Mixture-Model Clustering of Pathological Gait Patterns.

Area of Science:

• Biomechanics
• Rehabilitation Science
• Data Science

Background:

• Pathological gait patterns after stroke significantly impact mobility and recovery.
• Quantifying gait performance comprehensively is crucial for effective rehabilitation.
• Existing gait analysis methods may not fully capture the complexity of post-stroke gait variations.

Purpose of the Study:

• To apply mixture-model clustering to spatiotemporal gait parameters for characterizing pathological gait.
• To develop a composite measure of overall gait performance in adults post-stroke.
• To identify distinct gait patterns and their relationship with clinical outcomes.

Main Methods:

• Gait data were collected from 68 adults post-stroke and 20 healthy adults using a GAITRite mat.
• Mixture-model clustering was employed to group participants based on spatiotemporal gait features (symmetry, speed, variability).
• Different mixture-model configurations were evaluated to determine the optimal clustering approach.

Main Results:

• A selected clustering model successfully categorized gait data into three clinically meaningful groups.
• Group 1 comprised healthy individuals with symmetric, fast, and low-variability gait; Group 2 showed intermediate parameters; Group 3 exhibited the worst gait condition.
• Individual indexed memberships to each group were generated as a composite measure sensitive to gait changes over time.

Conclusions:

• Mixture-model clustering effectively characterizes distinct pathological gait patterns in stroke survivors.
• The derived composite measure, based on indexed memberships, shows potential for monitoring gait performance and rehabilitation progress.
• This approach provides a nuanced understanding of gait recovery and can inform personalized rehabilitation strategies.