AUTOMATED AVERAGING OF AUDITORY EVOKED RESPONSE WAVEFORMS USING WAVELET ANALYSIS

Abstract

The auditory brainstem response (ABR) is widely used as an index to assist hearing and brain function diagnoses. In particular, in clinical applications, the rapid detection of ABR peak characteristics is required. One approach to improving the speed of detection is to decrease the number of signal averaging procedures while denoising during the detection of ABR waveforms; another approach is to extract the characteristics of ABR waveform components. In our previous study, to represent ABR waveform components, we obtained not only the frequency characteristics of an ABR but also the frequency characteristics of each component of the ABR based on the time (latency). Using a one-dimensional discrete wavelet transform (DWT) in this latency-frequency analysis, we described an approximate method of reproducing ABR signals with a low SNR from observed values obtained with a smaller number of averaging procedures. At the same time, using this multiple-level frequency decomposition of ABR signals according to the known frequency content of the ABR, we extracted the peak latency of the fast component of the ABR using fewer averagings of the ABR data. From these decomposition and reconstruction results for ABR signals, we proposed the optimal decomposition level of the ABR and explained how we used the waveform of the ABR reconstructed by the inverse DWT (IDWT). In this paper, we propose a method of automated averaging of the ABR using the waveform reconstructed by discrete wavelet multiresolution analysis (MRA). Our proposed method will be useful for the fast detection of ABR latency characteristics in hearing screening test.