JMIR Publications

ABSTRACT

Background:

Recent advancements in rehabilitation sciences have progressively utilised computational techniques to improve diagnostic and treatment approaches. Notwithstanding these developments, the analysis of high-dimensional, time dependent data continues to pose a significant problem. Prior research has utilised clustering techniques on rehabilitation data to identify movement patterns and forecast recovery outcomes. Nonetheless, these initiatives have not yet employed force or motion data obtained outside of a clinical setting, constraining their capacity to guide therapeutic decisions. Biomechanical data analysis has demonstrated considerable potential in bridging these gaps and improving clinical decision-making in rehabilitation settings.

Objective:

This study presents a comprehensive clustering analysis of multi-dimensional movement datasets captured using a novel home exercise device, the “Slider®”. The aim is to identify clinically relevant movement patterns and provide answers to open research questions for the first time to inform personalized rehabilitation protocols, predict individual recovery trajectories, and assess the risks of potential post-operative complications.

Methods:

High-dimensional, time-dependent bilateral knee kinematics and force datasets were independently analyzed from 32 participants utilizing four unsupervised clustering techniques: K-means, hierarchical clustering, PAM (Partition Around Medoids), and CLARA (Clustering Large Applications). The data comprised kinetic variables including force, laser-measured distance, and optical tracker coordinates from lower limb activities. The optimal clusters identified through the unsupervised clustering methods were further evaluated and compared using silhouette analysis to quantify their performance, and key determinants of cluster membership were assessed, including demographic factors (e.g., gender, BMI, age) and pain levels, by employing a logistic regression model with Analysis of Covariance (ANCOVA) adjustment.

Results:

Three distinct time-varying movement patterns or clusters were identified for each knee. Hierarchical clustering performed best for the right knee datasets (with average silhouette score of 0.637), while CLARA was the most effective for the left knee datasets (with average silhouette score of 0.598). The clusters correlated with the mechanical and rotational dynamics of lower limb exercises. Key predictors of the time-dependent cluster membership were discovered for both knees. BMI was the most influential determinant of cluster membership for the right knee data (P<.001), whereas gender was found to be the most significant predictor for the left knee data (P<.001) based on the estimated odds ratios of all predictors.

Conclusions:

The discovered kinetic patterns offer significant insights for creating personalized rehabilitation procedures, potentially improving patient outcomes. These findings underscore the efficacy of unsupervised clustering techniques in the analysis of biomechanical data for clinical rehabilitation applications. Clinical Trial: Not applicable