Analysis of evoked EMG using wavelet transformation

Evoked EMG M-responses obtained from the thenar muscle in the palm by electrical stimulation of the median nerve demonstrate a well-established smooth bipolar shape for normal healthy subjects while kinks are observed in certain neurological disorders, particularly in cervical spondylotic neuropathy. A first differentiation failed to identify these kinks because of comparable values obtained for normally rising and falling segments of the smooth regions, and due to noise. In this study, the usefulness of the wavelet transform (WT), that provides localized measures of non-stationary signals is investigated. The Haar WT was used to analyze a total of 36 M-responses recorded from the median nerves of 6 normal subjects (having smooth shape) and 12 subjects with assumed neurological disorders (having kinks), for two points of stimulation on the same nerve. Features in the time-scale representation of the M-responses were studied using WT to distinguish smooth M-responses from ones with kinks. Variations in the coefficient line of the WT were also studied to allow visualization of WT at different scales (inverse of frequency). The high and low frequency regions in the WT came out distinctively which helped identifications of kinks even of very subtle ones in the M-responses which were difficult to obtain using the differentiated signal. In conclusion, the wavelet analysis may be a technique of choice in identifying kinks in M-responses in relation to time, thus enhancing the accuracy of neurological diagnosis.