Validation of a computer-aided EMG decomposition method

This paper presents an objective assessment of the accuracy of EMGLAB, a computer-aided EMG decomposition program that we developed. EMG signals were recorded simultaneously using monopolar needle and fine-wire electrodes from nearby sites in the tibialis anterior muscle during moderate isometric contractions. The signals were decomposed independently by an experienced operator, yielding 3-12 (mean 8.7) motor-unit action potential (MUAP) trains per signal. Decomposition accuracy was estimated by crosschecking 83 pairs of trains from different signals that corresponded to the same motor units. The results show that EMGLAB was able to decompose large MUAPs (peak amplitudes greater than 2.5 times rms signal amplitude) with 98-100% accuracy, and smaller MUAPs with 80-100% accuracy. Many of the errors involved misalignment of small MUAPs within superimpositions and amounted to less than 5 ms. These results validate the accuracy of EMGLAB for decomposing EMG signals of moderate complexity.