Elasticity imaging using physics-informed neural networks
Science
Firstly, elasticity imaging using physics-informed neural networks (PINNs) is a relatively new and innovative approach to imaging. Combining the advantages of physics-based models with the flexibility and expressiveness of machine learning algorithms. Furthermore, the goal of elasticity imaging using PINNs is to estimate the mechanical properties of a material or tissue. Such as its stiffness or compliance, by analyzing the response of the material to external loads or perturbations.
In elasticity imaging using PINNs, the relationship between the mechanical properties of the material and the resulting deformations is modeled using a physics-based elasticity equation. Such as the linearized Navier equation or the nonlinear Mooney-Rivlin equation. The unknown material properties are represented by a neural network. Which is trained using data from measurements of the deformation fields. The training process involves minimizing a loss function that penalizes the deviation of the neural network predictions from the physics-based model and the measurement data.
By combining the expressiveness and flexibility of neural networks with the constraints imposed by the physics-based models. Elasticity imaging using PINNs provides a powerful tool for estimating the mechanical properties of materials and tissues with high accuracy and robustness.
Ali Kamali’s recent post on his work on the subject”
“Are you interested in scientific machine learning, deep learning, optimization, mechanical characterization of materials, or tissue biomechanics?
Then you might find our recent paper titled “Elasticity imaging using physics-informed neural networks: Spatial discovery of elastic modulus and Poisson’s ratio” published in Acta Biomaterialia an interesting read.
Elasticity imaging is a technique to estimate the mechanical properties of tissues using available measurements and has clinical and materials science applications. Our method estimates the distribution of elastic modulus and Poisson’s ratio in a domain from available deformation field and boundary condition data using physics-informed neural networks (PINN). Moreover, this implementation builds on most other existing methods which only estimate elastic modulus assuming a uniform Poisson’s ratio distribution in the domain. We show the reliability of this model in domains with simple and complex heterogeneities.
Check out the paper and the GitHub repository if you want to learn more!
Paper link: Elasticity imaging using physics-informed neural networks: Spatial discovery of elastic modulus and Poisson’s ratio – ScienceDirect
Gitub repository: GitHub – klaksari/2D_hidden_elasticity_PINN“
