Leveraging explainable machine learning to identify gait biomechanical parameters associated with anterior cruciate ligament injury

Exploration and demonstration of explainable machine learning models in prosthetic rehabilitation-based gait analysis

Quantitative gait analysis is important for understanding the non-typical walking patterns associated with mobility impairments. Conventional linear statistical methods and machine learning (ML) models are commonly used to assess gait performance and related changes in the gait parameters. Nonetheless, explainable machine learning provides an alternative technique for distinguishing the significant and influential gait changes stemming from a given intervention. The goal of this work was to demonstrate the use of explainable ML models in gait analysis for prosthetic rehabilitation in both population- and sample-based interpretability analyses. Models were developed to classify amputee gait with two types of prosthetic knee joints. Sagittal plane gait patterns of 21 individuals with unilateral transfemoral amputations were video-recorded and 19 spatiotemporal and kinematic gait parameters were extracted and included in the models. Four ML models-logistic regression, support vector machine, random forest, and LightGBM-were assessed and tested for accuracy and precision. The Shapley Additive exPlanations (SHAP) framework was applied to examine global and local interpretability. Random Forest yielded the highest classification accuracy (98.3%). The SHAP framework quantified the level of influence of each gait parameter in the models where knee flexion-related parameters were found the most influential factors in yielding the outcomes of the models. The sample-based explainable ML provided additional insights over the population-based analyses, including an understanding of the effect of the knee type on the walking style of a specific sample, and whether or not it agreed with global interpretations. It was concluded that explainable ML models can be powerful tools for the assessment of gait-related clinical interventions, revealing important parameters that may be overlooked using conventional statistical methods.

Leveraging Emerging Technologies to Expand Accessibility and Improve Precision in Rehabilitation and Exercise for People with Disabilities

Physical rehabilitation and exercise training have emerged as promising solutions for improving health, restoring function, and preserving quality of life in populations that face disparate health challenges related to disability. Despite the immense potential for rehabilitation and exercise to help people with disabilities live longer, healthier, and more independent lives, people with disabilities can experience physical, psychosocial, environmental, and economic barriers that limit their ability to participate in rehabilitation, exercise, and other physical activities. Together, these barriers contribute to health inequities in people with disabilities, by disproportionately limiting their ability to participate in health-promoting physical activities, relative to people without disabilities. Therefore, there is great need for research and innovation focusing on the development of strategies to expand accessibility and promote participation in rehabilitation and exercise programs for people with disabilities. Here, we discuss how cutting-edge technologies related to telecommunications, wearables, virtual and augmented reality, artificial intelligence, and cloud computing are providing new opportunities to improve accessibility in rehabilitation and exercise for people with disabilities. In addition, we highlight new frontiers in digital health technology and emerging lines of scientific research that will shape the future of precision care strategies for people with disabilities.

Diagnosis of generalized joint hypermobility with gait patterns using a deep neural network

Generalized joint hypermobility (GJH) describes the situation that the range of joint motion exceeds the normal range. GJH is found to increase the risk of knee-related injury and osteoarthritis, challenging the athletic ability of the population. Gait signals are directly related to hip and knee athletic conditions, and have been shown to have significant changes with GJH by our previous research. But gait data are noisy, and vary with age, gender, weight, and ethnicity, which makes them hard to analyze with traditional statistical methods. In this study, we proposed an end-to-end deep learning model to recognize the patterns of the gait signals. The model consists of convolutional network blocks, residual network blocks, and attention blocks. Our dataset is composed of 452 samples of gait data obtained by a three-dimension motion capture system, with the six-degree-of-freedom kinematic data of hip, knee, and ankle joints during level walking, downhill, and uphill walking. The model achieves 95.77% accuracy and 98.68% specificity with a recall of 76.84% while is more efficient than traditional machine learning methods. The trained model can be run on economical friendly devices, and provide help for immediate and precise diagnosis of GJH. It is also meaningful to consider its application in large-scale GJH screening, which can contribute to sports medicine.

The Effects of a Single Session of a Rhythmic Movement Program on Selected Biopsychological Parameters in PD Patients: A Methodological Approach

The aim of the present study is to examine the acute effects of a specially designed musicokinetic (MSK) program for patients with Parkinson's disease (PD) on (a) anxiety levels, (b) select kinematic and kinetic parameters, and (c) frontal cortex hemodynamic responses, during gait initiation and steady-state walking. Methods: This is a blind cross-over randomized control trial (RCT) in which 13 volunteers with PD will attend a 45 min MSK program under the following conditions: (a) a synchronous learning format and (b) an asynchronous remote video-based format. Changes in gait biomechanics and frontal cortex hemodynamic responses will be examined using a 10-camera 3D motion analysis (Vicon T-series, Oxford, UK), and a functional near-infrared spectroscopy (f-NIRS-Portalite, Artinis NL) system, respectively, while anxiety levels will be evaluated using the Hamilton Anxiety Rating Scale. Expected results: Guided by the rules of music, where periodicity is distinct, our specially designed MSK program may eventually be beneficial in improving motor difficulties and, hence, reducing anxiety. The combined implementation of f-NIRS in parallel with 3D gait analysis has yet to be evaluated in Parkinsonian patients following a MSK intervention. It is expected that the aforementioned intervention, through better rhythmicity, may improve the automatization of motor control, gait kinematics, and kinetics-supported by decreased frontal cortex hemodynamic activity-which may be linked to reduced anxiety levels.

Leveraging Multivariable Linear Regression Analysis to Identify Patients with Anterior Cruciate Ligament Deficiency Using a Composite Index of the Knee Flexion and Muscle Force

Patients with anterior cruciate ligament (ACL) deficiency (ACLD) tend to have altered lower extremity kinematics and dynamics. Clinical diagnosis of ACLD requires more objective and convenient evaluation criteria. Twenty-five patients with ACLD before ACL reconstruction and nine healthy volunteers were recruited. Five experimental jogging data sets of each participant were collected and calculated using a musculoskeletal model. The resulting knee flexion and muscle force data were analyzed using a t-test for characteristic points, which were the time points in the gait cycle when the most significant difference between the two groups was observed. The data of the characteristic points were processed with principal component analysis to generate a composite index for multivariable linear regression. The accuracy rate of the regression model in diagnosing patients with ACLD was 81.4%. This study demonstrates that the multivariable linear regression model and composite index can be used to diagnose patients with ACLD. The composite index and characteristic points can be clinically objective and can be used to extract effective information quickly and conveniently.

Identifying Gait-Related Functional Outcomes in Post-Knee Surgery Patients Using Machine Learning: A Systematic Review

Modern lifestyles require new tools for determining a person's ability to return to daily activities after knee surgery. These quantitative instruments must feature high discrimination, be non-invasive, and be inexpensive. Machine learning is a revolutionary approach that has the potential to satisfy the aforementioned requirements and bridge the knowledge gap. The scope of this study is to summarize the results of a systematic literature review on the identification of gait-related changes and the determination of the functional recovery status of patients after knee surgery using advanced machine learning algorithms. The current systematic review was conducted using multiple databases in accordance with the PRISMA guidelines, including Scopus, PubMed, and Semantic Scholar. Six out of the 405 articles met our inclusion criteria and were directly related to the quantification of the recovery status using machine learning and gait data. The results were interpreted using appropriate metrics. The results demonstrated a recent increase in the use of sophisticated machine learning techniques that can provide robust decision-making support during personalized post-treatment interventions for knee-surgery patients.

Association between Self-Reported Prior Night's Sleep and Single-Task Gait in Healthy, Young Adults: A Study Using Machine Learning

Failure to obtain the recommended 7−9 h of sleep has been associated with injuries in youth and adults. However, most research on the influence of prior night’s sleep and gait has been conducted on older adults and clinical populations. Therefore, the objective of this study was to identify individuals who experience partial sleep deprivation and/or sleep extension the prior night using single task gait. Participants (n = 123, age 24.3 ± 4.0 years; 65% female) agreed to participate in this study. Self-reported sleep duration of the night prior to testing was collected. Gait data was collected with inertial sensors during a 2 min walk test. Group differences (<7 h and >9 h, poor sleepers; 7−9 h, good sleepers) in gait characteristics were assessed using machine learning and a post-hoc ANCOVA. Results indicated a correlation (r = 0.79) between gait parameters and prior night’s sleep. The most accurate machine learning model was a Random Forest Classifier using the top 9 features, which had a mean accuracy of 65.03%. Our findings suggest that good sleepers had more asymmetrical gait patterns and were better at maintaining gait speed than poor sleepers. Further research with larger subject sizes is needed to develop more accurate machine learning models to identify prior night’s sleep using single-task gait.