Pathological gait clustering in post-stroke patients using motion capture data

Construction of prediction models for novel subtypes in patients with arteriosclerosis obliterans undergoing endovascular therapy: an unsupervised machine learning study

BACKGROUND

Arteriosclerosis obliterans (ASO) is a chronic arterial disease that can lead to critical limb ischemia. Endovascular therapy is increasingly used for limb salvage in ASO patients, but the outcomes vary. The development of prediction models using unsupervised machine learning may lead to the identification of novel subtypes to guide patient prognosis and treatment.

METHODS

This retrospective study analyzed clinical data from 448 patients with ASOs who underwent endovascular therapy. Unsupervised machine learning algorithms were employed to identify subgroups. To validate the precision of the clustering outcomes, an analysis of the postoperative results of the clusters was conducted. A prediction model was constructed using binary logistic regression.

RESULTS

Two distinct subgroups were identified by unsupervised machine learning and characterized by differing patterns of clinical features. Patients in Cluster 2 had significantly worse conditions and prognoses than those in Cluster 1. For the novel ASO subtypes, a nomogram was developed using six predictive factors, namely, platelet count, ankle brachial index, Rutherford category, operation method, hypertension, and diabetes status. The nomogram achieved excellent discrimination for predicting membership in the two identified clusters, with an area under the curve of 0.96 and 0.95 in training cohort and internal test cohort.

CONCLUSION

This study demonstrated that unsupervised machine learning can reveal novel phenotypic subgroups of patients with varying prognostic risk who underwent endovascular therapy. The prediction model developed could support clinical decision-making and risk counseling for this complex patient population. Further external validation is warranted to assess the generalizability of the findings.

Clinical phenotypes of patients with acute stroke: a secondary analysis

INTRODUCTION

Stroke is a leading cause of mortality worldwide and a major cause of disability having a high burden on patients, society, and caregiving systems. This study was conducted to investigate the presence of clusters of in-hospital patients with acute stroke based on demographic and clinical data. Cluster analysis reveals patterns in patient characteristics without requiring knowledge of a predefined patient category or assumptions about likely groupings within the data.

METHODS

We performed a secondary analysis of open-access anonymized data from patients with acute stroke admitted to a hospital between December 2019 to June 2021. In total, 216 patients (78; 36.1% men) were included in the analytical dataset with a mean (SD) age of 60.3 (14.4). Many demographic and clinical features were included in the analysis and the Barthel Index on discharge was used for comparing the functional recovery of the identified clusters.

RESULTS

Hierarchical clustering based on the principal components identified two clusters of 109 and 107 patients. The clusters were different in the Barthel Index scores on discharge with the mean (SD) of 39.3 (29.3) versus 62.6 (29.4); t (213.87) = -5.818, P <0.001, Cohen's d (95%CI) = -0.80 (-1.07, -0.52). A logistic model showed that age, systolic blood pressure, pulse rate, D-dimer blood level, low-density lipoprotein, hemoglobin, creatinine concentration, the National Institute of Health Stroke Scale value, and the Barthel Index scores on admission were significant predictors of cluster profiles (all P ≤0.029).

CONCLUSION

There are two clusters in hospitalized patients with acute stroke with significantly different functional recovery. This allows prognostic grouping of hospitalized acute stroke patients for prioritization of care or resource allocation. The clusters can be recognized using easily measured demographic and clinical features.

Machine learning applied to gait analysis data in cerebral palsy and stroke: A systematic review

BACKGROUND

Among neurological pathologies, cerebral palsy and stroke are the main contributors to walking disorders. Machine learning methods have been proposed in the recent literature to analyze gait data from these patients. However, machine learning methods still fail to translate effectively into clinical applications. This systematic review addressed the gaps hindering the use of machine learning data analysis in the clinical assessment of cerebral palsy and stroke patients.

RESEARCH QUESTION

What are the main challenges in transferring proposed machine learning methods to clinical applications?

METHODS

PubMed, Web of Science, Scopus, and IEEE databases were searched for relevant publications on machine learning methods applied to gait analysis data from stroke and cerebral palsy patients until February the 23rd, 2023. Information related to the suitability, feasibility, and reliability of the proposed methods for their effective translation to clinical use was extracted, and quality was assessed based on a set of predefined questions.

RESULTS

From 4120 resulting references, 63 met the inclusion criteria. Thirty-one studies used supervised, and 32 used unsupervised machine learning methods. Artificial neural networks and k-means clustering were the most used methods in each category. The lack of rationale for features and algorithm selection, the use of unrepresentative datasets, and the lack of clinical interpretability of the clustering outputs were the main factors hindering the clinical reliability and applicability of these methods.

SIGNIFICANCE

The literature offers numerous machine learning methods for clustering gait data from cerebral palsy and stroke patients. However, the clinical significance of the proposed methods is still lacking, limiting their translation to real-world applications. The design of future studies must take into account clinical question, dataset significance, feature and model selection, and interpretability of the results, given their criticality for clinical translation.

Validity and reliability of inertial measurement units for gait assessment within a post stroke population

BACKGROUND

The ability to objectively measure spatiotemporal metrics within individuals post-stroke is integral to plan appropriate intervention, track recovery, and ultimately improve efficacy of rehabilitation programs. Inertial measurement units (IMUs) provide a means to systematically collect gait-specific metrics that could not otherwise be obtained from clinical outcomes. However, the use of IMUs to measure spatiotemporal parameters in stroke survivors has yet to be validated. The purpose of this study is to determine the validity and reliability of IMU-recorded spatiotemporal gait metrics as compared to a motion capture camera system (MCCS) in individuals post-stroke.

METHODS

Participants (n = 23, M/F = 12/11, mean (SD) age = 50.2(11.1) spatiotemporal data were collected simultaneously from a MCCS and APDM Opal IMUs during a five-minute treadmill walking task at a self-selected speed. Criterion validity and test-retest reliability were assessed using Lin's concordance correlation coefficients (CCCs) and intraclass correlation coefficients (ICCs), respectively. Spatiotemporal values from MCCS and IMU were used to calculate gait asymmetry, and a t-test was used to assess the difference between asymmetry values.

RESULTS

There were fair-to-excellent agreement between IMU and MCCS of temporal parameters (CCC 0.56-0.98), excellent agreement of spatial parameters (CCC >0.90), and excellent test-retest reliability for all parameters (ICC >0.90).

CONCLUSIONS

Compared to motion capture, the APDM Opal IMUs produced accurate and reliable measures of spatiotemporal parameters. Findings support the use of IMUs to assess spatiotemporal parameters in individual's post-stroke.

Optimizing Rehabilitation Outcomes for Stroke Survivors: The Impact of Speed and Slope Adjustments in Anti-Gravity Treadmill Training

Background and Objectives: This study explored the efficacy of customized anti-gravity treadmill (AGT) training, with adjustments in speed and incline, on rehabilitation outcomes for stroke patients, focusing on knee extensor muscle strength, joint angle, balance ability, and activities of daily living (ADLs). Materials and Methods: In this study, 30 individuals diagnosed with a stroke were divided into three groups. Experimental group 1 (EG1) underwent training without changes to speed and incline, experimental group 2 (EG2) received training with an increased incline, and experimental group 3 (EG3) underwent training with increased speed. Initially, all participants received AGT training under uniform conditions for two weeks. Subsequently, from the third to the sixth week, each group underwent their specified training intervention. Evaluations were conducted before the intervention and six weeks post-intervention using a manual muscle strength tester for knee strength, TETRAX for balance ability, Dartfish software for analyzing knee angle, and the Korean version of the Modified Barthel Index (K-MBI) for assessing activities of daily living. Results: Within-group comparisons revealed that AGT training led to enhancements in muscle strength, balance ability, joint angle, and ADLs across all participant groups. Between-group analyses indicated that EG2, which underwent increased incline training, demonstrated significant improvements in muscle strength and balance ability over EG1. EG3 not only showed significant enhancements in muscle strength, joint angle, and ADLs when compared to EG1 but also surpassed EG2 in terms of knee strength improvement. Conclusions: In conclusion, the application of customized AGT training positively impacts the rehabilitation of stroke patients, underscoring the importance of selecting a treatment method tailored to the specific needs of each patient.

Construction of a novel lower-extremity peripheral artery disease subtype prediction model using unsupervised machine learning and neutrophil-related biomarkers

Lower-extremity peripheral artery disease (LE-PAD) is a prevalent circulatory disorder with risks of critical limb ischemia and amputation. This study aimed to develop a prediction model for a novel LE-PAD subtype to predict the severity of the disease and guide personalized interventions. Additionally, LE-PAD pathogenesis involves altered immune microenvironment, we examined the immune differences to elucidate LE-PAD pathogenesis. A total of 460 patients with LE-PAD were enrolled and clustered using unsupervised machine learning algorithms (UMLAs). Logistic regression analyses were performed to screen and identify predictive factors for the novel subtype of LE-PAD and a prediction model was built. We performed a comparative analysis regarding neutrophil levels in different subgroups of patients and an immune cell infiltration analysis to explore the associations between neutrophil levels and LE-PAD. Through hematoxylin and eosin (H&E) staining of lower-extremity arteries, neutrophil infiltration in patients with and without LE-PAD was compared. We found that UMLAs can helped in constructing a prediction model for patients with novel LE-PAD subtypes which enabled risk stratification for patients with LE-PAD using routinely available clinical data to assist clinical decision-making and improve personalized management for patients with LE-PAD. Additionally, the results indicated the critical role of neutrophil infiltration in LE-PAD pathogenesis.

Multi-Site Identification and Generalization of Clusters of Walking Behaviors in Individuals With Chronic Stroke and Neurotypical Controls

BACKGROUND

Walking patterns in stroke survivors are highly heterogeneous, which poses a challenge in systematizing treatment prescriptions for walking rehabilitation interventions.

OBJECTIVES

We used bilateral spatiotemporal and force data during walking to create a multi-site research sample to: (1) identify clusters of walking behaviors in people post-stroke and neurotypical controls and (2) determine the generalizability of these walking clusters across different research sites. We hypothesized that participants post-stroke will have different walking impairments resulting in different clusters of walking behaviors, which are also different from control participants.

METHODS

We gathered data from 81 post-stroke participants across 4 research sites and collected data from 31 control participants. Using sparse K-means clustering, we identified walking clusters based on 17 spatiotemporal and force variables. We analyzed the biomechanical features within each cluster to characterize cluster-specific walking behaviors. We also assessed the generalizability of the clusters using a leave-one-out approach.

RESULTS

We identified 4 stroke clusters: a fast and asymmetric cluster, a moderate speed and asymmetric cluster, a slow cluster with frontal plane force asymmetries, and a slow and symmetric cluster. We also identified a moderate speed and symmetric gait cluster composed of controls and participants post-stroke. The moderate speed and asymmetric stroke cluster did not generalize across sites.

CONCLUSIONS

Although post-stroke walking patterns are heterogenous, these patterns can be systematically classified into distinct clusters based on spatiotemporal and force data. Future interventions could target the key features that characterize each cluster to increase the efficacy of interventions to improve mobility in people post-stroke.

Multi-site identification and generalization of clusters of walking behaviors in individuals with chronic stroke and neurotypical controls

Background

Walking patterns in stroke survivors are highly heterogeneous, which poses a challenge in systematizing treatment prescriptions for walking rehabilitation interventions.

Objective

We used bilateral spatiotemporal and force data during walking to create a multi-site research sample to: 1) identify clusters of walking behaviors in people post-stroke and neurotypical controls, and 2) determine the generalizability of these walking clusters across different research sites. We hypothesized that participants post-stroke will have different walking impairments resulting in different clusters of walking behaviors, which are also different from control participants.

Methods

We gathered data from 81 post-stroke participants across four research sites and collected data from 31 control participants. Using sparse K-means clustering, we identified walking clusters based on 17 spatiotemporal and force variables. We analyzed the biomechanical features within each cluster to characterize cluster-specific walking behaviors. We also assessed the generalizability of the clusters using a leave-one-out approach.

Results

We identified four stroke clusters: a fast and asymmetric cluster, a moderate speed and asymmetric cluster, a slow cluster with frontal plane force asymmetries, and a slow and symmetric cluster. We also identified a moderate speed and symmetric gait cluster composed of controls and participants post-stroke. The moderate speed and asymmetric stroke cluster did not generalize across sites.

Conclusions

Although post-stroke walking patterns are heterogenous, these patterns can be systematically classified into distinct clusters based on spatiotemporal and force data. Future interventions could target the key features that characterize each cluster to increase the efficacy of interventions to improve mobility in people post-stroke.

Discovering individual-specific gait signatures from data-driven models of neuromechanical dynamics

Locomotion results from the interactions of highly nonlinear neural and biomechanical dynamics. Accordingly, understanding gait dynamics across behavioral conditions and individuals based on detailed modeling of the underlying neuromechanical system has proven difficult. Here, we develop a data-driven and generative modeling approach that recapitulates the dynamical features of gait behaviors to enable more holistic and interpretable characterizations and comparisons of gait dynamics. Specifically, gait dynamics of multiple individuals are predicted by a dynamical model that defines a common, low-dimensional, latent space to compare group and individual differences. We find that highly individualized dynamics-i.e., gait signatures-for healthy older adults and stroke survivors during treadmill walking are conserved across gait speed. Gait signatures further reveal individual differences in gait dynamics, even in individuals with similar functional deficits. Moreover, components of gait signatures can be biomechanically interpreted and manipulated to reveal their relationships to observed spatiotemporal joint coordination patterns. Lastly, the gait dynamics model can predict the time evolution of joint coordination based on an initial static posture. Our gait signatures framework thus provides a generalizable, holistic method for characterizing and predicting cyclic, dynamical motor behavior that may generalize across species, pathologies, and gait perturbations.