Review of musculoskeletal modelling in a clinical setting: Current use in rehabilitation design, surgical decision making and healthcare interventions

NSF DARE-Transforming modeling in neurorehabilitation: Four threads for catalyzing progress

We present an overview of the Conference on Transformative Opportunities for Modeling in Neurorehabilitation held in March 2023. It was supported by the Disability and Rehabilitation Engineering (DARE) program from the National Science Foundation's Engineering Biology and Health Cluster. The conference brought together experts and trainees from around the world to discuss critical questions, challenges, and opportunities at the intersection of computational modeling and neurorehabilitation to understand, optimize, and improve clinical translation of neurorehabilitation. We organized the conference around four key, relevant, and promising Focus Areas for modeling: Adaptation & Plasticity, Personalization, Human-Device Interactions, and Modeling 'In-the-Wild'. We identified four common threads across the Focus Areas that, if addressed, can catalyze progress in the short, medium, and long terms. These were: (i) the need to capture and curate appropriate and useful data necessary to develop, validate, and deploy useful computational models (ii) the need to create multi-scale models that span the personalization spectrum from individuals to populations, and from cellular to behavioral levels (iii) the need for algorithms that extract as much information from available data, while requiring as little data as possible from each client (iv) the insistence on leveraging readily available sensors and data systems to push model-driven treatments from the lab, and into the clinic, home, workplace, and community. The conference archive can be found at (dare2023.usc.edu). These topics are also extended by three perspective papers prepared by trainees and junior faculty, clinician researchers, and federal funding agency representatives who attended the conference.

Evaluation of ground reaction forces and centers of pressure predicted by AnyBody Modeling System during load reaching/handling activities and effects of the prediction errors on model-estimated spinal loads

Full-body and lower-extremity human musculoskeletal models require feet ground reaction forces (GRFs) and centers of pressure (CoPs) as inputs to predict muscle forces and joint loads. GRFs/CoPs are traditionally measured via floor-mounted forceplates that are usually restricted to research laboratories thus limiting their applicability in real occupational and clinical setups. Alternatively, GRFs/CoPs can be estimated via inverse dynamic approaches as also implemented in the Anybody Modeling System (AnyBody Technology, Aalborg, Denmark). The accuracy of Anybody in estimating GRFs/CoPs during load-handling/reaching activities and the effect of its prediction errors on model-estimated spinal loads remain to be investigated. Twelve normal- and over-weight individuals performed total of 480 static load-handling/reaching activities while measuring (by forceplates) and predicting (by AnyBody) their GRFs/CoPs. Moreover, the effects of GRF/CoP prediction errors on the estimated spinal loads were evaluated by inputting measured or predicted GRFs/CoPs into subject-specific musculoskeletal models. Regardless of the subject groups (normal-weight or overweight) and tasks (load-reaching or load-handling), results indicated great agreements between the measured and predicted GRFs (normalized root-mean-squared error, nRMSEs < 14% and R2 > 0.90) and between their model-estimated spinal loads (nRMSEs < 14% and R2 > 0.83). These agreements were good but relatively less satisfactory for CoPs (nRMSEs < 17% and 0.57 < R2 < 0.68). The only exception, requiring a more throughout investigation, was the situation when the ground-foot contact was significantly reduced during the activity. It appears that occupational/clinical investigations performed in real workstation/clinical setups with no access to forceplates may benefit from the AnyBody GRF/CoP prediction tools for a wide range of load-reaching/handling activities.

An artificial neural network for full-body posture prediction in dynamic lifting activities and effects of its prediction errors on model-estimated spinal loads

Musculoskeletal models have indispensable applications in occupational risk assessment/management and clinical treatment/rehabilitation programs. To estimate muscle forces and joint loads, these models require body posture during the activity under consideration. Posture is usually measured via video-camera motion tracking approaches that are time-consuming, costly, and/or limited to laboratories. Alternatively, posture-prediction tools based on artificial intelligence can be trained using measured postures of several subjects performing many activities. We aimed to use our previous posture-prediction artificial neural network (ANN), developed based on many measured static postures, to predict posture during dynamic lifting activities. Moreover, effects of the ANN posture-prediction errors on dynamic spinal loads were investigated using subject-specific musculoskeletal models. Seven individuals each performed twenty-five lifting tasks while their full-body three-dimensional posture was measured by a 10-camera Vicon system and also predicted by the ANN as functions of the hand-load positions during the lifting activities. The measured and predicted postures (i.e., coordinates of 39 skin markers) and their model-estimated L5-S1 loads were compared. The overall root-mean-squared-error (RMSE) and normalized (by the range of measured values) RMSE (nRMSE) between the predicted and measured postures for all markers/tasks/subjects was equal to 7.4 cm and 4.1 %, respectively (R2 = 0.98 and p < 0.05). The model-estimated L5-S1 loads based on the predicted and measured postures were generally in close agreements as also confirmed by the Bland-Altman analyses; the nRMSE for all subjects/tasks was < 10 % (R2 > 0.7 and p > 0.05). In conclusion, the easy-to-use ANN can accurately predict posture in dynamic lifting activities and its predicted posture can drive musculoskeletal models.

Biomechanical effects of typical lower limb movements of Chen-style Tai Chi on knee joint

The load and stress distribution on cartilage and meniscus of the knee joint in typical lower limb movements of Chen-style Tai Chi (TC) and deep squat (DS) were analyzed using finite element (FE) analysis. The loadings for this analysis consisted of muscle forces and ground reaction force (GRF), which were calculated through the inverse dynamic approach based on kinematics and force plate measurements obtained from motion capture experiments. Thirteen experienced practitioners performed four typical TC movements, namely, single whip (SW), brush knee and twist step (BKTS), stretch down (SD), and part the wild horse's mane (PWHM), which exhibit lower posture and greater lower limb force compared to other TC styles. The results indicated that TC required greater lower limb muscle strength than DS, resulting in greater knee joint forces. The stress on the medial cartilage in SW and BKTS fell within a range conductive to maintaining the balance between anabolism and catabolism of cartilage matrix. This was due to the fact that SW and BKTS reduce the medial to total tibiofemoral contact force ratios through knee abduction, which may effectively alleviate mild medial knee osteoarthritis (KOA). However, the greater medial contact force ratios observed in SD and PWHM resulted in great contact stresses that may aggravate the pain of patients with KOA. To mitigate these effects, practitioners should consider elevating their postures appropriately to reduce knee flexion angles, especially during the single-leg support phase. This adjustment can decrease the required muscle strength, load and stress on the knee joint.

A neuromusculoskeletal modelling approach to bilateral hip mechanics due to unexpected lateral perturbations during overground walking

BACKGROUND

Current studies on how external perturbations impact gait dynamics have primarily focused on the changes in the body's center of mass (CoM) during treadmill walking. The biomechanical responses, in particular to the multi-planar hip joint coordination, following perturbations in overground walking conditions are not completely known.

METHODS

In this study, a customized gait-perturbing device was designed to impose controlled lateral forces onto the subject's pelvis during overground walking. The biomechanical responses of bilateral hips were simulated by subject-specific neuromusculoskeletal models (NMS) driven by in-vivo motion data, which were further evaluated by statistical parameter mapping (SPM) and muscle coactivation index (CI) analysis. The validity of the subject-specific NMS was confirmed through comparison with measured surface electromyographic signals.

RESULTS

Following perturbations, the sagittal-plane hip motions were reduced for the leading leg by 18.39° and for the trailing leg by 8.23°, while motions in the frontal and transverse plane were increased, with increased hip abduction for the leading leg by 10.71° and external rotation by 9.06°, respectively. For the hip kinetics, both the bilateral hip joints showed increased abductor moments during midstance (20%-30% gait cycle) and decreased values during terminal stance (38%-48%). Muscle CI in both sagittal and frontal planes was significantly decreased for perturbed walking (p < 0.05), except for the leading leg in the sagittal plane.

CONCLUSION

The distinctive phase-dependent biomechanical response of the hip demonstrated its coordinated control strategy for balance recovery due to gait perturbations. And the changes in muscle CI suggested a potential mechanism for rapid and precise control of foot placement through modulation of joint stiffness properties. These findings obtained during actual overground perturbation conditions could have implications for the improved design of wearable robotic devices for balance assistance.

Real-Time-Capable Muscle Force Estimation for Monitoring Robotic Rehabilitation Therapy in the Intensive Care Unit

In this paper, a method is proposed to enable real-time monitoring of muscle forces during robotic rehabilitation therapy in the ICU. This method is solely based on sensor information provided by the rehabilitation robot. In current clinical practice, monitoring primarily takes place in the later stages of rehabilitation, but it would also be highly beneficial during early stages. Musculoskeletal models have large, mostly unrealized potential to support and improve patient monitoring. The method presented in this paper is based on a state-of-the-art muscle-tendon path model, which is applied to the use case of the robotic rehabilitation device VEMOTION. The muscle force estimation is validated against surface electromyography measurements of lower limb muscles from 12 healthy volunteers The results show an overall correlation of R = 0.70 0.25 for the single-joint muscle m. iliopsoas, which has a ±major contribution to hip flexion. Given this correlation, the proposed model could be used for real-time monitoring of active patient participation.

Machine Learning for Optical Motion Capture-Driven Musculoskeletal Modelling from Inertial Motion Capture Data

Marker-based Optical Motion Capture (OMC) systems and associated musculoskeletal (MSK) modelling predictions offer non-invasively obtainable insights into muscle and joint loading at an in vivo level, aiding clinical decision-making. However, an OMC system is lab-based, expensive, and requires a line of sight. Inertial Motion Capture (IMC) techniques are widely-used alternatives, which are portable, user-friendly, and relatively low-cost, although with lesser accuracy. Irrespective of the choice of motion capture technique, one typically uses an MSK model to obtain the kinematic and kinetic outputs, which is a computationally expensive tool increasingly well approximated by machine learning (ML) methods. Here, an ML approach is presented that maps experimentally recorded IMC input data to the human upper-extremity MSK model outputs computed from ('gold standard') OMC input data. Essentially, this proof-of-concept study aims to predict higher-quality MSK outputs from the much easier-to-obtain IMC data. We use OMC and IMC data simultaneously collected for the same subjects to train different ML architectures that predict OMC-driven MSK outputs from IMC measurements. In particular, we employed various neural network (NN) architectures, such as Feed-Forward Neural Networks (FFNNs) and Recurrent Neural Networks (RNNs) (vanilla, Long Short-Term Memory, and Gated Recurrent Unit) and a comprehensive search for the best-fit model in the hyperparameters space in both subject-exposed (SE) as well as subject-naive (SN) settings. We observed a comparable performance for both FFNN and RNN models, which have a high degree of agreement (ravg,SE,FFNN=0.90±0.19, ravg,SE,RNN=0.89±0.17, ravg,SN,FFNN=0.84±0.23, and ravg,SN,RNN=0.78±0.23) with the desired OMC-driven MSK estimates for held-out test data. The findings demonstrate that mapping IMC inputs to OMC-driven MSK outputs using ML models could be instrumental in transitioning MSK modelling from 'lab to field'.

Differences in hip musculoskeletal loads between limbs during daily activities in patients with 3D-printed hemipelvic reconstructions following tumor surgery

BACKGROUND

Anatomical custom-made prostheses, thanks to computer-aided design and 3D-printing technology, help improve osseointegration and reduce mechanical complications in bone reconstructions following bone tumors. A recent quantitative analysis of long-term recovery in patients with 3D-printed reconstructions following pelvic tumor surgery showed asymmetries in ground reaction forces between limbs during different motor activities, while standing very good motor performance and quality of life.

RESEARCH QUESTION

We analyzed hip contact forces and muscle forces in that cohort of six patients with an innovative custom-made reconstruction of the hemipelvis, and we tested the hypothesis that asymmetries in ground reaction forces would result in more marked differences in musculoskeletal forces.

METHODS

State-of-the-art musculoskeletal modeling in an optimization-based inverse-dynamics workflow was used to calculate hip contact forces and muscle forces during five motor activities, and the differences between limbs were statistically evaluated across the motor activity cycles and on the force peaks.

RESULTS

The musculoskeletal loads were found to be not symmetric, as hip loads were generally higher in the contralateral limb. We found significant differences in considerable portions of the motor activities cycles except squat, load symmetry indices indicating a load increase (median up to 25%) on the contralateral limb, especially during stair descent and chair rise/sit, and significantly higher values in the contralateral limb at force peaks.

SIGNIFICANCE

We confirmed the hypothesis that residual asymmetries found in ground reaction forces were amplified when hip musculoskeletal loads were investigated, reflecting a shift of the loads toward the intact limb. Despite the general trend of higher loads found in the contralateral hip, this cannot be considered a risk of overloading, as both hips supported loads in a physiological range or lower, indicating a likely optimal recovery.

Prediction of ground reaction forces and moments during walking in children with cerebral palsy

Introduction

Gait analysis is increasingly used to support clinical decision-making regarding diagnosis and treatment planning for movement disorders. As a key part of gait analysis, inverse dynamics can be applied to estimate internal loading conditions during movement, which is essential for understanding pathological gait patterns. The inverse dynamics calculation uses external kinetic information, normally collected using force plates. However, collection of external ground reaction forces (GRFs) and moments (GRMs) can be challenging, especially in subjects with movement disorders. In recent years, a musculoskeletal modeling-based approach has been developed to predict external kinetics from kinematic data, but its performance has not yet been evaluated for altered locomotor patterns such as toe-walking. Therefore, the goal of this study was to investigate how well this prediction method performs for gait in children with cerebral palsy.

Methods

The method was applied to 25 subjects with various forms of hemiplegic spastic locomotor patterns. Predicted GRFs and GRMs, in addition to associated joint kinetics derived using inverse dynamics, were statistically compared against those based on force plate measurements.

Results

The results showed that the performance of the predictive method was similar for the affected and unaffected limbs, with Pearson correlation coefficients between predicted and measured GRFs of 0.71–0.96, similar to those previously reported for healthy adults, despite the motor pathology and the inclusion of toes-walkers within our cohort. However, errors were amplified when calculating the resulting joint moments to an extent that could influence clinical interpretation.

Conclusion

To conclude, the musculoskeletal modeling-based approach for estimating external kinetics is promising for pathological gait, offering the possibility of estimating GRFs and GRMs without the need for force plate data. However, further development is needed before implementation within clinical settings becomes possible.

Peak knee joint moments accurately predict medial and lateral knee contact forces in patients with valgus malalignment

Compressive knee joint contact force during walking is thought to be related to initiation and progression of knee osteoarthritis. However, joint loading is often evaluated with surrogate measures, like the external knee adduction moment, due to the complexity of computing joint contact forces. Statistical models have shown promising correlations between medial knee joint contact forces and knee adduction moments in particularly in individuals with knee osteoarthritis or after total knee replacements (R² = 0.44-0.60). The purpose of this study was to evaluate how accurately model-based predictions of peak medial and lateral knee joint contact forces during walking could be estimated by linear mixed-effects models including joint moments for children and adolescents with and without valgus malalignment. Peak knee joint moments were strongly correlated (R² > 0.85, p < 0.001) with both peak medial and lateral knee joint contact forces. The knee flexion and adduction moments were significant covariates in the models, strengthening the understanding of the statistical relationship between both moments and medial and lateral knee joint contact forces. In the future, these models could be used to evaluate peak knee joint contact forces from musculoskeletal simulations using peak joint moments from motion capture software, obviating the need for time-consuming musculoskeletal simulations.

Osteoarthritis year in review 2021: mechanics

Osteoarthritis (OA) has a complex, heterogeneous and only partly understood etiology. There is a definite role of joint cartilage pathomechanics in originating and progressing of the disease. Although it is still not identified precisely enough to design or select targeted treatments, the progress of this year's research demonstrates that this goal became much closer. On multiple scales - tissue, joint and whole body - an increasing number of studies were done, with impressive results. (1) Technology based instrument innovations, especially when combined with machine learning models, have broadened the applicability of biomechanics. (2) Combinations with imaging make biomechanics much more precise & personalized. (3) The combination of Musculoskeletal & Finite Element Models yield valid personalized cartilage loads. (4) Mechanical outcomes are becoming increasingly meaningful to inform and evaluate treatments, including predictive power from biomechanical models. Since most recent advancements in the field of biomechanics in OA are at the level of a proof op principle, future research should not only continue on this successful path of innovation, but also aim to develop clinical workflows that would facilitate including precision biomechanics in large scale studies. Eventually this will yield clinical tools for decision making and a rationale for new therapies in OA.

Altered Muscle Contributions are Required to Support the Stance Limb During Voluntary Toe-Walking

Toe-walking characterizes several neuromuscular conditions and is associated with a reduction in gait stability and efficiency, as well as in life quality. The optimal choice of treatment depends on a correct understanding of the underlying pathology and on the individual biomechanics of walking. The objective of this study was to describe gait deviations occurring in a cohort of healthy adult subjects when mimicking a unilateral toe-walking pattern compared to their normal heel-to-toe gait pattern. The focus was to characterize the functional adaptations of the major lower-limb muscles which are required in order to toe walk. Musculoskeletal modeling was used to estimate the required muscle contributions to the joint sagittal moments. The support moment, defined as the sum of the sagittal extensive moments at the ankle, knee, and hip joints, was used to evaluate the overall muscular effort necessary to maintain stance limb stability and prevent the collapse of the knee. Compared to a normal heel-to-toe gait pattern, toe-walking was characterized by significantly different lower-limb kinematics and kinetics. The altered kinetic demands at each joint translated into different necessary moment contributions from most muscles. In particular, an earlier and prolonged ankle plantarflexion contribution was required from the soleus and gastrocnemius during most of the stance phase. The hip extensors had to provide a higher extensive moment during loading response, while a significantly higher knee extension contribution from the vasti was necessary during mid-stance. Compensatory muscular activations are therefore functionally required at every joint level in order to toe walk. A higher support moment during toe-walking indicates an overall higher muscular effort necessary to maintain stance limb stability and prevent the collapse of the knee. Higher muscular demands during gait may lead to fatigue, pain, and reduced quality of life. Toe-walking is indeed associated with significantly larger muscle forces exerted by the quadriceps to the patella and prolonged force transmission through the Achilles tendon during stance phase. Optimal treatment options should therefore account for muscular demands and potential overloads associated with specific compensatory mechanisms.

Effectiveness of Global Optimisation and Direct Kinematics in Predicting Surgical Outcome in Children with Cerebral Palsy

Multibody optimisation approaches have not seen much use in routine clinical applications despite evidence of improvements in modelling through a reduction in soft tissue artifacts compared to the standard gait analysis technique of direct kinematics. To inform clinical use, this study investigated the consistency with which both approaches predicted post-surgical outcomes, using changes in Gait Profile Score (GPS) when compared to a clinical assessment of outcome that did not include the 3D gait data. Retrospective three-dimensional motion capture data were utilised from 34 typically developing children and 26 children with cerebral palsy who underwent femoral derotation osteotomies as part of Single Event Multi-Level Surgeries. Results indicated that while, as expected, the GPS estimated from the two methods were numerically different, they were strongly correlated (Spearman's ρ = 0.93), and no significant differences were observed between their estimations of change in GPS after surgery. The two scores equivalently classified a worsening or improvement in the gait quality in 93% of the cases. When compared with the clinical classification of responders versus non-responders to the intervention, an equivalent performance was found for the two approaches, with 27/41 and 28/41 cases in agreement with the clinical judgement for multibody optimisation and direct kinematics, respectively. With this equivalent performance to the direct kinematics approach and the benefit of being less sensitive to skin artefact and allowing additional analysis such as estimation of musculotendon lengths and joint contact forces, multibody optimisation has the potential to improve the clinical decision-making process in children with cerebral palsy.

Modeling Musculoskeletal Dynamics during Gait: Evaluating the Best Personalization Strategy through Model Anatomical Consistency

No consensus exists on how to model human articulations within MSK models for the analysis of gait dynamics. We propose a method to evaluate joint models and we apply it to three models with different levels of personalization. The method evaluates the joint model’s adherence to the MSK hypothesis of negligible joint work by quantifying ligament and cartilage deformations resulting from joint motion; to be anatomically consistent, these deformations should be minimum. The contrary would require considerable external work to move the joint, violating a strong working hypothesis and raising concerns about the credibility of the MSK outputs. Gait analysis and medical resonance imaging (MRI) from ten participants were combined to build lower limb subject-specific MSK models. MRI-reconstructed anatomy enabled three levels of personalization using different ankle joint models, in which motion corresponded to different ligament elongation and cartilage co-penetration. To estimate the impact of anatomical inconsistency in MSK outputs, joint internal forces resulting from tissue deformations were computed for each joint model and MSK simulations were performed ignoring or considering their contribution. The three models differed considerably for maximum ligament elongation and cartilage co-penetration (between 5.94 and 50.69% and between −0.53 and −5.36 mm, respectively). However, the model dynamic output from the gait simulations were similar. When accounting for the internal forces associated with tissue deformation, outputs changed considerably, the higher the personalization level the smaller the changes. Anatomical consistency provides a solid method to compare different joint models. Results suggest that consistency grows with personalization, which should be tailored according to the research question. A high level of anatomical consistency is recommended when individual specificity and the behavior of articular structures is under investigation.