Can Measured Synergy Excitations Accurately Construct Unmeasured Muscle Excitations?

Muscle synergies are shared across fundamental subtasks in complex movements of skateboarding

A common theory of motor control posits that movement is controlled by muscle synergies. However, the behavior of these synergies during highly complex movements remains largely unexplored. Skateboarding is a hardly researched sport that requires rapid motor control to perform tricks. The objectives of this study were to investigate three key areas: (i) whether motor complexity differs between skateboard tricks, (ii) the inter-participant variability in synergies, and (iii) whether synergies are shared between different tricks. Electromyography data from eight muscles per leg were collected from seven experienced skateboarders performing three different tricks (Ollie, Kickflip, 360°-flip). Synergies were extracted using non-negative matrix factorization. The number of synergies (NoS) was determined using two criteria based on the total variance accounted for (tVAF > 90% and adding an additional synergy does not increase tVAF > 1%). In summary: (i) NoS and tVAF did not significantly differ between tricks, indicating similar motor complexity. (ii) High inter-participant variability exists across participants, potentially caused by the low number of constraints given to perform the tricks. (iii) Shared synergies were observed in every comparison of two tricks. Furthermore, each participant exhibited at least one synergy vector, which corresponds to the fundamental 'jumping' task, that was shared through all three tricks.

Muscle synergy-informed neuromusculoskeletal modelling to estimate knee contact forces in children with cerebral palsy

Cerebral palsy (CP) includes a group of neurological conditions caused by damage to the developing brain, resulting in maladaptive alterations of muscle coordination and movement. Estimates of joint moments and contact forces during locomotion are important to establish the trajectory of disease progression and plan appropriate surgical interventions in children with CP. Joint moments and contact forces can be estimated using electromyogram (EMG)-informed neuromusculoskeletal models, but a reduced number of EMG sensors would facilitate translation of these computational methods to clinics. This study developed and evaluated a muscle synergy-informed neuromusculoskeletal modelling approach using EMG recordings from three to four muscles to estimate joint moments and knee contact forces of children with CP and typically developing (TD) children during walking. Using only three to four experimental EMG sensors attached to a single leg and leveraging an EMG database of walking data of TD children, the synergy-informed approach estimated total knee contact forces comparable to those estimated by EMG-assisted approaches that used 13 EMG sensors (children with CP, n = 3, R2 = 0.95 ± 0.01, RMSE = 0.40 ± 0.14 BW; TD controls, n = 3, R2 = 0.93 ± 0.07, RMSE = 0.19 ± 0.05 BW). The proposed synergy-informed neuromusculoskeletal modelling approach could enable rapid evaluation of joint biomechanics in children with unimpaired and impaired motor control within a clinical environment.

Changes in walking function and neural control following pelvic cancer surgery with reconstruction

Introduction: Surgical planning and custom prosthesis design for pelvic cancer patients are challenging due to the unique clinical characteristics of each patient and the significant amount of pelvic bone and hip musculature often removed. Limb-sparing internal hemipelvectomy surgery with custom prosthesis reconstruction has become a viable option for this patient population. However, little is known about how post-surgery walking function and neural control change from pre-surgery conditions. Methods: This case study combined comprehensive walking data (video motion capture, ground reaction, and electromyography) with personalized neuromusculoskeletal computer models to provide a thorough assessment of pre- to post-surgery changes in walking function (ground reactions, joint motions, and joint moments) and neural control (muscle synergies) for a single pelvic sarcoma patient who received internal hemipelvectomy surgery with custom prosthesis reconstruction. Pre- and post-surgery walking function and neural control were quantified using pre- and post-surgery neuromusculoskeletal models, respectively, whose pelvic anatomy, joint functional axes, muscle-tendon properties, and muscle synergy controls were personalized using the participant's pre-and post-surgery walking and imaging data. For the post-surgery model, virtual surgery was performed to emulate the implemented surgical decisions, including removal of hip muscles and implantation of a custom prosthesis with total hip replacement. Results: The participant's post-surgery walking function was marked by a slower self-selected walking speed coupled with several compensatory mechanisms necessitated by lost or impaired hip muscle function, while the participant's post-surgery neural control demonstrated a dramatic change in coordination strategy (as evidenced by modified time-invariant synergy vectors) with little change in recruitment timing (as evidenced by conserved time-varying synergy activations). Furthermore, the participant's post-surgery muscle activations were fitted accurately using his pre-surgery synergy activations but fitted poorly using his pre-surgery synergy vectors. Discussion: These results provide valuable information about which aspects of post-surgery walking function could potentially be improved through modifications to surgical decisions, custom prosthesis design, or rehabilitation protocol, as well as how computational simulations could be formulated to predict post-surgery walking function reliably given a patient's pre-surgery walking data and the planned surgical decisions and custom prosthesis design.

Comparison of Synergy Extrapolation and Static Optimization for Estimating Multiple Unmeasured Muscle Activations during Walking

Background

Calibrated electromyography (EMG)-driven musculoskeletal models can provide great insight into internal quantities (e.g., muscle forces) that are difficult or impossible to measure experimentally. However, the need for EMG data from all involved muscles presents a significant barrier to the widespread application of EMG-driven modeling methods. Synergy extrapolation (SynX) is a computational method that can estimate a single missing EMG signal with reasonable accuracy during the EMG-driven model calibration process, yet its performance in estimating a larger number of missing EMG signals remains unclear.

Methods

This study assessed the accuracy with which SynX can use eight measured EMG signals to estimate muscle activations and forces associated with eight missing EMG signals in the same leg during walking while simultaneously performing EMG-driven model calibration. Experimental gait data collected from two individuals post-stroke, including 16 channels of EMG data per leg, were used to calibrate an EMG-driven musculoskeletal model, providing "gold standard" muscle activations and forces for evaluation purposes. SynX was then used to predict the muscle activations and forces associated with the eight missing EMG signals while simultaneously calibrating EMG-driven model parameter values. Due to its widespread use, static optimization (SO) was also utilized to estimate the same muscle activations and forces. Estimation accuracy for SynX and SO was evaluated using root mean square errors (RMSE) to quantify amplitude errors and correlation coefficient r values to quantify shape similarity, each calculated with respect to "gold standard" muscle activations and forces.

Results

On average, SynX produced significantly more accurate amplitude and shape estimates for unmeasured muscle activations (RMSE 0.08 vs. 0.15 , r value 0.55 vs. 0.12) and forces (RMSE 101.3 N vs. 174.4   N , r value 0.53 vs. 0.07) compared to SO. SynX yielded calibrated Hill-type muscle-tendon model parameter values for all muscles and activation dynamics model parameter values for measured muscles that were similar to "gold standard" calibrated model parameter values.

Conclusions

These findings suggest that SynX could make it possible to calibrate EMG-driven musculoskeletal models for all important lower-extremity muscles with as few as eight carefully chosen EMG signals and eventually contribute to the design of personalized rehabilitation and surgical interventions for mobility impairments.

Influence of different calibration methods on surface electromyography-informed musculoskeletal models with few input signals

BACKGROUND

Although model personalization is critical when assessing individuals with morphological or neurological abnormalities, or even non-disabled subjects, its translation into routine clinical settings is hampered by the cumbersomeness of experimental data acquisition and lack of resources, which are linked to high costs and long processing pipelines. Quantifying the impact of neglecting subject-specific information in simulations that aim to estimate muscle forces with surface electromyography informed modeling approaches, can address their potential in relevant clinical questions. The present study investigates how different methods to fine-tune subject-specific neuromuscular parameters, reducing the number of electromyography input data, could affect the estimation of the unmeasured excitations and the musculotendon forces.

METHODS

Three-dimensional motion analysis was performed on 8 non-disabled adult subjects and 13 electromyographic signals captured. Four neuromusculoskeletal models were created for 8 participants: a reference model driven by a large set of sEMG signals; two models informed by four electromyographic signals but calibrated in different fashions; a model based on static optimization.

FINDINGS

The electromyography-informed models better predicted experimental excitations, including the unmeasured ones. The model based on static optimization obtained less reliable predictions of the experimental data. When comparing the different reduced models, no major differences were observed, suggesting that the less complex model may suffice for predicting muscle forces with a small set of input in clinical gait analysis tasks.

INTERPRETATION

Quantitative model performance evaluation in different conditions provides an objective indication of which method yields the most accurate prediction when a small set of electromyographic recordings is available.

Walking Speed Classification from Marker-Free Video Images in Two-Dimension Using Optimum Data and a Deep Learning Method

Walking speed is considered a reliable assessment tool for any movement-related functional activities of an individual (i.e., patients and healthy controls) by caregivers and clinicians. Traditional video surveillance gait monitoring in clinics and aged care homes may employ modern artificial intelligence techniques to utilize walking speed as a screening indicator of various physical outcomes or accidents in individuals. Specifically, ratio-based body measurements of walking individuals are extracted from marker-free and two-dimensional video images to create a walk pattern suitable for walking speed classification using deep learning based artificial intelligence techniques. However, the development of successful and highly predictive deep learning architecture depends on the optimal use of extracted data because redundant data may overburden the deep learning architecture and hinder the classification performance. The aim of this study was to investigate the optimal combination of ratio-based body measurements needed for presenting potential information to define and predict a walk pattern in terms of speed with high classification accuracy using a deep learning-based walking speed classification model. To this end, the performance of different combinations of five ratio-based body measurements was evaluated through a correlation analysis and a deep learning-based walking speed classification test. The results show that a combination of three ratio-based body measurements can potentially define and predict a walk pattern in terms of speed with classification accuracies greater than 92% using a bidirectional long short-term memory deep learning method.

EMG-driven musculoskeletal model calibration with estimation of unmeasured muscle excitations via synergy extrapolation

Subject-specific electromyography (EMG)-driven musculoskeletal models that predict muscle forces have the potential to enhance our knowledge of internal biomechanics and neural control of normal and pathological movements. However, technical gaps in experimental EMG measurement, such as inaccessibility of deep muscles using surface electrodes or an insufficient number of EMG channels, can cause difficulties in collecting EMG data from muscles that contribute substantially to joint moments, thereby hindering the ability of EMG-driven models to predict muscle forces and joint moments reliably. This study presents a novel computational approach to address the problem of a small number of missing EMG signals during EMG-driven model calibration. The approach (henceforth called "synergy extrapolation" or SynX) linearly combines time-varying synergy excitations extracted from measured muscle excitations to estimate 1) unmeasured muscle excitations and 2) residual muscle excitations added to measured muscle excitations. Time-invariant synergy vector weights defining the contribution of each measured synergy excitation to all unmeasured and residual muscle excitations were calibrated simultaneously with EMG-driven model parameters through a multi-objective optimization. The cost function was formulated as a trade-off between minimizing joint moment tracking errors and minimizing unmeasured and residual muscle activation magnitudes. We developed and evaluated the approach by treating a measured fine wire EMG signal (iliopsoas) as though it were "unmeasured" for walking datasets collected from two individuals post-stroke-one high functioning and one low functioning. How well unmeasured muscle excitations and activations could be predicted with SynX was assessed quantitatively for different combinations of SynX methodological choices, including the number of synergies and categories of variability in unmeasured and residual synergy vector weights across trials. The two best methodological combinations were identified, one for analyzing experimental walking trials used for calibration and another for analyzing experimental walking trials not used for calibration or for predicting new walking motions computationally. Both methodological combinations consistently provided reliable and efficient estimates of unmeasured muscle excitations and activations, muscle forces, and joint moments across both subjects. This approach broadens the possibilities for EMG-driven calibration of muscle-tendon properties in personalized neuromusculoskeletal models and may eventually contribute to the design of personalized treatments for mobility impairments.

A muscle synergy-based method to estimate muscle activation patterns of children with cerebral palsy using data collected from typically developing children

Preparing children with cerebral palsy prior to gait analysis may be a challenging and time-intensive task, especially when large number of sensors are involved. Collecting minimum number of electromyograms (EMG) and yet providing adequate information for clinical assessment might improve clinical workflow. The main goal of this study was to develop a method to estimate activation patterns of lower limb muscles from EMG measured from a small set of muscles in children with cerebral palsy. We developed and implemented a muscle synergy extrapolation method able to estimate the full set of lower limbs muscle activation patterns from only three experimentally measured EMG. Specifically, we extracted a set of hybrid muscle synergies from muscle activation patterns of children with cerebral palsy and their healthy counterparts. Next, those muscle synergies were used to estimate activation patterns of muscles, which were not initially measured in children with cerebral palsy. Two best combinations with three (medial gastrocnemius, semi membranous, and vastus lateralis) and four (lateral gastrocnemius, semi membranous, sartorius, and vastus medialis) experimental EMG were able to estimate the full set of 10 muscle activation patterns with mean (± standard deviation) variance accounted for of 79.93 (± 9.64)% and 79.15 (± 6.40)%, respectively, using only three muscle synergies. In conclusion, muscle activation patterns of unmeasured muscles in children with cerebral palsy can be estimated from EMG measured from three to four muscles using our muscle synergy extrapolation method. In the future, the proposed muscle synergy-based method could be employed in gait clinics to minimise the required preparation time.

Evaluation of Synergy Extrapolation for Predicting Unmeasured Muscle Excitations from Measured Muscle Synergies

Electromyography (EMG)-driven musculoskeletal modeling relies on high-quality measurements of muscle electrical activity to estimate muscle forces. However, a critical challenge for practical deployment of this approach is missing EMG data from muscles that contribute substantially to joint moments. This situation may arise due to either the inability to measure deep muscles with surface electrodes or the lack of a sufficient number of EMG channels. Muscle synergy analysis (MSA) is a dimensionality reduction approach that decomposes a large number of muscle excitations into a small number of time-varying synergy excitations along with time-invariant synergy weights that define the contribution of each synergy excitation to all muscle excitations. This study evaluates how well missing muscle excitations can be predicted using synergy excitations extracted from muscles with available EMG data (henceforth called “synergy extrapolation” or SynX). The method was evaluated using a gait data set collected from a stroke survivor walking on an instrumented treadmill at self-selected and fastest-comfortable speeds. The evaluation process started with full calibration of a lower-body EMG-driven model using 16 measured EMG channels (collected using surface and fine wire electrodes) per leg. One fine wire EMG channel (either iliopsoas or adductor longus) was then treated as unmeasured. The synergy weights associated with the unmeasured muscle excitation were predicted by solving a nonlinear optimization problem where the errors between inverse dynamics and EMG-driven joint moments were minimized. The prediction process was performed for different synergy analysis algorithms (principal component analysis and non-negative matrix factorization), EMG normalization methods, and numbers of synergies. SynX performance was most influenced by the choice of synergy analysis algorithm and number of synergies. Principal component analysis with five or six synergies consistently predicted unmeasured muscle excitations the most accurately and with the greatest robustness to EMG normalization method. Furthermore, the associated joint moment matching accuracy was comparable to that produced by initial EMG-driven model calibration using all 16 EMG channels per leg. SynX may facilitate the assessment of human neuromuscular control and biomechanics when important EMG signals are missing.

Modeling Muscle Synergies as a Gaussian Process: Estimating Unmeasured Muscle Excitations using a Measured Subset

Continuous observation of muscle activity could provide a comprehensive picture of the loads experienced by muscles and joints during daily life. However, a major limitation to the practical application of this approach is the need to have surface electromyography (sEMG) sensors on all involved muscles. In this work, we model the synergistic relationship between muscles as a Gaussian process enabling the inference of unmeasured muscle excitations using a subset of measured data. Specifically, we developed a model for a single subject which uses sEMG data from four leg muscles to estimate the muscle excitation time-series of six other leg muscles during level walking at a self-selected speed. The proposed technique was able to accurately estimate the held-out muscle excitation time-series of the six muscles with correlation coefficients ranging from 0.74 to 0.87 and with mean absolute error less than 3%.

A Gaussian Process Model of Muscle Synergy Functions for Estimating Unmeasured Muscle Excitations Using a Measured Subset

Estimation of muscle excitations from a reduced sensor array could greatly improve current techniques in remote patient monitoring. Such an approach could allow continuous monitoring of clinically relevant biomechanical variables that are ideal for personalizing rehabilitation. In this paper, we introduce the notion of a muscle synergy function which describes the synergistic relationship between a subset of muscles. We develop from first principles an approximation to their behavior using Gaussian process regression and demonstrate the utility of the technique for estimating the excitation time-series of leg muscles during normal walking for nine healthy subjects. Specifically, excitations for six muscles were estimated using surface electromyography (sEMG) data during a finite time interval (called the input window) from four different muscles (called the input muscles) with mean absolute error (MAE) less than 5.0% of the maximum voluntary contraction (MVC) and that accounts for 82-88% of the variance (VAF) in the true excitations. Further, these estimated excitations informed muscle activations with less than 4.0% MAE and 89-93% VAF. We also present a detailed analysis of a number of different modeling choices, including every possible combination of four-, three- and two-muscle input sets, the size and structure of the input window, and the stationarity of the Gaussian process covariance functions. Further, application specific modifications for future use are discussed. The proposed technique lays a foundation to explore the use of reduced wearable sensor arrays and muscle synergy functions for monitoring clinically relevant biomechanics during daily life.

Pre-treatment EMG can be used to model post-treatment muscle coordination during walking in children with cerebral palsy

When treating children with Cerebral Palsy (CP), computational simulations based on musculoskeletal models have a great potential in assisting the clinical decision-making process towards the most promising treatments. In particular, predictive simulations could be used to predict and compare the functional outcome of a series of candidate interventions. In order to be able to benefit from these predictive simulations however, it is important to know how much information about the post-treatment patient’s motor control could be gathered from data available before the intervention. Within this paper, we quantified how much of the muscle activity measured after a treatment could be explained by subject-specific muscle synergies computed from EMG data collected before the intervention. We also investigated whether generic synergies could be used, in case no EMG data is available when running predictive simulations, to reproduce both pre- and post-treatment muscle activity in children with CP. Subject-specific synergies proved to be a good indicator of the patient’s post-treatment motor control, explaining on average more than 85% of the post-treatment muscle activity, compared to an average of 94% when applied to the original pre-treatment data. Generic synergies explained 84% of the pre-treatment and 83% of the post-treatment muscle activity on average, but performed relatively well for patients with low selective motor control and poorly in patients with more selectivity. Our results suggest that subject-specific muscle synergies computed from pre-treatment EMG data could be used with confidence to represent the post-treatment motor control of children with CP during walking. In addition, when performing simulations involving patients with a low selective motor control, generic synergies could be a valid alternative.

Estimating Biomechanical Time-Series with Wearable Sensors: A Systematic Review of Machine Learning Techniques

Wearable sensors have the potential to enable comprehensive patient characterization and optimized clinical intervention. Critical to realizing this vision is accurate estimation of biomechanical time-series in daily-life, including joint, segment, and muscle kinetics and kinematics, from wearable sensor data. The use of physical models for estimation of these quantities often requires many wearable devices making practical implementation more difficult. However, regression techniques may provide a viable alternative by allowing the use of a reduced number of sensors for estimating biomechanical time-series. Herein, we review 46 articles that used regression algorithms to estimate joint, segment, and muscle kinematics and kinetics. We present a high-level comparison of the many different techniques identified and discuss the implications of our findings concerning practical implementation and further improving estimation accuracy. In particular, we found that several studies report the incorporation of domain knowledge often yielded superior performance. Further, most models were trained on small datasets in which case nonparametric regression often performed best. No models were open-sourced, and most were subject-specific and not validated on impaired populations. Future research should focus on developing open-source algorithms using complementary physics-based and machine learning techniques that are validated in clinically impaired populations. This approach may further improve estimation performance and reduce barriers to clinical adoption.

Computational Design of FastFES Treatment to Improve Propulsive Force Symmetry During Post-stroke Gait: A Feasibility Study

Stroke is a leading cause of long-term disability worldwide and often impairs walking ability. To improve recovery of walking function post-stroke, researchers have investigated the use of treatments such as fast functional electrical stimulation (FastFES). During FastFES treatments, individuals post-stroke walk on a treadmill at their fastest comfortable speed while electrical stimulation is delivered to two muscles of the paretic ankle, ideally to improve paretic leg propulsion and toe clearance. However, muscle selection and stimulation timing are currently standardized based on clinical intuition and a one-size-fits-all approach, which may explain in part why some patients respond to FastFES training while others do not. This study explores how personalized neuromusculoskeletal models could potentially be used to enable individual-specific selection of target muscles and stimulation timing to address unique functional limitations of individual patients post-stroke. Treadmill gait data, including EMG, surface marker positions, and ground reactions, were collected from an individual post-stroke who was a non-responder to FastFES treatment. The patient's gait data were used to personalize key aspects of a full-body neuromusculoskeletal walking model, including lower-body joint functional axes, lower-body muscle force generating properties, deformable foot-ground contact properties, and paretic and non-paretic leg neural control properties. The personalized model was utilized within a direct collocation optimal control framework to reproduce the patient's unstimulated treadmill gait data (verification problem) and to generate three stimulated walking predictions that sought to minimize inter-limb propulsive force asymmetry (prediction problems). The three predictions used: (1) Standard muscle selection (gastrocnemius and tibialis anterior) with standard stimulation timing, (2) Standard muscle selection with optimized stimulation timing, and (3) Optimized muscle selection (soleus and semimembranosus) with optimized stimulation timing. Relative to unstimulated walking, the optimal control problems predicted a 41% reduction in propulsive force asymmetry for scenario (1), a 45% reduction for scenario (2), and a 64% reduction for scenario (3), suggesting that non-standard muscle selection may be superior for this patient. Despite these predicted improvements, kinematic symmetry was not noticeably improved for any of the walking predictions. These results suggest that personalized neuromusculoskeletal models may be able to predict personalized FastFES training prescriptions that could improve propulsive force symmetry, though inclusion of kinematic requirements would be necessary to improve kinematic symmetry as well.

Muscle Synergies Modify Optimization Estimates of Joint Stiffness During Walking

Because of its simplicity, static optimization (SO) is frequently used to resolve the muscle redundancy problem (i.e., more muscles than degrees-of-freedom (DOF) in the human musculoskeletal system). However, SO minimizes antagonistic co-activation and likely joint stiffness as well, which may not be physiologically realistic since the body modulates joint stiffness during movements such as walking. Knowledge of joint stiffness is limited due to the difficulty of measuring it experimentally, leading researchers to estimate it using computational models. This study explores how imposing a synergy structure on the muscle activations estimated by optimization (termed "synergy optimization," or SynO) affects calculated lower body joint stiffnesses during walking. By limiting the achievable muscle activations and coupling all time frames together, a synergy structure provides a potential mechanism for reducing indeterminacy and improving physiological co-activation but at the cost of a larger optimization problem. To compare joint stiffnesses produced by SynO (2-6 synergies) and SO, we used both approaches to estimate lower body muscle activations and forces for sample experimental overground walking data obtained from the first knee grand challenge competition. Both optimizations used a custom Hill-type muscle model that permitted analytic calculation of individual muscle contributions to the stiffness of spanned joints. Both approaches reproduced inverse dynamic joint moments well over the entire gait cycle, though SynO with only two synergies exhibited the largest errors. Maximum and mean joint stiffnesses for hip and knee flexion in particular decreased as the number of synergies increased from 2 to 6, with SO producing the lowest joint stiffness values. Our results suggest that SynO increases joint stiffness by increasing muscle co-activation, and furthermore, that walking with a reduced number of synergies may result in increased joint stiffness and perhaps stability.