Biomechanical Posture Analysis in Healthy Adults with Machine Learning: Applicability and Reliability

Area of Science:

• Biomedical Engineering
• Computer Science
• Human Movement Science

Background:

• Musculoskeletal disorder prevention is crucial but often relies on subjective posture assessments.
• Objective and reliable human posture analysis methods are needed.

Purpose of the Study:

• To evaluate the applicability and reliability of a machine learning (ML) pose estimation model for human posture assessment.
• To explore data structure using principal component and cluster analyses for novel postural insights.
• To identify gender-specific postural differences.

Main Methods:

• Collected frontal, dorsal, and lateral photographs of 200 healthy individuals.
• Utilized ML pose estimation for posture analysis.
• Applied Student's t-test, Cohen's d, and Intraclass Correlation Coefficient (ICC) for statistical analysis.
• Performed principal component and cluster analyses on postural data.

Main Results:

• Identified significant sex differences in shoulder adduction angle (men: 16.1° ± 1.9°, women: 14.1° ± 1.5°) and hip adduction angle (men: 9.9° ± 2.2°, women: 6.7° ± 1.5°).
• Found no significant differences in horizontal inclinations between sexes.
• Achieved a high ICC value of 0.95, confirming the method's reliability.
• Unsupervised ML revealed new patterns, including significant shoulder-hip distance variations.

Conclusions:

• ML pose estimation offers a reliable and objective method for human posture analysis.
• The approach is promising for non-invasive, rapid screening in physical therapy, ergonomics, and sports.
• Unsupervised ML techniques can uncover novel insights into postural variations.