A Novel Approach to Movement Profiling: Multi-Task Classification for Enhanced Orthopedics Assessment

Human movement analysis is integral for diagnosing, managing, and treating orthopedic pathologies. Traditional marker-based Motion Capture (MoCap) systems have long been considered the gold standard for capturing kinematic data. However, they face several limitations in clinical scalability due to cost, complexity of setup procedures, and restricted ecological validity. These challenges have driven interest in markerless systems, which leverage advancements in computer vision to provide a scalable and cost-effective alternative. Despite their potential, they require rigorous validation to ensure their accuracy and reliability for clinical use. This study investigates the use of clustering techniques to identify movement-based subgroups in patients while evaluating the consistency between systems. To achieve this, marker-based and markerless kinematic data were simultaneously collected from participants during a series of functional tasks, chosen for their clinical relevance in assessing lower-limb biomechanics. Preprocessing steps, including noise filtering, temporal resampling, and gap-filling, were applied to the raw data to ensure consistency and comparability across systems. The resulting standardized datasets formed the basis for an analysis pipeline designed to extract meaningful insights from the data, which consisted of four key stages: 1. Dynamic Time Warping (DTW) was employed to quantify temporal alignment between marker-based and markerless MoCap systems; 2. Biomechanical features were selected based on thresholds derived from system agreement results and their clinical relevance; 3. K-means clustering was applied to group participants based on their task-specific kinematic profiles, revealing patterns within individual tasks; 4. Hierarchical clustering was subsequently used to identify broader movement subgroups across multiple tasks. Preliminary results showed a high level of agreement between marker-based and markerless systems for kinematic measurements, especially in the frontal and sagittal planes. In contrast, the reliability of the transverse plane was notably lower, highlighting an area for improvement in markerless systems. Even if markerless MoCap systems have the potential to revolutionize orthopedic diagnostics, further research is needed to address the identified limitations. The clustering analysis was conducted using two methods: one based on relevant data points from the kinematic curves, and the other considering the whole time series. The results revealed distinct subgroups within the cohorts, underscoring the presence of biomechanical heterogeneity and providing valuable insights into movement patterns that could inform personalized rehabilitation strategies. The validation of those findings will be essential to ensure their generalizability and clinical utility. In conclusion, this study demonstrates the feasibility of integrating markerless MoCap systems with clustering techniques to complement traditional methods in orthopedic diagnostics. By providing a scalable and personalized approach to movement analysis, these systems have the potential to enhance clinical decision-making and improve patient outcomes.

Author dc:creator

• Eleonora Cabai (22481647)

• Computer Science
• Engineering, Biomedical

Statement dc:rights

• In Copyright

DOI dc:identifier

https://doi.org/10.25417/uic.30424918.v1

OAI identifier oai:identifier

oai:figshare.com:article/30424918