Session: 06-02-01: Vibration and Acoustics in Biomedical Applications

Paper Number: 149908

149908 - Inversion in Shear Wave Elastography Using Traveling Wave Expansion 

              Shear wave elastography such as Magnetic Resonance Elastography (MRE) can measure viscoelastic mechanical parameters of soft tissues noninvasively. Based on shear wave propagation, estimation of the viscoelastic properties is equivalent to estimation of complex wavenumbers from the wave field. Numerous methods have been proposed. Typical algorithms include Direct Inversion (DI) using the Helmholtz equation, local frequency estimation (LFE), Multifrequency Dual Elasto-Visco inversion (MDEV), and Non-Linear Inversion (NLI) algorithms based on Finite Element (FE).

              Here, we proposed an inversion framework based on Traveling Wave Expansion (TWE). Wavefield of an isotropic medium generated using TWE was used for preparing training data for a DL based inversion algorithm. For anisotropic medium, the TWE model was derived for a nearly incompressible transversely isotropic (NITI) material. Simulation was shown to demonstrate the inversion performance of the TWE.

              To verify the TWE-based inversion framework, both isotropic and transversely isotropic phantoms were simulated. For isotropic medium, a cube with two circular inclusions were simulated. The two circular inclusions of 10 mm radius with complex shear moduli of 4 + i0.48 and 6 + i0.84 kPa. The complex shear modulus of background was 2 + i0.2 kPa. These values served as GT. A 28dB SNR of Gaussian noise was added to the wave images. For NITI material, a cube with a size of 505050mm³ was constructed. A total of 20 fiber directions were simulated. COMSOL (COMSOL, Stockholm, Sweden) was used for all simulation cases.

        Results of the estimated shear moduli showed TWENN could recover the inclusions (Fig. 1). Values of RMSE based on TWENN for background and the two inclusions are 0.60+0.10i kPa, 0.83+0.13i kPa, and 1.12+0.19i kPa, respectively. For NITI material, the 3 model parameters with different fiber directions can be estimated (Fig. 2). Although errors in some areas were observed to be larger than 5%, this is due to the existence of standing wave in the local wave field, and the simulation grid is not small enough. In addition, analytical solutions did include the noise reduction that could contribute to enlarged error. In this study, we presented a general framework using TWE for estimating biomechanical properties of soft tissues based on shear wave elastography. Both isotropic and anisotropic materials were simulated and tested. With complex-valued formulation, the TWE-based inversion showed robustness with accuracy for estimating viscoelastic properties. The TWENN showed promised in applying for a variety of cases with isotropy assumption. For fiber-reinforced material, especially transversely isotropic (TI) material, TWE-TI based on NITI model showed it can be used for a general type of cases with TI the assumption.  Future work include application of the algorithms for clinical studies.

Presenting Author: Yuan Feng Shanghai Jiao Tong University

Presenting Author Biography: Dr. Yuan Feng is an associate professor of biomedical engineering. Dr. Yuan Feng earned a B.S. degree in Thermo Energy and Power Engineering in 2006 and an M.S. degree in Mechatronics in 2008, both from Harbin Institute of Technology in China. In 2008, he earned a Ph.D. degree in Mechanical Engineering from Washington University in St. Louis with Drs. Philip Bayly and Guy Genin. His postdoctoral training was at Washington University School of Medicine with Dr. Yanle Hu in 2013, and at the University of Texas at Austin with Dr. Michael Sacks in 2014. He was a faculty member at Soochow University from late 2014 to 2017. Since 2018 he moved to Shanghai Jiao Tong University and joined the School of Biomedical Engineering. His research interests are brain biomechanics, MR elastography, and interventional MRI. He has authored or coauthored 41 referred papers, 1 book, 11 Chinese patents, and 4 US patents.

Authors:

Shengyuan Ma Shanghai Jiao Tong University
Guang-Zhong Yang Shanghai Jiao Tong University
Yuan Feng Shanghai Jiao Tong University