AddBiomechanics: Automating model scaling, inverse kinematics, and inverse dynamics from human motion data through...

Area of Science:

• Biomechanics
• Human Motion Analysis
• Data Science

Background:

• Large-scale human motion biomechanics datasets are crucial for advancing understanding of movement, diseases, and assistive technologies.
• Current manual processing of motion capture data is time-consuming and expensive, hindering data collection and sharing.

Purpose of the Study:

• To develop and validate AddBiomechanics, an automated method for quantifying human movement dynamics from motion capture data.
• To standardize and reduce the effort required for processing biomechanical data.

Main Methods:

• Utilized linear methods and non-convex bilevel optimization to scale musculoskeletal models, register markers, and compute kinematics.
• Employed optimization to determine body segment masses and refine kinematics by minimizing residual forces against ground reaction data.

Main Results:

• Automated processing takes 3-30 minutes per subject, compared to ~1 day of manual work.
• Achieved close correspondence with expert-calculated values, with marker errors <2 cm and residual forces <2% of peak external force.
• Successfully reproduced kinematics and kinetics from synthetic walking data.

Conclusions:

• AddBiomechanics offers a fast, standardized, and cost-effective solution for human movement dynamics quantification.
• The open-source cloud service lowers barriers to high-quality biomechanical analysis, promoting data sharing and research.
• Widespread adoption of AddBiomechanics can accelerate breakthroughs in understanding human motion and related conditions.