Three-Dimensional Motion Capture Data of a Movement Screen from 183 Athletes.

Movement screens are widely used to identify aberrant movement patterns in hopes of decreasing risk of injury, identifying talent, and/or improving performance. Motion capture data can provide quantitative, objective feedback regarding movement patterns. The dataset contains three-dimensional (3D) motion capture data of 183 athletes performing mobility tests (ankle, back bend, crossover adduction, crossover rotation, elbows, head, hip turn, scorpion, shoulder abduction, shoulder azimuth, shoulder rotation, side bends, side lunges and trunk rotation) and stability tests (drop jump, hop down, L-cut, lunge, rotary stability, step down and T-balance) bilaterally (where applicable), the athletes' injury history, and demographics. All data were collected at 120 Hz or 480 Hz using an 8-camera Raptor-E motion capture system with 45 passive reflective markers. A total of 5,493 trials were pre-processed and included in .c3d and .mat formats. This dataset will enable researchers and end users to explore movement patterns of athletes of varying demographics from different sports and competition levels; develop objective movement assessment tools; and gain new insights into the relationships between movement patterns and injury.

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.

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.

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.

Machine Learning and Deep Neural Network Architectures for 3D Motion Capture Datasets.

Biomechanical movement data are highly correlated multivariate time-series for which a variety of machine learning and deep neural network classification techniques are possible. For image classification, convolutional neural networks have reshaped the field, but have been challenging to apply to 3D movement data with its intrinsic multidimensional nonlinear correlations. Deep neural networks afford the opportunity to reduce feature engineering effort, remove model-based approximations that can introduce systematic errors, and reduce the manual data processing burden which is often a bottleneck in biomechanical data acquisition. What classification techniques are most appropriate for biomechanical movement data? Baseline performance for 3D joint centre trajectory classification using a number of traditional machine learning techniques are presented. Our framework and dataset support a robust comparison between classifier architectures over 416 athletes (professional, college, and amateur) from five primary and six non-primary sports performing thirteen non-sport-specific movements. A variety of deep neural networks specifically intended for time-series data are currently being evaluated.

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.

Classifying Elite From Novice Athletes Using Simulated Wearable Sensor Data.

Movement screens are frequently used to identify differences in movement patterns such as pathological abnormalities or skill related differences in sport; however, abnormalities are often visually detected by a human assessor resulting in poor reliability. Therefore, our previous research has focused on the development of an objective movement assessment tool to classify elite and novice athletes' kinematic data using machine learning algorithms. Classifying elite and novice athletes can be beneficial to objectively detect differences in movement patterns between the athletes, which can then be used to provide higher quality feedback to athletes and their coaches. Currently, the method requires optical motion capture, which is expensive and time-consuming to use, creating a barrier for adoption within industry. Therefore, the purpose of this study was to assess whether machine learning could classify athletes as elite or novice using data that can be collected easily and inexpensively in the field using inertial measurement units (IMUs). A secondary purpose of this study was to refine the architecture of the tool to optimize classification rates. Motion capture data from 542 athletes performing seven dynamic screening movements were analyzed. A principal component analysis (PCA)-based pattern recognition technique and machine learning algorithms with the Euclidean norm of the segment linear accelerations and angular velocities as inputs were used to classify athletes based on skill level. Depending on the movement, using metrics achievable with IMUs and a linear discriminant analysis (LDA), 75.1-84.7% of athletes were accurately classified as elite or novice. We have provided evidence that suggests our objective, data-driven method can detect meaningful differences during a movement screening battery when using data that can be collected using IMUs, thus providing a large methodological advance as these can be collected in the field using sensors. This method offers an objective, inexpensive tool that can be easily implemented in the field to potentially enhance screening, assessment, and rehabilitation in sport and clinical settings.

Considering movement competency within physical employment standards.

BACKGROUND: Physical employment standards (PES) ensure that candidates can demonstrate the physical capacity required to perform duties of work. However, movement competency, or an individual's movement strategy, can relate to injury risk and safety, and therefore should be considered in PES. OBJECTIVE: Demonstrate the utility of using artificial intelligence (AI) to detect risk-potential of different movement strategies within PES. METHODS: Biomechanical analysis was used to calculate peak flexion angles and peak extensor moment about the lumbar spine during participants' performance of a backboard lifting task. Lifts performed with relatively lower and higher exposure to postural and moment loading on the spine were characterized as "low" or "high" exposure, respectively. An AI model including principal component and linear discriminant analyses was then trained to detect and classify backboard lifts as "low" or "high". RESULTS: The AI model accurately classified over 85% of lifts as "low" or "high" exposure using only motion data as an input. CONCLUSIONS: This proof-of-principle demonstrates that movement competency can be assessed in PES using AI. Similar classification approaches could be used to improve the utility of PES as a musculoskeletal disorders (MSD) prevention tool by proactively identifying candidates at higher risk of MSD based on movement competency.

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.

Objectively Differentiating Movement Patterns between Elite and Novice Athletes.

INTRODUCTION: Movement screens are frequently used to identify abnormal movement patterns that may increase risk of injury or hinder performance. Abnormal patterns are often detected visually based on the observations of a coach or clinician. Quantitative or data-driven methods can increase objectivity, remove issues related to interrater reliability and offer the potential to detect new and important features that may not be observable by the human eye. Applying principal component analysis (PCA) to whole-body motion data may provide an objective data-driven method to identify unique and statistically important movement patterns, an important first step to objectively characterize optimal patterns or identify abnormalities. Therefore, the primary purpose of this study was to determine if PCA could detect meaningful differences in athletes' movement patterns when performing a non-sport-specific movement screen. As a proof of concept, athlete skill level was selected a priori as a factor likely to affect movement performance. METHODS: Motion capture data from 542 athletes performing seven dynamic screening movements (i.e., bird-dog, drop-jump, T-balance, step-down, L-hop, hop-down, and lunge) were analyzed. A PCA-based pattern recognition technique and a linear discriminant analysis with cross-validation were used to determine if skill level could be predicted objectively using whole-body motion data. RESULTS: Depending on the movement, the validated linear discriminant analysis models accurately classified 70.66% to 82.91% of athletes as either elite or novice. CONCLUSIONS: We have provided proof that an objective data-driven method can detect meaningful movement pattern differences during a movement screening battery based on a binary classifier (i.e., skill level in this case). Improving this method can enhance screening, assessment, and rehabilitation in sport, ergonomics, and medicine.

Pain catastrophizing moderates changes in spinal control in response to noxiously induced low back pain.

It is generally accepted that spine control and stability are relevant for the prevention and rehabilitation of low back pain (LBP). However, there are conflicting results in the literature in regards to how these variables are modified in the presence of LBP. The aims of the present work were twofold: (1) to use noxious stimulation to induce LBP in healthy individuals to assess the direct effects of pain on control (quantified by the time-dependent behavior of kinematic variance), and (2) to assess whether the relationship between pain and control is moderated by psychological features (i.e. pain catastrophizing (PC) and kinesiophobia). Participants completed three conditions (baseline, pain, recovery) during a task involving completion of 35 cycles of a repetitive unloaded spine flexion/extension movement. The neuromuscular control of spine movements was assessed during each condition using maximum finite-time Lyapunov exponents (λmax). Nociceptive stimulus involved injection of hypertonic saline into the interspinous ligament, eliciting pain that was greater than baseline and recovery (p<0.001). Although there was no overall main effect of the nociceptive stimulation (i.e. pain) on λmax when the whole group was included in the statistical model (p=0.564), when data were considered separately for those with high and low PC, two distinct and well established responses to the pain were observed. Specifically, those with high PC tightened their control (i.e. stabilized), whereas those with low PC loosened their control (i.e. destabilized). This study provides evidence that individuals' beliefs and attitudes towards pain are related to individual-specific motor behaviors, and suggests that future research studying spine control/stability and LBP should account for these variables.

The Effects of Experimentally Induced Low Back Pain on Spine Rotational Stiffness and Local Dynamic Stability.

Local dynamic stability, quantified using the maximum finite-time Lyapunov exponent (λ max), and the muscular contributions to spine rotational stiffness can provide pertinent information regarding the neuromuscular control of the spine during movement tasks. The primary goal of the present study was to assess if experimental capsaicin-induced low back pain (LBP) affects spine stability and the neuromuscular control of repetitive trunk movements in a group of healthy participants with no history of LBP. Fourteen healthy males were recruited for this investigation. Each participant was asked to complete three trials (baseline, in pain, and recovery) of 35 cycles of a repetitive trunk flexion/extension task at a rate of 0.25 Hz. Local dynamic stability and the muscular contributions to lumbar spine rotational stiffness were significantly impaired during the LBP trial compared to the baseline trial (p < 0.05); however, there was a trend for these measures to recover after a 1 h rest. This study provides evidence that capsaicin can effectively induce LBP, thereby altering spine rotational stiffness and local dynamic stability. Future research should directly compare the effects capsaicin-induced LBP and intramuscular/intraligamentous induced LBP on these same variables.

Comparing the local dynamic stability of trunk movements between varsity athletes with and without non-specific low back pain.

The local dynamic stability of trunk movements, quantified using the maximum Lyapunov exponent (λmax), can provide important information on the neuromuscular control of spine stability during movement tasks. Although previous research has displayed the promise of this technique, all studies were completed with healthy participants. Therefore the goal of this study was to compare the dynamic stability of spine kinematics and trunk muscle activations, as well as antagonistic muscle co-contraction, between athletes with and without low back pain (LBP). Twenty interuniversity varsity athletes (10 LBP, 10 healthy controls) were recruited to participate in the study. Each participant completed a repetitive trunk flexion task at 15 cycles per minute, both symmetrically and asymmetrically, while trunk kinematics and muscular activity (EMG) were monitored. The local dynamic stability of low back EMG was significantly higher (lower λmax) in healthy individuals (p=0.002), whereas the dynamic stability of kinematics, the dynamic stability of full trunk system EMG, and the amount of antagonistic co-contraction were significantly higher when moving asymmetrically (p<0.05 for all variables). Although non-significant, kinematic and trunk system EMG stability also tended to be impaired in LBP participants, whereas they also tended to co-contract their antagonist muscles more. This study provides evidence that Lyapunov analyses of kinematic and muscle activation data can provide insight into the neuromuscular control of spine stability in back pain participants. Future research will repeat these protocols in patients with higher levels of pain, with hopes of developing a tool to assess impairment and treatment effectiveness in clinical and workplace settings.