal-averaged subsets were compared to the overall mean EP, in terms of their mean-square estimation error (MSE). Wiener filtering produced smaller errors (median MSE ratio 0.5) in almost all cases (p less than 0.01), but variation in the estimated Wiener filter caused evoked potential changes that may be clinically unacceptable. Some deficiencies of the MSE approach to evoked potential estimation are described; these relate to estimation of ABR parameters, as opposed to waveforms, and to the underlying sources of the MSE.