Elasticity Imaging with Deep Learning
Deep learning was used as a surrogate route for elasticity imaging: instead of explicitly solving inverse elasticity problems from measured displacement fields, neural networks learned to infer clinically relevant mechanical information from physics-based simulation data.
The 2019 study used displacement fields from finite-element models as CNN inputs and trained classifiers to identify tumor elastic heterogeneity and nonlinear elastic response. The 2024 study extended this direction with a conditional generative adversarial network for quasi-static elastography, improving noisy shear-modulus reconstructions from synthetic displacement fields and showing how simulated data can support learning when clinical or experimental labels are limited.
Related publications:
- Patel, D., Tibrewala, R., Vega, A., Dong, L., Hugenberg, N., & Oberai, A. A. (2019). Circumventing the solution of inverse problems in mechanics through deep learning: Application to elasticity imaging. Computer Methods in Applied Mechanics and Engineering, 353, 448-466.
- Mei, Y., Deng, J., Zhao, D., Xiao, C., Wang, T., Dong, L., & Zhu, X. (2024). Toward improved accuracy in quasi-static elastography using deep learning. Computer Modeling in Engineering & Sciences, 139(1), 911-935.