Predictive constitutive modelling of arteries by deep learning.

Area of Science:

• Biomechanics
• Materials Science
• Computational Biology

Background:

• Constitutive modeling of soft biological tissues is crucial for understanding tissue mechanics.
• Current models can describe arterial tissue properties but struggle to predict them from microstructural data.
• Bridging microstructural information with macroscopic mechanical behavior remains a significant challenge.

Purpose of the Study:

• To introduce a novel hybrid modeling framework integrating theoretical concepts and deep learning.
• To predict the constitutive properties of soft biological tissues from microstructural information.
• To demonstrate the efficacy of this framework in a proof-of-concept study.

Main Methods:

• Developed a hybrid modeling framework combining advanced theoretical concepts with deep learning.
• Utilized data from mechanical tests, histological analysis, and second-harmonic generation imaging.
• Trained the model with a dataset comprising 27 soft tissue samples.

Main Results:

• The hybrid model successfully predicted stress-stretch curves from microstructural data.
• Achieved a median coefficient of determination (R²) of 0.97 for predictions.
• Effective prediction was demonstrated within the range of commonly encountered mechanical properties.

Conclusions:

• Deep learning offers a transformative potential for modeling constitutive properties of soft biological tissues.
• The hybrid framework shows promise for accurately linking microstructural features to tissue mechanical behavior.
• This approach could significantly advance the field of soft tissue biomechanics.