The comparison of wavelet- and Fourier-based electromyographic indices of back muscle fatigue during dynamic contractions: validity and reliability results

Abstract

The purpose of this study was to compare the electromyographic (EMG) fatigue indices computed from short-time Fourier transform (STFT) and wavelet transform (WAV), by analyzing their criterion validity and test-retest reliability. The effect of averaging spectral estimates within and between repeated contractions (cycles) on EMG fatigue indices was also demonstrated. Thirty-one healthy subjects performed trunk flexion-extension cycles until exhaustion on a Biodex dynamometer. The load was determined theoretically as twice the L5-S1 moment produced by the trunk mass. To assess reliability, 10 subjects performed the same experimental protocol after a two-week interval. EMG signals were recorded bilaterally with 12 pairs of electrodes placed on the back muscles (at L4, L3, L1 and T10 levels), as well as on the gluteus maximus and biceps femoris. The endurance time and perceived muscle fatigue (Borg CR-10 scale) were used as fatigue criteria. EMG signals were processed using STFT and WAV to extract global (e.g, median frequency and instantaneous median frequency, respectively) or local (e.g., intensity contained in 8 frequency bands) information from the power spectrum. The slope values of these variables over time, obtained from regression analyses, were retained as EMG fatigue indices. EMG fatigue indices (STFT vs. WAV) were not significantly different within each muscle, had a variable association (Pearson's r range.: 0.06 to 0.68) with our fatigue criteria, and showed comparable reliability (Intra-class correlation range: 0.00 to 0.88), although they varied between muscles. The effect of averaging, within and between cycles, contributed to the strong association between EMG fatigue indices computed from STFT and WAV. As for EMG spectral indices of muscle fatigue, the conclusion is that both transforms carry essentially the same information.