Objective: The primary aim of this study is to demonstrate the feasibility
of acquiring auditory evoked potentials (AEPs) by median averaging and stud
y its performance under various recording conditions. The auditory brain st
em response (ABR) was used as the AEP of choice because it has the poorest
signal to noise ratio (SNR) with inherent high susceptibility to extraneous
noise. Secondary aim is to evaluate the characteristics of the median ABRs
in comparison with the conventional mean averaged ABRs.
Design: Single sweep responses to clicks obtained from four subjects at 5 d
B steps were saved in hard disk and used for off-line mean and median avera
ging. The characteristics of the median averaging technique were investigat
ed by manipulating the averaging procedure using the same set of single swe
ep recordings and comparing them with the mean averaged responses. The effe
cts of analog to digital input resolution (bit size) was simulated computat
ionally by increasing quantization.
Results: The results showed that AEPs with low SNRs such as the ABR can be
successfully acquired using median averaging with about the same number of
sweeps as was required for mean averaging, provided the EEG signal is digit
ized with a high number of bits. The resulting waveform generally contained
more identifiable waves than the corresponding mean average and had a high
-frequency noise superimposed on it. This high-frequency noise was successf
ully filtered out using a digital, running mean smoothing filter. The media
n average showed an advantage over the mean average when occasional artifac
ts were recorded.
Conclusion: The results showed that ABRs can be acquired successfully by me
dian averaging provided EEG is digitized with high bit size. Compared with
conventional mean averaging, median averaging is less sensitive to infreque
nt, externally and internally generated noise that plagues conventional tec
hniques and may help improve wave identification.