The influence of muscle shapes on HDsEMG decomposition yield and accuracy

The application of high-density surface electromyography in conjunction with decomposition algorithms has enabled decoding of motor unit (MU) firings. In developing such algorithms, model-based simulation is an established means to assess yield and accuracy. In this work, we therefore investigate how the simulated muscle shape impacts the decomposition performance. To this end, an arm model with an anatomical biceps brachii shape was developed, and its plausibility in representing MU action potential waveforms was verified with decomposition of experimental, submaximal ramp contractions. From this anatomical muscle shape model, two simplified shapes were derived. In a simulation study, comparable fiber pathways and MU properties were used in anatomical and simplified muscle shape models, and a blind source separation-based decomposition was performed. The results were evaluated with respect to the number of identified MUs and their rate of agreement (RoA). Across all models, MUs with the highest energy contribution were identified. For the two simplified shapes, statistically significantly more MUs (p <0.05) were decomposed due to changed electrode-fiber distances, while the RoA remained consistently high above 94.63±12.16% indicating reliable MU property extraction with all three models. The results emphasize the benefit of anatomically accurate muscle shape models since small simplifications can affect decomposition yield.