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

Agnese Grison, Irene Méndez Guerra, Alexander Kenneth Clarke, Silvia Muceli, Jaime Ibáñez, Dario Farina

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.

Introduction

To understand how the nervous system orchestrates movement, researchers focus on the motor unit—the fundamental functional block consisting of a single motor neuron and the muscle fibers it innervates. Traditionally, capturing the firing patterns of these units required invasive needle electrodes (intramuscular EMG). While accurate, these methods are painful and sample only a localized area of the muscle.

High-Density Surface EMG (HD-sEMG) offers a non-invasive alternative, using grids of electrodes placed on the skin to capture the electrical activity of the underlying muscle. However, extracting individual motor unit activity from these surface signals presents a massive "cocktail party problem." Current algorithms often struggle to distinguish individual "voices" when signals overlap in time or are muffled by anatomical factors like subcutaneous fat. This study introduces Swarm-Contrastive Decomposition (SCD), a novel approach designed to maximize the "yield" of identified motor units even in the most challenging signal environments.

The Limitations of Current Decomposition

Most existing HD-sEMG decomposition methods rely on Blind Source Separation (BSS). These algorithms use a fixed mathematical tool, known as a contrast function, to isolate individual sources from the mixed signal. While effective for high-amplitude, superficial motor units, these "one-size-fits-all" functions often fail to detect smaller, deeper, or lower-amplitude signals. This leads to a low yield, where only a fraction of the active motor units are successfully identified, particularly during high-force or rapid movements where signal interference is high.

The SCD Framework: Adaptive and Iterative

SCD overcomes these limitations through two primary innovations:

  1. Adaptive Contrast Functions via Particle Swarm Optimization (PSO): Instead of using a static mathematical exponent to identify signal peaks, SCD employs PSO to dynamically tune the contrast function for each specific motor unit. This allows the algorithm to adapt its search criteria to the unique statistical signature of an individual source, significantly improving the sensitivity to low-amplitude potentials.
  2. The "Peel-off" Strategy: Traditional methods can suffer from "convergence bias," where the algorithm repeatedly identifies the same dominant, high-amplitude motor units. SCD implements a sequential strategy: once a motor unit is identified with high confidence, its contribution is mathematically "peeled off" (removed) from the raw signal. This clears the "interference" and allows the algorithm to see the more subtle motor units that were previously overshadowed.

Experimental Validation

The researchers validated SCD using a multi-faceted approach:

  • Simulated Data: Using the NeuroMotion musculoskeletal simulator, the team generated signals with known "ground truth" firing patterns. This allowed for precise measurement of accuracy and yield across varying force levels (10–100% Maximum Voluntary Contraction) and noise intensities.
  • Experimental Two-Source Validation: To confirm real-world efficacy, signals were recorded from the Tibialis Anterior (TA) and forearm muscles using surface grids while simultaneously recording with invasive intramuscular wires. This "gold standard" setup verified that the SCD-identified units matched the physical reality inside the muscle.
  • Diverse Demographics: The study included both male and female participants to ensure the algorithm could handle the anatomical variations (such as skin-fold thickness) that typically degrade surface EMG quality.

Key Results and Performance

SCD demonstrated a significant performance leap over current state-of-the-art methods (such as convolutional BSS):

  • Increased Yield: In standard simulations, SCD identified an average of 25.9 motor units per contraction, nearly doubling the 13.9 units found by traditional methods.
  • Ballistic Task Superiority: During rapid, high-intensity muscle bursts—notoriously difficult to decompose due to simultaneous firing—SCD detected three times as many motor units as previous techniques.
  • Precision and Accuracy: The Rate of Agreement (RoA) with ground truth was consistently higher for SCD, often reaching near 100% in simulations. In experimental TA muscle tests, SCD identified significantly more units that matched the intramuscular truth (e.g., 41.6 units vs. 13.7 units in a representative subject).

Conclusion and Future Directions

By moving away from rigid mathematical filters and adopting an adaptive, iterative approach, Swarm-Contrastive Decomposition unlocks a much deeper level of information from non-invasive muscle sensors. This high-yield capability is essential for studying complex motor control strategies that were previously invisible to surface EMG.

While currently validated for isometric (static) contractions, future work will focus on adapting SCD for dynamic movements, where muscle architecture shifts during activity. Additionally, optimizing the PSO process for real-time application could revolutionize the control of high-tech prosthetics and human-machine interfaces, providing a high-fidelity neural bridge without the need for surgery.

Summary updated

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.
BibTeX
@article{grison2025unlocking,
  author = {Agnese Grison and Irene Méndez Guerra and Alexander Kenneth Clarke and Silvia Muceli and Jaime Ibáñez and Dario Farina},
  title = {{Unlocking the Full Potential of High-Density Surface EMG: Novel Non-Invasive High-Yield Motor Unit Decomposition}},
  journal = {The Journal of Physiology},
  year = {2025},
  doi = {10.1113/jp287913},
}
Figure