R Mühler1, H von Specht, J Pethe. 1. Otto-von-Guericke-Universität Magdeburg, Univeritätsklinik für Hals-, Nasen- und Ohrenheilkunde, Abteilung für Experimentelle Audiologie und Medizinische Physik.
Abstract
BACKGROUND: The level of residual noise in auditory brainstem responses (ABR) depends not only on the number of averages but also on the amplitude of background noise and on the frequency of artifacts. This paper describes the influence of digital filtering and of different methods of artifact suppression on the residual noise of ABRs. METHOD: Amplitude of background noise was estimated for 1033 ABRs recorded under suprathreshold stimulation (70 and 90 dB nHL) in 251 subjects. In 45 ABR recordings in 15 subjects, all 4000 individual sweeps were stored for off-line simulation. The power spectrum of background noise was investigated using an FFT analyzer. RESULTS: A great variability of mean noise amplitude was found both between subjects and in the recordings for each subject. Depending on the slope of the analogue 100-Hz high-pass filter, mean RMS values of background noise of 4.2 microV (6 dB/Oct.) and 2.5 microV (12 dB/Oct.), respectively, were found. Digital high pass filtering before averaging was found to increase the signal-to-noise ratio (SNR) considerably. CONCLUSIONS: Results indicate that (i) effective suppression of low-frequency noise components can only be achieved by zero phase digital filtering and (ii) if clipping of noise amplitude to 25 microV is used, optimized artifact rejection as weighted averaging or adopted artifact rejection levels have only small effect on the SNR.
BACKGROUND: The level of residual noise in auditory brainstem responses (ABR) depends not only on the number of averages but also on the amplitude of background noise and on the frequency of artifacts. This paper describes the influence of digital filtering and of different methods of artifact suppression on the residual noise of ABRs. METHOD: Amplitude of background noise was estimated for 1033 ABRs recorded under suprathreshold stimulation (70 and 90 dB nHL) in 251 subjects. In 45 ABR recordings in 15 subjects, all 4000 individual sweeps were stored for off-line simulation. The power spectrum of background noise was investigated using an FFT analyzer. RESULTS: A great variability of mean noise amplitude was found both between subjects and in the recordings for each subject. Depending on the slope of the analogue 100-Hz high-pass filter, mean RMS values of background noise of 4.2 microV (6 dB/Oct.) and 2.5 microV (12 dB/Oct.), respectively, were found. Digital high pass filtering before averaging was found to increase the signal-to-noise ratio (SNR) considerably. CONCLUSIONS: Results indicate that (i) effective suppression of low-frequency noise components can only be achieved by zero phase digital filtering and (ii) if clipping of noise amplitude to 25 microV is used, optimized artifact rejection as weighted averaging or adopted artifact rejection levels have only small effect on the SNR.