Essential Tremor accentuates the pattern of tremor-band coherence between upper-limb muscles.

Although Essential Tremor is one of the most common movement disorders, current treatment options are relatively limited. Peripheral tremor suppression methods have shown potential, but we do not currently know which muscles are most responsible for patients' tremor, making it difficult to optimize suppression methods. The purpose of this study was to quantify the relationships between the tremorogenic activity in muscles throughout the upper limb. Muscle activity was recorded from the 15 major superficial upper-limb muscles in 24 subjects with Essential Tremor while they held various postures or made upper-limb movements. We calculated the coherence in the tremor band (4-12 Hz) between the activity of all muscle pairs and the time-varying phase difference between sufficiently coherent muscle pairs. Overall, the observed pattern somewhat mirrored functional relationships: agonistic muscle pairs were most coherent and in phase, whereas antagonist and unrelated muscle pairs exhibited less coherence and were either consistently in phase, consistently antiphase, consistently out of phase (unrelated pairs only), or else inconsistent. Patients exhibited significantly more coherence than control subjects (p<0.001) in the vast majority of muscle pairs (95 out of 105). Furthermore, differences between patients and controls were most pronounced among agonists; thus, the coherence pattern existing in control subjects was accentuated in patients with ET. We conclude that tremor-band activity is broadly distributed among the muscles of the upper limb, challenging efforts to determine which muscles are most responsible for a patient's tremor.

Synergistic Organization of Neural Inputs from Spinal Motor Neurons to Extrinsic and Intrinsic Hand Muscles.

Our current understanding of synergistic muscle control is based on the analysis of muscle activities. Modules (synergies) in muscle coordination are extracted from electromyographic (EMG) signal envelopes. Each envelope indirectly reflects the neural drive received by a muscle; therefore, it carries information on the overall activity of the innervating motor neurons. However, it is not known whether the output of spinal motor neurons, whose number is orders of magnitude greater than the muscles they innervate, is organized in a low-dimensional fashion when performing complex tasks. Here, we hypothesized that motor neuron activities exhibit a synergistic organization in complex tasks and therefore that the common input to motor neurons results in a large dimensionality reduction in motor neuron outputs. To test this hypothesis, we factorized the output spike trains of motor neurons innervating 14 intrinsic and extrinsic hand muscles and analyzed the dimensionality of control when healthy individuals exerted isometric forces using seven grip types. We identified four motor neuron synergies, accounting for >70% of the variance of the activity of 54.1 ± 12.9 motor neurons, and we identified four functionally similar muscle synergies. However, motor neuron synergies better discriminated individual finger forces than muscle synergies and were more consistent with the expected role of muscles actuating each finger. Moreover, in a few cases, motor neurons innervating the same muscle were active in separate synergies. Our findings suggest a highly divergent net neural inputs to spinal motor neurons from spinal and supraspinal structures, contributing to the dimensionality reduction captured by muscle synergies.SIGNIFICANCE STATEMENT We addressed whether the output of spinal motor neurons innervating multiple hand muscles could be accounted for by a modular organization, i.e., synergies, previously described to account for the coordination of multiple muscles. We found that motor neuron synergies presented similar dimensionality (implying a >10-fold reduction in dimensionality) and structure as muscle synergies. Nonetheless, the synergistic behavior of subsets of motor neurons within a muscle was also observed. These results advance our understanding of how neuromuscular control arises from mapping descending inputs to muscle activation signals. We provide, for the first time, insights into the organization of neural inputs to spinal motor neurons which, to date, has been inferred through analysis of muscle synergies.

Development of functional organization within the sensorimotor network across the perinatal period.

In the mature human brain, the neural processing related to different body parts is reflected in patterns of functional connectivity, which is strongest between functional homologs in opposite cortical hemispheres. To understand how this organization is first established, we investigated functional connectivity between limb regions in the sensorimotor cortex in 400 preterm and term infants aged across the equivalent period to the third trimester of gestation (32-45 weeks postmenstrual age). Masks were obtained from empirically derived functional responses in neonates from an independent data set. We demonstrate the early presence of a crude but spatially organized functional connectivity, that rapidly matures across the preterm period to achieve an adult-like configuration by the normal time of birth. Specifically, connectivity was strongest between homolog regions, followed by connectivity between adjacent regions (different limbs but same hemisphere) already in the preterm brain, and increased with age. These changes were specific to the sensorimotor network. Crucially, these trajectories were strongly dependent on age more than age of birth. This demonstrates that during the perinatal period the sensorimotor cortex undergoes preprogrammed changes determining the functional movement organization that are not altered by preterm birth in absence of brain injury.

Fundamental Concepts of Bipolar and High-Density Surface EMG Understanding and Teaching for Clinical, Occupational, and Sport Applications: Origin, Detection, and Main Errors.

Surface electromyography (sEMG) has been the subject of thousands of scientific articles, but many barriers limit its clinical applications. Previous work has indicated that the lack of time, competence, training, and teaching is the main barrier to the clinical application of sEMG. This work follows up and presents a number of analogies, metaphors, and simulations using physical and mathematical models that provide tools for teaching sEMG detection by means of electrode pairs (1D signals) and electrode grids (2D and 3D signals). The basic mechanisms of sEMG generation are summarized and the features of the sensing system (electrode location, size, interelectrode distance, crosstalk, etc.) are illustrated (mostly by animations) with examples that teachers can use. The most common, as well as some potential, applications are illustrated in the areas of signal presentation, gait analysis, the optimal injection of botulinum toxin, neurorehabilitation, ergonomics, obstetrics, occupational medicine, and sport sciences. The work is primarily focused on correct sEMG detection and on crosstalk. Issues related to the clinical transfer of innovations are also discussed, as well as the need for training new clinical and/or technical operators in the field of sEMG.

An fMRI Compatible Smart Device for Measuring Palmar Grasping Actions in Newborns.

Grasping is one of the first dominant motor behaviors that enable interaction of a newborn infant with its surroundings. Although atypical grasping patterns are considered predictive of neuromotor disorders and injuries, their clinical assessment suffers from examiner subjectivity, and the neuropathophysiology is poorly understood. Therefore, the combination of technology with functional magnetic resonance imaging (fMRI) may help to precisely map the brain activity associated with grasping and thus provide important insights into how functional outcomes can be improved following cerebral injury. This work introduces an MR-compatible device (i.e., smart graspable device (SGD)) for detecting grasping actions in newborn infants. Electromagnetic interference immunity (EMI) is achieved using a fiber Bragg grating sensor. Its biocompatibility and absence of electrical signals propagating through the fiber make the safety profile of the SGD particularly favorable for use with fragile infants. Firstly, the SGD design, fabrication, and metrological characterization are described, followed by preliminary assessments on a preterm newborn infant and an adult during an fMRI experiment. The results demonstrate that the combination of the SGD and fMRI can safely and precisely identify the brain activity associated with grasping behavior, which may enable early diagnosis of motor impairment and help guide tailored rehabilitation programs.

Voluntary and tremorogenic inputs to motor neuron pools of agonist/antagonist muscles in essential tremor patients.

Pathological tremor is an oscillation of body parts at 3-10 Hz, determined by the output of spinal motor neurons (MNs), which receive synaptic inputs from supraspinal centers and muscle afferents. The behavior of spinal MNs during tremor is not well understood, especially in relation to the activation of the multiple muscles involved. Recent studies on patients with essential tremor have shown that antagonist MN pools receive shared input at the tremor frequency. In this study, we investigated the synaptic inputs related to tremor and voluntary movement, and their coordination across antagonist muscles. We analyzed the spike trains of motor units (MUs) identified from high-density surface electromyography from the forearm extensor and flexor muscles in 15 patients with essential tremor during postural tremor. The shared synaptic input was quantified by coherence and phase difference analysis of the spike trains. All pairs of spike trains in each muscle showed coherence peaks at the voluntary drive frequency (1-3 Hz, 0.2 ± 0.2, mean ± SD) and tremor frequency (3-10 Hz, 0.6 ± 0.3) and were synchronized with small phase differences (3.3 ± 25.2° and 3.9 ± 22.0° for the voluntary drive and tremor frequencies, respectively). The coherence between MN spike trains of antagonist muscle groups at the tremor frequency was significantly smaller than intramuscular coherence. We predominantly observed in-phase activation of MUs between agonist/antagonist muscles at the voluntary frequency band (0.6 ± 48.8°) and out-of-phase activation at the tremor frequency band (126.9 ± 75.6°). Thus MNs innervating agonist/antagonist muscles concurrently receive synaptic inputs with different phase shifts in the voluntary and tremor frequency bands.NEW & NOTEWORTHY Although the mechanical characteristics of tremor have been widely studied, the activation of the affected muscles is still poorly understood. We analyzed the behavior of motor units of pairs of antagonistic wrist muscle groups in patients with essential tremor and studied their activity at voluntary movement- and tremor-related frequencies. We found that the phase relation between inputs to antagonistic muscles is different at the voluntary and tremor frequency bands.

Principles of Motor Unit Physiology Evolve With Advances in Technology.

Movements are generated by the coordinated activation of motor units. Recent technological advances have made it possible to identify the concurrent activity of several tens of motor units, in contrast with much smaller samples available in classic studies. We discuss how these advances in technology have enabled the development of a population perspective of how the central nervous system controls motor unit activity and thereby the forces exerted by muscles.

Accurate and representative decoding of the neural drive to muscles in humans with multi-channel intramuscular thin-film electrodes.

Intramuscular electrodes developed over the past 80 years can record the concurrent activity of only a few motor units active during a muscle contraction. We designed, produced and tested a novel multi-channel intramuscular wire electrode that allows in vivo concurrent recordings of a substantially greater number of motor units than with conventional methods. The electrode has been extensively tested in deep and superficial human muscles. The performed tests indicate the applicability of the proposed technology in a variety of conditions. The electrode represents an important novel technology that opens new avenues in the study of the neural control of muscles in humans. We describe the design, fabrication and testing of a novel multi-channel thin-film electrode for detection of the output of motoneurones in vivo and in humans, through muscle signals. The structure includes a linear array of 16 detection sites that can sample intramuscular electromyographic activity from the entire muscle cross-section. The structure was tested in two superficial muscles (the abductor digiti minimi (ADM) and the tibialis anterior (TA)) and a deep muscle (the genioglossus (GG)) during contractions at various forces. Moreover, surface electromyogram (EMG) signals were concurrently detected from the TA muscle with a grid of 64 electrodes. Surface and intramuscular signals were decomposed into the constituent motor unit (MU) action potential trains. With the intramuscular electrode, up to 31 MUs were identified from the ADM muscle during an isometric contraction at 15% of the maximal force (MVC) and 50 MUs were identified for a 30% MVC contraction of TA. The new electrode detects different sources from a surface EMG system, as only one MU spike train was found to be common in the decomposition of the intramuscular and surface signals acquired from the TA. The system also allowed access to the GG muscle, which cannot be analysed with surface EMG, with successful identification of MU activity. With respect to classic detection systems, the presented thin-film structure enables recording from large populations of active MUs of deep and superficial muscles and thus can provide a faithful representation of the neural drive sent to a muscle.

Reorganization of muscle synergies during multidirectional reaching in the horizontal plane with experimental muscle pain.

Muscle pain induces a complex reorganization of the motor strategy which cannot be fully explained by current theories. We tested the hypothesis that the neural control of muscles during reaching in the presence of nociceptive input is determined by a reorganization of muscle synergies with respect to control conditions. Muscle pain was induced by injection of hypertonic saline into the anterior deltoid muscle of eight men. Electromyographic (EMG) signals were recorded from 12 upper limb muscles as subjects performed a reaching task before (baseline) and after the injection of hypertonic (pain) saline, and after the pain sensation vanished. The EMG envelopes were factorized in muscle synergies, and activation signals extracted for each condition. Nociceptive stimulation resulted in a complex muscle reorganization without changes in the kinematic output. The anterior deltoid muscle activity decreased in all subjects while the changes in other muscles were subject specific. Three synergies sufficed to describe the EMG patterns in each condition, suggesting that reaching movements remain modular in the presence of experimental pain. Muscle reorganization in all subjects was accompanied by a change in the activation signals compatible with a change in the central drive to muscles. One, two or three synergies were shared between the baseline and painful conditions, depending on the subject. These results indicate that nociceptive stimulation may induce a reorganization of modular control in reaching. We speculate that such reorganization may be due to the recruitment of synergies specific to the painful condition.

Hand and distal joint tremor are most coherent with the activity of elbow flexors and wrist extensors in persons with essential tremor.

Essential tremor (ET) affects millions of people. Although frontline treatment options (medication, deep brain stimulation, and focused ultrasound ablation) have provided significant relief, many patients are unsatisfied with the outcomes. Peripheral suppression techniques, such as injections of botulinum toxin or sensory electrical stimulation of muscles, are gaining popularity, but could be optimized if the muscles most responsible for a patient's tremor were identified. The purpose of this study was to quantify the relationship between the activity in various upper limb muscles and the resulting tremor in patients with ET. Surface electromyogram (sEMG) from the 15 major superficial muscles of the upper limb and displacement of the hand and upper limb joints were recorded from 22 persons with ET while they performed kinetic and postural tasks representative of activities of daily living. We calculated the peak coherence (frequency-dependent correlation) in the tremor band (4-8 Hz) between the sEMG of each muscle and the displacement in each major degree of freedom (DOF). Averaged across subjects with ET, the highest coherence was found between elbow flexors (particularly biceps brachii and brachioradialis) and the distal DOF (forearm, wrist, and hand motion), and between wrist extensors (extensor carpi radialis and ulnaris) and the same distal DOF. These coherence values represent the upper bound on the proportion of the tremor caused by each muscle. We conclude that, without further information, elbow flexors and wrist extensors should be among the first muscles considered for peripheral suppression techniques in persons with ET.NEW & NOTEWORTHY We characterized the relationships between activity in upper limb muscles and tremor in persons with essential tremor using coherence, which provides an upper bound on the proportion of the tremor due to each muscle. Averaged across subjects and various tasks, tremor in the hand and distal joints was most coherent with elbow flexors and wrist extensors. We conclude that, without further information, these muscle groups should be among the first considered for peripheral suppression techniques.

A Particle Swarm Optimised Independence Estimator for Blind Source Separation of Neurophysiological Time Series.

The decomposition of neurophysiological recordings into their constituent neural sources is of major importance to a diverse range of neuroscientific fields and neuroengineering applications. The advent of high density electrode probes and arrays has driven a major need for novel semi-automated and automated blind source separation methodologies that take advantage of the increased spatial resolution and coverage these new devices offer. Independent component analysis (ICA) offers a principled theoretical framework for such algorithms, but implementation inefficiencies often drive poor performance in practice, particularly for sparse sources. Here we observe that the use of a single non-linear optimization function to identify spiking sources with ICA often has a detrimental effect that precludes the recovery and correct separation of all spiking sources in the signal. We go on to propose a projection-pursuit ICA algorithm designed specifically for spiking sources, which uses a particle swarm methodology to adaptively traverse a polynomial family of non-linearities approximating the asymmetric cumulants of the sources. We robustly prove state-of-the-art decomposition performance on recordings from high density intramuscular probes and demonstrate how the particle swarm quickly finds optimal contrast non-linearities across a range of neurophysiological datasets.

An Accurate and Real-Time Method for Resolving Superimposed Action Potentials in MultiUnit Recordings.

OBJECTIVE: Spike sorting of muscular and neural recordings requires separating action potentials that overlap in time (superimposed action potentials (APs)). We propose a new algorithm for resolving superimposed action potentials, and we test it on intramuscular EMG (iEMG) and intracortical recordings. METHODS: Discrete-time shifts of the involved APs are first selected based on a heuristic extension of the peel-off algorithm. Then, the time shifts that provide the minimal residual Euclidean norm are identified (Discrete Brute force Correlation (DBC)). The optimal continuous-time shifts are then estimated (High-Resolution BC (HRBC)). In Fusion HRBC (FHRBC), two other cost functions are used. A parallel implementation of the DBC and HRBC algorithms was developed. The performance of the algorithms was assessed on 11,000 simulated iEMG and 14,000 neural recording superpositions, including two to eight APs, and eight experimental iEMG signals containing four to eleven active motor units. The performance of the proposed algorithms was compared with that of the Branch-and-Bound (BB) algorithm using the Rank-Product (RP) method in terms of accuracy and efficiency. RESULTS: The average accuracy of the DBC, HRBC and FHRBC methods on the entire simulated datasets was 92.16±17.70, 93.65±16.89, and 94.90±15.15 (%). The DBC algorithm outperformed the other algorithms based on the RP method. The average accuracy and running time of the DBC algorithm on 10.5 ms superimposed spikes of the experimental signals were 92.1±21.7 (%) and 2.3±15.3 (ms). CONCLUSION AND SIGNIFICANCE: The proposed algorithm is promising for real-time neural decoding, a central problem in neural and muscular decoding and interfacing.

Biomechanical Analysis of Body Movements of Myoelectric Prosthesis Users During Standardized Clinical Tests.

OBJECTIVE: The objective clinical evaluation of user's capabilities to handle their prosthesis is done using various tests which primarily focus on the task completion speed and do not explicitly account for the potential presence of compensatory motions. Given that the excessive body compensation is a common indicator of inadequate prosthesis control, tests which include subjective observations on the quality of performed motions have been introduced. However, these metrics are then influenced by the examiner's opinions, skills, and training making them harder to standardize across patient pools and compare across different prosthetic technologies. Here we aim to objectively quantify the severity of body compensations present in myoelectric prosthetic hand users and evaluate the extent to which traditional objective clinical scores are still able to capture them. METHODS: We have instructed 9 below-elbow prosthesis users and 9 able-bodied participants to complete three established objective clinical tests: Box-and-Blocks-Test, Clothespin-Relocation-Test, and Southampton-Hand-Assessment-Procedure. During all tests, upper-body kinematics has been recorded. RESULTS: While the analysis showed that there are some correlations between the achieved clinical scores and the individual body segment travel distances and average speeds, there were only weak correlations between the clinical scores and the observed ranges of motion. At the same time, the compensations were observed in all prosthesis users and, for the most part, they were substantial across the tests. CONCLUSION: The sole reliance on the currently available objective clinical assessment methods seems inadequate as the compensatory movements are prominent in prosthesis users and yet not sufficiently accounted for.

Recruitment of motor units in the medial gastrocnemius muscle during human quiet standing: is recruitment intermittent? What triggers recruitment?

The recruitment and the rate of discharge of motor units are determinants of muscle force. Within a motoneuron pool, recruitment and rate coding of individual motor units might be controlled independently, depending on the circumstances. In this study, we tested whether, during human quiet standing, the force of the medial gastrocnemius (MG) muscle is predominantly controlled by recruitment or rate coding. If MG control during standing was mainly due to recruitment, then we further asked what the trigger mechanism is. Is it determined internally, or is it related to body kinematics? While seven healthy subjects stood quietly, intramuscular electromyograms were recorded from the MG muscle with three pairs of wire electrodes. The number of active motor units and their mean discharge rate were compared for different sway velocities and positions. Motor unit discharges occurred more frequently when the body swayed faster and forward (Pearson R = 0.63; P < 0.0001). This higher likelihood of observing motor unit potentials was explained chiefly by the recruitment of additional units. During forward body shifts, the median number of units detected increased from 3 to 11 (P < 0.0001), whereas the discharge rate changed from 8 ± 1.1 (mean ± SD) to 10 ± 0.9 pulses/s (P = 0.001). Strikingly, motor units did not discharge continuously throughout standing. They were recruited within individual, forward sways and intermittently, with a modal rate of two recruitments per second. This modal rate is consistent with previous circumstantial evidence relating the control of standing to an intrinsic, higher level planning process.

EMG-based simultaneous and proportional estimation of wrist/hand kinematics in uni-lateral trans-radial amputees.

We propose a method for estimating wrist kinematics during dynamic wrist contractions from multi-channel surface electromyography (EMG). The algorithm extracts features from the surface EMG and uses dedicated multi-layer perceptron networks to estimate individual joint angles of the 3 degrees of freedom (DoFs) of the wrist. The method was designed with the aim of proportional and simultaneous control of multiple DoFs of active prostheses by unilateral amputees. Therefore, the proposed approach was tested in both unilateral transradial amputees and in intact-limbed control subjects. It was shown that the joint angles at the 3 DoFs of amputees can be estimated from surface EMG recordings , during mirrored bi-lateral contractions that simultaneously and proportionally articulated the 3 DoFs. The estimation accuracies of amputee subjects with long stumps were 62.5% ± 8.50% across all 3 DoFs, while accuracies of the intact-limbed control subjects were 72.0% ± 8.29%. The estimation results from intact-limbed subjects were consistent with earlier studies. The results from the current study demonstrated the feasibility of the proposed myoelectric control approach to provide a more intuitive myoelectric control strategy for unilateral transradial amputees.

Blind identification of the spinal cord output in humans with high-density electrode arrays implanted in muscles.

Invasive electromyography opened a new window to explore motoneuron behavior in vivo. However, the technique is limited by the small fraction of active motoneurons that can be concurrently detected, precluding a population analysis in natural tasks. Here, we developed a high-density intramuscular electrode for in vivo human recordings along with a fully automatic methodology that could detect the discharges of action potentials of up to 67 concurrently active motoneurons with 99% accuracy. These data revealed that motoneurons of the same pool receive common synaptic input at frequencies up to 75 Hz and that late-recruited motoneurons inhibit the discharges of those recruited earlier. These results constitute an important step in the population coding analysis of the human motor system in vivo.

Online Tracking of the Phase Difference Between Neural Drives to Antagonist Muscle Pairs in Essential Tremor Patients.

Transcutaneous electrical stimulation has been applied in tremor suppression applications. Out-of-phase stimulation strategies applied above or below motor threshold result in a significant attenuation of pathological tremor. For stimulation to be properly timed, the varying phase relationship between agonist-antagonist muscle activity during tremor needs to be accurately estimated in real-time. Here we propose an online tremor phase and frequency tracking technique for the customized control of electrical stimulation, based on a phase-locked loop (PLL) system applied to the estimated neural drive to muscles. Surface electromyography signals were recorded from the wrist extensor and flexor muscle groups of 13 essential tremor patients during postural tremor. The EMG signals were pre-processed and decomposed online and offline via the convolution kernel compensation algorithm to discriminate motor unit spike trains. The summation of motor unit spike trains detected for each muscle was bandpass filtered between 3 to 10 Hz to isolate the tremor related components of the neural drive to muscles. The estimated tremorogenic neural drive was used as input to a PLL that tracked the phase differences between the two muscle groups. The online estimated phase difference was compared with the phase calculated offline using a Hilbert Transform as a ground truth. The results showed a rate of agreement of 0.88 ± 0.22 between offline and online EMG decomposition. The PLL tracked the phase difference of tremor signals in real-time with an average correlation of 0.86 ± 0.16 with the ground truth (average error of 6.40° ± 3.49°). Finally, the online decomposition and phase estimation components were integrated with an electrical stimulator and applied in closed-loop on one patient, to representatively demonstrate the working principle of the full tremor suppression system. The results of this study support the feasibility of real-time estimation of the phase of tremorogenic neural drive to muscles, providing a methodology for future tremor-suppression neuroprostheses.

Intramuscular EMG-Driven Musculoskeletal Modelling: Towards Implanted Muscle Interfacing in Spinal Cord Injury Patients.

OBJECTIVE: Surface EMG-driven modelling has been proposed as a means to control assistive devices by estimating joint torques. Implanted EMG sensors have several advantages over wearable sensors but provide a more localized information on muscle activity, which may impact torque estimates. Here, we tested and compared the use of surface and intramuscular EMG measurements for the estimation of required assistive joint torques using EMG driven modelling. METHODS: Four healthy subjects and three incomplete spinal cord injury (SCI) patients performed walking trials at varying speeds. Motion capture marker trajectories, surface and intramuscular EMG, and ground reaction forces were measured concurrently. Subject-specific musculoskeletal models were developed for all subjects, and inverse dynamics analysis was performed for all individual trials. EMG-driven modelling based joint torque estimates were obtained from surface and intramuscular EMG. RESULTS: The correlation between the experimental and predicted joint torques was similar when using intramuscular or surface EMG as input to the EMG-driven modelling estimator in both healthy individuals and patients. CONCLUSION: We have provided the first comparison of non-invasive and implanted EMG sensors as input signals for torque estimates in healthy individuals and SCI patients. SIGNIFICANCE: Implanted EMG sensors have the potential to be used as a reliable input for assistive exoskeleton joint torque actuation.

Postural activation of the human medial gastrocnemius muscle: are the muscle units spatially localised?

In cat medial gastrocnemius (MG), fibres supplied by individual motoneurones (muscle units) distribute extensively along the muscle longitudinal axis. In the human MG, the size of motor unit territory is unknown. It is uncertain if the absolute size of muscle unit territory or the size relative to the whole muscle is most comparable with the cat. By comparing intramuscular and surface electromyograms we tested whether muscle units extend narrowly or widely along the human MG muscle. Due to the pennation of the MG, if individual motoneurones supply fibres scattered along the muscle, then action potentials of single motor units are expected to appear sparsely on the surface of the skin. In nine healthy subjects, pairs of wire electrodes were inserted in three locations along the MG muscle (MG60%, MG75% and MG90%). A longitudinal array of 16 surface electrodes was positioned alongside the intramuscular electrodes. While subjects stood quietly, 55 motor units were identified, of which, significantly more units were detected in the most distal sites. The surface action potentials had maximum amplitude at 4.40 ±1.67 (mean±S.D.), 8.02±2.16 and 11.63±2.09 cm (P <0.001) from the most proximal surface electrode, for motor units in the MG60%, MG75% and MG90% locations, respectively. Single motor unit potentials were recorded by five consecutive surface electrodes, at most, indicating that muscle units extend shortly along the MG longitudinal axis. It is concluded that relative to the whole muscle, and compared with the cat, muscle units in human MG are localised. The localisation of muscle units might have implications for the regional control of muscle activity.

Deep Learning for Robust Decomposition of High-Density Surface EMG Signals.

Blind source separation (BSS) algorithms, such as gradient convolution kernel compensation (gCKC), can efficiently and accurately decompose high-density surface electromyography (HD-sEMG) signals into constituent motor unit (MU) action potential trains. Once the separation matrix is blindly estimated on a signal interval, it is also possible to apply the same matrix to subsequent signal segments. Nonetheless, the trained separation matrices are sub-optimal in noisy conditions and require that incoming data undergo computationally expensive whitening. One unexplored alternative is to instead use the paired HD-sEMG signal and BSS output to train a model to predict MU activations within a supervised learning framework. A gated recurrent unit (GRU) network was trained to decompose both simulated and experimental unwhitened HD-sEMG signal using the output of the gCKC algorithm. The results on the experimental data were validated by comparison with the decomposition of concurrently recorded intramuscular EMG signals. The GRU network outperformed gCKC at low signal-to-noise ratios, proving superior performance in generalising to new data. Using 12 seconds of experimental data per recording, the GRU performed similarly to gCKC, at rates of agreement of 92.5% (84.5%-97.5%) and 94.9% (88.8%-100.0%) respectively for GRU and gCKC against matched intramuscular sources.