Nonlinear stochastic system identification of skin using volterra kernels

Area of Science:

• Biomechanics
• System Identification
• Nonlinear Dynamics

Background:

• Biological tissues exhibit complex nonlinear mechanical properties.
• Accurate modeling of these nonlinearities is crucial for understanding tissue function and disease.
• Existing linear models often fail to capture the intricate dynamics of biological systems like skin.

Purpose of the Study:

• To apply Volterra kernel stochastic system identification for modeling nonlinear skin mechanics in vivo.
• To compare the efficacy of Volterra kernel methods against simple linear models.
• To investigate the correlation of identified kernel parameters with subject-specific factors.

Main Methods:

• Developed a high-bandwidth Lorentz force linear actuator system for in vivo skin testing.
• Utilized indentation and extension configurations with non-white input forces.
• Applied Volterra kernel solution methods, including fast least squares and orthogonalization.
• Incorporated frequency domain filtering for low-pass filtered inputs.

Main Results:

• Volterra kernel models achieved significantly higher variance accounted for (90-97%) compared to linear models (<75%).
• The second Volterra kernel correlated well with a dynamic-parameter nonlinearity model.
• Kernel peak values showed low coefficients of variation (3-8%) and correlations with subject demographics.

Conclusions:

• Volterra kernel stochastic system identification is a robust and efficient method for characterizing biological tissue nonlinearities.
• This technique shows promise for applications in skin disease diagnosis and consumer product efficacy assessment.
• The identified parameters provide insights into the depth-dependent mechanical properties of skin.