Abstract
Detection of an auditory brainstem response, ABR, usually relies on visual evaluation of two or more data acquisition runs of a fixed number of sweeps to determine if there is sufficient replication of the averaged waveforms to indicate a response. Visual interpretation can be difficult when the signal-to-noise ratio is poor because of either a small response or high levels of physiological background noise. Moreover, variations in the background noise from run to run can result in poor or spurious replications of component peaks and troughs in the waveform. A previous study (Elberling & Don, 1984) described a statistical approach for objective evaluation of the quality of an ABR recording. The method uses variance analysis in calculating the ratio of the magnitude of the ABR to the estimated averaged background noise. This study further applies this method to obtain a quantitative definition of the ABR threshold, to demonstrate its application in automatic threshold detection, and to estimate the number of sweeps required to reach detection criterion. Application of this method is valuable in reducing the variability of test interpretation and in maximizing the efficiency of recording ABRs by avoiding the averaging of excessive or insufficient numbers of sweeps. These improvements enhance the cost-benefit of ABR testing to the patient.