Advances in Soft Matter Rheology

Researchers have introduced a novel method using Physics-Informed Neural Networks (PINNs) to model elastoviscoplastic (EVP) materials. These materials, which include biological tissues and various pharmaceutical gels, typically present significant challenges for predictive modeling. Specifically, their nonlinear behavior under large amplitude oscillatory shear (LAOS) conditions often results in complex datasets. Traditional models frequently struggle with computational intensity and noisy experimental measurements. Consequently, this new PINN-based approach offers a robust alternative by embedding physical laws directly into the machine learning framework.

The Role of Physics-Informed Neural Networks in Advanced Rheology

The study re-evaluates EVP modeling by integrating a modified Saramito model into the training process. By doing so, the Physics-Informed Neural Networks can fit time-dependent stress data without requiring manual gradient estimation. This differentiable framework significantly enhances data efficiency and generalizability. Furthermore, the researchers introduced a shear-thinning formulation to better reflect decreasing viscosity at high strain amplitudes. As a result, the model provides a more accurate representation of how materials yield and transition between viscoelastic states.

Clinical and Diagnostic Implications

Understanding the rheology of soft matter is essential for several medical applications. For instance, magnetic resonance elastography (MRE) relies on precise modeling of tissue stiffness to diagnose conditions like liver fibrosis or identify malignant tumors. This framework bridges microstructural deformation modes with macro-level rheological modeling. Therefore, it provides a powerful tool for predicting nonlinear viscoelastic behavior in biological systems. Stable recovery of physical parameters suggests that PINNs could soon enhance the interpretability of diagnostic imaging and personalized tissue characterization.

Frequently Asked Questions

What are Physics-Informed Neural Networks?

Physics-Informed Neural Networks are a class of deep learning models that incorporate physical laws, such as partial differential equations, into their loss functions to ensure physically consistent predictions.

How do PINNs improve upon traditional rheological fitting?

Unlike conventional methods, PINNs are differentiable and highly resistant to noise. They eliminate the need for complex gradient estimations, making them more efficient for fitting experimental data from soft matter.

Are these models applicable to medical diagnostics?

Yes. By accurately modeling the elastoviscoplastic properties of tissues, these models can improve the accuracy of elastography and other biomechanical assessments used in oncology and hepatology.

Disclaimer: This content is for informational and educational purposes only. It does not constitute medical advice or a professional relationship. Refer to the latest local and national guidelines for clinical practice.

References

Valipour Goodarzi B et al. Modeling elastoviscoplastic materials using physics-informed neural networks. Soft Matter. 2026 May 26. doi: 10.1039/d6sm00198j. PMID: 42189592.

Ragoza M, Batmanghelich K. Physics-Informed Neural Networks for Tissue Elasticity Reconstruction in Magnetic Resonance Elastography. Med Image Comput Comput Assist Interv. 2023 Oct. doi: 10.1007/978-3-031-43999-5_32.

Michaloglou A et al. Physics-Informed Neural Networks in Materials Modeling and Design: A Review. ResearchGate. 2025 Sep.