Motor Units Recruitment and Modulation Patterns in Variable-Force Single Digit Tasks

Previous attempts to achieve volitional control of individual motor units, critical for developing future neural interfaces, have primarily relied on visual feedback of motor units’ spiking activity and subject-driven empirical control through trial and error. Enhancing control strategies at the motor unit level requires a deeper understanding of force-dependent motor unit activity.

In this study, we seek to characterise motor units’ behaviour during a variable-force task to inform improved motor unit control training protocols. We analysed a publicly available dataset consisting of 256 EMG channels over forearm flexor and extensor muscles (128 on each side) while participants (n=20) tracked a triangular force trace using individual finger isometric contractions. The force trace required variable forces in both finger flexion and extension directions, with 6 trials per finger for all 20 subjects, totalling 600 trials.

Motor unit decomposition, based on convolutive kernel compensation, was applied to all trials. On average, 79.9 ± 47.1 motor units were decoded per trial, showing significant inter-subject variability (p<0.01). Motor unit discharge rates were analysed after smoothing spike trains using a Hann window of 400ms. The coefficient of variation (CoV) of smoothed discharge rates served as a measure of modulation intensity for each motor unit in flexion or extension directions. Heatmaps of motor unit territories were generated based on discharge-weighted amplitude values of motor unit action potential waveforms.

Analysis of a histogram depicting motor units against their modulation values revealed a bimodal distribution, indicating two populations of motor neurons: one with high (CoV>0.5) and another with low (CoV≤0.5) modulation, controlling variable-force tasks. Distinct differences were observed in the spread and activation intensities of motor unit territories covered by high- and low-modulation motor units during finger flexion or extension. Although high-modulation motor units displayed greater temporal variability and correlation with force amplitude and direction, neither motor unit population nor all motor units together could be adequately modelled by a low number of synergies, highlighting the complexity of motor unit modulations in variable force scenarios.

In conclusion, our study unveils critical insights into motor unit control, laying the groundwork for unlocking the full potential of neural interfaces. The identification of distinct modulation patterns by two motor unit populations sets the stage for targeted training strategies. Filtering only signals coming from previously identified sub-samples of motor units could lead to improved control of force-dependent or even force-independent motoneuron activity.

Recommended citation: R. Mio, J. Narayan and A. A. Faisal. “Motor Units Recruitment and Modulation Patterns in Variable-Force Single Digit Tasks”, Neural Control of Movement 2024.