Unlocking the Full Potential of High-Density Surface EMG: Novel Non-Invasive High-Yield Motor Unit Decomposition

TL;DR

The proposed Swarm‐Contrastive Decomposition (SCD) method consistently outperformed existing techniques in both the quantity of decoded motor units and the precision of their firing time identification, and represents a step forward in non‐invasive EMG technology for studying motor unit activity in complex scenarios.

Abstract

The decomposition of high-density surface electromyography (HD-sEMG) signals into motor unit discharge patterns has become a powerful tool for investigating the neural control of movement, providing insights into motor neuron recruitment and discharge behavior. However, current algorithms, while very effective under certain conditions, face significant challenges in complex scenarios, as their accuracy and motor unit yield are highly dependent on anatomical differences among individuals. This can limit the number of decomposed motor units, particularly in challenging conditions. To address this issue, we recently introduced Swarm-Contrastive Decomposition (SCD), which dynamically adjusts the separation function based on the distribution of the data and prevents convergence to the same source. Initially applied to intramuscular EMG signals, SCD is here adapted for HD-sEMG signals. We demonstrated its ability to address key challenges faced by existing methods, particularly in identifying low-amplitude motor unit action potentials and effectively handling complex decomposition scenarios, like high-interference signals. We extensively validated SCD using simulated and experimental HD-sEMG recordings and compared it with current state-of-the-art decomposition methods under varying conditions, including different excitation levels, noise intensities, force profiles, sexes, and muscle groups. The proposed method consistently outperformed existing techniques in both the quantity of decoded motor units and the precision of their firing time identification. For instance, under certain experimental conditions, SCD detected more than three times as many motor units compared to previous methods, while also significantly improving accuracy. These advancements represent a major step forward in non-invasive EMG technology for studying motor unit activity in complex scenarios.

Figures (9)

• Figure 1: Schematics of the HD-sEMG data used in the analysis. a Simulated HD-sEMG data from forearm muscles during index finger flexion, with the electrode grid positioned over the proximal third of the forearm. The same setup was applied for the experimental recordings from the forearm. Representative data (simulations) is shown for a 30 %MVC contraction. b Experimental HD-sEMG data were also recorded from the TA muscle during ankle dorsiflexion. Representative data for a 20 %MVC contraction is displayed.
• Figure 2: a Median number of motor units per bootstrap iteration for SCD (blue) and cBSS (pink). b Distribution of exponents for SCD. c Distribution of the RoA between the automatic methods and the simulated ground truth for the motor units commonly identified by SCD and cBSS. d-g Distributions of the peak-to-peak MUAP amplitudes (d), conduction velocity (e), number of fibers innervated (f), and depth (measured with respect to the skin, higher is deeper) (g) for motor units common to both SCD and cBSS, and those uniquely identified by SCD.
• Figure 3: a Effect of the method used to prevent source convergence on the number of motor units found. Three methods are reported: 1) activity (use the activity index to initialize the separation vectors), 2) deflation (activity index to initialize the separation vectors and orthogonalize the separation vectors), 3) peel-off (remove found sources from the EMG). b Effect of the exponent of the contrast function on the number of motor units found. Peel-off approach was used for all the three methods reported: 1) exponent fixed at 2, 2) exponent fixed at 3, 3) exponents starting at [2,3,4,5,6,7] and updated with particle swarm optimisation.
• Figure 4: Effect of noise on the number of motor units found and the RoA with the ground truth for 30 %MVC force level. a Number of motor units against noise level, for SCD and cBSS. b Distribution of the RoA between decomposed motor units and their simulated ground truth for the motor units commonly identified by SCD and cBSS.
• Figure 5: Representative example of a unit in a ballistic contraction decomposed with 100% accuracy. a Spatial distribution of the MUAP arranged in the 10x32 electrode configuration. b A one-second zoom-in of the innervation pulse train of the full source. The discharge times of the clustered source are shown in red. c A one-second zoom-in of the EMG.