Myoelectric Control Signals Research Articles

Is Intermittent Control the Source of the Non-Linear Oscillatory Component (0.2-2Hz) in Human Balance Control?

To explain the 0.2-2Hz oscillation in human balance. Oscillation (0.2-2 Hz) in the control signal (ankle moment) is sustained independently of external disturbances and exaggerated in Parkinson's disease. Does resonance or limit cycles in the neurophysiological feedback loop cause this oscillation? We investigate two linear (non-predictive, predictive) and one non-linear (intermittent-predictive) control model (NPC, PC, IPC). Fourteen healthy participants, strapped to an actuated single segment robot with dynamics of upright standing, used natural haptic-visual feedback and myoelectric control signals from lower leg muscles to maintain balance. An input disturbance applied stepwise changes in external force. A linear time invariant model (ARX) extracted the delayed component of the control signal related linearly to the disturbance, leaving the remaining, larger, oscillatory non-linear component. We optimized model parameters and noise (observation, motor) to replicate concurrently (i) estimated-delay, (ii) time-series of the linear component, and (iii) magnitude-frequency spectrum and transient magnitude response of the non-linear component. Results (mean±S.D., p<0.05): NPC produced estimated delays (0.116±0.03s) significantly lower than experiment (0.145±0.04s). Overall fit (i)-(iii) was (79±7%, 83±7%, 84±6% for NPC, PC, IPC). IPC required little or no noise. Mean frequency of experimental oscillation (0.99±0.16 Hz) correlated trial by trial with closed loop resonant frequency (fres), not limit cycles, nor sampling rate. NPC (0.36±0.08Hz) and PC (0.86±0.4Hz) showed fres significantly lower than IPC (0.98±0.2Hz). Human balance control requires short-term prediction. IPC mechanisms (prediction error, threshold related sampling, sequential re-initialization of open-loop predictive control) explain resonant gain without uncontrolled oscillation for healthy balance.

What is the contribution of voluntary and reflex processes to sensorimotor control of balance?

The contribution to balance of spinal and transcortical processes including the long-latency reflex is well known. The control of balance has been modelled previously as a continuous, state feedback controller representing, long-latency reflexes. However, the contribution of slower, variable delay processes has not been quantified. Compared with fixed delay processes (spinal, transcortical), we hypothesize that variable delay processes provide the largest contribution to balance and are sensitive to historical context as well as current states. Twenty-two healthy participants used a myoelectric control signal from their leg muscles to maintain balance of their own body while strapped to an actuated, inverted pendulum. We study the myoelectric control signal (u) in relation to the independent disturbance (d) comprising paired, discrete perturbations of varying inter-stimulus-interval (ISI). We fit the closed loop response, u from d, using one linear and two non-linear non-parametric (many parameter) models. Model M1 (ARX) is a generalized, high-order linear-time-invariant (LTI) process with fixed delay. Model M1 is equivalent to any parametric, closed-loop, continuous, linear-time-invariant (LTI), state feedback model. Model M2, a single non-linear process (fixed delay, time-varying amplitude), adds an optimized response amplitude to each stimulus. Model M3, two non-linear processes (one fixed delay, one variable delay, each of time-varying amplitude), add a second process of optimized delay and optimized response amplitude to each stimulus. At short ISI, the myoelectric control signals deviated systematically both from the fixed delay LTI process (M1), and also from the fixed delay, time-varying amplitude process (M2) and not from the two-process model (M3). Analysis of M3 (all fixed delay and variable delay response amplitudes) showed the variable (compared with fixed) delay process 1) made the largest contribution to the response, 2) exhibited refractoriness (increased delay related to short ISI) and 3) was sensitive to stimulus history (stimulus direction 2 relative to stimulus 1). For this whole-body balance task and for these impulsive stimuli, non-linear processes at variable delay are central to control of balance. Compared with fixed delay processes (spinal, transcortical), variable delay processes provided the largest contribution to balance and were sensitive to historical context as well as current states.

Estimating muscle activation from EMG using deep learning-based dynamical systems models

Objective. To study the neural control of movement, it is often necessary to estimate how muscles are activated across a variety of behavioral conditions. One approach is to try extracting the underlying neural command signal to muscles by applying latent variable modeling methods to electromyographic (EMG) recordings. However, estimating the latent command signal that underlies muscle activation is challenging due to its complex relation with recorded EMG signals. Common approaches estimate each muscle’s activation independently or require manual tuning of model hyperparameters to preserve behaviorally-relevant features. Approach. Here, we adapted AutoLFADS, a large-scale, unsupervised deep learning approach originally designed to de-noise cortical spiking data, to estimate muscle activation from multi-muscle EMG signals. AutoLFADS uses recurrent neural networks to model the spatial and temporal regularities that underlie multi-muscle activation. Main results. We first tested AutoLFADS on muscle activity from the rat hindlimb during locomotion and found that it dynamically adjusts its frequency response characteristics across different phases of behavior. The model produced single-trial estimates of muscle activation that improved prediction of joint kinematics as compared to low-pass or Bayesian filtering. We also applied AutoLFADS to monkey forearm muscle activity recorded during an isometric wrist force task. AutoLFADS uncovered previously uncharacterized high-frequency oscillations in the EMG that enhanced the correlation with measured force. The AutoLFADS-inferred estimates of muscle activation were also more closely correlated with simultaneously-recorded motor cortical activity than were other tested approaches. Significance. This method leverages dynamical systems modeling and artificial neural networks to provide estimates of muscle activation for multiple muscles. Ultimately, the approach can be used for further studies of multi-muscle coordination and its control by upstream brain areas, and for improving brain-machine interfaces that rely on myoelectric control signals.

Therapy Interventions for Upper Limb Amputees Undergoing Selective Nerve Transfers

Targeted Muscle Reinnervation (TMR) improves the biological control interface for myoelectric prostheses after above-elbow amputation. Selective activation of muscle units is made possible by surgically re-routing nerves, yielding a high number of independent myoelectric control signals. However, this intervention requires careful patient selection and specific rehabilitation therapy. Here a rehabilitation protocol is presented for high-level upper limb amputees undergoing TMR, based on an expert Delphi study. Interventions before surgery include detailed patient assessment and general measures for pain control, muscle endurance and strength, balance, and range of motion of the remaining joints. After surgery, additional therapeutic interventions focus on edema control and scar treatment and the selective activation of cortical areas responsible for upper limb control. Following successful reinnervation of target muscles, surface electromyographic (sEMG) biofeedback is used to train the activation of the novel muscular units. Later on, a table-top prosthesis may provide the first experience of prosthetic control. After fitting the actual prosthesis, training includes repetitive drills without objects, object manipulation, and finally, activities of daily living. Ultimately, regular patient appointments and functional assessments allow tracking prosthetic function and enabling early interventions if malfunctioning.

Therapy Interventions for Upper Limb Amputees Undergoing Selective Nerve Transfers.

Targeted Muscle Reinnervation (TMR) improves the biological control interface for myoelectric prostheses after above-elbow amputation. Selective activation of muscle units is made possible by surgically re-routing nerves, yielding a high number of independent myoelectric control signals. However, this intervention requires careful patient selection and specific rehabilitation therapy. Here a rehabilitation protocol is presented for high-level upper limb amputees undergoing TMR, based on an expert Delphi study. Interventions before surgery include detailed patient assessment and general measures for pain control, muscle endurance and strength, balance, and range of motion of the remaining joints. After surgery, additional therapeutic interventions focus on edema control and scar treatment and the selective activation of cortical areas responsible for upper limb control. Following successfulreinnervation of target muscles, surface electromyographic (sEMG) biofeedback is used to train the activation of the novel muscular units. Later on, a table-top prosthesis may provide the first experience of prosthetic control. After fitting the actual prosthesis, training includes repetitive drills without objects, object manipulation, and finally, activities of daily living. Ultimately, regular patient appointments and functional assessments allow tracking prosthetic function and enabling early interventions if malfunctioning.

Voluntary Control of Residual Antagonistic Muscles in Transtibial Amputees: Reciprocal Activation, Coactivation, and Implications for Direct Neural Control of Powered Lower Limb Prostheses.

Residual ankle muscles (i.e., previously antagonistic ankle muscles) of transtibial amputees are a potential source for continuous feedforward control of powered ankle prostheses using proportional myoelectric control. The ability for transtibial amputees to use their residual ankle muscles for two control input degrees of freedom (i.e., two independent myoelectric control input sources) for direct neural control depends on the ability for amputees to generate varying magnitudes of reciprocal activation and coactivation using their residual ankle muscles, which is not well understood. In this paper, we aimed to fill this knowledge gap. We asked 12 transtibial amputees to control the 2-D movement of a computer cursor using continuous proportional myoelectric control via their residual plantar flexor and residual dorsiflexor muscles to define their reachable 2-D control input space. The x-y position of the computer cursor was directly proportional to the independent continuous myoelectric control signals from the residual lateral gastrocnemius (x-axis) and the residual tibialis anterior (y-axis) where the limits of each axis were 0%-100% maximum voluntary activation of the corresponding residual muscle. Our results show that the reachable control input space varied widely across amputee subjects ranging from 38% to 81% of the maximum possible control input space. The cumulative time for the amputee subjects to saturate their reachable control input space ranged from 1.95 to 6.85 min. The amputee subjects used different residual muscle activation patterns and coordination strategies to expand their reachable control input space depending on their ability to perform coactivation and reciprocal activation using their residual plantar flexor and dorsiflexor muscles. The future development of powered lower limb prostheses using direct continuous proportional myoelectric control via residual muscles (e.g., for direct voluntary control of prosthesis joint impedance) should consider how an amputee user's immediately accessible residual muscle activation patterns and reachable 2-D control input space may affect their learning and performance.

Extracting Signals Robust to Electrode Number and Shift for Online Simultaneous and Proportional Myoelectric Control by Factorization Algorithms

Previous research proposed the extraction of myoelectric control signals by linear factorization of multi-channel electromyogram (EMG) recordings from forearm muscles. This paper further analyses the theoretical basis for dimensionality reduction in high-density EMG signals from forearm muscles. Moreover, it shows that the factorization of muscular activation patterns in weights and activation signals by non-negative matrix factorization (NMF) is robust with respect to the channel configuration from where the EMG signals are obtained. High-density surface EMG signals were recorded from the forearm muscles of six individuals. Weights and activation signals extracted offline from 10 channel configurations with varying channel numbers (6, 8, 16, 192 channels) were highly similar. Additionally, the method proved to be robust against electrode shifts in both transversal and longitudinal direction with respect to the muscle fibers. In a second experiment, six subjects directly used the activation signals extracted from high-density EMG for online goal-directed control tasks involving simultaneous and proportional control of two degrees-of-freedom of the wrist. The synergy weights for this control task were extracted from a reference configuration and activation signals were calculated online from the reference configuration as well as from the two shifted configurations, simulating electrode shift. Despite the electrode shift, the task completion rate, task completion time, and execution efficiency were generally not statistically different among electrode configurations. Online performances were also mostly similar when using either 6, 8, or 16 EMG channels. The robustness of the method to the number and location of channels, proved both offline and online, indicates that EMG signals recorded from forearm muscles can be approximated as linear instantaneous mixtures of activation signals and justifies the use of linear factorization algorithms for extracting, in a minimally supervised way, control signals for simultaneous multi-degree of freedom prosthesis control.

Eliminating cardiac contamination from myoelectric control signals developed by targeted muscle reinnervation

The electrocardiogram (ECG) artifact is a major noise contaminating the myoelectric control signals when using shoulder disarticulation prosthesis. This is an even more significant problem with targeted muscle reinnervation to develop additional myoelectric sites for improved prosthesis control in a bilateral amputee at shoulder disarticulation level. This study aims at removal of ECG artifacts from the myoelectric prosthesis control signals produced from targeted muscle reinnervation. Three ECG artifact removal methods based on template subtracting, wavelet thresholding and adaptive filtering were investigated, respectively. Surface EMG signals were recorded from the reinnervated pectoralis muscles of the amputee. As a key parameter for clinical myoelectric prosthesis control, the amplitude measurement of the signal was used as a performance indicator to evaluate the proposed methods. The feasibility of the different methods for clinical application was also investigated with consideration of the clinical speed requirements and memory limitations of commercial prosthesis controllers.

Poster 149: Use of nerve-muscle grafts to improve the control of a myoelectric shoulder disarticulation prosthesis: a case report.

Objective: To develop additional electromyographic control signals in a shoulder disarticulation amputee. Design: Single case study. Setting: Academic acute hospital and rehabilitation hospital. Participant: Patient with bilateral shoulder disarticulation amputation. Interventions: During revision surgery, the patient’s 4 major brachial plexus nerves were surgically grafted to 3 different areas of his pectoralis major muscle and the pectoralis minor (which was moved to the lateral chest wall). After reinnervation occurred a myoelectric prosthesis that used these electromyographic control sites was fit on the patient. Main Outcome Measures: Number of independent electromyographic control sites obtained and prosthetic function. Results: After reinnervation occurred, 4 separate electromyographic sites could be identified and 3 were usable for the control of the prosthesis. The musculocutaneous nerve-muscle graft was used to control elbow flexion. 2 independent sites were recordable from the median nerve-muscle graft and were used to control terminal device opening and closing. A radial nerve-muscle graft had a recordable signal, but it was small, had large electromyographic interference, and was difficult to use for controlling a prosthesis. The ulnar nerve to pectoralis muscle graft was nonviable. The patient was able to operate his terminal device, elbow, and, with cocontraction switching, a powered wrist. He reported that the control was much easier and better than with his previous externally powered prosthesis, which used shoulder touch-pad control. Conclusion: Nerve-muscle grafts can be used to develop additional myoelectric control signals for improved prosthesis control in high level upper-limb amputees.

Training program for a myo-electrically controlled prosthesis with sensory feedback system.

A myo-electrically controlled prosthesis that incorporates a sensory feedback mechanism to provide the wearer with a sense of touch or gripping force in the prosthetic hand was developed for patients with below-elbow amputations. A training protocol was established in which patients were taught to produce myo-electric control signals and to interpret the sensory feedback. The protocol includes functional training in daily living skills. The experiences of four patients trained with these limbs are described.