Continuous wavelet transform in the evaluation of stretch reflex responses from surface EMG

Objective: This is the first reported use of the continuous wavelet transform (CWT) of the surface EMG (sEMG) to extract the reflex response to muscle stretch. We used a modulus-based method to estimate instantaneous amplitude-envelopes from ridges of the CWT (referred in this work as sEMG intensity) to extract the dynamic reflex response from sEMG. We tested the method on tendon reflexes where excellent temporal resolution is required to identify the different latency components, and on the tonic stretch reflex (tonic SR) response to an ongoing perturbation that characteristically has a low signal to noise ratio. Methods: Eight subjects without neurological impairment were subjected to a series of archilles tendon taps and a 2min continuous perturbation of the ankle using a pseudo-sinusoidal stretch profile containing frequencies from 0.1 to 8.0Hz. The tendon reflexes were assessed in the soleus muscle at 10% of MVC and the tonic SR in tibialis anterior while the muscle was relaxed, at 5 and 10% of maximal voluntary contraction. Root mean square (RMS) and wavelet ridge extraction was applied to the sEMG signal to extract sEMG amplitudes (RMS) and intensities for all reflexes. To obtain the tonic SR, these estimates and those from the sEMG-RMS were subsequently cross-correlated with the perturbation record to yield 2 sets of estimates of reflex gain and coherence for comparison. Results: The sEMG intensities were highly correlated with the torques resulting from a ramped voluntary contraction. Following tendon taps, the method resolved the M1, M2, M3 response components at accurate latencies and with more complete reconstruction of the components than RMS-derived estimates. The wavelet ridge estimates extracted the tonic SR from resting and contracting muscles with significantly higher coherence than RMS estimates. Reflex gain, when estimated from sEMG intensity or sEMG-RMS, demonstrated similar relationships to the perturbation frequency and background contraction level. When the sEMG intensity reflex gain estimates from different subjects were pooled, they showed significantly lower variance about the mean than gain estimates derived from the rectified sEMG. Conclusions: Wavelet-ridge extraction provides a valid approach to reflex evaluation from sEMG that does not depend on the absolute amplitude of the potentials measured at the EMG electrodes. This may have substantial advantages in more directly comparing responses between subjects on an absolute frequency scale without the need for normalisation against maximal contraction levels