Towards characterizing soft tissue viscoelasticity using ultrasound-based harmonic shear wave elastography and physics-informed neural networks

Toward the characterization of viscoelasticity of the soft tissue, which is an important biomarker, this study aims to investigate the effectiveness of an ultrasound-based novel approach called the Harmonic Shear Wave Elastography (HSWE) framework. We developed and validated a 3-D finite-element simulation framework to model tissue displacement under a multi-frequency HSWE setting. The HSWE results were compared with those from the widely used Pulsed Shear Wave Elastography (PSWE) method through both simulation and phantom experiments simulating breast tumor problems. We focused on analyzing frequency-dependent phase velocity maps within the 300–800 Hz range, examining the effects of inclusion size, stiffness, and viscosity. The HSWE framework demonstrated accurate measurements of group and phase velocities, with mean differences from PSWE of 5.47% and 1.04% in simulations, and 9.72% and 2.8% in phantom experiments. Notably, HSWE excelled in identifying the size and location of smaller inclusions and effectively captured variations in inclusion elasticity and viscosity. Additionally, we applied a physics-informed neural network approach to the HSWE problem as an inverse problem toward the estimation of viscoelasticity of soft tissue, and corresponding results will be presented in the talk, highlighting the potential of these methods in clinical applications.