On the Crosstalk in Motor Unit Spike Train Identification from High-Density Surface Electromyograms

Abstract

We analyzed the problem of motor unit (MU) crosstalk in the spike trains that are identified from high-density surface electromyograms (HDsEMG). We used advanced volume conductor model to simulate five muscles with 500 MUs distributed in the 60 × 30 mm elliptical muscle cross-section. HDsEMG signals were generated during 20 s long isometric muscle contractions at 10%, 30%, 50% and 70% of the maximum excitation. Noiseless and noisy HDsEMG signals were recorded by an array of 10 × 9 electrodes and decomposed by the Convolution Kernel Compensation (CKC) method to yield estimated MU spike trains (MUSTs). Results show that 70.4 ± 26.1% of MUSTs exhibit crosstalk from other MUs, but with relatively low spike amplitudes. Across all the simulated contraction and noise levels, the crosstalk in the identified MUSTs came mostly from nearby MUs (average distance from the dominant MU with the largest spikes in MUST <3.5 mm) and relatively small MUs (average number of MU fibers <150).