EMG-Driven Musculoskeletal Model Calibration With Wrapping Surface Personalization.

Area of Science:

• Biomechanics
• Musculoskeletal Modeling
• Computational Physiology

Background:

• Electromyography (EMG)-driven musculoskeletal models estimate muscle forces and joint moments.
• Model accuracy is sensitive to muscle wrapping surface geometry, which is often not personalized.
• Personalization is computationally expensive, limiting its application in EMG-driven model calibration.

Purpose of the Study:

• To develop a novel, computationally efficient method for personalizing cylindrical wrapping surfaces in OpenSim during EMG-driven model calibration.
• To reduce the computational cost associated with repeated OpenSim muscle analyses.
• To improve the accuracy of muscle force and joint moment predictions.

Main Methods:

• Developed a two-level polynomial surrogate model approach to approximate muscle-tendon lengths and moment arms.
• Outer-level models predict time-varying muscle properties based on joint angles.
• Inner-level models predict outer-level coefficients based on wrapping surface parameters, enabling efficient calibration.

Main Results:

• Personalization of wrapping surfaces significantly reduced lower extremity joint moment matching errors compared to generic surfaces.
• The proposed method (PGA) reduced peak hip joint contact force predictions by 47% bodyweight, aligning with in vivo data.
• Calibration strategies involving wrapping surface parameter optimization yielded the most accurate and physically realistic model predictions.

Conclusions:

• The novel method enables effective personalization of wrapping surfaces in EMG-driven musculoskeletal models.
• This approach enhances model accuracy by reducing joint moment matching errors and improving joint contact force predictions.
• The findings support the use of personalized wrapping surfaces for creating more realistic and reliable OpenSim models.