Clinical gait analysis using video-based pose estimation: Multiple perspectives, clinical populations, and measuring...

Area of Science:

• Biomechanics
• Rehabilitation Engineering
• Computer Vision

Background:

• Gait dysfunction significantly impacts independence and quality of life in clinical populations.
• Current gait analysis methods (e.g., motion capture) are costly and inaccessible.
• Accessible gait analysis is crucial for effective rehabilitation and treatment planning.

Purpose of the Study:

• To develop and validate a novel, low-cost workflow for quantitative gait analysis using video recordings.
• To demonstrate the clinical relevance and accessibility of this new gait analysis approach.

Main Methods:

• Utilized an open-source human pose estimation algorithm (OpenPose) for gait analysis from multi-perspective videos.
• Validated the workflow in adults without impairment, post-stroke, and Parkinson's disease patients.
• Compared results against gold-standard 3D motion capture for accuracy.

Main Results:

• Successfully performed quantitative gait analyses in diverse clinical populations using low-cost devices.
• Captured clinically relevant, condition-specific gait parameters.
• Demonstrated the ability to track within-participant gait changes for rehabilitation monitoring.

Conclusions:

• A novel, accessible workflow for quantitative gait analysis using videos has been developed and validated.
• This computer vision-based approach significantly enhances the accessibility of gait analysis across various clinical settings.
• The freely available workflow empowers gait assessment without requiring specialized expertise or expensive equipment.