Metabolic cost calculations of gait using musculoskeletal energy models, a comparison study

Marker Data Enhancement for Markerless Motion Capture

OBJECTIVE

Human pose estimation models can measure movement from videos at a large scale and low cost; however, open-source pose estimation models typically detect only sparse keypoints, which leads to inaccurate joint kinematics. OpenCap, a freely available service for researchers to measure movement from videos, mitigates this issue using a deep learning model-the marker enhancer-that transforms sparse video keypoints into dense anatomical markers. However, OpenCap performs poorly on movements not included in the training data. Here, we create a much larger and more diverse training dataset and develop a more accurate and generalizable marker enhancer.

METHODS

We compiled marker-based motion capture data from 1176 subjects and synthesized 1433 hours of video keypoints and anatomical markers to train the marker enhancer. We evaluated its accuracy in computing kinematics using both benchmark movement videos and synthetic data representing unseen, diverse movements.

RESULTS

The marker enhancer improved kinematic accuracy on benchmark movements (mean error: 4.1$^\circ$, max: 8.7$^\circ$) compared to using video keypoints (mean: 9.6$^\circ$, max: 43.1$^\circ$) and OpenCap's original enhancer (mean: 5.3$^\circ$, max: 11.5$^\circ$). It also better generalized to unseen, diverse movements (mean: 4.1$^\circ$, max: 6.7$^\circ$) than OpenCap's original enhancer (mean: 40.4$^\circ$, max: 252.0$^\circ$).

CONCLUSION

Our marker enhancer demonstrates both improved accuracy and generalizability across diverse movements.

SIGNIFICANCE

We integrated the marker enhancer into OpenCap, thereby offering its thousands of users more accurate measurements across a broader range of movements.

Abnormal activity in the brainstem affects gait in a neuromusculoskeletal model

BACKGROUND

The ability to start and stop locomotion in response to different situations is an essential survival strategy in mammals. Mammalian locomotion is controlled by central pattern generators in the spinal cord, which are modulated by higher centers, particularly by the stimulation of the midbrain locomotor region. The midbrain locomotor region consists of the pedunculopontine nucleus and cuneiform nucleus, each having different roles in animals. Optogenetic activation of the cuneiform nucleus increases locomotion activities, whereas that of pedunculopontine nucleus decreases them. In neurological disorders such as Parkinson's disease, patients exhibit disturbed locomotion controls, including freezing of gait, which is defined as "a brief, episodic absence or marked reduction in the forward progression of the feet despite the intention to walk." However, the details and pathophysiological mechanisms of freezing of gait remain unclear.

METHODS

In this study, we aimed to elucidate the mechanisms underlying freezing of gait using a two-dimensional neuromusculoskeletal model fixed on the sagittal plane. This model consisted of a body with seven links and 18 muscles as well as a neural system including the brainstem and spinal cord. We developed a normal condition model and then derived a model of abnormal brainstem activity by modifying the parameters of the pedunculopontine nucleus and cuneiform nucleus during the initial 3 s of walking.

RESULTS

The normal models walked successfully following internal parameter optimization using standard genetic algorithms. In an abnormal model, 156 freezing of gait events were detected among 40,000 parameter sets using a freezing of gait-identifying algorithm. Hierarchical cluster analysis identified four clusters of parameters, based on the intensities of the pedunculopontine nucleus and cuneiform nucleus activity, differentiated in physiological movement types during freezing of gait events that were similar to the clinical classification types of freezing of gait.

CONCLUSIONS

Our results indicate that pedunculopontine nucleus and cuneiform nucleus activities could be linked with freezing of gait and that different modifications of those activities could generate observed freezing of gait subtypes. Our models can provide insights relevant for understanding the pathophysiological mechanisms of freezing of gait and are expected to assist in the classification of freezing of gait subtypes.

Using physiologically based models to predict in vivo skeletal muscle energetics

Understanding how muscles use energy is essential for elucidating the role of skeletal muscle in animal locomotion. Yet, experimental measures of in vivo muscle energetics are challenging to obtain, so physiologically based muscle models are often used to estimate energy use. These predictions of individual muscle energy expenditure are not often compared with indirect whole-body measures of energetic cost. Here, we examined and illustrated the capability of physiologically based muscle models to predict in vivo measures of energy use, which rely on fundamental relationships between muscle mechanical state and energy consumption. To improve model predictions and ensure a physiological basis for model parameters, we refined our model to include data from isolated muscle experiments and account for inefficiencies in ATP recovery processes. Simulations were performed to capture three different experimental protocols, which involved varying contraction frequency, duty cycle and muscle fascicle length. Our results demonstrated the ability of the model to capture the dependence of energetic cost on mechanical state across contractile conditions, but tended to underpredict the magnitude of energetic cost. Our analysis revealed that the model was most sensitive to the force-velocity parameters and the data informing the energetic parameters when predicting in vivo energetic rates. This work highlights that it is the mechanics of skeletal muscle contraction that govern muscle energy use, although the precise physiological parameters for human muscle likely require detailed investigation.

Empowering human-like walking with a bio-inspired gait controller for an under-actuated torque-driven human model

Human gait simulation plays a crucial role in providing insights into various aspects of locomotion, such as diagnosing injuries and impairments, assessing abnormal gait patterns, and developing assistive and rehabilitation technologies. To achieve more realistic gait simulation results, it is essential to use a comprehensive model that accurately replicates the kinematics and kinetics of human movement. Human skeletal models in OpenSim software provide anatomically accurate and anthropomorphic structures, enabling users to create personalized models that accurately replicate individual human behavior. However, these torque-driven models encounter challenges in stabilizing unactuated degree of freedom of pelvis tilt in forward dynamic simulations Adopting a bio-inspired strategy that ensures human balance with a minimized energy expenditure during walking, this paper addresses a gait controller for a torque-driven human skeletal model to achieve stable walking. The proposed controller employs a nonlinear model-based approach to calculate a balance-equivalent control torque and utilizes the hip-ankle strategy to distribute this torque across the lower-limb joints during the stance phase. To optimize the parameters of the trajectory tracking controller and the balance distribution coefficients, we developed a forward dynamic simulation interface established between MATLAB and OpenSim. The simulation results indicated that the torque-driven model achieves a natural gait, with joint torques closely aligning with the experimental data. The robustness of the bio-inspired gait controller was further evaluated by applying a range of external forces on the skeletal model. The robustness analysis demonstrated efficient balance recovery mechanism of the proposed bio-inspired gait controller in response to external disturbances.

Predicting lower body joint moments and electromyography signals using ground reaction forces during walking and running: An artificial neural network approach

BACKGROUND

This study leverages Artificial Neural Networks (ANNs) to predict lower limb joint moments and electromyography (EMG) signals from Ground Reaction Forces (GRF), providing a novel perspective on human gait analysis. This approach aims to enhance the accessibility and affordability of biomechanical assessments using GRF data, thus eliminating the need for costly motion capture systems.

RESEARCH QUESTION

Can ANNs use GRF data to accurately predict joint moments in the lower limbs and EMG signals?

METHODS

We employed ANNs to analyze GRF data and to use them to predict joint moments (363-trials; 4-datasets) and EMG signals (63-trials; 2-datasets). We selected the EMG timeseries of 6 muscles (Biceps Femoris, Gluteus Maximus, Rectus Femoris, Medial Gastrocnemius, Soleus, and Tibialis Anterior) and joint moment timeseries in the lower limbs (ankle, knee, and hip).

RESULTS

The ANN models demonstrated high predictive accuracy for joint moments (R-value: 0.97, p < 0.0001) and EMG signals (R-value: 0.95, p < 0.0001) across various gait activities, including walking and running. This underscores the potential of using GRF data to understand complex biomechanical interactions, offering significant insights into human locomotion.

SIGNIFICANCE

The significance of this research extends broadly, touching upon the development of portable devices, assistive technologies, and personalized rehabilitation programs. Our findings have the potential to broaden the accessibility of advanced biomechanical analysis with implications spanning disciplines such as sports science, rehabilitation, and the advancement of innovative assistive devices and exoskeletons.

Behavioural energetics in human locomotion: how energy use influences how we move

Nearly a century of research has shown that humans, and other animals, tend to move in ways that minimize energy use. A growing body of evidence suggests that energetic cost is not only an outcome of our movement, but also plays a central role in continuously shaping it. This has led to an emerging research area, at the nexus between biomechanics and neuroscience, termed behavioural energetics, which is focused on understanding the mechanisms of energy optimization and how this shapes our coordination and behaviour. In this Review, we first summarize the existing evidence for and against our preferred locomotor behaviours coinciding with energy optima. Although evidence of our preference for energetically optimal gaits has existed for decades, new research is revealing its relevance across a surprising array of dynamic locomotor tasks and complex environments. We next discuss evidence that we adapt our gait toward energy optima over short timescales and in novel environments, which we view as a more stringent test that energy expenditure is optimized in real-time. This necessitates that we sense energy use, or proxies for it, on similar timescales. We therefore next provide an overview of candidate sensory mechanisms of energy expenditure. Finally, we discuss how behavioural energetics can be applied to novel wearable assistive technologies and rehabilitation paradigms, and conclude the Review by outlining what we see as the most important future challenges and opportunities in behavioural energetics.

Integrating muscle energetics into biomechanical models to understand variance in the cost of movement

In this Review, we explore the state of the art of biomechanical models for estimating energy consumption during terrestrial locomotion. We consider different mechanical models that provide a solid framework to understand movement energetics from the perspective of force and work requirements. Whilst such models are highly informative, they lack specificity for predicting absolute metabolic rates across a range of species or variations in movement patterns. Muscles consume energy when they activate to generate tension, as well as when they shorten to generate positive work. Phenomenological muscle models incorporating steady-state parameters have been developed and are able to reproduce how muscle fibre energy consumption changes under different contractile conditions; however, such models are difficult to validate when scaled up to whole muscle. This is, in part, owing to limited availability of data that relate muscle dynamics to energetic rates during contraction of large mammalian muscles. Furthermore, factors including the compliance of tendinous tissue, dynamic shape changes and motor unit recruitment can alter the dynamics of muscle contractile tissue and potentially improve muscle efficiency under some locomotion conditions. Despite the many challenges, energetic cost estimates derived from musculoskeletal models that simulate muscle function required to generate movement have been shown to reasonably predict changes in human metabolic rates under different movement conditions. However, accurate predictions of absolute metabolic rate are still elusive. We suggest that conceptual models may be adapted based on our understanding of muscle energetics to better predict the variance in movement energetics both within and between terrestrial species.

An in-silico investigation of the effect of changing cycling crank power and cadence on muscle energetics and active muscle volume

This study used musculoskeletal modelling to explore the relationship between cycling conditions (power output and cadence) and muscle activation and metabolic power. We hypothesized that the cadence that minimized the simulated average active muscle volume would be higher than the cadence that minimized the simulated metabolic power. We validated the simulation by comparing the predicted muscle activation and fascicle velocities with experimental electromyography and ultrasound images. We found strong correlations for averaged muscle activations and moderate to good correlations for fascicle dynamics. These correlations tended to weaken when analyzed at the individual participant level. Our study revealed a curvilinear relationship between the average active muscle volume and cadence, with the minimum active volume being aligned to the self-selected cadence. The simulated metabolic power was consistent with previous results and was minimized at lower cadences than that which minimized active muscle volume across power outputs. Although there are some limitations to the musculoskeletal modelling approach, the findings suggest that minimizing active muscle volume may be a more important factor than minimizing metabolic power for self-selected cycling cadence preferences. Further research is warranted to explore the potential of an active muscle volume-based objective function for control schemes across a wider range of cycling conditions.

A novel rigid Foot-Ground contact model for Predicting ground reaction forces and center of pressure during normal gait

Ground reaction forces (GRFs) and center of pressure (COP) are essential for understanding human motion and evaluating biomechanical parameters, but measuring them with force plates is often limited in many scenarios. In this study, we propose a novel methodology for estimating GRFs and COP during normal gait based on a rigid foot-ground contact model, referred to as the COP phase transition continuity model (COP-PTCM). The GRFs and COP are calculated based on the Newton-Euler Equations during the single support phase (SSP). Considering the spatiotemporal continuity of the COP trajectory during normal gait, the COP data for the double support phase (DSP) is obtained by an improved logistic function fitted using the COP data from the SSP. GRFs during the DSP are optimized using the minimum energy hypothesis. The COP-PTCM method is used to estimate the GRFs and COP of ten participants during normal gait, and the results are compared with simultaneously measured force plate data, yielding the relative root mean square error (rRMSE) between measured and estimated GRFs in the anterior-posterior, vertical, and medial-lateral directions are 10.90±2.09 %, 4.73±1.44 %, and 15.17±1.69 %, respectively. Additionally, the rRMSE between measured and estimated COP in the anterior-posterior direction is 11.23±0.03 %. The above comparison validates the effectiveness and accuracy of the proposed method.

Influence of restricted visual input on lower limb joint works of female children during sit-to-stand

BACKGROUND

The ankle-knee-hip joint systems have structures that can produce mechanical work through elastic, viscoelastic mechanisms or muscle activity. This study aimed to compute sit-to-stand (STS) joint works in lower limbs between blind and sighted children to find the relationship between visual memory and STS joint work variables.

METHODS

This study included fifteen female children with congenitally blind (CB) and 30 healthy girls without visual impairments. The children with no visual impairments were randomly divided into two condition groups with 15 each, the eyes open (EO) and the eyes closed (EC). Inverse dynamics calculated joint works by integrating multiple the moment and angular velocity (F1) and force and velocity (F2). They were normalized to body mass and body height.

RESULTS

Generally, the sensitivity of F1 (on both sides in the sagittal and frontal planes) was more than F2 (on the non-dominant side in the mediolateral and vertical axes). In the ML axis, the EC group had insufficient maximal non-dominant hip work relative to the EO group (p = 0.002). In addition, the CB group suffered from low hip efficiency (p = 0.003) and high knee (p < 0.001) mechanical work.

CONCLUSIONS

Numerous differences between CB and EC groups (on knee and hip works) showed that the time of visual input deprivation could change the type of human body's strategies to reach the consolidation process and keep adequate balance during STS. Therefore, rehabilitation programs should be aimed at addressing the impairments in the management of restricted visual input during STS performance.

Insights into muscle metabolic energetics: Modelling muscle-tendon mechanics and metabolic rates during walking across speeds

The metabolic energy rate of individual muscles is impossible to measure without invasive procedures. Prior studies have produced models to predict metabolic rates based on experimental observations of isolated muscle contraction from various species. Such models can provide reliable predictions of metabolic rates in humans if muscle properties and control are accurately modeled. This study aimed to examine how muscle-tendon model individualization and metabolic energy models influenced estimation of muscle-tendon states and time-series metabolic rates, to evaluate the agreement with empirical data, and to provide predictions of the metabolic rate of muscle groups and gait phases across walking speeds. Three-dimensional musculoskeletal simulations with prescribed kinematics and dynamics were performed. An optimal control formulation was used to compute muscle-tendon states with four levels of individualization, ranging from a scaled generic model and muscle controls based on minimal activations, inclusion of calibrated muscle passive forces, personalization of Achilles and quadriceps tendon stiffnesses, to finally informing muscle controls with electromyography. We computed metabolic rates based on existing models. Simulations with calibrated passive forces and personalized tendon stiffness most accurately estimate muscle excitations and fiber lengths. Interestingly, the inclusion of electromyography did not improve our estimates. The whole-body average metabolic cost was better estimated with a subset of metabolic energy models. We estimated metabolic rate peaks near early stance, pre-swing, and initial swing at all walking speeds. Plantarflexors accounted for the highest cost among muscle groups at the preferred speed and were similar to the cost of hip adductors and abductors combined. Also, the swing phase accounted for slightly more than one-quarter of the total cost in a gait cycle, and its relative cost decreased with walking speed. Our prediction might inform the design of assistive devices and rehabilitation treatment. The code and experimental data are available online.

Relationship between metabolic cost, muscle moments and co-contraction during walking and running

BACKGROUND

The metabolic cost of locomotion is a key factor in walking and running performance. It has been studied by analysing the activation and co-activation of the muscles of the lower limbs. However, these measures do not comprehensively address muscle mechanics, in contrast to approaches using muscle moments and co-contraction.

RESEARCH QUESTION

What is the effect of speed and type of locomotion on muscle moments and co-contraction, and their relationship with metabolic cost during walking and running?

METHODS

Eleven recreational athletes (60.5 ± 7.1 kg; 169.0 ± 6.6 cm; 23.6 ± 3.3 years) walked and ran on a treadmill at different speeds, including a similar speed of 1.75 m.s-1. Metabolic cost was estimated from gas exchange measurements. Muscle moments and co-contraction of ankle and knee flexors and extensors during the stance and swing phases were estimated using an electromyographic-driven model.

RESULTS

Both the slowest and fastest walking speeds had significantly higher metabolic costs than intermediate ones (p < 0.05). The metabolic cost of walking was correlated with plantarflexors moment during swing phase (r = 0.62 at 0.5 m.s-1, r = 0.67 at 1,25 m.s-1), dorsiflexors moment during stance phase (r = 0.65 at 1.25 m.s-1, r = 0.67 at 1.5 and 1.75 m.s-1), and ankle co-contraction during the stance phase (r = 0.63 at 1.25 and 1.75 m.s-1). The metabolic cost of running at 3.25 m.s-1 during the swing phase was correlated with the dorsiflexors moment (r = 0.63), plantarflexors moment (r = 0.61) and ankle co-contraction (r = 0.60).

DISCUSSION AND CONCLUSION

Fluctuations in metabolic cost of walking and running could be explained, at least in part, by increased ankle antagonist moments and co-contraction.

Perturbation-based estimation of within-stride cycle metabolic cost

Metabolic cost greatly impacts trade-offs within a variety of human movements. Standard respiratory measurements only obtain the mean cost of a movement cycle, preventing understanding of the contributions of different phases in, for example, walking. We present a method that estimates the within-stride cost of walking by leveraging measurements under different force perturbations. The method reproduces time series with greater consistency (r = 0.55 and 0.80 in two datasets) than previous model-based estimations (r = 0.29). This perturbation-based method reveals how the cost of push-off (10%) is much smaller than would be expected from positive mechanical work (~ 70%). This work elucidates the costliest phases during walking, offering new targets for assistive devices and rehabilitation strategies.

Marker Data Enhancement For Markerless Motion Capture

Objective

Human pose estimation models can measure movement from videos at a large scale and low cost; however, open-source pose estimation models typically detect only sparse keypoints, which leads to inaccurate joint kinematics. OpenCap, a freely available service for researchers to measure movement from videos, addresses this issue using a deep learning model-the marker enhancer-that transforms sparse keypoints into dense anatomical markers. However, OpenCap performs poorly on movements not included in the training data. Here, we create a much larger and more diverse training dataset and develop a more accurate and generalizable marker enhancer.

Methods

We compiled marker-based motion capture data from 1176 subjects and synthesized 1433 hours of keypoints and anatomical markers to train the marker enhancer. We evaluated its accuracy in computing kinematics using both benchmark movement videos and synthetic data representing unseen, diverse movements.

Results

The marker enhancer improved kinematic accuracy on benchmark movements (mean error: 4.1°, max: 8.7°) compared to using video keypoints (mean: 9.6°, max: 43.1°) and OpenCap's original enhancer (mean: 5.3°, max: 11.5°). It also better generalized to unseen, diverse movements (mean: 4.1°, max: 6.7°) than OpenCap's original enhancer (mean: 40.4°, max: 252.0°).

Conclusion

Our marker enhancer demonstrates both accuracy and generalizability across diverse movements.

Significance

We integrated the marker enhancer into OpenCap, thereby offering its thousands of users more accurate measurements across a broader range of movements.

Modelling based Approach towards Evaluation and Selection of Ankle Foot Orthosis for Crouch Gait

Ankle foot orthosis (AFO) are commonly prescribed for correcting crouch gait in children with cerebral palsy (CP). There are multiple AFO variants and selecting an optimal AFO for CP subject is often challenging. In this work, we have analyzed the effect of two passive AFO, naming Ground reaction AFO (GRAFO) and Leaf spring AFO (LSAFO) on varying severity of crouch gait in a musculoskeletal simulation environment. Impact of GRAFO and LSAFO on muscle loading and recruitment, in terms of muscle 'impulse', 'yank' and 'coactivation' was investigated for children with crouch gait along with normal gait of typically developing children. Simulation results show increased activation of the dorsiflexor muscles, indicating better ankle control and a net reduction in ankle coactivation for both types of AFO. Compared to GRAFO, LSAFO produced lower energetic cost of walking. The simulation pipeline enables to quantify the effect of different AFO configuration on muscle behavior and recruitment, and can be used to design personalized AFO as per subject-specific gait dynamics.

Ground Reaction Forces and Joint Moments Predict Metabolic Cost in Physical Performance: Harnessing the Power of Artificial Neural Networks

Understanding metabolic cost through biomechanical data, including ground reaction forces (GRFs) and joint moments, is vital for health, sports, and rehabilitation. The long stabilization time (2-5 min) of indirect calorimetry poses challenges in prolonged tests. This study investigated using artificial neural networks (ANNs) to predict metabolic costs from the GRF and joint moment time series. Data from 20 participants collected over 270 walking trials, including the GRF and joint moments, formed a detailed dataset. Two ANN models were crafted, netGRF for the GRF and netMoment for joint moments, and both underwent training, validation, and testing to validate their predictive accuracy for metabolic cost. NetGRF (six hidden layers, two input delays) showed significant correlations: 0.963 (training), 0.927 (validation), 0.883 (testing), p < 0.001. NetMoment (three hidden layers, one input delay) had correlations of 0.920 (training), 0.956 (validation), 0.874 (testing), p < 0.001. The models' low mean squared errors reflect their precision. Using Partial Dependence Plots, we demonstrated how gait cycle phases affect metabolic cost predictions, pinpointing key phases. Our findings show that the GRF and joint moments data can accurately predict metabolic costs via ANN models, with netGRF being notably consistent. This emphasizes ANNs' role in biomechanics as a crucial method for estimating metabolic costs, impacting sports science, rehabilitation, assistive technology development, and fostering personalized advancements.

Contributing Components of Metabolic Energy Models to Metabolic Cost Estimations in Gait

OBJECTIVE

As metabolic cost is a primary factor influencing humans' gait, we want to deepen our understanding of metabolic energy expenditure models. Therefore, this paper identifies the parameters and input variables, such as muscle or joint states, that contribute to accurate metabolic cost estimations.

METHODS

We explored the parameters of four metabolic energy expenditure models in a Monte Carlo sensitivity analysis. Then, we analysed the model parameters by their calculated sensitivity indices, physiological context, and the resulting metabolic rates during the gait cycle. The parameter combination with the highest accuracy in the Monte Carlo simulations represented a quasi-optimized model. In the second step, we investigated the importance of input parameters and variables by analysing the accuracy of neural networks trained with different input features.

RESULTS

Power-related parameters were most influential in the sensitivity analysis and the neural network-based feature selection. We observed that the quasi-optimized models produced negative metabolic rates, contradicting muscle physiology. Neural network-based models showed promising abilities but have been unable to match the accuracy of traditional metabolic energy expenditure models.

CONCLUSION

We showed that power-related metabolic energy expenditure model parameters and inputs are most influential during gait. Furthermore, our results suggest that neural network-based metabolic energy expenditure models are viable. However, bigger datasets are required to achieve better accuracy.

SIGNIFICANCE

As there is a need for more accurate metabolic energy expenditure models, we explored which musculoskeletal parameters are essential when developing a model to estimate metabolic energy.

Muscle-inspired bi-planar cable routing: a novel framework for designing cable driven lower limb rehabilitation exoskeletons (C-LREX)

There is a growing interest in the research and development of Cable Driven Rehabilitation Devices (CDRDs) due to multiple inherent features attractive to clinical applications, including low inertia, lightweight, high payload-to-weight ratio, large workspace, and modular design. However, previous CDRDs have mainly focused on modifying motor impairment in the sagittal plane, despite the fact that neurological disorders, such as stroke, often involve postural control and gait impairment in multiple planes. To address this gap, this work introduces a novel framework for designing a cable-driven lower limb rehabilitation exoskeleton which can assist with bi-planar impaired posture and gait. The framework used a lower limb model to analyze different cable routings inspired by human muscle architecture and attachment schemes to identify optimal routing and associated parameters. The selected cable routings were safeguarded for non-interference with the human body while generating bi-directional joint moments. The subsequent optimal cable routing model was then implemented in simulations of tracking reference healthy trajectory with bi-planar impaired gait (both in the sagittal and frontal planes). The results showed that controlling joints independently via cables yielded better performance compared to dependent control. Routing long cables through intermediate hinges reduced the peak tensions in the cables, however, at a cost of induced additional joint forces. Overall, this study provides a systematic and quantitative in silico approach, featured with accessible graphical user interface (GUI), for designing subject-specific, safe, and effective lower limb cable-driven exoskeletons for rehabilitation with options for multi-planar personalized impairment-specific features.

NSF DARE-transforming modeling in neurorehabilitation: a patient-in-the-loop framework

In 2023, the National Science Foundation (NSF) and the National Institute of Health (NIH) brought together engineers, scientists, and clinicians by sponsoring a conference on computational modelling in neurorehabiilitation. To facilitate multidisciplinary collaborations and improve patient care, in this perspective piece we identify where and how computational modelling can support neurorehabilitation. To address the where, we developed a patient-in-the-loop framework that uses multiple and/or continual measurements to update diagnostic and treatment model parameters, treatment type, and treatment prescription, with the goal of maximizing clinically-relevant functional outcomes. This patient-in-the-loop framework has several key features: (i) it includes diagnostic and treatment models, (ii) it is clinically-grounded with the International Classification of Functioning, Disability and Health (ICF) and patient involvement, (iii) it uses multiple or continual data measurements over time, and (iv) it is applicable to a range of neurological and neurodevelopmental conditions. To address the how, we identify state-of-the-art and highlight promising avenues of future research across the realms of sensorimotor adaptation, neuroplasticity, musculoskeletal, and sensory & pain computational modelling. We also discuss both the importance of and how to perform model validation, as well as challenges to overcome when implementing computational models within a clinical setting. The patient-in-the-loop approach offers a unifying framework to guide multidisciplinary collaboration between computational and clinical stakeholders in the field of neurorehabilitation.

Validation of Metabolic Models for Estimation of Energy Expenditure During Isolated Concentric and Eccentric Muscle Contractions

Musculoskeletal modeling uses metabolic models to estimate energy expenditure of human locomotion. However, accurate estimation of energy expenditure is challenging, which may be due to uncertainty about the true energy cost of eccentric and concentric muscle contractions. The purpose of this study was to validate three commonly used metabolic models, using isolated isokinetic concentric and eccentric knee extensions/flexions. Five resistance-trained adult males (25.6 ± 2.4 year, 90.6 ± 7.5 kg, 1.81 ± 0.09 m) performed 150 repetitions at four different torques in a dynamometer. Indirect calorimetry was used to measure energy expenditure during these muscle contractions. All three models underestimated the energy expenditure (compared with indirect calorimetry) for up to 55.8% and 78.5% for concentric and eccentric contractions, respectively. Further, the coefficient of determination was in general low for eccentric contractions (R2 < 0.46) indicating increases in the absolute error with increases in load. These results show that the metabolic models perform better when predicting energy expenditure of concentric contractions compared with eccentric contractions. Thus, more knowledge about the relationship between energy expenditure and eccentric work is needed to optimize the metabolic models for musculoskeletal modeling of human locomotion.

Validation of a musculoskeletal model for simulating muscle mechanics and energetics during diverse human hopping tasks

Computational musculoskeletal modelling has emerged as an alternative, less-constrained technique to indirect calorimetry for estimating energy expenditure. However, predictions from modelling tools depend on many assumptions around muscle architecture and function and motor control. Therefore, these tools need to continue to be validated if we are to eventually develop subject-specific simulations that can accurately and reliably model rates of energy consumption for any given task. In this study, we used OpenSim software and experimental motion capture data to simulate muscle activations, muscle fascicle dynamics and whole-body metabolic power across mechanically and energetically disparate hopping tasks, and then evaluated these outputs at a group- and individual-level against experimental electromyography, ultrasound and indirect calorimetry data. Comparing simulated and experimental outcomes, we found weak to strong correlations for peak muscle activations, moderate to strong correlations for absolute fascicle shortening and mean shortening velocity, and strong correlations for gross metabolic power. These correlations tended to be stronger on a group-level rather than individual-level. We encourage the community to use our publicly available dataset from SimTK.org to experiment with different musculoskeletal models, muscle models, metabolic cost models, optimal control policies, modelling tools and algorithms, data filtering etc. with subject-specific simulations being a focal goal.

Simulating Muscle-Level Energetic Cost Savings When Humans Run with a Passive Assistive Device

Connecting the legs with a spring attached to the shoelaces, called an exotendon, can reduce the energetic cost of running, but how the exotendon reduces the energetic burden of individual muscles remains unknown. We generated muscle-driven simulations of seven individuals running with and without the exotendon to discern whether savings occurred during the stance phase or the swing phase, and to identify which muscles contributed to energy savings. We computed differences in muscle-level energy consumption, muscle activations, and changes in muscle-fiber velocity and force between running with and without the exotendon. The seven of nine participants who reduced energy cost when running with the exotendon reduced their measured energy expenditure rate by 0.9 W/kg (8.3%). Simulations predicted a 1.4 W/kg (12.0%) reduction in the average rate of energy expenditure and correctly identified that the exotendon reduced rates of energy expenditure for all seven individuals. Simulations showed most of the savings occurred during stance (1.5 W/kg), though the rate of energy expenditure was also reduced during swing (0.3 W/kg). The energetic savings were distributed across the quadriceps, hip flexor, hip abductor, hamstring, hip adductor, and hip extensor muscle groups, whereas no changes were observed in the plantarflexor or dorsiflexor muscles. Energetic savings were facilitated by reductions in the rate of mechanical work performed by muscles and their estimated rate of heat production. By modeling muscle-level energetics, this simulation framework accurately captured measured changes in whole-body energetics when using an assistive device. This is a useful first step towards using simulation to accelerate device design by predicting how humans will interact with assistive devices that have yet to be built.

Predictive Simulations with OpenSim Moco to Investigate the Interaction Between Human and Assistive Exoskeleton

The purpose of this work was to investigate the interaction between human and lower limbs assistive exoskeleton under different levels of assistance, by using computational simulations. To this, a human-exoskeleton interaction model was used and three predictive simulations were carried out with the OpenSim Moco. The results proved that the increase in the level of robot assistance causes a reduction in human effort. In addition, it was possible to verify the RMS torque of both the robot and the human, as well as the muscle activations, for the different levels of assistance simulated. For future work, we intend to run predictive simulations with more complex movements, such as gait free and with obstacles, in addition to using models that can represent a human being with muscle weakness on one side of the body (hemiparesis).

A forefoot strike pattern during 18° uphill walking leads to greater ankle joint and plantar flexor loading

BACKGROUND

The ankle joint is one of the most involved joints in uphill walking. Furthermore, it is well known that toe walking increases the external dorsiflexion moment in the first half of stance during level walking. However, the effects of different foot-strike patterns on plantar flexor muscle forces, ankle joint forces, and other lower limb joint and muscle forces are unknown.

RESEARCH QUESTION

Do foot-strike patterns during 18° uphill walking affect lower limb sagittal joint angles and moments, as well as joint contact and muscle forces?

METHODS

This study was based on a data subset from previous publications, analysing uphill walking on an 18° ramp at a preset speed of 1.1 m/s in 18 male participants (34 limbs analyzed, 27 ± 5 years). Participants were divided into two groups based on their foot-strike pattern at initial contact: heel (HC) and forefoot (FC). Lower limb sagittal joint angles and moments as well as joint contact and muscle forces were assessed. Differences between the groups were assessed using two-sample t-tests.

RESULTS

FC showed increased soleus and gastrocnemius muscle forces as well as ankle joint forces during loading response and mid stance compared to HC. The soleus muscle force impulse was 51.1% higher in the FC group than in the HC group (p < 0.001). On the other hand, FC had a lower absolute centre of mass vertical displacement and reduced knee and hip joint, as well as iliopsoas and hamstring muscle force impulses.

SIGNIFICANCE

In terms of plantar flexor and ankle joint loading, it is advantageous to exhibit a heel strike pattern. The current results can be used to recommend foot-strike patterns for uphill walking, particularly in the presence or prevention of musculoskeletal issues.

How Connecting the Legs with a Spring Improves Human Running Economy

Connecting the legs with a spring attached to the shoelaces reduces the energy cost of running, but how the spring reduces the energy burden of individual muscles remains unknown. We generated muscle-driven simulations of seven individuals running with and without the spring to discern whether savings occurred during the stance phase or the swing phase, and to identify which muscles contributed to energy savings. We computed differences in muscle-level energy consumption, muscle activations, and changes in muscle-fiber velocity and force between running with and without the spring. Across participants, running with the spring reduced the measured rate of energy expenditure by 0.9 W/kg (8.3%). Simulations predicted a 1.4 W/kg (12.0%) reduction in the average rate of energy expenditure and correctly identified that the spring reduced rates of energy expenditure for all participants. Simulations showed most of the savings occurred during stance (1.5 W/kg), though the rate of energy expenditure was also reduced during swing (0.3 W/kg). The energetic savings were distributed across the quadriceps, hip flexor, hip abductor, hamstring, hip adductor, and hip extensor muscle groups, whereas no changes in the rate of energy expenditure were observed in the plantarflexor or dorsiflexor muscles. Energetic savings were facilitated by reductions in the rate of mechanical work performed by muscles and their estimated rate of heat production. The simulations provide insight into muscle-level changes that occur when utilizing an assistive device and the mechanisms by which a spring connecting the legs improves running economy.

Acquisition of bipedal locomotion in a neuromusculoskeletal model with unilateral transtibial amputation

Adaptive locomotion is an essential behavior for animals to survive. The central pattern generator in the spinal cord is responsible for the basic rhythm of locomotion through sensory feedback coordination, resulting in energy-efficient locomotor patterns. Individuals with symmetrical body proportions exhibit an energy-efficient symmetrical gait on flat ground. In contrast, individuals with lower limb amputation, who have morphologically asymmetrical body proportions, exhibit asymmetrical gait patterns. However, it remains unclear how the nervous system adjusts the control of the lower limbs. Thus, in this study, we investigated how individuals with unilateral transtibial amputation control their left and right lower limbs during locomotion using a two-dimensional neuromusculoskeletal model. The model included a musculoskeletal model with 7 segments and 18 muscles, as well as a neural model with a central pattern generator and sensory feedback systems. Specifically, we examined whether individuals with unilateral transtibial amputation acquire prosthetic gait through a symmetric or asymmetric feedback control for the left and right lower limbs. After acquiring locomotion, the metabolic costs of transport and the symmetry of the spatiotemporal gait factors were evaluated. Regarding the metabolic costs of transportation, the symmetric control model showed values approximately twice those of the asymmetric control model, whereas both scenarios showed asymmetry of spatiotemporal gait patterns. Our results suggest that individuals with unilateral transtibial amputation can reacquire locomotion by modifying sensory feedback parameters. In particular, the model reacquired reasonable locomotion for activities of daily living by re-searching asymmetric feedback parameters for each lower limb. These results could provide insight into effective gait assessment and rehabilitation methods to reacquire locomotion in individuals with unilateral transtibial amputation.

Generating Human Arm Kinematics Using Reinforcement Learning to Train Active Muscle Behavior in Automotive Research

Computational human body models (HBMs) are important tools for predicting human biomechanical responses under automotive crash environments. In many scenarios, the prediction of the occupant response will be improved by incorporating active muscle control into the HBMs to generate biofidelic kinematics during different vehicle maneuvers. In this study, we have proposed an approach to develop an active muscle controller based on reinforcement learning (RL). The RL muscle activation control (RL-MAC) approach is a shift from using traditional closed-loop feedback controllers, which can mimic accurate active muscle behavior under a limited range of loading conditions for which the controller has been tuned. Conversely, the RL-MAC uses an iterative training approach to generate active muscle forces for desired joint motion and is analogous to how a child develops gross motor skills. In this study, the ability of a deep deterministic policy gradient (DDPG) RL controller to generate accurate human kinematics is demonstrated using a multibody model of the human arm. The arm model was trained to perform goal-directed elbow rotation by activating the responsible muscles and investigated using two recruitment schemes: as independent muscles or as antagonistic muscle groups. Simulations with the trained controller show that the arm can move to the target position in the presence or absence of externally applied loads. The RL-MAC trained under constant external loads was able to maintain the desired elbow joint angle under a simplified automotive impact scenario, implying the robustness of the motor control approach.

Effects of powered versus passive-elastic ankle foot prostheses on leg muscle activity during level, uphill and downhill walking

People with transtibial amputation (TTA) using passive-elastic prostheses have greater leg muscle activity and metabolic cost during level-ground and sloped walking than non-amputees. Use of a stance-phase powered (BiOM) versus passive-elastic prosthesis reduces metabolic cost for people with TTA during level-ground, +3° and +6° walking. Metabolic cost is associated with muscle activity, which may provide insight into differences between prostheses. We measured affected leg (AL) and unaffected leg (UL) muscle activity from ten people with TTA (6 males, 4 females) walking at 1.25 m s^-1 on a dual-belt force-measuring treadmill at 0°, ±3°, ±6° and ±9° using their own passive-elastic and the BiOM prosthesis. We compared stride average integrated EMG (iEMG), peak EMG and muscle activity burst duration. Use of the BiOM increased UL lateral gastrocnemius iEMG on downhill slopes and AL biceps femoris on +6° and +9° slopes, and decreased UL rectus femoris on uphill slopes, UL vastus lateralis on +6° and +9°, and soleus and tibialis anterior on a +9° slope compared to a passive-elastic prosthesis. Differences in leg muscle activity for people with TTA using a passive-elastic versus stance-phase powered prosthesis do not clearly explain differences in metabolic cost during walking on level ground and slopes.

Model-Based Biomechanical Exoskeleton Concept Optimization for a Representative Lifting Task in Logistics

Occupational exoskeletons are a promising solution to prevent work-related musculoskeletal disorders (WMSDs). However, there are no established systems that support heavy lifting to shoulder height. Thus, this work presents a model-based analysis of heavy lifting activities and subsequent exoskeleton concept optimization. Six motion sequences were captured in the laboratory for three subjects and analyzed in multibody simulations with respect to muscle activities (MAs) and joint forces (JFs). The most strenuous sequence was selected and utilized in further simulations of a human model connected to 32 exoskeleton concept variants. Six simulated concepts were compared concerning occurring JFs and MAs as well as interaction loads in the exoskeleton arm interfaces. Symmetric uplifting of a 21 kg box from hip to shoulder height was identified as the most strenuous motion sequence with highly loaded arms, shoulders, and back. Six concept variants reduced mean JFs (spine: >70%, glenohumeral joint: >69%) and MAs (back: >63%, shoulder: >59% in five concepts). Parasitic loads in the arm bracing varied strongly among variants. An exoskeleton design was identified that effectively supports heavy lifting, combining high musculoskeletal relief and low parasitic loads. The applied workflow can help developers in the optimization of exoskeletons.

Biomechanical knee energy harvester: Design optimization and testing

Biomechanical energy harvesters are designed to generate electrical energy from human locomotion (e.g., walking) with minimal or no additional effort by the users. These harvesters aim to carry out the work of the muscles during phases in locomotion where the muscles are acting as brakes. Currently, many harvesters focus on the knee joint during late swing, which is only one of three phases available during the gait cycle. For the device to be successful, there is a need to consider design components such as the motor/generator and the gear ratio. These components influence the amount of electrical energy that could be harvested, metabolic power during harvesting, and more. These various components make it challenging to achieve the optimal design. This paper presents a design of a knee harvester with a direct drive that enables harvesting both in flexion and extension using optimization. Subsequently, two knee devices were built and tested using five different harvesting levels. Results show that the 30% level was the best, harvesting approximately 5 W of electricity and redacting 8 W of metabolic energy compared to walking with the device as a dead weight. Evaluation of the models used in the optimization showed a good match to the system model but less for the metabolic power model. These results could pave the way for an energy harvester that could utilize more of the negative joint power during the gait cycle while reducing metabolic effort.

Patterns of asymmetry and energy cost generated from predictive simulations of hemiparetic gait

Hemiparesis, defined as unilateral muscle weakness, often occurs in people post-stroke or people with cerebral palsy, however it is difficult to understand how this hemiparesis affects movement patterns as it often presents alongside a variety of other neuromuscular impairments. Predictive musculoskeletal modeling presents an opportunity to investigate how impairments affect gait performance assuming a particular cost function. Here, we use predictive simulation to quantify the spatiotemporal asymmetries and changes to metabolic cost that emerge when muscle strength is unilaterally reduced and how reducing spatiotemporal symmetry affects metabolic cost. We modified a 2-D musculoskeletal model by uniformly reducing the peak isometric muscle force unilaterally. We then solved optimal control simulations of walking across a range of speeds by minimizing the sum of the cubed muscle excitations. Lastly, we ran additional optimizations to test if reducing spatiotemporal asymmetry would result in an increase in metabolic cost. Our results showed that the magnitude and direction of effort-optimal spatiotemporal asymmetries depends on both the gait speed and level of weakness. Also, the optimal speed was 1.25 m/s for the symmetrical and 20% weakness models but slower (1.00 m/s) for the 40% and 60% weakness models, suggesting that hemiparesis can account for a portion of the slower gait speed seen in people with hemiparesis. Modifying the cost function to minimize spatiotemporal asymmetry resulted in small increases (~4%) in metabolic cost. Overall, our results indicate that spatiotemporal asymmetry may be optimal for people with hemiparesis. Additionally, the effect of speed and the level of weakness on spatiotemporal asymmetry may help explain the well-known heterogenous distribution of spatiotemporal asymmetries observed in the clinic. Future work could extend our results by testing the effects of other neuromuscular impairments on optimal gait strategies, and therefore build a more comprehensive understanding of the gait patterns observed in clinical populations.

Biomechanical Analysis of Stoop and Free-Style Squat Lifting and Lowering with a Generic Back-Support Exoskeleton Model

Musculoskeletal disorders (MSDs) induced by industrial manual handling tasks are a major issue for workers and companies. As flexible ergonomic solutions, occupational exoskeletons can decrease critically high body stress in situations of awkward postures and motions. Biomechanical models with detailed anthropometrics and motions help us to acquire a comprehension of person- and application-specifics by considering the intended and unintended effects, which is crucial for effective implementation. In the present model-based analysis, a generic back-support exoskeleton model was introduced and applied to the motion data of one male subject performing symmetric and asymmetric dynamic manual handling tasks. Different support modes were implemented with this model, including support profiles typical of passive and active systems and an unconstrained optimal support mode used for reference to compare and quantify their biomechanical effects. The conducted simulations indicate that there is a high potential to decrease the peak compression forces in L4/L5 during the investigated heavy loaded tasks for all motion sequences and exoskeleton support modes (mean reduction of 16.0% without the optimal support mode). In particular, asymmetric motions (mean reduction of 11.9%) can be relieved more than symmetric ones (mean reduction of 8.9%) by the exoskeleton support modes without the optimal assistance. The analysis of metabolic energy consumption indicates a high dependency on lifting techniques for the effectiveness of the exoskeleton support. While the exoskeleton support substantially reduces the metabolic cost for the free-squat motions, a slightly higher energy consumption was found for the symmetric stoop motion technique with the active and optimal support mode.

Predictive Simulations to Replicate Human Gait Adaptations and Energetics With Exoskeletons

Robotic exoskeletons have the potential to restore and enhance human mobility. However, optimally controlling these devices, to work in concert with human users, is challenging. Accurate model simulations of the interaction between exoskeletons and users may expedite the design process and improve control. Here, as a proof of principle, we tested if we could use predictive simulations to replicate human gait adaptations and changes in energy expenditure from an experiment where participants walked with exoskeletons. We recreated a past experimental paradigm, where robotic exoskeletons were used to shift people's energetically optimal step frequency to frequencies higher and lower than normally preferred. To match the experimental controller, we modelled knee-worn exoskeletons that applied resistive torques, either proportional or inversely proportional to step frequency-decreasing or increasing the energy optimal step frequency, respectively. We were able to replicate the experiment, finding higher and lower optimal step frequencies than in natural walking under each respective condition. Our simulated resistive torques and objective landscapes resembled the measured experimental resistive torque and energy landscapes. Individual muscle energetics revealed distinct coordination strategies consistent with each exoskeleton controller condition. Increasing the accuracy of step frequency and energetic predictions was best achieved by increasing the number of virtual participants (varying whole-body anthropometrics), rather than the number of muscle parameter sets (varying muscle anthropometrics). In future, our approach can be used to design controllers in advance of human testing, to help identify reasonable solution spaces or tailor design to individual users.

The Added Value of Musculoskeletal Simulation for the Study of Physical Performance in Military Tasks

The performance of military tasks is often exacerbated by additional load carriage, leading to increased physical demand. Previous studies showed that load carriage may lead to increased risk of developing musculoskeletal injuries, a reduction in task speed and mobility, and overall performance degradation. However, these studies were limited to a non-ambulatory setting, and the underlying causes of performance degradation remain unclear. To obtain insights into the underlying mechanisms of reduced physical performance during load-carrying military activities, this study proposes a combination of IMUs and musculoskeletal modeling. Motion data of military subjects was captured using an Xsens suit during the performance of an agility run under three different load-carrying conditions (no load, 16 kg, and 31 kg). The physical performance of one subject was assessed by means of inertial motion-capture driven musculoskeletal analysis. Our results showed that increased load carriage led to an increase in metabolic power and energy, changes in muscle parameters, a significant increase in completion time and heart rate, and changes in kinematic parameters. Despite the exploratory nature of this study, the proposed approach seems promising to obtain insight into the underlying mechanisms that result in performance degradation during load-carrying military activities.

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

Muscle forces and the demands of human walking

Reconstructing the locomotor behavior of extinct animals depends on elucidating the principles that link behavior, function, and morphology, which can only be done using extant animals. Within the human lineage, the evolution of bipedalism represents a critical transition, and evaluating fossil hominins depends on understanding the relationship between lower limb forces and skeletal morphology in living humans. As a step toward that goal, here we use a musculoskeletal model to estimate forces in the lower limb muscles of ten individuals during walking. The purpose is to quantify the consistency, timing, and magnitude of these muscle forces during the stance phase of walking. We find that muscles which act to support or propel the body during walking demonstrate the greatest force magnitudes as well as the highest consistency in the shape of force curves among individuals. Muscles that generate moments in the same direction as, or orthogonal to, the ground reaction force show lower forces of greater variability. These data can be used to define the envelope of load cases that need to be examined in order to understand human lower limb skeletal load bearing.

Sensing leg movement enhances wearable monitoring of energy expenditure

Physical inactivity is the fourth leading cause of global mortality. Health organizations have requested a tool to objectively measure physical activity. Respirometry and doubly labeled water accurately estimate energy expenditure, but are infeasible for everyday use. Smartwatches are portable, but have significant errors. Existing wearable methods poorly estimate time-varying activity, which comprises 40% of daily steps. Here, we present a Wearable System that estimates metabolic energy expenditure in real-time during common steady-state and time-varying activities with substantially lower error than state-of-the-art methods. We perform experiments to select sensors, collect training data, and validate the Wearable System with new subjects and new conditions for walking, running, stair climbing, and biking. The Wearable System uses inertial measurement units worn on the shank and thigh as they distinguish lower-limb activity better than wrist or trunk kinematics and converge more quickly than physiological signals. When evaluated with a diverse group of new subjects, the Wearable System has a cumulative error of 13% across common activities, significantly less than 42% for a smartwatch and 44% for an activity-specific smartwatch. This approach enables accurate physical activity monitoring which could enable new energy balance systems for weight management or large-scale activity monitoring.

Identification of Postural Controllers in Human Standing Balance

Standing balance is a simple motion task for healthy humans but the actions of the central nervous system (CNS) have not been described by generalized and sufficiently sophisticated control laws. While system identification approaches have been used to extracted models of the CNS, they either focus on short balance motions, leading to task-specific control laws, or assume that the standing balance system is linear. To obtain comprehensive control laws for human standing balance, complex balance motions, long duration tests, and nonlinear controller models are all needed. In this paper, we demonstrate that trajectory optimization with the direct collocation method can achieve these goals to identify complex CNS models for the human standing balance task. We first examined this identification method using synthetic motion data and showed that correct control parameters can be extracted. Then, six types of controllers, from simple linear to complex nonlinear, were identified from 100 s of motion data from randomly perturbed standing. Results showed that multiple time-delay paths and nonlinear properties are both needed in order to fully explain human feedback control of standing balance.

Muscle metabolic energy costs while modifying propulsive force generation during walking

We pose that an age-related increase in the metabolic cost of walking arises in part from a redistribution of joint power where muscles spanning the hip compensate for insufficient ankle push-off and smaller peak propulsive forces (F_P). Young adults elicit a similar redistribution when walking with smaller F_P via biofeedback. We used targeted F_P biofeedback and musculoskeletal models to estimate the metabolic costs of operating lower limb muscles in young adults walking across a range of F_P. Our simulations support the theory of distal-to-proximal redistribution of joint power as a determinant of increased metabolic cost in older adults during walking.

OpenSim Moco: Musculoskeletal optimal control

Musculoskeletal simulations are used in many different applications, ranging from the design of wearable robots that interact with humans to the analysis of patients with impaired movement. Here, we introduce OpenSim Moco, a software toolkit for optimizing the motion and control of musculoskeletal models built in the OpenSim modeling and simulation package. OpenSim Moco uses the direct collocation method, which is often faster and can handle more diverse problems than other methods for musculoskeletal simulation. Moco frees researchers from implementing direct collocation themselves-which typically requires extensive technical expertise-and allows them to focus on their scientific questions. The software can handle a wide range of problems that interest biomechanists, including motion tracking, motion prediction, parameter optimization, model fitting, electromyography-driven simulation, and device design. Moco is the first musculoskeletal direct collocation tool to handle kinematic constraints, which enable modeling of kinematic loops (e.g., cycling models) and complex anatomy (e.g., patellar motion). To show the abilities of Moco, we first solved for muscle activity that produced an observed walking motion while minimizing squared muscle excitations and knee joint loading. Next, we predicted how muscle weakness may cause deviations from a normal walking motion. Lastly, we predicted a squat-to-stand motion and optimized the stiffness of an assistive device placed at the knee. We designed Moco to be easy to use, customizable, and extensible, thereby accelerating the use of simulations to understand the movement of humans and other animals.

Musculoskeletal Model Personalization Affects Metabolic Cost Estimates for Walking

Assessment of metabolic cost as a metric for human performance has expanded across various fields within the scientific, clinical, and engineering communities. As an alternative to measuring metabolic cost experimentally, musculoskeletal models incorporating metabolic cost models have been developed. However, to utilize these models for practical applications, the accuracy of their metabolic cost predictions requires improvement. Previous studies have reported the benefits of using personalized musculoskeletal models for various applications, yet no study has evaluated how model personalization affects metabolic cost estimation. This study investigated the effect of musculoskeletal model personalization on estimates of metabolic cost of transport (CoT) during post-stroke walking using three commonly used metabolic cost models. We analyzed walking data previously collected from two male stroke survivors with right-sided hemiparesis. The three metabolic cost models were implemented within three musculoskeletal modeling approaches involving different levels of personalization. The first approach used a scaled generic OpenSim model and found muscle activations via static optimization (SOGen). The second approach used a personalized electromyographic (EMG)-driven musculoskeletal model with personalized functional axes but found muscle activations via static optimization (SOCal). The third approach used the same personalized EMG-driven model but calculated muscle activations directly from EMG data (EMGCal). For each approach, the muscle activation estimates were used to calculate each subject's CoT at different gait speeds using three metabolic cost models (Umberger et al., 2003; Bhargava et al., 2004; Umberger, 2010). The calculated CoT values were compared with published CoT data as a function of walking speed, step length asymmetry, stance time asymmetry, double support time asymmetry, and severity of motor impairment (i.e., Fugl-Meyer score). Overall, only SOCal and EMGCal with the Bhargava metabolic cost model were able to reproduce accurately published experimental trends between CoT and various clinical measures of walking asymmetry post-stroke. Tuning of the parameters in the different metabolic cost models could potentially resolve the observed CoT magnitude differences between model predictions and experimental measurements. Realistic CoT predictions may allow researchers to predict human performance, surgical outcomes, and rehabilitation outcomes reliably using computational simulations.

Differences between joint-space and musculoskeletal estimations of metabolic rate time profiles

Motion capture laboratories can measure multiple variables at high frame rates, but we can only measure the average metabolic rate of a stride using respiratory measurements. Biomechanical simulations with equations for calculating metabolic rate can estimate the time profile of metabolic rate within the stride cycle. A variety of methods and metabolic equations have been proposed, including metabolic time profile estimations based on joint parameters. It is unclear whether differences in estimations are due to differences in experimental data or due to methodological differences. This study aimed to compare two methods for estimating the time profile of metabolic rate, within a single dataset. Knowledge about the consistency of different methods could be useful for applications such as detecting which part of the gait cycle causes increased metabolic cost in patients. Here we compare estimations of metabolic rate time profiles using a musculoskeletal and a joint-space method. The musculoskeletal method was driven by kinematics and electromyography data and used muscle metabolic rate equations, whereas the joint-space method used metabolic rate equations based on joint parameters. Both estimations of changes in stride average metabolic rate correlated significantly with large changes in indirect calorimetry from walking on different grades showing that both methods accurately track changes. However, estimations of changes in stride average metabolic rate did not correlate significantly with more subtle changes in indirect calorimetry due to walking with different shoe inclinations, and both the musculoskeletal and joint-space time profile estimations did not correlate significantly with each other except in the most downward shoe inclination. Estimations of the relative cost of stance and swing matched well with previous simulations with similar methods and estimations from experimental perturbations. Rich experimental datasets could further advance time profile estimations. This knowledge could be useful to develop therapies and assistive devices that target the least metabolically economic part of the gait cycle.

Exploring the Contribution of Proprioceptive Reflexes to Balance Control in Perturbed Standing

Humans control balance using different feedback loops involving the vestibular system, the visual system, and proprioception. In this article, we focus on proprioception and explore the contribution of reflexes based on force and length feedback to standing balance. In particular, we address the questions of how much proprioception alone could explain balance control, and whether one modality, force or length feedback, is more important than the other. A sagittal plane neuro-musculoskeletal model was developed with six degrees of freedom and nine muscles in each leg. A controller was designed using proprioceptive reflexes and a dead zone. No feedback control was applied inside the dead zone. Reflexes were active once the center of mass moved outside the dead zone. Controller parameters were found by solving an optimization problem, where effort was minimized while the neuro-musculoskeletal model should remain standing upright on a perturbed platform. The ground was perturbed with random square pulses in the sagittal plane with different amplitudes and durations. The optimization was solved for three controllers: using force and length feedback (base model), using only force feedback, and using only length feedback. Simulations were compared to human data from previous work, where an experiment with the same perturbation signal was performed. The optimized controller yielded a similar posture, since average joint angles were within 5 degrees of the experimental average joint angles. The joint angles of the base model, the length only model, and the force only model correlated weakly (ankle) to moderately with the experimental joint angles. The ankle moment correlated weakly to moderately with the experimental ankle moment, while the hip and knee moment were only weakly correlated, or not at all. The time series of the joint angles showed that the length feedback model was better able to explain the experimental joint angles than the force feedback model. Changes in time delay affected the correlation of the joint angles and joint moments. The objective of effort minimization yielded lower joint moments than in the experiment, suggesting that other objectives are also important in balance control, which cause an increase in effort and thus larger joint moments.

Energy Expenditure of Dynamic Submaximal Human Plantarflexion Movements: Model Prediction and Validation by in-vivo Magnetic Resonance Spectroscopy

To understand the organization and efficiency of biological movement, it is important to evaluate the energy requirements on the level of individual muscles. To this end, predicting energy expenditure with musculoskeletal models in forward-dynamic computer simulations is currently the most promising approach. However, it is challenging to validate muscle models in-vivo in humans, because access to the energy expenditure of single muscles is difficult. Previous approaches focused on whole body energy expenditure, e.g., oxygen consumption (VO2), or on thermal measurements of individual muscles by tracking blood flow and heat release (through measurements of the skin temperature). This study proposes to validate models of muscular energy expenditure by using functional phosphorus magnetic resonance spectroscopy (³¹P-MRS). ³¹P-MRS allows to measure phosphocreatine (PCr) concentration which changes in relation to energy expenditure. In the first 25 s of an exercise, PCr breakdown rate reflects ATP hydrolysis, and is therefore a direct measure of muscular enthalpy rate. This method was applied to the gastrocnemius medialis muscle of one healthy subject during repetitive dynamic plantarflexion movements at submaximal contraction, i.e., 20% of the maximum plantarflexion force using a MR compatible ergometer. Furthermore, muscle activity was measured by surface electromyography (EMG). A model (provided as open source) that combines previous models for muscle contraction dynamics and energy expenditure was used to reproduce the experiment in simulation. All parameters (e.g., muscle length and volume, pennation angle) in the model were determined from magnetic resonance imaging or literature (e.g., fiber composition), leaving no free parameters to fit the experimental data. Model prediction and experimental data on the energy supply rates are in good agreement with the validation phase (<25 s) of the dynamic movements. After 25 s, the experimental data differs from the model prediction as the change in PCr does not reflect all metabolic contributions to the energy expenditure anymore and therefore underestimates the energy consumption. This shows that this new approach allows to validate models of muscular energy expenditure in dynamic movements in vivo.

Characterizing Motor Control of Mastication With Soft Actor-Critic

The human masticatory system is a complex functional unit characterized by a multitude of skeletal components, muscles, soft tissues, and teeth. Muscle activation dynamics cannot be directly measured on live human subjects due to ethical, safety, and accessibility limitations. Therefore, estimation of muscle activations and their resultant forces is a longstanding and active area of research. Reinforcement learning (RL) is an adaptive learning strategy which is inspired by the behavioral psychology and enables an agent to learn the dynamics of an unknown system via policy-driven explorations. The RL framework is a well-formulated closed-loop system where high capacity neural networks are trained with the feedback mechanism of rewards to learn relatively complex actuation patterns. In this work, we are building on a deep RL algorithm, known as the Soft Actor-Critic, to learn the inverse dynamics of a simulated masticatory system, i.e., learn the activation patterns that drive the jaw to its desired location. The outcome of the proposed training procedure is a parametric neural model which acts as the brain of the biomechanical system. We demonstrate the model's ability to navigate the feasible three-dimensional (3D) envelope of motion with sub-millimeter accuracies. We also introduce a performance analysis platform consisting of a set of quantitative metrics to assess the functionalities of a given simulated masticatory system. This platform assesses the range of motion, metabolic efficiency, the agility of motion, the symmetry of activations, and the accuracy of reaching the desired target positions. We demonstrate how the model learns more metabolically efficient policies by integrating a force regularization term in the RL reward. We also demonstrate the inverse correlation between the metabolic efficiency of the models and their agility and range of motion. The presented masticatory model and the proposed RL training mechanism are valuable tools for the analysis of mastication and other biomechanical systems. We see this framework's potential in facilitating the functional analyses aspects of surgical treatment planning and predicting the rehabilitation performance in post-operative subjects.

Habitual foot strike pattern does not affect simulated triceps surae muscle metabolic energy consumption during running

Foot strike pattern affects ankle joint work and triceps surae muscle-tendon dynamics during running. Whether these changes in muscle-tendon dynamics also affect triceps surae muscle energy consumption is still unknown. In addition, as the triceps surae muscle accounts for a substantial amount of the whole-body metabolic energy consumption, changes in triceps surae energy consumption may affect whole-body metabolic energy consumption. However, direct measurements of muscle metabolic energy consumption during dynamic movements is difficult. Model-based approaches can be used to estimate individual muscle and whole-body metabolic energy consumption based on Hill type muscle models. In this study, we use an integrated experimental and dynamic optimization approach to compute muscle states (muscle forces, lengths, velocities, excitations and activations) of 10 habitual midfoot/forefoot striking and nine habitual rearfoot striking runners while running at 10 and 14 km h^-1 The Achilles tendon stiffness of the musculoskeletal model was adapted to fit experimental ultrasound data of the gastrocnemius medialis muscle during ground contact. Next, we calculated triceps surae muscle and whole-body metabolic energy consumption using four different metabolic energy models provided in the literature. Neither triceps surae metabolic energy consumption (P>0.35) nor whole-body metabolic energy consumption (P>0.14) was different between foot strike patterns, regardless of the energy model used or running speed tested. Our results provide new evidence that midfoot/forefoot and rearfoot strike patterns are metabolically equivalent.

Metabolic cost calculations of gait using musculoskeletal energy models, a comparison study

This paper compares predictions of metabolic energy expenditure in gait using seven metabolic energy expenditure models to assess their correlation with experimental data. Ground reaction forces, marker data, and pulmonary gas exchange data were recorded for six walking trials at combinations of two speeds, 0.8 m/s and 1.3 m/s, and three inclines, -8% (downhill), level, and 8% (uphill). The metabolic cost, calculated with the metabolic energy models was compared to the metabolic cost from the pulmonary gas exchange rates. A repeated measures correlation showed that all models correlated well with experimental data, with correlations of at least 0.9. The model by Bhargava et al. (J Biomech, 2004: 81-88) and the model by Lichtwark and Wilson (J Exp Biol, 2005: 2831-3843) had the highest correlation, 0.95. The model by Margaria (Int Z Angew Physiol Einschl Arbeitsphysiol, 1968: 339-351) predicted the increase in metabolic cost following a change in dynamics best in absolute terms.