Discovering dispersion: How robust is automated model discovery for human myocardial tissue?

Area of Science:

• Computational mechanics
• Biomaterials science
• Tissue engineering

Background:

• Computational modeling is crucial for understanding biological tissues like the human myocardium.
• Traditional models often assume perfect fiber alignment, which contradicts experimental findings of local fiber dispersion.
• Fiber dispersion significantly impacts the mechanical behavior of tissues.

Purpose of the Study:

• To integrate the generalized structure tensor approach into automated material model discovery.
• To represent and quantify fiber orientation dispersion in biological tissues.
• To explore the robustness of discovered models under varying dispersion and noise levels.

Main Methods:

• Utilized the generalized structure tensor approach for probabilistic fiber orientation description.
• Applied automated material model discovery to human myocardium biaxial and triaxial shear data.
• Systematically varied directional dispersion and stress measurement noise to test model robustness.

Main Results:

• Moderate fiber direction dispersion and arbitrary sheet/normal dispersion improved model fit.
• Successfully recovered a previously proposed four-term model.
• The approach demonstrated robustness, identifying similar model terms even with 7% stress data noise.

Conclusions:

• Automated model discovery using generalized structure tensors is robust to noise.
• This method effectively captures microstructural uncertainty and heterogeneity in a physiologically relevant manner.
• The findings enhance the understanding of mechanical behavior in fiber-reinforced tissues.