A comparison of wavelet and wavelet-Bayesian approaches for the classification of auditory brainstem response

The auditory brainstem response (ABR) is the early portion of the electrical activity of the brain in response to the auditory stimulus and is recorded from electrodes attached to the scalp. It has become a very useful and routine clinical tool for hearing assessment. Since the ABR is easily obscured by the background EEG activities, averaging of many repeated trials is necessary and it typically requires up to 2000 repetitions. This number of repetitions can be very difficult, time consuming and uncomfortable for some subjects. Therefore, reducing the required number of repetitions offers a great advantage in the clinical situation. In this study the ABR data (574 ABRs with 128 repetitions and 1160 ABRs with 64 repetitions) recorded from 20 subjects are used. Two classification approaches, wavelet approach and wavelet-Bayesian approach for response detection are developed and a comparison of these two methods is considered. A 10-fold cross-validation method and t-test are also applied in this paper for performance comparison purposes.