A method based on wavelets to analyse overlapped and dependant M-Waves

Implanted stimulation restores hand movement in patients with complete spinal cord injuries. However, assessing the response by surface evoked EMG recordings is challenging because the forearm muscles are small and overlapping. Moreover, M-waves are dependent because they are induced by a single stimulation paradigm. We hypothesized that the M-waves of each muscle has a specific time-frequency signature and we have developed a method to reconstruct the recruitment curves using the energy of this specific time-frequency signature. Orthogonal wavelets are used to analyze individual M-waves. As the selection of the wavelet family and the determination of the time-frequency signature were not trivial, the impact of these choices was evaluated. First, we were able to discriminate the 2 relevant M-waves related to the studied muscles thanks to their specific time-frequency representations. Second, the Meyer family, compared to the Daubechies 2 and 4 families, is the most robust choice against the uncertainty of the time-frequency region definition. Finally, the results are consistent with the semi-quantitative evaluation performed with the MRC scoring. The Meyer wavelet transform combined with the definition of a specific area of interest for each individual muscle allows us to quantitatively and objectively evaluate the evoked EMG in a robust manner.