Abstract:Three-dimensional variation in structural components or fiber alignments results in complex mechanical property distribution in tissues and biomaterials. In this paper, we use a physics-informed UNet-based neural network model (El-UNet) to discover the three-dimensional (3D) internal composition and space-dependent material properties of heterogeneous isotropic and transversely isotropic materials without a priori knowledge of the composition. We then show the capabilities of El-UNet by validating against data obtained from finite-element simulations of two soft tissues, namely, brain tissue and articular cartilage, under various loading conditions. We first simulated compressive loading of 3D brain tissue comprising of distinct white matter and gray matter mechanical properties undergoing small strains with isotropic linear elastic behavior, where El-UNet reached mean absolute relative errors under 1.5 % for elastic modulus and Poisson's ratio estimations across the 3D volume. We showed that the 3D solution achieved by El-UNet was superior to relative stiffness mapping by inverse of axial strain and two-dimensional plane stress/plane strain approximations. Additionally, we simulated a transversely isotropic articular cartilage with known fiber orientations undergoing compressive loading, and accurately estimated the spatial distribution of all five material parameters, with mean absolute relative errors under 5 %. Our work demonstrates the application of the computationally efficient physics-informed El-UNet in 3D elasticity imaging and provides methods for translation to experimental 3D characterization of soft tissues and other materials. The proposed El-UNet offers a powerful tool for both in vitro and ex vivo tissue analysis, with potential extensions to in vivo diagnostics. STATEMENT OF SIGNIFICANCE: Elasticity imaging is a technique that reconstructs mechanical properties of tissue using deformation and force measurements. Given the complexity of this reconstruction, most existing methods have mostly focused on 2D problems. Our work is the first implementation of physics-informed UNets to reconstruct three-dimensional material parameter distributions for isotropic and transversely isotropic linear elastic materials by having deformation and force measurements. We comprehensively validate our model using synthetic data generated using finite element models of biological tissues with high bio-fidelity-the brain and articular cartilage. Our method can be implemented in elasticity imaging scenarios for in vitro and ex vivo mechanical characterization of biomaterials and biological tissues, with potential extensions to in vivo diagnostics.

Learning hidden elasticity with deep neural networks

Physics‐Informed Deep‐Learning For Elasticity: Forward, Inverse, and Mixed Problems

Displacement-based Reconstruction of Elasticity Distribution with Deep Neural Network

Deep Manifold Computing and Visualization Using Elastic Locally Isometric Smoothness

Physics-informed UNets for Discovering Hidden Elasticity in Heterogeneous Materials

[Reconstruction of Elasticity Modulus Distribution Base on Semi-Supervised Neural Network].

Elasticity imaging using physics-informed neural networks: Spatial discovery of elastic modulus and Poisson's ratio

Combining physics-based modeling and deep learning for ultrasound elastography

Global Ultrasound Elastography Using Convolutional Neural Network

An ℋ∞ Approach for Elasticity Properties Reconstruction

Ultrasound Elasticity Imaging Using Physics-based Models And Learning-based Plug-And-Play Priors

Deep Learning for High Speed Optical Coherence Elastography

Reconstruction of elasticity: a stochastic model-based approach in ultrasound elastography

NN-EUCLID: Deep-learning hyperelasticity without stress data

Toward Improved Accuracy in Quasi-Static Elastography Using Deep Learning

An Unsupervised Approach to Ultrasound Elastography with End-to-end Strain Regularisation

Discovering 3D hidden elasticity in isotropic and transversely isotropic materials with physics-informed UNets

Integrating Learning-based Priors with Physics-based Models in Ultrasound Elasticity Reconstruction

Identifying heterogeneous micromechanical properties of biological tissues via physics-informed neural networks

Deep Convolutional Neural Networks Predict Elasticity Tensors and their Bounds in Homogenization

SELFNet: Denoising Shear Wave Elastography Using Spatial-temporal Fourier Feature Networks