Personalized neuromusculoskeletal modeling to improve treatment of mobility impairments: a perspective from European research sites

Inertia-Constrained Reinforcement Learning to Enhance Human Motor Control Modeling

Locomotor impairment is a highly prevalent and significant source of disability and significantly impacts the quality of life of a large portion of the population. Despite decades of research on human locomotion, challenges remain in simulating human movement to study the features of musculoskeletal drivers and clinical conditions. Most recent efforts to utilize reinforcement learning (RL) techniques are promising in the simulation of human locomotion and reveal musculoskeletal drives. However, these simulations often fail to mimic natural human locomotion because most reinforcement strategies have yet to consider any reference data regarding human movement. To address these challenges, in this study, we designed a reward function based on the trajectory optimization rewards (TOR) and bio-inspired rewards, which includes the rewards obtained from reference motion data captured by a single Inertial Moment Unit (IMU) sensor. The sensor was equipped on the participants' pelvis to capture reference motion data. We also adapted the reward function by leveraging previous research on walking simulations for TOR. The experimental results showed that the simulated agents with the modified reward function performed better in mimicking the collected IMU data from participants, which means that the simulated human locomotion was more realistic. As a bio-inspired defined cost, IMU data enhanced the agent's capacity to converge during the training process. As a result, the models' convergence was faster than those developed without reference motion data. Consequently, human locomotion can be simulated more quickly and in a broader range of environments, with a better simulation performance.

Abnormal synergistic gait mitigation in acute stroke using an innovative ankle-knee-hip interlimb humanoid robot: a preliminary randomized controlled trial

Abnormal spasticity and associated synergistic patterns are the most common neuromuscular impairments affecting ankle–knee–hip interlimb coordinated gait kinematics and kinetics in patients with hemiparetic stroke. Although patients with hemiparetic stroke undergo various treatments to improve gait and movement, it remains unknown how spasticity and associated synergistic patterns change after robot-assisted and conventional treatment. We developed an innovative ankle–knee–hip interlimb coordinated humanoid robot (ICT) to mitigate abnormal spasticity and synergistic patterns. The objective of the preliminary clinical trial was to compare the effects of ICT combined with conventional physical therapy (ICT-C) and conventional physical therapy and gait training (CPT-G) on abnormal spasticity and synergistic gait patterns in 20 patients with acute hemiparesis. We performed secondary analyses aimed at elucidating the biomechanical effects of Walkbot ICT on kinematic (spatiotemporal parameters and angles) and kinetic (active force, resistive force, and stiffness) gait parameters before and after ICT in the ICT-C group. The intervention for this group comprised 60-min conventional physical therapy plus 30-min robot-assisted training, 7 days/week, for 2 weeks. Significant biomechanical effects in knee joint kinematics; hip, knee, and ankle active forces; hip, knee, and ankle resistive forces; and hip, knee, and ankle stiffness were associated with ICT-C. Our novel findings provide promising evidence for conventional therapy supplemented by robot-assisted therapy for abnormal spasticity, synergistic, and altered biomechanical gait impairments in patients in the acute post-stroke recovery phase.

Trial Registration: Clinical Trials.gov identifier NCT03554642 (14/01/2020).

Torso Kinematics in Human Rolling Do Not Change When Upper Extremity Motion Is Constrained

Human rolling, as turning in bed, is a fundamental activity of daily living. A quantitative analysis of rolling could help identify the neuromusculoskeletal disorders that prohibit rolling and develop interventions for individuals who cannot roll. This study sought to determine whether crossing the arms over the chest would alter fundamental coordination patterns when rolling. Kinematic data were collected from 24 subjects as they rolled with and without their arms crossed over their chest. Crossing the arms decreased the mean peak angular velocities of the shoulders (p = .001) and pelvis (p = .013) and influenced the mean duration of the roll (p = .057). There were no fundamental differences in shoulder and pelvis coordination when rolling with the arms crossed over the chest, implying that the arms may not have a major role in rolling.

Deep reinforcement learning for modeling human locomotion control in neuromechanical simulation

Modeling human motor control and predicting how humans will move in novel environments is a grand scientific challenge. Researchers in the fields of biomechanics and motor control have proposed and evaluated motor control models via neuromechanical simulations, which produce physically correct motions of a musculoskeletal model. Typically, researchers have developed control models that encode physiologically plausible motor control hypotheses and compared the resulting simulation behaviors to measurable human motion data. While such plausible control models were able to simulate and explain many basic locomotion behaviors (e.g. walking, running, and climbing stairs), modeling higher layer controls (e.g. processing environment cues, planning long-term motion strategies, and coordinating basic motor skills to navigate in dynamic and complex environments) remains a challenge. Recent advances in deep reinforcement learning lay a foundation for modeling these complex control processes and controlling a diverse repertoire of human movement; however, reinforcement learning has been rarely applied in neuromechanical simulation to model human control. In this paper, we review the current state of neuromechanical simulations, along with the fundamentals of reinforcement learning, as it applies to human locomotion. We also present a scientific competition and accompanying software platform, which we have organized to accelerate the use of reinforcement learning in neuromechanical simulations. This “Learn to Move” competition was an official competition at the NeurIPS conference from 2017 to 2019 and attracted over 1300 teams from around the world. Top teams adapted state-of-the-art deep reinforcement learning techniques and produced motions, such as quick turning and walk-to-stand transitions, that have not been demonstrated before in neuromechanical simulations without utilizing reference motion data. We close with a discussion of future opportunities at the intersection of human movement simulation and reinforcement learning and our plans to extend the Learn to Move competition to further facilitate interdisciplinary collaboration in modeling human motor control for biomechanics and rehabilitation research

A Conceptual Blueprint for Making Neuromusculoskeletal Models Clinically Useful

The ultimate goal of most neuromusculoskeletal modeling research is to improve the treatment of movement impairments. However, even though neuromusculoskeletal models have become more realistic anatomically, physiologically, and neurologically over the past 25 years, they have yet to make a positive impact on the design of clinical treatments for movement impairments. Such impairments are caused by common conditions such as stroke, osteoarthritis, Parkinson’s disease, spinal cord injury, cerebral palsy, limb amputation, and even cancer. The lack of clinical impact is somewhat surprising given that comparable computational technology has transformed the design of airplanes, automobiles, and other commercial products over the same time period. This paper provides the author’s personal perspective for how neuromusculoskeletal models can become clinically useful. First, the paper motivates the potential value of neuromusculoskeletal models for clinical treatment design. Next, it highlights five challenges to achieving clinical utility and provides suggestions for how to overcome them. After that, it describes clinical, technical, collaboration, and practical needs that must be addressed for neuromusculoskeletal models to fulfill their clinical potential, along with recommendations for meeting them. Finally, it discusses how more complex modeling and experimental methods could enhance neuromusculoskeletal model fidelity, personalization, and utilization. The author hopes that these ideas will provide a conceptual blueprint that will help the neuromusculoskeletal modeling research community work toward clinical utility.

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.

Patients' perceptions with musculoskeletal disorders regarding their experience with healthcare providers and health services: an overview of reviews

Objectives

This overview of reviews aimed to identify (1) aspects of the patient experience when seeking care for musculoskeletal disorders from healthcare providers and the healthcare system, and (2) which mechanisms are used to measure aspects of the patient experience.

Data sources

Four databases were searched from inception to December 20th, 2019.

Review methods

Systematic or scoping reviews examining patient experience in seeking care for musculoskeletal from healthcare providers and the healthcare system were included. Independent authors screened and selected studies, extracted data, and assessed the methodological quality of the reviews. Patient experience concepts were compiled into five themes from a perspective of a) relational and b) functional aspects. A list of mechanisms used to capture the patient experience was also collected.

Results

Thirty reviews were included (18 systematic and 12 scoping reviews). Relational aspects were reported in 29 reviews and functional aspects in 25 reviews. For relational aspects, the most prevalent themes were “information needs” (education and explanation on diseases, symptoms, and self-management strategies) and “understanding patient expectations” (respect and empathy). For functional aspects, the most prevalent themes were patient’s “physical and environmental needs,” (cleanliness, safety, and accessibility of clinics), and “trusted expertise,” (healthcare providers’ competence and clinical skills to provide holistic care). Interviews were the most frequent mechanism identified to collect patient experience.

Conclusions

Measuring patient experience provides direct insights about the patient’s perspectives and may help to promote better patient-centered health services and increase the quality of care. Areas of improvement identified were interpersonal skills of healthcare providers and logistics of health delivery, which may lead to a more desirable patient-perceived experience and thus better overall healthcare outcomes.

Trial registration

Systematic review registration: PROSPERO (CRD42019136500).

Modelling the loading mechanics of anterior cruciate ligament

BACKGROUND AND OBJECTIVES

The anterior cruciate ligament (ACL) plays a crucial role in knee stability and is the most commonly injured knee ligament. Although ACL loading patterns have been investigated previously, the interactions between knee loadings transmitted to ACL remain elusive. Understanding the loading mechanism of ACL during dynamic tasks is essential to prevent ACL injuries. Therefore, we propose a computational model that predicts the force applied to ACL in response to knee loading in three planes of motion.

METHODS

First, a three-dimensional (3D) computational model was developed and validated using available cadaveric experimental data to predict ACL force. This 3D model was then combined with a neuromusculoskeletal model of lower limb and used to estimate in vivo ACL forces during a standardised drop-landing task. The neuromusculoskeletal model utilised movement data collected from female participants during a dynamic task and calculated lower limb joint kinematics and kinetics, as well as muscle forces.

RESULTS

The total ACL force predicted by the 3D computational ACL force model was in good agreement with cadaveric data, as strong correlation (r² = 0.96 and P < 0.001), minimal bias, and narrow limits of agreement were observed. The combined model further illustrated that the ACL is primarily loaded through the sagittal plane, mainly due to muscle loading.

CONCLUSIONS

The proposed computational model is the first validated model that can provide an accessible tool to develop and test knee ACL injury prevention programs for people with normal ACL. This method can be extended to study the abnormal ACL upon the availability of relevant experimental data.

Gait alteration strategies for knee osteoarthritis: a comparison of joint loading via generic and patient-specific musculoskeletal model scaling techniques

Gait modifications and laterally wedged insoles are non-invasive approaches used to treat medial compartment knee osteoarthritis. However, the outcome of these alterations is still a controversial topic. This study investigates how gait alteration techniques may have a unique effect on individual patients; and furthermore, the way we scale our musculoskeletal models to estimate the medial joint contact force may influence knee loading conditions. Five patients with clinical evidence of medial knee osteoarthritis were asked to walk at a normal walking speed over force plates and simultaneously 3D motion was captured during seven conditions (0°-, 5°-, 10°-insoles, shod, toe-in, toe-out, and wide stance). We developed patient-specific musculoskeletal models, using segmentations from magnetic resonance imaging to morph a generic model to patient-specific bone geometries and applied this morphing to estimate muscle insertion sites. Additionally, models were created of these patients using a simple linear scaling method. When examining the patients' medial compartment contact force (peak and impulse) during stance phase, a 'one-size-fits-all' gait alteration aimed to reduce medial knee loading did not exist. Moreover, the different scaling methods lead to differences in medial contact forces; highlighting the importance of further investigation of musculoskeletal modeling methods prior to use in the clinical setting.

JNER at 15 years: analysis of the state of neuroengineering and rehabilitation

On JNER's 15th anniversary, this editorial analyzes the state of the field of neuroengineering and rehabilitation. I first discuss some ways that the nature of neurorehabilitation research has evolved in the past 15 years based on my perspective as editor-in-chief of JNER and a researcher in the field. I highlight increasing reliance on advanced technologies, improved rigor and openness of research, and three, related, new paradigms - wearable devices, the Cybathlon competition, and human augmentation studies - indicators that neurorehabilitation is squarely in the age of wearability. Then, I briefly speculate on how the field might make progress going forward, highlighting the need for new models of training and learning driven by big data, better personalization and targeting, and an increase in the quantity and quality of usability and uptake studies to improve translation.

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.

Integrating an openEHR-based personalized virtual model for the ageing population within HBase

Background

Frailty is a common clinical syndrome in ageing population that carries an increased risk for adverse health outcomes including falls, hospitalization, disability, and mortality. As these outcomes affect the health and social care planning, during the last years there is a tendency of investing in monitoring and preventing strategies. Although a number of electronic health record (EHR) systems have been developed, including personalized virtual patient models, there are limited ageing population oriented systems.

Methods

We exploit the openEHR framework for the representation of frailty in ageing population in order to attain semantic interoperability, and we present the methodology for adoption or development of archetypes. We also propose a framework for a one-to-one mapping between openEHR archetypes and a column-family NoSQL database (HBase) aiming at the integration of existing and newly developed archetypes into it.

Results

The requirement analysis of our study resulted in the definition of 22 coherent and clinically meaningful parameters for the description of frailty in older adults. The implemented openEHR methodology led to the direct use of 22 archetypes, the modification and reuse of two archetypes, and the development of 28 new archetypes. Additionally, the mapping procedure led to two different HBase tables for the storage of the data.

Conclusions

In this work, an openEHR-based virtual patient model has been designed and integrated into an HBase storage system, exploiting the advantages of the underlying technologies. This framework can serve as a base for the development of a decision support system using the openEHR’s Guideline Definition Language in the future.

OpenSim Versus Human Body Model: A Comparison Study for the Lower Limbs During Gait

Musculoskeletal modeling and simulations have become popular tools for analyzing human movements. However, end users are often not aware of underlying modeling and computational assumptions. This study investigates how these assumptions affect biomechanical gait analysis outcomes performed with Human Body Model and the OpenSim gait2392 model. The authors compared joint kinematics, kinetics, and muscle forces resulting from processing data from 7 healthy adults with both models. Although outcome variables had similar patterns, there were statistically significant differences in joint kinematics (maximal difference: 9.8° [1.5°] in sagittal plane hip rotation), kinetics (maximal difference: 0.36 [0.10] N·m/kg in sagittal plane hip moment), and muscle forces (maximal difference: 8.51 [1.80] N/kg for psoas). These differences might be explained by differences in hip and knee joint center locations up to 2.4 (0.5) and 1.9 (0.2) cm in the posteroanterior and inferosuperior directions, respectively, and by the offset in pelvic reference frames of about 10° around the mediolateral axis. The choice of model may not influence the conclusions in clinical settings, where the focus is on interpreting deviations from the reference data, but it will affect the conclusions of mechanical analyses in which the goal is to obtain accurate estimates of kinematics and loading.

A Muscle-Specific Rehabilitation Training Method Based on Muscle Activation and the Optimal Load Orientation Concept

Training based on muscle-oriented repetitive movements has been shown to be beneficial for the improvement of movement abilities in human limbs in relation to fitness, athletic training, and rehabilitation training. In this paper, a muscle-specific rehabilitation training method based on the optimal load orientation concept (OLOC) was proposed for patients whose motor neurons are injured, but whose muscles and tendons are intact, to implement high-efficiency resistance training for the shoulder muscles, which is one of the most complex joints in the human body. A three-dimensional musculoskeletal model of the human shoulder was used to predict muscle forces experienced during shoulder movements, in which muscles that contributed to shoulder motion were divided into 31 muscle bundles, and the Hill model was used to characterize the force-length properties of the muscle. According to the musculoskeletal model, muscle activation was calculated to represent the muscle force. Thus, training based on OLOC was proposed by maximizing the activation of a specific muscle under each posture of the training process. The analysis indicated that the muscle-specific rehabilitation training method based on the OLOC significantly improved the training efficiency for specific muscles. The method could also be used for trajectory planning, load magnitude planning, and evaluation of training effects.

Construction and evaluation of a model for wheelchair propulsion in an individual with tetraplegia

Upper limb overuse injuries are common in manual wheelchair users with spinal cord injury. Patient-specific in silico models enhance experimental biomechanical analyses by estimating in vivo shoulder muscle and joint contact forces. Current models exclude deep shoulder muscles that have important roles in wheelchair propulsion. Freely accessible patient-specific models have not been generated for persons with tetraplegia, who have a greater risk for shoulder pain and injury. The objectives of this work were to (i) construct a freely accessible, in silico, musculoskeletal model capable of generating patient-specific dynamic simulations of wheelchair propulsion and (ii) establish proof-of-concept with data obtained from an individual with tetraplegia. Constructed with OpenSim, the model features muscles excluded in existing models. Shoulder muscle forces and activations were estimated via inverse dynamics. Mean absolute error of estimated muscle activations and fine-wire electromyography (EMG) recordings was computed. Mean muscle activation for five consecutive stroke cycles demonstrated good correlation (0.15-0.17) with fine-wire EMG. These findings, comparable to other studies, suggest that the model is capable of estimating shoulder muscle forces during wheelchair propulsion. The additional muscles may provide a greater understanding of shoulder muscle contribution to wheelchair propulsion. The model may ultimately serve as a powerful clinical tool. Graphical abstract ᅟ.

Sensitivity analysis of the estimated muscle forces during gait with respect to the musculoskeletal model parameters and dynamic simulation techniques

The purpose of the current study was to investigate the robustness of dynamic simulation results in the presence of uncertainties resulting from application of a scaled-generic musculoskeletal model instead of a subject-specific model as well as the effect of the choice of simulation method on the obtained muscle forces. The performed sensitivity analysis consisted of the following multibody parameter modifications: maximum isometric muscle forces, number of muscles, the hip joint centre location, segment masses as well as different dynamic simulation methods, namely static optimization with three different criteria and a computed muscle control algorithm (hybrid approach combining forward and inverse dynamics). Twenty-four different models and fifty-five resultant dynamic simulation data sets were analysed. The effects of model perturbation on the magnitude and profile of muscle forces were compared. It has been shown that estimated muscle forces are very sensitive to model parameters. The greatest impact was observed in the case of the force magnitude of the muscles generating high forces during gait (regardless of the modification introduced). However, the force profiles of those muscles were preserved. Relatively large differences in muscle forces were observed for different simulation techniques, which included both magnitude and profile of muscle forces. Personalization of model parameters would affect the resultant muscle forces and seems to be necessary to improve general accuracy of the estimated parameters. However, personalization alone will not ensure high accuracy due to the still unresolved muscle force sharing problem.

In Vivo Neuromechanics: Decoding Causal Motor Neuron Behavior with Resulting Musculoskeletal Function

Human motor function emerges from the interaction between the neuromuscular and the musculoskeletal systems. Despite the knowledge of the mechanisms underlying neural and mechanical functions, there is no relevant understanding of the neuro-mechanical interplay in the neuro-musculo-skeletal system. This currently represents the major challenge to the understanding of human movement. We address this challenge by proposing a paradigm for investigating spinal motor neuron contribution to skeletal joint mechanical function in the intact human in vivo. We employ multi-muscle spatial sampling and deconvolution of high-density fiber electrical activity to decode accurate α-motor neuron discharges across five lumbosacral segments in the human spinal cord. We use complete α-motor neuron discharge series to drive forward subject-specific models of the musculoskeletal system in open-loop with no corrective feedback. We perform validation tests where mechanical moments are estimated with no knowledge of reference data over unseen conditions. This enables accurate blinded estimation of ankle function purely from motor neuron information. Remarkably, this enables observing causal associations between spinal motor neuron activity and joint moment control. We provide a new class of neural data-driven musculoskeletal modeling formulations for bridging between movement neural and mechanical levels in vivo with implications for understanding motor physiology, pathology, and recovery.

Lower extremity EMG-driven modeling of walking with automated adjustment of musculoskeletal geometry

Neuromusculoskeletal disorders affecting walking ability are often difficult to manage, in part due to limited understanding of how a patient’s lower extremity muscle excitations contribute to the patient’s lower extremity joint moments. To assist in the study of these disorders, researchers have developed electromyography (EMG) driven neuromusculoskeletal models utilizing scaled generic musculoskeletal geometry. While these models can predict individual muscle contributions to lower extremity joint moments during walking, the accuracy of the predictions can be hindered by errors in the scaled geometry. This study presents a novel EMG-driven modeling method that automatically adjusts surrogate representations of the patient’s musculoskeletal geometry to improve prediction of lower extremity joint moments during walking. In addition to commonly adjusted neuromusculoskeletal model parameters, the proposed method adjusts model parameters defining muscle-tendon lengths, velocities, and moment arms. We evaluated our EMG-driven modeling method using data collected from a high-functioning hemiparetic subject walking on an instrumented treadmill at speeds ranging from 0.4 to 0.8 m/s. EMG-driven model parameter values were calibrated to match inverse dynamic moments for five degrees of freedom in each leg while keeping musculoskeletal geometry close to that of an initial scaled musculoskeletal model. We found that our EMG-driven modeling method incorporating automated adjustment of musculoskeletal geometry predicted net joint moments during walking more accurately than did the same method without geometric adjustments. Geometric adjustments improved moment prediction errors by 25% on average and up to 52%, with the largest improvements occurring at the hip. Predicted adjustments to musculoskeletal geometry were comparable to errors reported in the literature between scaled generic geometric models and measurements made from imaging data. Our results demonstrate that with appropriate experimental data, joint moment predictions for walking generated by an EMG-driven model can be improved significantly when automated adjustment of musculoskeletal geometry is included in the model calibration process.

Biofeedback for Gait Retraining Based on Real-Time Estimation of Tibiofemoral Joint Contact Forces

Biofeedback assisted rehabilitation and intervention technologies have the potential to modify clinically relevant biomechanics. Gait retraining has been used to reduce the knee adduction moment, a surrogate of medial tibiofemoral joint loading often used in knee osteoarthritis research. In this paper, we present an electromyogram-driven neuromusculoskeletal model of the lower-limb to estimate, in real-time, the tibiofemoral joint loads. The model included 34 musculotendon units spanning the hip, knee, and ankle joints. Full-body inverse kinematics, inverse dynamics, and musculotendon kinematics were solved in real-time from motion capture and force plate data to estimate the knee medial tibiofemoral contact force (MTFF). We analyzed five healthy subjects while they were walking on an instrumented treadmill with visual biofeedback of their MTFF. Each subject was asked to modify their gait in order to vary the magnitude of their MTFF. All subjects were able to increase their MTFF, whereas only three subjects could decrease it, and only after receiving verbal suggestions about possible gait modification strategies. Results indicate the important role of knee muscle activation patterns in modulating the MTFF. While this paper focused on the knee, the technology can be extended to examine the musculoskeletal tissue loads at different sites of the human body.

Neuromusculoskeletal model self-calibration for on-line sequential bayesian moment estimation

OBJECTIVE

Neuromusculoskeletal models involve many subject-specific physiological parameters that need to be adjusted to adequately represent muscle properties. Traditionally, neuromusculoskeletal models have been calibrated with a forward-inverse dynamic optimization which is time-consuming and unfeasible for rehabilitation therapy. Non self-calibration algorithms have been applied to these models. To the best of our knowledge, the algorithm proposed in this work is the first on-line calibration algorithm for muscle models that allows a generic model to be adjusted to different subjects in a few steps.

APPROACH

In this paper we propose a reformulation of the traditional muscle models that is able to sequentially estimate the kinetics (net joint moments), and also its full self-calibration (subject-specific internal parameters of the muscle from a set of arbitrary uncalibrated data), based on the unscented Kalman filter. The nonlinearity of the model as well as its calibration problem have obliged us to adopt the sum of Gaussians filter suitable for nonlinear systems.

MAIN RESULTS

This sequential Bayesian self-calibration algorithm achieves a complete muscle model calibration using as input only a dataset of uncalibrated sEMG and kinematics data. The approach is validated experimentally using data from the upper limbs of 21 subjects.

SIGNIFICANCE

The results show the feasibility of neuromusculoskeletal model self-calibration. This study will contribute to a better understanding of the generalization of muscle models for subject-specific rehabilitation therapies. Moreover, this work is very promising for rehabilitation devices such as electromyography-driven exoskeletons or prostheses.

Gait analysis in children with cerebral palsy

Cerebral palsy (CP) children present complex and heterogeneous motor disorders that cause gait deviations.

Clinical gait analysis (CGA) is needed to identify, understand and support the management of gait deviations in CP. CGA assesses a large amount of quantitative data concerning patients’ gait characteristics, such as video, kinematics, kinetics, electromyography and plantar pressure data.

Common gait deviations in CP can be grouped into the gait patterns of spastic hemiplegia (drop foot, equinus with different knee positions) and spastic diplegia (true equinus, jump, apparent equinus and crouch) to facilitate communication. However, gait deviations in CP tend to be a continuum of deviations rather than well delineated groups. To interpret CGA, it is necessary to link gait deviations to clinical impairments and to distinguish primary gait deviations from compensatory strategies.

CGA does not tell us how to treat a CP patient, but can provide objective identification of gait deviations and further the understanding of gait deviations. Numerous treatment options are available to manage gait deviations in CP. Generally, treatments strive to limit secondary deformations, re-establish the lever arm function and preserve muscle strength.

Additional roles of CGA are to better understand the effects of treatments on gait deviations.

Cite this article: Armand S, Decoulon G, Bonnefoy-Mazure A. Gait analysis in children with cerebral palsy. EFORT Open Rev 2016;1:448-460. DOI: 10.1302/2058-5241.1.000052.

Real-time inverse kinematics and inverse dynamics for lower limb applications using OpenSim

Real-time estimation of joint angles and moments can be used for rapid evaluation in clinical, sport, and rehabilitation contexts. However, real-time calculation of kinematics and kinetics is currently based on approximate solutions or generic anatomical models. We present a real-time system based on OpenSim solving inverse kinematics and dynamics without simplifications at 2000 frame per seconds with less than 31.5 ms of delay. We describe the software architecture, sensitivity analyses to minimise delays and errors, and compare offline and real-time results. This system has the potential to strongly impact current rehabilitation practices enabling the use of personalised musculoskeletal models in real-time.

Neural Data-Driven Musculoskeletal Modeling for Personalized Neurorehabilitation Technologies

OBJECTIVES

The development of neurorehabilitation technologies requires the profound understanding of the mechanisms underlying an individual's motor ability and impairment. A major factor limiting this understanding is the difficulty of bridging between events taking place at the neurophysiologic level (i.e., motor neuron firings) with those emerging at the musculoskeletal level (i.e. joint actuation), in vivo in the intact moving human. This review presents emerging model-based methodologies for filling this gap that are promising for developing clinically viable technologies.

METHODS

We provide a design overview of musculoskeletal modeling formulations driven by recordings of neuromuscular activity with a critical comparison to alternative model-free approaches in the context of neurorehabilitation technologies. We present advanced electromyography-based techniques for interfacing with the human nervous system and model-based techniques for translating the extracted neural information into estimates of motor function.

RESULTS

We introduce representative application areas where modeling is relevant for accessing neuromuscular variables that could not be measured experimentally. We then show how these variables are used for designing personalized rehabilitation interventions, biologically inspired limbs, and human-machine interfaces.

CONCLUSION

The ability of using electrophysiological recordings to inform biomechanical models enables accessing a broader range of neuromechanical variables than analyzing electrophysiological data or movement data individually. This enables understanding the neuromechanical interplay underlying in vivo movement function, pathology, and robot-assisted motor recovery.

SIGNIFICANCE

Filling the gap between our understandings of movement neural and mechanical functions is central for addressing one of the major challenges in neurorehabilitation: personalizing current technologies and interventions to an individual's anatomy and impairment.

Upper limb strength estimation of physically impaired persons using a musculoskeletal model: A sensitivity analysis

Sensitivity of upper limb strength calculated from a musculoskeletal model was analyzed, with focus on how the sensitivity is affected when the model is adapted to represent a person with physical impairment. Sensitivity was calculated with respect to four muscle-tendon parameters: muscle peak isometric force, muscle optimal length, muscle pennation, and tendon slack length. Results obtained from a musculoskeletal model of average strength showed highest sensitivity to tendon slack length, followed by muscle optimal length and peak isometric force, which is consistent with existing studies. Muscle pennation angle was relatively insensitive. The analysis was repeated after adapting the musculoskeletal model to represent persons with varying severities of physical impairment. Results showed that utilizing the weakened model significantly increased the sensitivity of the calculated strength at the hand, with parameters previously insensitive becoming highly sensitive. This increased sensitivity presents a significant challenge in applications utilizing musculoskeletal models to represent impaired individuals.

MOtoNMS: A MATLAB toolbox to process motion data for neuromusculoskeletal modeling and simulation

BACKGROUND

Neuromusculoskeletal modeling and simulation enable investigation of the neuromusculoskeletal system and its role in human movement dynamics. These methods are progressively introduced into daily clinical practice. However, a major factor limiting this translation is the lack of robust tools for the pre-processing of experimental movement data for their use in neuromusculoskeletal modeling software.

RESULTS

This paper presents MOtoNMS (matlab MOtion data elaboration TOolbox for NeuroMusculoSkeletal applications), a toolbox freely available to the community, that aims to fill this lack. MOtoNMS processes experimental data from different motion analysis devices and generates input data for neuromusculoskeletal modeling and simulation software, such as OpenSim and CEINMS (Calibrated EMG-Informed NMS Modelling Toolbox). MOtoNMS implements commonly required processing steps and its generic architecture simplifies the integration of new user-defined processing components. MOtoNMS allows users to setup their laboratory configurations and processing procedures through user-friendly graphical interfaces, without requiring advanced computer skills. Finally, configuration choices can be stored enabling the full reproduction of the processing steps. MOtoNMS is released under GNU General Public License and it is available at the SimTK website and from the GitHub repository. Motion data collected at four institutions demonstrate that, despite differences in laboratory instrumentation and procedures, MOtoNMS succeeds in processing data and producing consistent inputs for OpenSim and CEINMS.

CONCLUSIONS

MOtoNMS fills the gap between motion analysis and neuromusculoskeletal modeling and simulation. Its support to several devices, a complete implementation of the pre-processing procedures, its simple extensibility, the available user interfaces, and its free availability can boost the translation of neuromusculoskeletal methods in daily and clinical practice.

Effect of lower-limb joint models on subject-specific musculoskeletal models and simulations of daily motor activities

Understanding the validity of using musculoskeletal models is critical, making important to assess how model parameters affect predictions. In particular, assumptions on joint models can affect predictions from simulations of movement, and the identification of image-based joints is unavoidably affected by uncertainty that can decrease the benefits of increasing model complexity. We evaluated the effect of different lower-limb joint models on muscle and joint contact forces during four motor tasks, and assessed the sensitivity to the uncertainties in the identification of anatomical four-bar-linkage joints. Three MRI-based musculoskeletal models having different knee and ankle joint models were created and used for the purpose. Model predictions were compared against a baseline model including simpler and widely-adopted joints. In addition, a probabilistic analysis was performed by perturbing four-bar-linkage joint parameters according to their uncertainty. The differences between models depended on the motor task analyzed, and there could be marked differences at peak loading (up to 2.40 BW at the knee and 1.54 BW at the ankle), although they were rather small over the motor task cycles (up to 0.59 BW at the knee and 0.31 BW at the ankle). The model including more degrees of freedom showed more discrepancies in predicted muscle activations compared to measured muscle activity. Further, including image-based four-bar-linkages was robust to simulate walking, chair rise and stair ascent, but not stair descent (peak standard deviation of 2.66 BW), suggesting that joint model complexity should be set according to the imaging dataset available and the intended application, performing sensitivity analyses.

CEINMS: A toolbox to investigate the influence of different neural control solutions on the prediction of muscle excitation and joint moments during dynamic motor tasks

Personalized neuromusculoskeletal (NMS) models can represent the neurological, physiological, and anatomical characteristics of an individual and can be used to estimate the forces generated inside the human body. Currently, publicly available software to calculate muscle forces are restricted to static and dynamic optimisation methods, or limited to isometric tasks only. We have created and made freely available for the research community the Calibrated EMG-Informed NMS Modelling Toolbox (CEINMS), an OpenSim plug-in that enables investigators to predict different neural control solutions for the same musculoskeletal geometry and measured movements. CEINMS comprises EMG-driven and EMG-informed algorithms that have been previously published and tested. It operates on dynamic skeletal models possessing any number of degrees of freedom and musculotendon units and can be calibrated to the individual to predict measured joint moments and EMG patterns. In this paper we describe the components of CEINMS and its integration with OpenSim. We then analyse how EMG-driven, EMG-assisted, and static optimisation neural control solutions affect the estimated joint moments, muscle forces, and muscle excitations, including muscle co-contraction.

Selective lateral muscle activation in moderate medial knee osteoarthritis subjects does not unload medial knee condyle

There is some debate in the literature regarding the role of quadriceps-hamstrings co-contraction in the onset and progression of knee osteoarthritis. Does co-contraction during walking increase knee contact loads, thereby causing knee osteoarthritis, or might it be a compensatory mechanism to unload the medial tibial condyle? We used a detailed musculoskeletal model of the lower limb to test the hypothesis that selective activation of lateral hamstrings and quadriceps, in conjunction with inhibited medial gastrocnemius, can actually reduce the joint contact force on the medial compartment of the knee, independent of changes in kinematics or external forces. "Baseline" joint loads were computed for eight subjects with moderate medial knee osteoarthritis (OA) during level walking, using static optimization to resolve the system of muscle forces for each subject's scaled model. Holding all external loads and kinematics constant, each subject's model was then perturbed to represent non-optimal "OA-type" activation based on mean differences detected between electromyograms (EMG) of control and osteoarthritis subjects. Knee joint contact forces were greater for the "OA-type" than the "Baseline" distribution of muscle forces, particularly during early stance. The early-stance increase in medial contact load due to the "OA-type" perturbation could implicate this selective activation strategy as a cause of knee osteoarthritis. However, the largest increase in the contact load was found at the lateral condyle, and the "OA-type" lateral activation strategy did not increase the overall (greater of the first or second) medial peak contact load. While "OA-type" selective activation of lateral muscles does not appear to reduce the medial knee contact load, it could allow subjects to increase knee joint stiffness without any further increase to the peak medial contact load.

Admittance control scheme for implementing model-based assistance-as-needed on a robot

A model-based assistance-as-needed paradigm has been developed to govern the assistance provided by an assistive robot to its operator. This paradigm has advantages over existing methods of providing assistance-as-needed for applications such as robotic rehabilitation. However, implementation of the model-based paradigm requires a control scheme to be developed which controls the robot so as to provide the assistance calculated by the model-based paradigm to its operator. In this paper an admittance control scheme for providing model-based assistance-as-needed is presented. It is developed considering its suitability for human-robot interaction, and its role within the model-based assistance-as-needed framework. Results from the control implemented on an example robot showed it is capable of providing the operator with the desired level of assistance as governed by the model-based paradigm. This is an essential requirement for the paradigm to be capable of providing efficacious assistance-as-needed in applications such as robotic rehabilitation.

Neuromotor recovery from stroke: computational models at central, functional, and muscle synergy level

Computational models of neuromotor recovery after a stroke might help to unveil the underlying physiological mechanisms and might suggest how to make recovery faster and more effective. At least in principle, these models could serve: (i) To provide testable hypotheses on the nature of recovery; (ii) To predict the recovery of individual patients; (iii) To design patient-specific “optimal” therapy, by setting the treatment variables for maximizing the amount of recovery or for achieving a better generalization of the learned abilities across different tasks. Here we review the state of the art of computational models for neuromotor recovery through exercise, and their implications for treatment. We show that to properly account for the computational mechanisms of neuromotor recovery, multiple levels of description need to be taken into account. The review specifically covers models of recovery at central, functional and muscle synergy level.

A musculoskeletal model of human locomotion driven by a low dimensional set of impulsive excitation primitives

Human locomotion has been described as being generated by an impulsive (burst-like) excitation of groups of musculotendon units, with timing dependent on the biomechanical goal of the task. Despite this view being supported by many experimental observations on specific locomotion tasks, it is still unknown if the same impulsive controller (i.e., a low-dimensional set of time-delayed excitastion primitives) can be used as input drive for large musculoskeletal models across different human locomotion tasks. For this purpose, we extracted, with non-negative matrix factorization, five non-negative factors from a large sample of muscle electromyograms in two healthy subjects during four motor tasks. These included walking, running, sidestepping, and crossover cutting maneuvers. The extracted non-negative factors were then averaged and parameterized to obtain task-generic Gaussian-shaped impulsive excitation curves or primitives. These were used to drive a subject-specific musculoskeletal model of the human lower extremity. Results showed that the same set of five impulsive excitation primitives could be used to predict the dynamics of 34 musculotendon units and the resulting hip, knee and ankle joint moments (i.e., NRMSE = 0.18 ± 0.08, and R² = 0.73 ± 0.22 across all tasks and subjects) without substantial loss of accuracy with respect to using experimental electromyograms (i.e., NRMSE = 0.16 ± 0.07, and R² = 0.78 ± 0.18 across all tasks and subjects). Results support the hypothesis that biomechanically different motor tasks might share similar neuromuscular control strategies. This might have implications in neurorehabilitation technologies such as human-machine interfaces for the torque-driven, proportional control of powered prostheses and orthoses. In this, device control commands (i.e., predicted joint torque) could be derived without direct experimental data but relying on simple parameterized Gaussian-shaped curves, thus decreasing the input drive complexity and the number of needed sensors.

Experimental evaluation of a model-based assistance-as-needed paradigm using an assistive robot

In robotic rehabilitation a promising paradigm is assistance-as-needed. This is because it promotes patient active participation which is essential for neuro-rehabilitation. A model-based assistance-as-needed paradigm has been developed which utilizes a musculoskeletal model representing the subject to calculate their assistance needs. In this paper we experimentally evaluate this model-based paradigm to control an assistive robot and provide a subject with assistance-as-needed at the muscular level. A subject with impairments defined in specific muscle groups performs a number of upper limb tasks, whilst receiving assistance from a robotic exoskeleton. The paradigm is evaluated on its ability to provide assistance only as the subject needs, depending on the tasks being performed and the impairments defined. Results show that the model-based assistance-as-needed paradigm was relatively successful in providing assistance when it was needed.