Comparison of machine learning classifiers for differentiating level and sport using movement data.

The purposes of this study were to determine if 1) recurrent neural networks designed for multivariate, time-series analyses outperform traditional linear and non-linear machine learning classifiers when classifying athletes based on competition level and sport played, and 2) athletes of different sports move differently during non-sport-specific movement screens. Optical-based kinematic data from 542 athletes were used as input data for nine different machine learning algorithms to classify athletes based on competition level and sport played. For the traditional machine learning classifiers, principal component analysis and feature selection were used to reduce the data dimensionality and to determine the best principal components to retain. Across tasks, recurrent neural networks and linear machine learning classifiers tended to outperform the non-linear machine learning classifiers. For all tasks, reservoir computing took the least amount of time to train. Across tasks, reservoir computing had one of the highest classification rates and took the least amount of time to train; however, interpreting the results is more difficult compared to linear classifiers. In addition, athletes were successfully classified based on sport suggesting that athletes competing in different sports move differently during non-sport specific movements. Therefore, movement assessment screens should incorporate sport-specific scoring criteria.

Influence of Articular Geometry and Tibial Tubercle Location on Patellofemoral Kinematics and Contact Mechanics.

Trochlear groove geometry and the location of the tibial tubercle, where the patellar tendon inserts, have both been associated with patellofemoral instability and can be modified surgically. Although their effects on patellofemoral biomechanics have been investigated individually, the interaction between the two is unclear. The authors' aim was to use statistical shape modeling and musculoskeletal simulation to examine the effect of patellofemoral geometry on the relationship between tibial tubercle location and patellofemoral function. A statistical shape model was used to generate new knee geometries with trochlear grooves ranging from shallow to deep. A Monte Carlo approach was used to create 750 knee models by randomly selecting a geometry and randomly translating the tibial tubercle medially/laterally and anteriorly. Each knee model was incorporated into a musculoskeletal model, and an overground walking trial was simulated. Knees with shallow trochlear geometry were more sensitive to tubercle medialization with greater changes in lateral patella position (-3.0 mm/cm medialization shallow vs -0.6 mm/cm deep) and cartilage contact pressure (-0.51 MPa/cm medialization shallow vs 0.04 MPa/cm deep). However, knees with deep trochlear geometry experienced greater increases in medial cartilage contact pressure with medialization. This modeling framework has the potential to aid in surgical decision making.

Patella Apex Influences Patellar Ligament Forces and Ratio.

The relationship between three-dimensional shape and patellofemoral mechanics is complicated. The Wiberg patella classification is a method of distinguishing shape differences in the axial plane of the patella that can be used to connect shape differences to observed mechanics. This study uses the Wiberg patella classification to differentiate variance in a statistical shape model describing changes in patella morphology and height. We investigate how patella morphology influences force distribution within the patellofemoral joint. The Wiberg type I patella has a more symmetrical medial and lateral facet while the type III patella has a larger lateral facet compared to medial. The second principal component of the statistical shape model described shape variation that qualitatively resembled the different Wiberg patellas. We generated patellofemoral morphologies from the statistical shape model and integrated them into a musculoskeletal model with a twelve degrees-of-freedom knee. We simulated an overground walking trial with these morphologies and recorded patellofemoral mechanics and ligament forces. An increase in patellar ligament force corresponded with an increase in patella height. Wiberg type III patellas had a sharper patella apex which related to lower ratios of quadriceps tendon forces to patellar ligament forces. The change in pivot point of the patella affects the ratio of forces as well as the patellofemoral reaction force. This study provides a better understanding of how patella morphology affects fundamental patella mechanics which may help identify at-risk populations for pathology development.

Validation of an IMU Suit for Military-Based Tasks.

Investigating the effects of load carriage on military soldiers using optical motion capture is challenging. However, inertial measurement units (IMUs) provide a promising alternative. Our purpose was to compare optical motion capture with an Xsens IMU system in terms of movement reconstruction using principal component analysis (PCA) using correlation coefficients and joint kinematics using root mean squared error (RMSE). Eighteen civilians performed military-type movements while their motion was recorded using both optical and IMU-based systems. Tasks included walking, running, and transitioning between running, kneeling, and prone positions. PCA was applied to both the optical and virtual IMU markers, and the correlations between the principal component (PC) scores were assessed. Full-body joint angles were calculated and compared using RMSE between optical markers, IMU data, and virtual markers generated from IMU data with and without coordinate system alignment. There was good agreement in movement reconstruction using PCA; the average correlation coefficient was 0.81 ± 0.14. RMSE values between the optical markers and IMU data for flexion-extension were less than 9°, and 15° for the lower and upper limbs, respectively, across all tasks. The underlying biomechanical model and associated coordinate systems appear to influence RMSE values the most. The IMU system appears appropriate for capturing and reconstructing full-body motion variability for military-based movements.

Relationship Between Lateral Patellar Stability and Tibial Tubercle Location for Varying Patellofemoral Geometries.

The geometry of the patellofemoral joint affects function and pathology. However, the impact of trochlear groove depth on treatments for patellar instability and pain is not clear. Tibial tubercle osteotomy is a common surgical intervention for patellar instability where the tibial insertion of the patellar tendon (PT) is translated to align the extensor mechanism and stabilize the joint. The aim of this work was to investigate the interaction between trochlear groove depth and PT insertion and their effect on patellar stability. Patellofemoral geometry was modified based on a statistical shape model to create knees with a range of trochlear groove depths. A Monte Carlo approach was used and 750 instances of a musculoskeletal model were generated with varying geometry and anterior and medial transfer of the PT. Stability was assessed by applying a lateral perturbation force to the patella during simulation of overground walking. In knees with deep trochlear grooves, a medialized PT increased stability. However, in knees with shallow trochlear grooves, stability was maximized for tendon insertion ∼1 mm medial to its neutral location. This PT insertion also corresponded to the best alignment of the patella in the trochlear groove in these knees, indicating that good alignment may be important to maximizing stability. Anterior PT transfer had minimal effect on stability for all geometries. A better understanding of the effects of articular geometry and tubercle location on stability may aid clinicians in patient-specific surgical planning.

Prediction of Knee Joint Contact Forces From External Measures Using Principal Component Prediction and Reconstruction.

Abnormal loading of the knee joint contributes to the pathogenesis of knee osteoarthritis. Gait retraining is a noninvasive intervention that aims to reduce knee loads by providing audible, visual, or haptic feedback of gait parameters. The computational expense of joint contact force prediction has limited real-time feedback to surrogate measures of the contact force, such as the knee adduction moment. We developed a method to predict knee joint contact forces using motion analysis and a statistical regression model that can be implemented in near real-time. Gait waveform variables were deconstructed using principal component analysis, and a linear regression was used to predict the principal component scores of the contact force waveforms. Knee joint contact force waveforms were reconstructed using the predicted scores. We tested our method using a heterogenous population of asymptomatic controls and subjects with knee osteoarthritis. The reconstructed contact force waveforms had mean (SD) root mean square differences of 0.17 (0.05) bodyweight compared with the contact forces predicted by a musculoskeletal model. Our method successfully predicted subject-specific shape features of contact force waveforms and is a potentially powerful tool in biofeedback and clinical gait analysis.

Experimental validation of finite element analysis of human vertebral collapse under large compressive strains.

Osteoporosis-related vertebral fractures represent a major health problem in elderly populations. Such fractures can often only be diagnosed after a substantial deformation history of the vertebral body. Therefore, it remains a challenge for clinicians to distinguish between stable and progressive potentially harmful fractures. Accordingly, novel criteria for selection of the appropriate conservative or surgical treatment are urgently needed. Computer tomography-based finite element analysis is an increasingly accepted method to predict the quasi-static vertebral strength and to follow up this small strain property longitudinally in time. A recent development in constitutive modeling allows us to simulate strain localization and densification in trabecular bone under large compressive strains without mesh dependence. The aim of this work was to validate this recently developed constitutive model of trabecular bone for the prediction of strain localization and densification in the human vertebral body subjected to large compressive deformation. A custom-made stepwise loading device mounted in a high resolution peripheral computer tomography system was used to describe the progressive collapse of 13 human vertebrae under axial compression. Continuum finite element analyses of the 13 compression tests were realized and the zones of high volumetric strain were compared with the experiments. A fair qualitative correspondence of the strain localization zone between the experiment and finite element analysis was achieved in 9 out of 13 tests and significant correlations of the volumetric strains were obtained throughout the range of applied axial compression. Interestingly, the stepwise propagating localization zones in trabecular bone converged to the buckling locations in the cortical shell. While the adopted continuum finite element approach still suffers from several limitations, these encouraging preliminary results towards the prediction of extended vertebral collapse may help in assessing fracture stability in future work.

Factors affecting the stability of reverse shoulder arthroplasty: a biomechanical study.

BACKGROUND: Despite the success of reverse shoulder arthroplasty (RSA) in treating patients with painful pseudoparalytic shoulders, instability is a common complication and currently the factors affecting stability are not well understood. The objective of this study was to investigate a number of factors as well as the interactions between factors to determine how they affect the stability of the prosthesis. These factors included: active arm posture (abduction and abduction plane angles), loading direction, glenosphere diameter and eccentricity, and humeral socket constraint. METHODS: Force required to dislocate the joint, determined using a biomechanical shoulder simulator, was used as a measure of stability. A factorial design experiment was implemented to examine the factors and interactions. RESULTS: Actively increasing the abduction angle by 15° leads to a 30% increase in stability and use of an inferior-offset rather than a centered glenosphere improved stability by 17%. Use of a more constrained humeral socket also increased stability; but the effect was dependent on loading direction, with a 88% improvement for superior loading, 66% for posterior, 36% for anterior, and no change for inferior loading. Abduction plane angle and glenosphere diameter had no effect on stability. CONCLUSION: Increased glenohumeral abduction and the use of an inferior-offset glenosphere were found to increase the stability of RSA. Additionally, use of a more constrained humeral socket increased stability for anterior, posterior, and superior loading. These identified factor effects have the potential to decrease the risk of dislocation following RSA.

Morphable models of the lumbar spine to vary geometry based on pathology, demographics, and anatomical measurements.

The shape of the lumbar spine influences its function and dysfunction. Yet examining the influence of geometric differences associated with pathology or demographics on lumbar biomechanics is challenging in vivo where these effects cannot be isolated, and the use of simple anatomical measurements does not fully capture the complex three-dimensional geometry. The goal of this work was to develop and share morphable models of the lumbar spine that allow geometry to be varied according to pathology, demographics, or anatomical measurements. Partial least squares regression was used to generate statistical shape models that quantify geometric differences associated with pathology, demographics, and anatomical measurements from the lumbar spines of 87 patients. To determine if the morphable models detected meaningful geometric differences, the ability of the morphable models to classify spines was compared with models generated from random labels. The models for disc herniation (p < 0.04), spondylolisthesis (p < 0.001), and sex (p < 0.01) all performed significantly better than the random models. Age was predicted with a root mean square error of 14.1 years using the age-based model. The morphable models for anatomical measurements were able to produce instances with root mean square errors less than 0.8°, 0.3 cm2, and 0.7 mm between desired and resulting measurements. This method can be used to produce morphable models that enable further analysis of the relationship among shape, pathology, demographics, and function through computational simulations. The morphable models and code are available at https://github.com/aclouthier/morphable-lumbar-model.

Predicting vertical and shear ground reaction forces during walking and jogging using wearable plantar pressure insoles.

BACKGROUND: The development of plantar pressure insoles has made them a potential replacement for force plates. These wearable devices can measure multiple steps and might be used outside of the lab environment for rehabilitation and evaluation of sport performance. However, they can only measure the vertical force which does not completely represent the vertical ground reaction force. In addition, they are not able to measure shear forces which play an import role in the dynamic performance of individuals. Indirect approaches might be implemented to improve the accuracy of the force estimated by plantar pressure systems. RESEARCH QUESTION: The aim of this study was to predict the vertical and shear components of ground reaction force from plantar pressure data using recurrent neural networks. METHODS: Ground reaction force and plantar pressure data were collected from 16 healthy individuals during 10 trials of walking and five trials of jogging using Bertec force plates at 1000 Hz and FScan plantar pressure insoles at 100 Hz. A long short-term memory neural network was built to consider the time dependency of pressure and force data in predictions. The data were split into three subsets of train, to train the model, evaluate, to optimize the model hyperparameters, and test sets, to assess the accuracy of the model predictions. RESULTS: The results of this study showed that our long short-term memory model could accurately predict the shear and vertical force components during walking and jogging. The predictions were more accurate during walking compared to jogging. In addition, the predictions of mediolateral force had higher error and lower correlation compared to vertical and anteroposterior components. SIGNIFICANCE: The long short-term memory model developed in this study may be an acceptable option for accurate estimation of ground reaction force during outdoor activities which can have significant impacts in rehabilitation, sport performance, and gaming.

The role of machine learning in the primary prevention of work-related musculoskeletal disorders: A scoping review.

To determine the applications of machine learning (ML) techniques used for the primary prevention of work-related musculoskeletal disorders (WMSDs), a scoping review was conducted using seven literature databases. Of the 4,639 initial results, 130 primary research studies were deemed relevant for inclusion. Studies were reviewed and classified as a contribution to one of six steps within the primary WMSD prevention research framework by van der Beek et al. (2017). ML techniques provided the greatest contributions to the development of interventions (48 studies), followed by risk factor identification (33 studies), underlying mechanisms (29 studies), incidence of WMSDs (14 studies), evaluation of interventions (6 studies), and implementation of effective interventions (0 studies). Nearly a quarter (23.8%) of all included studies were published in 2020. These findings provide insight into the breadth of ML techniques used for primary WMSD prevention and can help identify areas for future research and development.

A data-driven framework for assessing soldier performance, health, and survivability.

Presented is a framework that uses pattern classification methods to incrementally morph whole-body movement patterns to investigate how personal (sex, military experience, and body mass) and load characteristics affect the survivability tradespace: performance, musculoskeletal health, and susceptibility to enemy action. Sixteen civilians and 12 soldiers performed eight military-based movement patterns under three body-borne loads: ∼5.5 kg, ∼22 kg, and ∼38 kg. Our framework reduces dimensionality using principal component analysis and uses linear discriminant analysis to classify groups and morph movement patterns. Our framework produces morphed whole-body movement patterns that emulate previously published changes to the survivability tradespace caused by body-borne loads. Additionally, we identified that personal characteristics can greatly impact the tradespace when carrying heavy body-borne loads. Using our framework, military leaders can make decisions based on objective information for armour procurement, employment of armour, and battlefield performance, which can positively impact operational readiness and increase overall mission success.

Patellofemoral Mechanics: a Review of Pathomechanics and Research Approaches.

PURPOSE OF REVIEW: The patellofemoral joint is a complicated articulation of the patella and femur that is prone to pathologies. The purpose of this review is to report on the current methods of investigating patellofemoral mechanics, factors that affect joint function, and future directions in patellofemoral joint research with emerging technologies and techniques. RECENT FINDINGS: While previous hypotheses have suggested that the patella is only a moment arm extender, recent literature has suggested that the patella influences the control of knee moments and forces acting on the tibia as well as contributes to various aspects of patellar function with minimal neural input. With advancements in simulating a six-degrees-of-freedom patellofemoral joint, we have gained a better understanding of patella motion and have shown that geometry and muscle activations impact patella mechanics. Research into influences on patella mechanics from other joints such as the hip and foot has become more prevalent. In this review, we report current in vivo, in vitro, and in silico approaches to studying the patellofemoral joint. Kinematic and anatomical factors that affect patellofemoral joint function such as patella alta and tilt or bone morphology and ligaments are discussed. Moving forward, we suggest that advanced in vivo dynamic imaging methods coupled to musculoskeletal simulation will provide further understanding of patellofemoral pathomechanics and allow engineers and clinicians to design interventions to mitigate or prevent patellofemoral pathologies.

Three-dimensional scapular morphology is associated with rotator cuff tears and alters the abduction moment arm of the supraspinatus.

BACKGROUND: Numerous studies have reported an association between rotator cuff injury and two-dimensional measures of scapular morphology. However, the mechanical underpinnings explaining how these shape features affect glenohumeral joint function and lead to injury are poorly understood. We hypothesized that three-dimensional features of scapular morphology differentiate asymptomatic shoulders from those with rotator cuff tears, and that these features would alter the mechanical advantage of the supraspinatus. METHODS: Twenty-four individuals with supraspinatus tears and twenty-seven age-matched controls were recruited. A statistical shape analysis identified scapular features distinguishing symptomatic patients from asymptomatic controls. We examined the effect of injury-associated morphology on mechanics by developing a morphable model driven by six degree-of-freedom biplanar videoradiography data. We used the model to simulate abduction for a range of shapes and computed the supraspinatus moment arm. FINDINGS: Rotator cuff injury was associated with a cranial orientation of the glenoid and scapular spine (P = .011, d = 0.75) and/or decreased subacromial space (P = .001, d = 0.94). The shape analysis also identified previously undocumented features associated with superior inclination and subacromial narrowing. In our computational model, warping the scapula from a cranial to a lateral orientation increased the supraspinatus moment arm at 20° of abduction and decreased the moment arm at 160° of abduction. INTERPRETATIONS: Three-dimensional analysis of scapular morphology indicates a stronger relationship between morphology and cuff tears than two-dimensional measures. Insight into how morphological features affect rotator cuff mechanics may improve patient-specific strategies for prevention and treatment of cuff tears.

Sensor Data Required for Automatic Recognition of Athletic Tasks Using Deep Neural Networks.

Movement screens are used to assess the overall movement quality of an athlete. However, these rely on visual observation of a series of movements and subjective scoring. Data-driven methods to provide objective scoring of these movements are being developed. These currently use optical motion capture and require manual pre-processing of data to identify the start and end points of each movement. Therefore, we aimed to use deep learning techniques to automatically identify movements typically found in movement screens and assess the feasibility of performing the classification based on wearable sensor data. Optical motion capture data were collected on 417 athletes performing 13 athletic movements. We trained an existing deep neural network architecture that combines convolutional and recurrent layers on a subset of 278 athletes. A validation subset of 69 athletes was used to tune the hyperparameters and the final network was tested on the remaining 70 athletes. Simulated inertial measurement data were generated based on the optical motion capture data and the network was trained on this data for different combinations of body segments. Classification accuracy was similar for networks trained using the optical and full-body simulated inertial measurement unit data at 90.1 and 90.2%, respectively. A good classification accuracy of 85.9% was obtained using as few as three simulated sensors placed on the torso and shanks. However, using three simulated sensors on the torso and upper arms or fewer than three sensors resulted in poor accuracy. These results for simulated sensor data indicate the feasibility of classifying athletic movements using a small number of wearable sensors. This could facilitate objective data-driven methods that automatically score overall movement quality using wearable sensors to be easily implemented in the field.

The effect of articular geometry features identified using statistical shape modelling on knee biomechanics.

Articular geometry in the knee varies widely among people which has implications for risk of injury and pathology. The goals of this work were to develop a framework to systematically vary geometry in a multibody knee model and to use this framework to investigate the effect of morphological features on dynamic knee kinematics and contact mechanics. A statistical shape model of the tibiofemoral and patellofemoral joints was created from magnetic resonance images of 14 asymptomatic knees. The shape model was then used to generate 37 unique multibody knee models based on -3 to +3 standard deviations of the scores for the first six principal components identified. Each multibody model was then incorporated into a lower extremity musculoskeletal model and the Concurrent Optimization of Muscle Activations and Kinematics (COMAK) routine was used to simulate knee mechanics for overground walking. Changes in articular geometry affected knee function, resulting in differences up to 17° in orientation, 8 mm in translation, 0.7 BW in contact force, and 2.0 MPa in mean cartilage contact pressure. Understanding the relationship between shape and function in a joint could provide insight into the mechanisms behind injury and pathology and the variability in response to treatment.

Contributions of muscles and external forces to medial knee load reduction due to osteoarthritis braces.

BACKGROUND: Braces for medial knee osteoarthritis can reduce medial joint loads through a combination of three mechanisms: application of an external brace abduction moment, alteration of gait dynamics, and reduced activation of antagonistic muscles. Although the effect of knee bracing has been reported independently for each of these parameters, no previous study has quantified their relative contributions to reducing medial knee loads. METHODS: In this study, we used a detailed musculoskeletal model to investigate immediate changes in medial and lateral loads caused by two different knee braces: OA Assist and OA Adjuster 3 (DJO Global). Seventeen osteoarthritis subjects and eighteen healthy controls performed overground gait trials in unbraced and braced conditions. RESULTS: Across all subjects, bracing reduced medial loads by 0.1 to 0.3 times bodyweight (BW), or roughly 10%, and increased lateral loads by 0.03 to 0.2 BW. Changes in gait kinematics due to bracing were subtle, and had little effect on medial and lateral joint loads. The knee adduction moment was unaltered unless the brace moment was included in its computation. Only one muscle, biceps femoris, showed a significant change in EMG with bracing, but this did not contribute to altered peak medial contact loads. CONCLUSIONS: Knee braces reduced medial tibiofemoral loads primarily by applying a direct, and substantial, abduction moment to each subject's knee. To further enhance brace effectiveness, future brace designs should seek to enhance the magnitude of this unloader moment, and possibly exploit additional kinematic or neuromuscular gait modifications.

Identification of good candidates for valgus bracing as a treatment for medial knee osteoarthritis.

Valgus unloader braces are a conservative treatment option for medial compartment knee osteoarthritis that aim to unload the damaged medial compartment through application of an external abduction moment. Patient response to bracing is highly variable, however. While some experience improvements in pain, function, and joint loading, others receive little to no benefit. The objective of this work was to analyze clinical measures and biomechanical characteristics of unbraced walking to identify variables that are associated with the mechanical effectiveness of valgus unloader bracing. Seventeen patients with medial knee osteoarthritis walked overground with and without a valgus unloader brace. A musculoskeletal model was used to estimate the contact forces in the medial compartment of the tibiofemoral joint and brace effectiveness was defined as the decrease in peak medial contact force between unbraced and braced conditions. Stepwise linear regression was used to identify clinical and biomechanical measures that predicted brace effectiveness. The final regression model explained 77% of the variance in brace effectiveness using two variables. Bracing was more effective for those with greater peak external hip adduction moments and for those with higher Kellgren-Lawrence grades, indicating more severe radiographic osteoarthritis. The hip adduction moment was the best predictor of brace effectiveness and was well correlated with several other measures indicating that it may be functioning as a "biomarker" for good bracing candidates. CLINICAL SIGNIFICANCE: The ability to predict good candidates for valgus bracing may improve issues of patient compliance and could enable the ability to train patients to respond better to bracing. © 2017 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 36:351-356, 2018.

Correcting waveform bias using principal component analysis: Applications in multicentre motion analysis studies.

Multicentre studies are rare in three dimensional motion analyses due to challenges associated with combining waveform data from different centres. Principal component analysis (PCA) is a statistical technique that can be used to quantify variability in waveform data and identify group differences. A correction technique based on PCA is proposed that can be used in post processing to remove nuisance variation introduced by the differences between centres. Using this technique, the waveform bias that exists between the two datasets is corrected such that the means agree. No information is lost in the individual datasets, but the overall variability in the combined data is reduced. The correction is demonstrated on gait kinematics with synthesized crosstalk and on gait data from knee arthroplasty patients collected in two centres. The induced crosstalk was successfully removed from the knee joint angle data. In the second example, the removal of the nuisance variation due to the multicentre data collection allowed significant differences in implant type to be identified. This PCA-based technique can be used to correct for differences between waveform datasets in post processing and has the potential to enable multicentre motion analysis studies.

Finite element analysis predicts experimental failure patterns in vertebral bodies loaded via intervertebral discs up to large deformation.

Vertebral compression fractures are becoming increasingly common. Patient-specific nonlinear finite element (FE) models have shown promise in predicting yield strength and damage pattern but have not been experimentally validated for clinically relevant vertebral fractures, which involve loading through intervertebral discs with varying degrees of degeneration up to large compressive strains. Therefore, stepwise axial compression was applied in vitro on segments and performed in silico on their FE equivalents using a nonlocal damage-plastic model including densification at large compression for bone and a time-independent hyperelastic model for the disc. The ability of the nonlinear FE models to predict the failure pattern in large compression was evaluated for three boundary conditions: healthy and degenerated intervertebral discs and embedded endplates. Bone compaction and fracture patterns were predicted using the local volume change as an indicator and the best correspondence was obtained for the healthy intervertebral discs. These preliminary results show that nonlinear finite element models enable prediction of bone localisation and compaction. To the best of our knowledge, this is the first study to predict the collapse of osteoporotic vertebral bodies up to large compression using realistic loading via the intervertebral discs.