Estimating Muscle Parameters via Hierarchical Bayesian Neuromechanics

Area of Science:

• Biomechanics
• Computational Biology
• Human Movement Science

Background:

• Musculoskeletal models are crucial for understanding human movement.
• Current models often use generic parameters, limiting accuracy for individual subjects.
• Subject-specific parameters are vital for precise force generation estimation.

Purpose of the Study:

• To develop and validate a hierarchical Bayesian framework for estimating subject-specific musculoskeletal model parameters.
• To overcome limitations of generic parameters derived from cadaveric data.
• To leverage surface electromyography (EMG) and torque data for personalized biomechanical modeling.

Main Methods:

• A hierarchical Bayesian framework was implemented using surface EMG and torque data from isometric elbow tasks.
• Six key subject-specific muscle parameters were inferred, including muscle strength and moment arm geometry.
• The model was applied to 14 healthy adults performing isometric elbow flexion and extension.

Main Results:

• The hierarchical Bayesian model accurately reproduced measured elbow torque (R² = 0.96).
• Subject-specific muscle strength varied significantly (1.0-3.5), averaging twice that of generic models.
• Tendon slack length estimates showed minimal inter-individual variation, and bilateral testing indicated subject-specific parameter capture.

Conclusions:

• Hierarchical Bayesian inference provides a robust method for personalizing musculoskeletal models.
• This framework quantifies uncertainty and avoids invasive measurements, offering advantages over traditional methods.
• The approach is extensible to dynamic tasks and adaptable for clinical populations, advancing biomechanical understanding.