Bayesian Inference of Vocal Fold Material Properties from Glottal Area Waveforms Using a 2D Finite Element Model

Area of Science:

• Biomechanics of the vocal folds
• Computational modeling in acoustics
• Biomaterial property estimation

Background:

• Previous Bayesian estimation methods for vocal fold models relied on lumped-element fitting and synthetic data.
• Extracting physical tissue properties from lumped-element parameters is challenging due to indirect relationships.
• Need for subject-specific vocal fold models with direct estimation of material properties.

Purpose of the Study:

• To propose and validate a finite element (FE) fitting model for vocal fold analysis.
• To directly estimate material properties (elastic moduli, density) of vocal folds using Bayesian inference.
• To assess the model's sensitivity to experimental variations and its ability to extract unmeasured parameters like contact pressures.

Main Methods:

• Developed a finite element model representing vocal folds as a three-layered viscoelastic body.
• Utilized glottal area waveforms from self-oscillating silicone vocal folds as observational data.
• Employed Bayesian importance sampling for direct estimation of material properties and contact pressures.

Main Results:

• Estimated material properties of silicone vocal folds agreed with experimental ground truth values within 3%.
• The FE model coupled with Bayesian estimation demonstrated sensitivity to variations in subglottal pressure and medial compression.
• Contact pressures were successfully extracted, showing an expected increase with higher subglottal pressure and medial compression.

Conclusions:

• The proposed finite element fitting model offers a robust method for directly estimating vocal fold material properties.
• This approach overcomes limitations of previous lumped-element models and synthetic data requirements.
• The model provides valuable insights into vocal fold biomechanics, including contact pressures, under different physiological conditions.