Multiresolution analysis is a potentially useful tool to enhance the brain's electrical fields (spatial distributions of event-related potentials (ERP)), and to bring out spatial features which may not be seen in the fields before enhancement. For comparing different images (slices from ERP of different subjects or from the same subject but evoked by different stimuli), we define a measure (surface energy) at each decomposition scale and for different wavelets. The best wavelet and the best level for comparing the given images can be chosen based on this measure. Our experiments show that for very similar images, their difference can be brought out at some scale level. Three preprocessing steps are needed in order to carry out this wavelet analysis. First, a wavelet denoising step is needed to remove noise from the raw ERP. Secondly, a one-to-one mapping is needed to map scalp surface into a square, because the current wavelet analysis theory and algorithm are constructed on regular domains. Finally, a fitting or interpolation step is needed to construct an image on a regular grid in order to apply the fast wavelet transform algorithms.
[1]
P. Nunez,et al.
Neocortical Dynamics and Human EEG Rhythms
,
1995
.
[2]
Jean-Michel Poggi,et al.
Wavelet Toolbox User s Guide
,
1996
.
[3]
M. Victor Wickerhauser,et al.
Wavelets: Algorithms and Applications (Yves Meyer)
,
1994,
SIAM Rev..
[4]
V. Menon,et al.
Local estimate of surface Laplacian derivation on a realistically shaped scalp surface and its performance on noisy data.
,
1994,
Electroencephalography and clinical neurophysiology.
[5]
P. Nunez,et al.
A theoretical and experimental study of high resolution EEG based on surface Laplacians and cortical imaging.
,
1994,
Electroencephalography and clinical neurophysiology.