Dylan Kobsar

Classification accuracy of a single tri-axial accelerometer for training background and experience level in runners.

Accelerometers are increasingly used tools for gait analysis, but there remains a lack of research on their application to running and their ability to classify running patterns. The purpose of this study was to conduct an exploratory examination into the capability of a tri-axial accelerometer to classify runners of different training backgrounds and experience levels, according to their 3-dimensional (3D) accelerometer data patterns. Training background was examined with 14 competitive soccer players and 12 experienced marathon runners, and experience level was examined with 16 first-time and the same 12 experienced marathon runners. Discrete variables were extracted from 3D accelerations during a short run using root mean square, wavelet transformation, and autocorrelation procedures. A principal component analysis (PCA) was conducted on all variables, including gait speed to account for covariance. Eight PCs were retained, explaining 88% of the variance in the data. A stepwise discriminant analysis of PCs was used to determine the binary classification accuracy for training background and experience level, with and without the PC of Speed. With Speed, the accelerometer correctly classified 96% of runners for both training background and experience level. Without Speed, the accelerometer correctly classified 85% of runners based on training background, but only 68% based on experience level. These findings suggest that the accelerometer is effective in classifying athletes of different training backgrounds, but is less effective for classifying runners of different experience levels where gait speed is the primary discriminator.

Gait Biomechanics and Patient-Reported Function as Predictors of Response to a Hip Strengthening Exercise Intervention in Patients with Knee Osteoarthritis

Objective Muscle strengthening exercises have been shown to improve pain and function in adults with mild-to-moderate knee osteoarthritis, but individual response rates can vary greatly. Predicting individuals who respond and those who do not is important in developing a more efficient and effective model of care for knee osteoarthritis (OA). Therefore, the purpose of this study was to use pre-intervention gait kinematics and patient-reported outcome measures to predict post-intervention response to a 6-week hip strengthening exercise intervention in patients with mild-to-moderate knee OA. Methods Thirty-nine patients with mild-to-moderate knee osteoarthritis completed a 6-week hip-strengthening program and were subgrouped as Non-Responders, Low-Responders, or High-Responders following the intervention based on their change in Knee injury Osteoarthritis Outcome Score (KOOS). Predictors of responder subgroups were retrospectively determined from baseline patient-reported outcome measures and kinematic gait parameters in a discriminant analysis of principal components. A 3–4 year follow-up on 16 of the patients with knee OA was also done to examine long-term changes in these parameters. Results A unique combination of patient-reported outcome measures and kinematic factors was able to successfully subgroup patients with knee osteoarthritis with a cross-validated classification accuracy of 85.4%. Lower patient-reported function in daily living (ADL) scores and hip frontal plane kinematics during the loading response were most important in classifying High-Responders from other sub-groups, while a combination of hip, knee, ankle kinematics were used to classify Non-Responders from Low-Responders. Conclusion Patient-reported outcome measures and objective biomechanical gait data can be an effective method of predicting individual treatment success to an exercise intervention. Measuring gait kinematics, along with patient-reported outcome measures in a clinical setting can be useful in helping make evidence-based decisions regarding optimal treatment for patients with knee OA.

Determination of patellofemoral pain sub-groups and development of a method for predicting treatment outcome using running gait kinematics

BACKGROUND Not all patients with patellofemoral pain exhibit successful outcomes following exercise therapy. Thus, the ability to identify patellofemoral pain subgroups related to treatment response is important for the development of optimal therapeutic strategies to improve rehabilitation outcomes. The purpose of this study was to use baseline running gait kinematic and clinical outcome variables to classify patellofemoral pain patients on treatment response retrospectively. METHODS Forty-one individuals with patellofemoral pain that underwent a 6-week exercise intervention program were sub-grouped as treatment Responders (n=28) and Non-responders (n=13) based on self-reported measures of pain and function. Baseline three-dimensional running kinematics, and self-reported measures underwent a linear discriminant analysis of the principal components of the variables to retrospectively classify participants based on treatment response. The significance of the discriminant function was verified with a Wilks lambda test (α=0.05). FINDINGS The model selected 2 gait principal components and had a 78.1% classification accuracy. Overall, Non-responders exhibited greater ankle dorsiflexion, knee abduction and hip flexion during the swing phase and greater ankle inversion during the stance phase, compared to Responders. INTERPRETATION This is the first study to investigate an objective method to use baseline kinematic and self-report outcome variables to classify on patellofemoral pain treatment outcome. This study represents a significant first step towards a method to help clinicians make evidence-informed decisions regarding optimal treatment strategies for patients with patellofemoral pain.

Reliability of gait analysis using wearable sensors in patients with knee osteoarthritis

The aim of this study was to determine the test-retest reliability of linear acceleration waveforms collected at the low back, thigh, shank, and foot during walking, in a cohort of knee osteoarthritis patients, by applying two separate sensor attitude correction methods (static attitude correction and dynamic attitude correction). Linear acceleration data were collected on the subjects׳ most affected limb during treadmill walking on two separate days. Results reveal all attitude corrected acceleration waveforms displayed high repeatability, with coefficient of multiple determination values ranging from 0.82 to 0.99. Overall, mediolateral accelerations and the thigh sensor demonstrated the lowest reliabilities, but interaction effects revealed only mediolateral accelerations at the thigh and foot sensors were different than other axes and sensor locations. Both attitude correction methods led to improved reliability of linear acceleration waveforms. These findings suggest that while all sensor locations and axes display acceptable reliability in a clinical knee osteoarthritis population, the back and shank locations, and the vertical and anteroposterior acceleration directions, are most reliable.

Accelerometer-Based Step Regularity Is Lower in Older Adults with Bilateral Knee Osteoarthritis

Purpose: To compare the regularity and symmetry of gait between a cohort of older adults with bilateral knee osteoarthritis (OA) and an age and sex-matched control group of older adults with healthy knees. Methods: Fifteen (8 females) older adults with knee OA (64.7 ± 6.7 years) and fifteen (8 females) pain-free controls (66.1 ± 10.0 years) completed a 9-min. walk at a self-selected, comfortable speed while wearing a single waist-mounted tri-axial accelerometer. The following gait parameters were compared between the two groups according to sex: mean step time, mean stride time, stride and step regularity (defined as the consistency of the stride-to-stride or step-to-step pattern) and the symmetry of gait (defined as the difference between step and stride regularity) as determined by an unbiased autocorrelation procedure that analyzed the pattern of acceleration in the vertical, mediolateral and anteroposterior directions. Results: Older adults with knee OA displayed significantly less step regularity in the vertical (p < 0.05) and anteroposterior (p < 0.05) directions than controls. Females with knee OA were also found to have significantly less mediolateral step regularity than female controls (p < 0.05), whereas no difference was found between males. Conclusion: The results showed that the regularity of the step pattern in individuals with bilateral knee OA was less consistent compared to similarly-aged older adults with healthy knees. The findings suggest that future studies should investigate the relationship between step regularity, sex and movement direction as well as the application of these methods to the clinical assessment of knee OA.

Biomechanical Features of Running Gait Data Associated with Iliotibial Band Syndrome: Discrete Variables Versus Principal Component Analysis

The features associated with temporal gait biomechanical data are complex and multivariate and it is therefore necessary to identify methods that reduce the difficulty underlying the interpretation and identification of differences between groups of interest. Discrete variables and principal component analysis (PCA) are feature extraction methods that have been widely used. However, a comprehensive understanding of the relationship between discrete variables and PCA features has never been completed. The objectives of this study were to (1) determine the relationships between the two feature methods and (2) compare the performance of each for the identification and discrimination of between-group differences for injured and non-injured subjects. Running gait kinematic data of 48 patients experiencing iliotibial band syndrome (ITBS) were compared to a group of 48 asymptomatic control subjects for transverse plane hip and ankle joint and frontal plane hip joint waveform data. Twenty-two discrete variables and three to four PCA features were extracted from each waveform and divided into three subgroups: magnitude features, difference operator features, and phase shift features. The following key results were obtained: (1) strong correlations were found between discrete variables; (2) the first PCA feature captured the magnitude information and thus showed strong correlation with the discrete variables in the magnitude group; (3) there was no consistent result that showed all discrete variables were found in the first few principal components; (4) the performance of the PCA features in identifying between-group differences decreased (reduced the effect size) as compared to using the discrete variables, but this does not necessarily result in a decrease in the performance of the PCA features to discriminate between ITBS and controls using a support vector machine classifier. These results suggest care must be taken when selecting features of gait waveforms for both identification and discrimination of between-group differences for injured and non-injured runners.

Relationship between lower limb muscle strength, self-reported pain and function, and frontal plane gait kinematics in knee osteoarthritis

BACKGROUND The relationship between muscle strength, gait biomechanics, and self-reported physical function and pain for patients with knee osteoarthritis is not well known. The objective of this study was to investigate these relationships in this population. METHODS Twenty-four patients with knee osteoarthritis and 24 healthy controls were recruited. Self-reported pain and function, lower-limb maximum isometric force, and frontal plane gait kinematics during treadmill walking were collected on all patients. Between-group differences were assessed for 1) muscle strength and 2) gait biomechanics. Linear regressions were computed within the knee osteoarthritis group to examine the effect of muscle strength on 1) self-reported pain and function, and 2) gait kinematics. FINDINGS Patients with knee osteoarthritis exhibited reduced hip external rotator, knee extensor, and ankle inversion muscle force output compared with healthy controls, as well as increased peak knee adduction angles (effect size=0.770; p=0.013). Hip abductor strength was a significant predictor of function, but not after controlling for covariates. Ankle inversion, hip abduction, and knee flexion strength were significant predictors of peak pelvic drop angle after controlling for covariates (34.4% unique variance explained). INTERPRETATION Patients with knee osteoarthritis exhibit deficits in muscle strength and while they play an important role in the self-reported function of patients with knee osteoarthritis, the effect of covariates such as sex, age, mass, and height was more important in this relationship. Similar relationships were observed from gait variables, except for peak pelvic drop, where hip, knee, and ankle strength remained important predictors of this variable after controlling for covariates.

Wearable Sensor Data to Track Subject-Specific Movement Patterns Related to Clinical Outcomes Using a Machine Learning Approach

Wearable sensors can provide detailed information on human movement but the clinical impact of this information remains limited. We propose a machine learning approach, using wearable sensor data, to identify subject-specific changes in gait patterns related to improvements in clinical outcomes. Eight patients with knee osteoarthritis (OA) completed two gait trials before and one following an exercise intervention. Wearable sensor data (e.g., 3-dimensional (3D) linear accelerations) were collected from a sensor located near the lower back, lateral thigh and lateral shank during level treadmill walking at a preferred speed. Wearable sensor data from the 2 pre-intervention gait trials were used to define each individual’s typical movement pattern using a one-class support vector machine (OCSVM). The percentage of strides defined as outliers, based on the pre-intervention gait data and the OCSVM, were used to define the overall change in an individual’s movement pattern. The correlation between the change in movement patterns following the intervention (i.e., percentage of outliers) and improvement in self-reported clinical outcomes (e.g., pain and function) was assessed using a Spearman rank correlation. The number of outliers observed post-intervention exhibited a large association (ρ = 0.78) with improvements in self-reported clinical outcomes. These findings demonstrate a proof-of-concept and a novel methodological approach for integrating machine learning and wearable sensor data. This approach provides an objective and evidence-informed way to understand clinically important changes in human movement patterns in response to exercise therapy.

Classifying running speed conditions using a single wearable sensor: Optimal segmentation and feature extraction methods

Accelerometers have been used to classify running patterns, but classification accuracy and computational load depends on signal segmentation and feature extraction. Stride-based segmentation relies on identifying gait events, a step avoided by using window-based segmentation. For each segment, discrete points can be extracted from the accelerometer signal, or advanced features can be computed. Therefore, the purpose of this study was to examine how different segmentation and feature extraction methods influence the accuracy and computational load of classifying running conditions. Forty-four runners ran at their preferred speed and 25% faster than preferred while an accelerometer at the lower back recorded 3D accelerations. Computational load was determined as the accelerometer signal was segmented into single and five strides, and corresponding small and large windows, with discrete points extracted from the single stride segments and advanced features computed from all four segment types. Each feature set was used to classify speed conditions and classification accuracy was recorded. Computational load and classification accuracy were compared across all feature sets using a repeated-measures MANOVA, with follow-up t-tests to compare feature type (discrete vs. advanced), segmentation method (stride- vs. window-based), and segment size (small vs. large), using a Bonferroni-adjusted α = 0.003. The five-stride (97.49 (±4.57)%) and large-window advanced (97.23 (±5.51)%) feature sets produced the greatest classification accuracy, but the large-window advanced feature set had a lower computational load (0.0041 (±0.0002)s) than the stride-based feature sets. Therefore, using a few advanced features and large overlapping window sizes yields the best performance of both classification accuracy and computational load.

Using wearable sensors to classify subject-specific running biomechanical gait patterns based on changes in environmental weather conditions

Running-related overuse injuries can result from a combination of various intrinsic (e.g., gait biomechanics) and extrinsic (e.g., running surface) risk factors. However, it is unknown how changes in environmental weather conditions affect running gait biomechanical patterns since these data cannot be collected in a laboratory setting. Therefore, the purpose of this study was to develop a classification model based on subject-specific changes in biomechanical running patterns across two different environmental weather conditions using data obtained from wearable sensors in real-world environments. Running gait data were recorded during winter and spring sessions, with recorded average air temperatures of -10° C and +6° C, respectively. Classification was performed based on measurements of pelvic drop, ground contact time, braking, vertical oscillation of pelvis, pelvic rotation, and cadence obtained from 66,370 strides (~11,000/runner) from a group of recreational runners. A non-linear and ensemble machine learning algorithm, random forest (RF), was used to classify and compute a heuristic for determining the importance of each variable in the prediction model. To validate the developed subject-specific model, two cross-validation methods (one-against-another and partitioning datasets) were used to obtain experimental mean classification accuracies of 87.18% and 95.42%, respectively, indicating an excellent discriminatory ability of the RF-based model. Additionally, the ranked order of variable importance differed across the individual runners. The results from the RF-based machine-learning algorithm demonstrates that processing gait biomechanical signals from a single wearable sensor can successfully detect changes to an individual’s running patterns based on data obtained in real-world environments.