Visualizing Single-Trial Evoked Potentials with Wavelet Denoising
Abstract number :
1.086
Submission category :
Year :
2001
Submission ID :
425
Source :
www.aesnet.org
Presentation date :
12/1/2001 12:00:00 AM
Published date :
Dec 1, 2001, 06:00 AM
Authors :
R. Quian Quiroga, Dr., NIC, Forschungszentrum Juelich, Juelich, Germany
RATIONALE: Due to the low amplitude of evoked potentials (EPs), responses to several stimuli need to be averaged in order to distinguish them from the background EEG. However, when averaging, information about the variations between the single trials is lost. This study presents a new method based on the Wavelet Transform to obtain the EPs at the single-trial level.
METHODS: The outcomes of the method will be shown in auditory and pattern (checkerboard) visual EPs obtained with an oddball paradigm.
Briefly, the method works as follows: 1) The average EP signal is decomposed by using the Wavelet Multiresolution Decomposition scheme. This gives a set of wavelet coefficients reflecting the amount of activity at different frequency bands and times. 2) Wavelet coefficients not related with the EPs are set to zero. But we remark that this choice of coefficients should consider latency variations between trials. 3) The inverse Wavelet Transform is applied, thus obtaining a filtered (denoised) average EP. 4) The previous proceedure is applied to the single trials,thus eliminating activity not related with the EPs.
RESULTS: In the average EP three main responses were identified, namely the P100, the N200 and the P300. In the original single-trial signals the evoked responses were masked by the background EEG activity. On the other hand, after denoising the EPs were easily visualized. We remark that a conventional filtering would have not been suitable due to the overlapping time and frequency localizations of the different EPs.
The same denoising scheme when applied to other visual EP signals, showed different degrees of variability in the evoked responses not seen in the average EP.
The method was also suitable for auditory EPs, in which the visualization of the single-trial responses allowed the calculation of selective and jitter corrected averages.
CONCLUSIONS: Wavelet denoising significantly improves the visualization of the single-trial EPs. The method allows the study of the variability between single-trials as well as the obtention of better resolved averages.
Since the method is relatively fast and parameter free, it could complement routine EEG/EP analysis.