Related Experiment Video
Updated: Aug 5, 2026

3D Ultrasound Imaging: Fast and Cost-effective Morphometry of Musculoskeletal Tissue
Published on: November 27, 2017
Stress-Constrained Physics-Informed UNet for Voxel-Wise Multiparameter Hyperelastic Inversion in Volumetric Medical
Amirreza Asadi1, Kaveh Laksari2
1Department of Mechanical Engineering, Marlan and Rosemary Bourns College of Engineering, University of California, Riverside, CA, USA.
This study introduces a novel physics-informed UNet (PI-UNet) for detailed 3D soft tissue material property mapping. The framework accurately characterizes heterogeneous tissues using displacement data, advancing hyperelastic modeling.
Area of Science:
- * Computational mechanics and biomechanics.
- * Medical imaging and image analysis.
- * Machine learning for scientific applications.
Background:
- * Accurate characterization of soft tissue mechanics is crucial for understanding physiological processes and disease states.
- * Voxel-wise multiparameter hyperelastic modeling from experimental data remains a significant challenge.
- * Physics-informed machine learning offers a promising avenue for integrating physical laws into data-driven models.
Purpose of the Study:
- * To develop and validate a stress-constrained physics-informed UNet (PI-UNet) framework.
- * To enable voxel-wise, multiparameter hyperelastic characterization of heterogeneous soft tissues.
- * To utilize volumetric displacement data and boundary reaction information in controlled 3D synthetic benchmarks.
Main Methods:
- * A physics-informed UNet (PI-UNet) was developed to estimate voxel-wise Mooney-Rivlin parameter maps.
- * The framework processed multi-loading, strain-derived volumetric inputs from synthetic finite-element data.
- * Static equilibrium was enforced via stress divergence, incorporating boundary reaction data; noise robustness was assessed.
Main Results:
- * PI-UNet successfully reconstructed heterogeneous Mooney-Rivlin fields with high spatial fidelity across diverse synthetic benchmarks.
- * The framework differentiated material signatures in gray/white matter brain models and improved lesion identification in tumor models.
- * Moderate smoothing enhanced noise sensitivity while preserving critical geometric features.
Conclusions:
- * The stress-constrained PI-UNet offers a scalable computational framework for 3D hyperelastic inversion.
- * It effectively utilizes volumetric deformation and boundary reaction measurements for material characterization.
- * The approach supports future validation, advanced constitutive models, uncertainty quantification, and in vivo elastography.
Related Concept Videos
Elastic Strain Energy for Shearing Stresses
Conservation of Mass in Moving, Nondeforming Control Volume
In the context of a detention basin, the conservation of mass states that the total mass of water entering the basin must equal the mass leaving the basin plus any accumulation of...
Elastic Strain Energy for Normal Stresses
If...
Conservation of Mass in Fixed, Nondeforming Control Volume
In the case of a sewer pipe, which can be modeled...
Generalized Hooke's Law
Deformation of Member under Multiple Loadings
In the case of a member with a variable cross-section, the strain is not constant but depends on the position. The deformation of an...
