A mobile tool for static standing posture assessment using accurate 3D body reconstruction.

Area of Science:

• Biomedical Engineering
• Computer Vision
• Human Biomechanics

Background:

• Accurate posture assessment is crucial for managing postural disorders.
• Current mobile apps use 2D imaging and markers, lacking a comprehensive body view.
• Existing methods are limited in providing detailed 3D postural analysis.

Purpose of the Study:

• To develop a novel mobile tool for 3D body reconstruction-based posture assessment.
• To enable accurate and comprehensive posture analysis using smartphone video.
• To provide a user-friendly and visualized tool for screening postural abnormalities.

Main Methods:

• Utilized smartphone video capture of a static person in an A-pose.
• Employed OpenPose and U2Net for 2D joint and contour extraction.
• Reconstructed 3D joints from 2D data and estimated camera poses.
• Optimized SMPL parameters using multi-view projection contour consistency for 3D body reconstruction.

Main Results:

• Achieved over 90% accuracy in identifying common postural abnormalities (e.g., uneven shoulders, forward head posture).
• Demonstrated close agreement between 3D reconstruction results and existing methods.
• Validated reliability using Cohen's d and Intraclass Correlation Coefficient (ICC).

Conclusions:

• 3D body reconstruction from smartphone video is an effective method for posture assessment.
• The proposed approach offers a user-friendly, visualized tool for abnormal posture screening.
• This technology has potential for widespread clinical and personal use in posture evaluation.