Automated Analysis of the Auditory Brainstem Response

In this paper we describe an algorithm for automated peak-trough labelling of the auditory brainstem response (ABR). The algorithm finds the peaks of clinical interest by estimating the first and second derivatives of the ABR waveform. Prior knowledge regarding the expected latencies and amplitudes of the various peaks is then used to incrementally label peaks I through to VII. The performance of the algorithm is estimated on a set of 240 ABR waveforms captured at a stimulus intensity of 90dBnHL. The proposed algorithm not only offers insight into the manual process of ABR peak detection, but is shown to be extremely accurate (96-98%) at finding the primary peaks of clinical interest (I, III, and V). It is also shown that although peak IV can only be found with moderate accuracy (77%), almost one third of these correct detections come as a direct result of locating the 'peak' using the second derivative