Selecting Sensitive Parameter Subsets in Dynamical Models With Application to Biomechanical System Identification.

Area of Science:

• Biomechanics
• Computational Modeling
• Parameter Estimation

Background:

• Estimating numerous biomechanical system parameters with limited data can lead to poor parameter identifiability and wide confidence intervals.
• This challenge hinders the reliability of biomechanical model predictions.

Purpose of the Study:

• To develop and validate a novel method for selecting sensitive biomechanical model parameters for estimation.
• To improve parameter identifiability and reduce confidence intervals in biomechanical models.

Main Methods:

• The proposed method utilizes the Fisher Information Matrix (FIM) to identify parameters most sensitive to measurement outputs.
• A comparative analysis was conducted against the nonlinear least absolute shrinkage and selection operator (LASSO) method.
• The methods were applied to a head position-tracking task in ten human subjects.

Main Results:

• The FIM method reduced model complexity by selecting 5 out of 12 parameters for estimation.
• Parameter confidence intervals were significantly reduced by up to 89%.
• Goodness of fit (Variance Accounted For) was maintained at 82%, with a 164-fold reduction in computation time compared to LASSO.
• The FIM method identified parameters comparable to those identified by LASSO.

Conclusions:

• The FIM-based approach effectively identifies sensitive biomechanical parameters, enhancing model identifiability and efficiency.
• This method offers a computationally faster and more accurate alternative for parameter estimation in biomechanics.