Comparison of techniques for decomposing surface EMG signals into motor unit action potential trains

Abstract

Advancements in surface electromyography (sEMG) have led to many discrepancies in techniques used for signal decomposition. Specifically, the capabilities of well-established recording systems, and the methods involved in identifying motor unit (MU) action potentials and respective firing behaviors.

PURPOSE: To examine the differences in MU identification and validation procedures, and firing behaviors between a four-channel (4-ch) sensor and a sixty-four channel (64-ch) high-density sEMG array.

METHODS: Following 2 maximal voluntary contractions (MVC), ten (23 ± 3 yrs.; 178.64 ± 5.82 cm; 177.8 ± 17.37 kg) lower body resistance trained males performed 10 sec submaximal isometric ramp contractions of the knee extension exercise at 10%, 20%, and 50% MVC. During testing sEMG was recorded from the vastus lateralis using both 4-ch and 64-ch sensors. Signals were separately decomposed into their constituent MU action potential trains and were further validated for subsequent analysis of firing behaviors. The slope and y-intercept were calculated across the relationships between recruitment threshold versus mean firing rate (RT/MFR). A 2-way mixed factorial ANOVA (sensor [4-ch vs 64-ch] x contraction intensity [10% vs 20% vs 50%]) was used to examine mean differences in MU yield during all contraction. For validated MUs, the RT/MFR relationships were compared between sensors at each intensity and a paired samples t test was used to compare differences in RTs.

RESULTS: There was a significant interaction between sensor and intensity, as well as a main effect for intensity, with follow up analysis revealing a significant difference between MUs validated at 10% and 50% MVC (p < 0.05). There was a significant difference in slopes at 10% and 50% MVC, and y-intercepts at 20% MVC for RT/MFR relationships (p < 0.10) and the RT of validated MUs were significantly different (p <0.5) between sensors at each intensity.

CONCLUSION: MUs validated using the 4-ch sensor yielded a greater numbers during higher contraction intensities versus the 64-ch sensor. The inability of the 64-ch sensor to yield a greater amount of MUs at 50% MVC may have been due to the subjectivity of the manual editing procedures. However, both validation procedures eliminated a high amount of decomposed MUs.