The detection of age groups by dynamic gait outcomes using machine learning approaches

Leveraging feature selection for enhanced fall risk prediction in elderly using gait analysis

There is no effective fall risk screening tool for the elderly that can be integrated into clinical practice. Developing a system that can be easily used in primary care services is a current need. Current studies focus on the use of multiple sensors or activities to achieve higher accuracy. However, multiple sensors and activities reduce the availability of these systems. This study aims to develop a system to perform fall prediction for the elderly by using signals recorded from a single sensor during a short-term activity. A total of 168 features in the time and frequency domains were created using acceleration signals obtained from 71 elderly people. The features were weighted based on the ReliefF algorithm, and the artificial neural networks model was developed using the most important features. The best classification result was obtained using the 17 most important features of those weighted for K = 20 nearest neighbors. The highest accuracy was 82.2% (82.9% Sensitivity, 81.6% Specificity). The partially high accuracy obtained in our study shows that falling can be detected early with a sensor and a simple activity by determining the right features and can be easily applied in the assessment of the elderly during routine follow-ups.

A validation study demonstrating portable motion capture cameras accurately characterize gait metrics when compared to a pressure-sensitive walkway

Digital quantification of gait can be used to measure aging- and disease-related decline in mobility. Gait performance also predicts prognosis, disease progression, and response to therapies. Most gait analysis systems require large amounts of space, resources, and expertise to implement and are not widely accessible. Thus, there is a need for a portable system that accurately characterizes gait. Here, depth video from two portable cameras accurately reconstructed gait metrics comparable to those reported by a pressure-sensitive walkway. 392 research participants walked across a four-meter pressure-sensitive walkway while depth video was recorded. Gait speed, cadence, and step and stride durations and lengths strongly correlated (r > 0.9) between modalities, with root-mean-squared-errors (RMSE) of 0.04 m/s, 2.3 steps/min, 0.03 s, and 0.05-0.08 m for speed, cadence, step/stride duration, and step/stride length, respectively. Step, stance, and double support durations (gait cycle percentage) significantly correlated (r > 0.6) between modalities, with 5% RMSE for step and stance and 10% RMSE for double support. In an exploratory analysis, gait speed from both modalities significantly related to healthy, mild, moderate, or severe categorizations of Charleson Comorbidity Indices (ANOVA, Tukey's HSD, p < 0.0125). These findings demonstrate the viability of using depth video to expand access to quantitative gait assessments.

Association between gait video information and general cardiovascular diseases: a prospective cross-sectional study

Aims

Cardiovascular disease (CVD) may not be detected in time with conventional clinical approaches. Abnormal gait patterns have been associated with pathological conditions and can be monitored continuously by gait video. We aim to test the association between non-contact, video-based gait information and general CVD status.

Methods and results

Individuals undergoing confirmatory CVD evaluation were included in a prospective, cross-sectional study. Gait videos were recorded with a Kinect camera. Gait features were extracted from gait videos to correlate with the composite and individual components of CVD, including coronary artery disease, peripheral artery disease, heart failure, and cerebrovascular events. The incremental value of incorporating gait information with traditional CVD clinical variables was also evaluated. Three hundred fifty-two participants were included in the final analysis [mean (standard deviation) age, 59.4 (9.8) years; 25.3% were female]. Compared with the baseline clinical variable model [area under the receiver operating curve (AUC) 0.717, (0.690-0.743)], the gait feature model demonstrated statistically better performance [AUC 0.753, (0.726-0.780)] in predicting the composite CVD, with further incremental value when incorporated with the clinical variables [AUC 0.764, (0.741-0.786)]. Notably, gait features exhibited varied association with different CVD component conditions, especially for peripheral artery disease [AUC 0.752, (0.728-0.775)] and heart failure [0.733, (0.707-0.758)]. Additional analyses also revealed association of gait information with CVD risk factors and the established CVD risk score.

Conclusion

We demonstrated the association and predictive value of non-contact, video-based gait information for general CVD status. Further studies for gait video-based daily living CVD monitoring are promising.

Walking on a Balance Beam as a New Measure of Dynamic Balance to Predict Falls in Older Adults and Patients with Neurological Conditions

BACKGROUND

Beam walking is a new test to estimate dynamic balance. We characterized dynamic balance measured by the distance walked on beams of different widths in five age groups of healthy adults (20, 30, 40, 50, 60 years) and individuals with neurological conditions (i.e., Parkinson, multiple sclerosis, stroke, age: 66.9 years) and determined if beam walking distance predicted prospective falls over 12 months.

METHODS

Individuals with (n = 97) and without neurological conditions (n = 99, healthy adults, age 20-60) participated in this prospective longitudinal study. Falls analyses over 12 months were conducted. The summed distance walked under single (walking only) and dual-task conditions (walking and serial subtraction by 7 between 300 to 900) on three beams (4, 8, and 12-cm wide) was used in the analyses. Additional functional tests comprised grip strength and the Short Physical Performance Battery.

RESULTS

Beam walking distance was unaffected on the 12-cm-wide beam in the healthy adult groups. The distance walked on the 8-cm-wide beam decreased by 0.34 m in the 20-year-old group. This reduction was ~ 3 × greater, 1.1 m, in the 60-year-old group. In patients, beam walking distances decreased sharply by 0.8 m on the 8 versus 12 cm beam and by additional 1.6 m on the 4 versus 8 cm beam. Beam walking distance under single and dual-task conditions was linearly but weakly associated with age (R2 = 0.21 for single task, R2 = 0.27 for dual-task). Age, disease, and beam width affected distance walked on the beam. Beam walking distance predicted future falls in the combined population of healthy adults and patients with neurological conditions. Based on receiver operating characteristic curve analyses using data from the entire study population, walking ~ 8.0 of the 12 m maximum on low-lying beams predicted future fallers with reasonable accuracy.

CONCLUSION

Balance beam walking is a new but worthwhile measure of dynamic balance to predict falls in the combined population of healthy adults and patients with neurological conditions. Future studies are needed to evaluate the predictive capability of beam walking separately in more homogenous populations. Clinical Trial Registration Number NCT03532984.

Towards a diagnostic tool for neurological gait disorders in childhood combining 3D gait kinematics and deep learning

Gait abnormalities are frequent in children and can be caused by different pathologies, such as cerebral palsy, neuromuscular disease, toe walker syndrome, etc. Analysis of the "gait pattern" (i.e., the way the person walks) using 3D analysis provides highly relevant clinical information. This information is used to guide therapeutic choices; however, it is underused in diagnostic processes, probably because of the lack of standardization of data collection methods. Therefore, 3D gait analysis is currently used as an assessment rather than a diagnostic tool. In this work, we aimed to determine if deep learning could be combined with 3D gait analysis data to diagnose gait disorders in children. We tested the diagnostic accuracy of deep learning methods combined with 3D gait analysis data from 371 children (148 with unilateral cerebral palsy, 60 with neuromuscular disease, 19 toe walkers, 60 with bilateral cerebral palsy, 25 stroke, and 59 typically developing children), with a total of 6400 gait cycles. We evaluated the accuracy, sensitivity, specificity, F1 score, Area Under the Curve (AUC) score, and confusion matrix of the predictions by ResNet, LSTM, and InceptionTime deep learning architectures for time series data. The deep learning-based models had good to excellent diagnostic accuracy (ranging from 0.77 to 0.99) for discrimination between healthy and pathological gait, discrimination between different etiologies of pathological gait (binary and multi-classification); and determining stroke onset time. LSTM performed best overall. This study revealed that the gait pattern contains specific, pathology-related information. These results open the way for an extension of 3D gait analysis from evaluation to diagnosis. Furthermore, the method we propose is a data-driven diagnostic model that can be trained and used without human intervention or expert knowledge. Furthermore, the method could be used to distinguish gait-related pathologies and their onset times beyond those studied in this research.

Adaptive Control Method for Gait Detection and Classification Devices with Inertial Measurement Unit

Cueing and feedback training can be effective in maintaining or improving gait in individuals with Parkinson's disease. We previously designed a rehabilitation assist device that can detect and classify a user's gait at only the swing phase of the gait cycle, for the ease of data processing. In this study, we analyzed the impact of various factors in a gait detection algorithm on the gait detection and classification rate (GDCR). We collected acceleration and angular velocity data from 25 participants (1 male and 24 females with an average age of 62 ± 6 years) using our device and analyzed the data using statistical methods. Based on these results, we developed an adaptive GDCR control algorithm using several equations and functions. We tested the algorithm under various virtual exercise scenarios using two control methods, based on acceleration and angular velocity, and found that the acceleration threshold was more effective in controlling the GDCR (average Spearman correlation -0.9996, p < 0.001) than the gyroscopic threshold. Our adaptive control algorithm was more effective in maintaining the target GDCR than the other algorithms (p < 0.001) with an average error of 0.10, while other tested methods showed average errors of 0.16 and 0.28. This algorithm has good scalability and can be adapted for future gait detection and classification applications.

Systematic Review on the Applicability of Principal Component Analysis for the Study of Movement in the Older Adult Population

Principal component analysis (PCA) is a dimensionality reduction method that has identified significant differences in older adults' motion analysis previously not detected by the discrete exploration of biomechanical variables. This systematic review aims to synthesize the current evidence regarding PCA use in the study of movement in older adults (kinematics and kinetics), summarizing the tasks and biomechanical variables studied. From the search results, 1685 studies were retrieved, and 19 studies were included for review. Most of the included studies evaluated gait or quiet standing. The main variables considered included spatiotemporal parameters, range of motion, and ground reaction forces. A limited number of studies analyzed other tasks. Further research should focus on the PCA application in tasks other than gait to understand older adults' movement characteristics that have not been identified by discrete analysis.

Computer Vision and Machine Learning-Based Gait Pattern Recognition for Flat Fall Prediction

BACKGROUND

Gait recognition has been applied in the prediction of the probability of elderly flat ground fall, functional evaluation during rehabilitation, and the training of patients with lower extremity motor dysfunction. Gait distinguishing between seemingly similar kinematic patterns associated with different pathological entities is a challenge for the clinician. How to realize automatic identification and judgment of abnormal gait is a significant challenge in clinical practice. The long-term goal of our study is to develop a gait recognition computer vision system using artificial intelligence (AI) and machine learning (ML) computing. This study aims to find an optimal ML algorithm using computer vision techniques and measure variables from lower limbs to classify gait patterns in healthy people. The purpose of this study is to determine the feasibility of computer vision and machine learning (ML) computing in discriminating different gait patterns associated with flat-ground falls.

METHODS

We used the Kinect® Motion system to capture the spatiotemporal gait data from seven healthy subjects in three walking trials, including normal gait, pelvic-obliquity-gait, and knee-hyperextension-gait walking. Four different classification methods including convolutional neural network (CNN), support vector machine (SVM), K-nearest neighbors (KNN), and long short-term memory (LSTM) neural networks were used to automatically classify three gait patterns. Overall, 750 sets of data were collected, and the dataset was divided into 80% for algorithm training and 20% for evaluation.

RESULTS

The SVM and KNN had a higher accuracy than CNN and LSTM. The SVM (94.9 ± 3.36%) had the highest accuracy in the classification of gait patterns, followed by KNN (94.0 ± 4.22%). The accuracy of CNN was 87.6 ± 7.50% and that of LSTM 83.6 ± 5.35%.

CONCLUSIONS

This study revealed that the proposed AI machine learning (ML) techniques can be used to design gait biometric systems and machine vision for gait pattern recognition. Potentially, this method can be used to remotely evaluate elderly patients and help clinicians make decisions regarding disposition, follow-up, and treatment.

Random forest algorithms to classify frailty and falling history in seniors using plantar pressure measurement insoles: a large-scale feasibility study

BACKGROUND

Frailty and falls are two adverse characteristics of aging that impair the quality of life of senior people and increase the burden on the healthcare system. Various methods exist to evaluate frailty, but none of them are considered the gold standard. Technological methods have also been proposed to assess the risk of falling in seniors. This study aims to propose an objective method for complementing existing methods used to identify the frail state and risk of falling in older adults.

METHOD

A total of 712 subjects (age: 71.3 ± 8.2 years, including 505 women and 207 men) were recruited from two Japanese cities. Two hundred and three people were classified as frail according to the Kihon Checklist. One hundred and forty-two people presented with a history of falling during the previous 12 months. The subjects performed a 45 s standing balance test and a 20 m round walking trial. The plantar pressure data were collected using a 7-sensor insole. One hundred and eighty-four data features were extracted. Automatic learning random forest algorithms were used to build the frailty and faller classifiers. The discrimination capabilities of the features in the classification models were explored.

RESULTS

The overall balanced accuracy for the recognition of frail subjects was 0.75 ± 0.04 (F1-score: 0.77 ± 0.03). One sub-analysis using data collected for men aged > 65 years only revealed accuracies as high as 0.78 ± 0.07 (F1-score: 0.79 ± 0.05). The overall balanced accuracy for classifying subjects with a recent history of falling was 0.57 ± 0.05 (F1-score: 0.62 ± 0.04). The classification of subjects relative to their frailty state primarily relied on features extracted from the plantar pressure series collected during the walking test.

CONCLUSION

In the future, plantar pressures measured with smart insoles inserted in the shoes of senior people may be used to evaluate aspects of frailty related to the physical dimension (e.g., gait and balance alterations), thus allowing assisting clinicians in the early identification of frail individuals.

GaitRec-Net: A Deep Neural Network for Gait Disorder Detection Using Ground Reaction Force

Walking (gait) irregularities and abnormalities are predictors and symptoms of disorder and disability. In the past, elaborate video (camera-based) systems, pressure mats, or a mix of the two has been used in clinical settings to monitor and evaluate gait. This article presents an artificial intelligence-based comprehensive investigation of ground reaction force (GRF) pattern to classify the healthy control and gait disorders using the large-scale ground reaction force. The used dataset comprised GRF measurements from different patients. The article includes machine learning- and deep learning-based models to classify healthy and gait disorder patients using ground reaction force. A deep learning-based architecture GaitRec-Net is proposed for this classification. The classification results were evaluated using various metrics, and each experiment was analysed using a fivefold cross-validation approach. Compared to machine learning classifiers, the proposed deep learning model is found better for feature extraction resulting in high accuracy of classification. As a result, the proposed framework presents a promising step in the direction of automatic categorization of abnormal gait pattern.

Stable Sparse Classifiers predict cognitive impairment from gait patterns

Background

Although gait patterns disturbances are known to be related to cognitive decline, there is no consensus on the possibility of predicting one from the other. It is necessary to find the optimal gait features, experimental protocols, and computational algorithms to achieve this purpose.

Purposes

To assess the efficacy of the Stable Sparse Classifiers procedure (SSC) for discriminating young and healthy older adults (YA vs. HE), as well as healthy and cognitively impaired elderly groups (HE vs. MCI-E) from their gait patterns. To identify the walking tasks or combinations of tasks and specific spatio-temporal gait features (STGF) that allow the best prediction with SSC.

Methods

A sample of 125 participants (40 young- and 85 older-adults) was studied. They underwent assessment with five neuropsychological tests that explore different cognitive domains. A summarized cognitive index (MDCog), based on the Mahalanobis distance from normative data, was calculated. The sample was divided into three groups (young adults, healthy and cognitively impaired elderly adults) using k-means clustering of MDCog in addition to Age. The participants executed four walking tasks (normal, fast, easy- and hard-dual tasks) and their gait patterns, measured with a body-fixed Inertial Measurement Unit, were used to calculate 16 STGF and dual-task costs. SSC was then employed to predict which group the participants belonged to. The classification's performance was assessed using the area under the receiver operating curves (AUC) and the stable biomarkers were identified.

Results

The discrimination HE vs. MCI-E revealed that the combination of the easy dual-task and the fast walking task had the best prediction performance (AUC = 0.86, sensitivity: 90.1%, specificity: 96.9%, accuracy: 95.8%). The features related to gait variability and to the amplitude of vertical acceleration had the largest predictive power. SSC prediction accuracy was better than the accuracies obtained with linear discriminant analysis and support vector machine classifiers.

Conclusions

The study corroborated that the changes in gait patterns can be used to discriminate between young and healthy older adults and more importantly between healthy and cognitively impaired adults. A subset of gait tasks and STGF optimal for achieving this goal with SSC were identified, with the latter method superior to other classification techniques.

A machine-vision-based frailty index for mice

Heterogeneity in biological aging manifests itself in health status and mortality. Frailty indices (FIs) capture health status in humans and model organisms. To accelerate our understanding of biological aging and carry out scalable interventional studies, high-throughput approaches are necessary. Here we introduce a machine-learning-based visual FI for mice that operates on video data from an open-field assay. We use machine vision to extract morphometric, gait and other behavioral features that correlate with FI score and age. We use these features to train a regression model that accurately predicts the normalized FI score within 0.04 ± 0.002 (mean absolute error). Unnormalized, this error is 1.08 ± 0.05, which is comparable to one FI item being mis-scored by 1 point or two FI items mis-scored by 0.5 points. This visual FI provides increased reproducibility and scalability that will enable large-scale mechanistic and interventional studies of aging in mice.

A Survey of Human Gait-Based Artificial Intelligence Applications

We performed an electronic database search of published works from 2012 to mid-2021 that focus on human gait studies and apply machine learning techniques. We identified six key applications of machine learning using gait data: 1) Gait analysis where analyzing techniques and certain biomechanical analysis factors are improved by utilizing artificial intelligence algorithms, 2) Health and Wellness, with applications in gait monitoring for abnormal gait detection, recognition of human activities, fall detection and sports performance, 3) Human Pose Tracking using one-person or multi-person tracking and localization systems such as OpenPose, Simultaneous Localization and Mapping (SLAM), etc., 4) Gait-based biometrics with applications in person identification, authentication, and re-identification as well as gender and age recognition 5) “Smart gait” applications ranging from smart socks, shoes, and other wearables to smart homes and smart retail stores that incorporate continuous monitoring and control systems and 6) Animation that reconstructs human motion utilizing gait data, simulation and machine learning techniques. Our goal is to provide a single broad-based survey of the applications of machine learning technology in gait analysis and identify future areas of potential study and growth. We discuss the machine learning techniques that have been used with a focus on the tasks they perform, the problems they attempt to solve, and the trade-offs they navigate.

Anticholinergic and Sedative Medications and Dynamic Gait Parameters in Older Patients

BACKGROUND

Anticholinergic and sedative medications are associated with poorer physical function in older age. Gait and physical function have traditionally been assessed with the time needed to execute objective function tests. Accelerometer-based gait parameters provide a precise capturing of gait dynamics and patterns and as such have added value.

OBJECTIVES

This study examined the associations between cumulative exposure to anticholinergic and sedative medications and gait dimensions as assessed with accelerometer-based dynamic gait parameters.

METHODS

Data were collected from outpatients of a diagnostic geriatric day clinic who underwent a comprehensive geriatric assessment (CGA). Cumulative exposure to anticholinergic and sedative medications was quantified with the Drug Burden Index (DBI), a linear additive pharmacological dose-response model. From a total of 22 dynamic gait parameters, the gait dimensions 'Regularity', 'Complexity', 'Stability', 'Pace', and 'Postural Control' were derived using factor analysis (and standardized total scores for these dimensions were calculated accordingly). Data were analyzed with multivariable linear regression analysis, in which adjustment was made for the covariates age, gender, body mass index (BMI), Mini Mental State Examination (MMSE) score, Charlson Comorbidity Index (CCI) including dementia, and number of medications not included in the DBI.

RESULTS

A total of 184 patients participated, whose mean age was 79.8 years (± SD 5.8), of whom 110 (60%) were women and of whom 88 (48%) had polypharmacy (i.e., received treatment with ≥5 medications). Of the 893 medications that were prescribed in total, 157 medications (17.6%) had anticholinergic and/or sedative properties. Of the patients, 100 (54%) had no exposure (DBI = 0), 42 (23%) had moderate exposure (0 > DBI ≤ 1), while another 42 (23%) had high exposure (DBI >1) to anticholinergic and sedative medications. Findings showed that high cumulative exposure to anticholinergic and sedative medications was related with poorer function on the Regularity and Pace dimensions. Furthermore, moderate and high exposure were associated with poorer function on the Complexity dimension.

CONCLUSIONS

These findings show that in older patients with comorbidities, cumulative anticholinergic and sedative exposure is associated with poorer function on multiple gait dimensions.

Wearable Inertial Gait Algorithms: Impact of Wear Location and Environment in Healthy and Parkinson's Populations

Wearable inertial measurement units (IMUs) are used in gait analysis due to their discrete wearable attachment and long data recording possibilities within indoor and outdoor environments. Previously, lower back and shin/shank-based IMU algorithms detecting initial and final contact events (ICs-FCs) were developed and validated on a limited number of healthy young adults (YA), reporting that both IMU wear locations are suitable to use during indoor and outdoor gait analysis. However, the impact of age (e.g., older adults, OA), pathology (e.g., Parkinson's Disease, PD) and/or environment (e.g., indoor vs. outdoor) on algorithm accuracy have not been fully investigated. Here, we examined IMU gait data from 128 participants (72-YA, 20-OA, and 36-PD) to thoroughly investigate the suitability of ICs-FCs detection algorithms (1 × lower back and 1 × shin/shank-based) for quantifying temporal gait characteristics depending on IMU wear location and walking environment. The level of agreement between algorithms was investigated for different cohorts and walking environments. Although mean temporal characteristics from both algorithms were significantly correlated for all groups and environments, subtle but characteristically nuanced differences were observed between cohorts and environments. The lowest absolute agreement level was observed in PD (ICC_2,1 = 0.979, 0.806, 0.730, 0.980) whereas highest in YA (ICC_2,1 = 0.987, 0.936, 0.909, 0.989) for mean stride, stance, swing, and step times, respectively. Absolute agreement during treadmill walking (ICC_2,1 = 0.975, 0.914, 0.684, 0.945), indoor walking (ICC_2,1 = 0.987, 0.936, 0.909, 0.989) and outdoor walking (ICC_2,1 = 0.998, 0.940, 0.856, 0.998) was found for mean stride, stance, swing, and step times, respectively. Findings of this study suggest that agreements between algorithms are sensitive to the target cohort and environment. Therefore, researchers/clinicians should be cautious while interpreting temporal parameters that are extracted from inertial sensors-based algorithms especially for those with a neurological condition.

Basic Spatiotemporal Gait Variables of Young and Older Healthy Volunteers Walking Along a Novel Figure-of-8 Path

Background: Locomotion along curved trajectories requires fine coordination among body segments. Elderly people may adopt a cautious attitude when steering. A simple, expeditious, patient-friendly walking protocol can be a tool to help clinicians. We evaluated the feasibility of a procedure based upon a newly designed Figure-of-eight (nFo8) path and an easy measurement operation. Methods: Sixty healthy volunteers, aged from 20 to 86 years, walked three times at self-selected speed along a 20 m linear (LIN) and the 20 m nFo8 path. Number of steps, mean speed and walk ratio (step length/cadence) were collected. Data were analysed for the entire cohort and for the groups aged 20-45, 46-65, and >65 years. Results: There was no difference in mean LIN walking speed between the two younger groups but the oldest was slower. During nFo8, all groups were slower (about 16%) than during LIN. Cadence was not different across groups but lower during nFo8 in each group. Step length was about 8% shorter in the two younger groups and 14% shorter in the oldest during nFo8 compared to LIN. Walk ratio was the smallest in the oldest group for both LIN and nFo8. Conclusions: A complex nFo8 walking path, with fast and easy measurement of a simple set of variables, detects significant differences with moderate and large effects in gait variables in people >65 years. This challenging trajectory is more revealing than LIN. Further studies are needed to develop a quick clinical tool for assessment of gait conditions or outcome of rehabilitative treatments.

Assessment Methods of Post-stroke Gait: A Scoping Review of Technology-Driven Approaches to Gait Characterization and Analysis

Background: Gait dysfunction or impairment is considered one of the most common and devastating physiological consequences of stroke, and achieving optimal gait is a key goal for stroke victims with gait disability along with their clinical teams. Many researchers have explored post stroke gait, including assessment tools and techniques, key gait parameters and significance on functional recovery, as well as data mining, modeling and analyses methods.

Research Question: This study aimed to review and summarize research efforts applicable to quantification and analyses of post-stroke gait with focus on recent technology-driven gait characterization and analysis approaches, including the integration of smart low cost wearables and Artificial Intelligence (AI), as well as feasibility and potential value in clinical settings.

Methods: A comprehensive literature search was conducted within Google Scholar, PubMed, and ScienceDirect using a set of keywords, including lower extremity, walking, post-stroke, and kinematics. Original articles that met the selection criteria were included.

Results and Significance: This scoping review aimed to shed light on tools and technologies employed in post stroke gait assessment toward bridging the existing gap between the research and clinical communities. Conventional qualitative gait analysis, typically used in clinics is mainly based on observational gait and is hence subjective and largely impacted by the observer's experience. Quantitative gait analysis, however, provides measured parameters, with good accuracy and repeatability for the diagnosis and comparative assessment throughout rehabilitation. Rapidly emerging smart wearable technology and AI, including Machine Learning, Support Vector Machine, and Neural Network approaches, are increasingly commanding greater attention in gait research. Although their use in clinical settings are not yet well leveraged, these tools promise a paradigm shift in stroke gait quantification, as they provide means for acquiring, storing and analyzing multifactorial complex gait data, while capturing its non-linear dynamic variability and offering the invaluable benefits of predictive analytics.

Detecting Sensitive Mobility Features for Parkinson's Disease Stages Via Machine Learning

BACKGROUND

It is not clear how specific gait measures reflect disease severity across the disease spectrum in Parkinson's disease (PD).

OBJECTIVE

To identify the gait and mobility measures that are most sensitive and reflective of PD motor stages and determine the optimal sensor location in each disease stage.

METHODS

Cross-sectional wearable-sensor records were collected in 332 patients with PD (Hoehn and Yahr scale I-III) and 100 age-matched healthy controls. Sensors were adhered to the participant's lower back, bilateral ankles, and wrists. Study participants walked in a ~15-meter corridor for 1 minute under two walking conditions: (1) preferred, usual walking speed and (2) walking while engaging in a cognitive task (dual-task). A subgroup (n = 303, 67% PD) also performed the Timed Up and Go test. Multiple machine-learning feature selection and classification algorithms were applied to discriminate between controls and PD and between the different PD severity stages.

RESULTS

High discriminatory values were found between motor disease stages with mean sensitivity in the range 72%-83%, specificity 69%-80%, and area under the curve (AUC) 0.76-0.90. Measures from upper-limb sensors best discriminated controls from early PD, turning measures obtained from the trunk sensor were prominent in mid-stage PD, and stride timing and regularity were discriminative in more advanced stages.

CONCLUSIONS

Applying machine-learning to multiple, wearable-derived features reveals that different measures of gait and mobility are associated with and discriminate distinct stages of PD. These disparate feature sets can augment the objective monitoring of disease progression and may be useful for cohort selection and power analyses in clinical trials of PD. © 2021 International Parkinson and Movement Disorder Society.

Detection of Postural Control in Young and Elderly Adults Using Deep and Machine Learning Methods with Joint-Node Plots

Postural control decreases with aging. Thus, an efficient and accurate method of detecting postural control is needed. We enrolled 35 elderly adults (aged 82.06 ± 8.74 years) and 20 healthy young adults (aged 21.60 ± 0.60 years) who performed standing tasks for 40 s, performed six times. The coordinates of 15 joint nodes were captured using a Kinect device (30 Hz). We plotted joint positions into a single 2D figure (named a joint–node plot, JNP) once per second for up to 40 s. A total of 15 methods combining deep and machine learning for postural control classification were investigated. The accuracy, sensitivity, specificity, positive predicted value (PPV), negative predicted value (NPV), and kappa values of the selected methods were assessed. The highest PPV, NPV, accuracy, sensitivity, specificity, and kappa values were higher than 0.9 in validation testing. The presented method using JNPs demonstrated strong performance in detecting the postural control ability of young and elderly adults.

Identification of Patients with Sarcopenia Using Gait Parameters Based on Inertial Sensors

Sarcopenia can cause various senile diseases and is a major factor associated with the quality of life in old age. To diagnose, assess, and monitor muscle loss in daily life, 10 sarcopenia and 10 normal subjects were selected using lean mass index and grip strength, and their gait signals obtained from inertial sensor-based gait devices were analyzed. Given that the inertial sensor can measure the acceleration and angular velocity, it is highly useful in the kinematic analysis of walking. This study detected spatial-temporal parameters used in clinical practice and descriptive statistical parameters for all seven gait phases for detailed analyses. To increase the accuracy of sarcopenia identification, we used Shapley Additive explanations to select important parameters that facilitated high classification accuracy. Support vector machines (SVM), random forest, and multilayer perceptron are classification methods that require traditional feature extraction, whereas deep learning methods use raw data as input to identify sarcopenia. As a result, the input that used the descriptive statistical parameters for the seven gait phases obtained higher accuracy. The knowledge-based gait parameter detection was more accurate in identifying sarcopenia than automatic feature selection using deep learning. The highest accuracy of 95% was achieved using an SVM model with 20 descriptive statistical parameters. Our results indicate that sarcopenia can be monitored with a wearable device in daily life.

Using Different Combinations of Body-Mounted IMU Sensors to Estimate Speed of Horses-A Machine Learning Approach

Speed is an essential parameter in biomechanical analysis and general locomotion research. It is possible to estimate the speed using global positioning systems (GPS) or inertial measurement units (IMUs). However, GPS requires a consistent signal connection to satellites, and errors accumulate during IMU signals integration. In an attempt to overcome these issues, we have investigated the possibility of estimating the horse speed by developing machine learning (ML) models using the signals from seven body-mounted IMUs. Since motion patterns extracted from IMU signals are different between breeds and gaits, we trained the models based on data from 40 Icelandic and Franches-Montagnes horses during walk, trot, tölt, pace, and canter. In addition, we studied the estimation accuracy between IMU locations on the body (sacrum, withers, head, and limbs). The models were evaluated per gait and were compared between ML algorithms and IMU location. The model yielded the highest estimation accuracy of speed (RMSE = 0.25 m/s) within equine and most of human speed estimation literature. In conclusion, highly accurate horse speed estimation models, independent of IMU(s) location on-body and gait, were developed using ML.

Classification of Neurological Patients to Identify Fallers Based on Spatial-Temporal Gait Characteristics Measured by a Wearable Device

Neurological patients can have severe gait impairments that contribute to fall risks. Predicting falls from gait abnormalities could aid clinicians and patients mitigate fall risk. The aim of this study was to predict fall status from spatial-temporal gait characteristics measured by a wearable device in a heterogeneous population of neurological patients. Participants (n = 384, age 49–80 s) were recruited from a neurology ward of a University hospital. They walked 20 m at a comfortable speed (single task: ST) and while performing a dual task with a motor component (DT1) and a dual task with a cognitive component (DT2). Twenty-seven spatial-temporal gait variables were measured with wearable sensors placed at the lower back and both ankles. Partial least square discriminant analysis (PLS-DA) was then applied to classify fallers and non-fallers. The PLS-DA classification model performed well for all three gait tasks (ST, DT1, and DT2) with an evaluation of classification performance Area under the receiver operating characteristic Curve (AUC) of 0.7, 0.6 and 0.7, respectively. Fallers differed from non-fallers in their specific gait patterns. Results from this study improve our understanding of how falls risk-related gait impairments in neurological patients could aid the design of tailored fall-prevention interventions.