Adaptive multiresolution analysis based evoked potential filtering

Evoked potentials (EPs) are electrical activities of the brain synchronised with external stimuli. They have proved valuable for the understanding of the functioning of the brain and in investigating several brain related disorders. EPs are usually obscured by the background electroencephalogram (EEG) and thus require appropriate filtering. As the frequency spectra of the EEG and EPs overlap, the application of deterministic filters on their own is usually inadequate. Synchronised averaging improves the signal-to-noise ratio; however it inhibits measurement of the important variations which develop from one EP recording or trial to the next. The presence of these variations also makes an averaged EP a distorted version of an EP which evolves with time. A novel adaptive filtering algorithm based on the wavelet transform method of multiresolution analysis (MRA) was developed and was successfully used for single trial recovery of a type of EP known as the contingent negative variation (CNV). Both simulated and real CNV waveforms were processed. A technique to evaluate the effectiveness of the developed method was devised and was used to select the best orthogonal filter among Daubechies, Coifman and Symmlet for the adaptive MRA based filtering operation. The technique enabled the magnitude of the background EEG to be reduced by a factor of 5 while preserving the main features of the CNV waveform.