Addressing Occlusions and Pose Challenges in Clinical Gait Analysis: A Robust 3D-to-2D Motion Pipeline

Area of Science:

• Biomedical Engineering
• Computer Vision
• Clinical Biomechanics

Background:

• Cerebral palsy (CP) is a leading cause of childhood physical disability, impacting movement and posture.
• Clinical gait analysis (CGA) is crucial for CP diagnosis and management, often utilizing motion capture systems.
• Markerless motion capture offers advantages like patient comfort but faces accuracy challenges (pose ambiguity, occlusions).

Purpose of the Study:

• To develop and validate a robust workflow for markerless motion capture in clinical gait analysis (CGA).
• To enhance the accuracy and reliability of markerless systems for assessing movement patterns in children with cerebral palsy (CP).

Main Methods:

• Utilized machine learning and computer vision techniques for markerless motion capture.
• Implemented a workflow projecting 3D onto 2D points for subject tracking via distance metrics.
• Integrated multiple data modalities to improve system robustness and accuracy.

Main Results:

• The proposed workflow addresses challenges like pose ambiguities and occlusions in markerless motion capture.
• Enabled detailed comparisons between markerless and marker-based motion capture systems.
• Demonstrated improved reliability and accuracy of kinematic data in clinical gait analysis.

Conclusions:

• The developed workflow enhances the performance and applicability of markerless motion capture for CGA.
• Contributes to more accessible and effective clinical assessments for children with CP.
• Advances the use of markerless technology in pediatric movement analysis.