OBJECTIVE: Somatosensory evoked potentials (SEP) have been shown to be a useful tool in monitoring of the central nervous system (CNS) during anaesthesia. SEP analysis is usually performed by an experienced human operator. For automatic analysis, appropriate parameter extraction and signal representation methods are required. The aim of this work is to evaluate the discrete wavelet transform (DWT) as such a method for an SEP representation. METHODS: Median nerve SEP were derived in 52 female patients, scheduled for elective surgery with SEP monitoring, under clinically proven conditions in the awake state. The discrete wavelet transform implemented as the multiresolution analysis was adopted for evaluating SEP. The suitability of the wavelet coefficients was investigated by calculating the error between the averaged response and the corresponding wavelet reconstructions. RESULTS: SEP can be represented by a very small number of wavelet coefficients. Although the individual SEP waveform has an influence on the number and selection of wavelet coefficients, in all subjects more than 84% of the SEP waveform energy can be represented by a set 16 wavelet coefficients. CONCLUSIONS: The discrete wavelet transformation provides an efficient tool for SEP representation and parameterisation. Depending on the specific problem the DWT, can be adjusted to the desired accuracy, which is important for the subsequent development of automatic SEP analysers.
OBJECTIVE: Somatosensory evoked potentials (SEP) have been shown to be a useful tool in monitoring of the central nervous system (CNS) during anaesthesia. SEP analysis is usually performed by an experienced human operator. For automatic analysis, appropriate parameter extraction and signal representation methods are required. The aim of this work is to evaluate the discrete wavelet transform (DWT) as such a method for an SEP representation. METHODS: Median nerve SEP were derived in 52 female patients, scheduled for elective surgery with SEP monitoring, under clinically proven conditions in the awake state. The discrete wavelet transform implemented as the multiresolution analysis was adopted for evaluating SEP. The suitability of the wavelet coefficients was investigated by calculating the error between the averaged response and the corresponding wavelet reconstructions. RESULTS: SEP can be represented by a very small number of wavelet coefficients. Although the individual SEP waveform has an influence on the number and selection of wavelet coefficients, in all subjects more than 84% of the SEP waveform energy can be represented by a set 16 wavelet coefficients. CONCLUSIONS: The discrete wavelet transformation provides an efficient tool for SEP representation and parameterisation. Depending on the specific problem the DWT, can be adjusted to the desired accuracy, which is important for the subsequent development of automatic SEP analysers.