Nonlinearity parameter imaging in the frequency domain

Barbara Kaltenbacher,William Rundell

2023-03-30

Abstract:Nonlinearity parameter tomography leads to the problem of identifying a coefficient in a nonlinear wave equation (such as the Westervelt equation) modeling ultrasound propagation. In this paper we transfer this into frequency domain, where the Westervelt equation gets replaced by a coupled system of Helmholtz equations with quadratic nonlinearities. For the case of the to-be-determined nonlinearity coefficient being a characteristic function of an unknown, not necessarily connected domain $D$, we devise and test a reconstruction algorithm based on weighted point source approximations combined with Newton's method. In a more abstract setting, convergence of a regularised Newton type method for this inverse problem is proven by verifying a range invariance condition of the forward operator and establishing injectivity of its linearisation.

Numerical Analysis

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to identify the nonlinear coefficients in ultrasonic propagation models (such as the Westervelt equation) through nonlinear parametric tomography in the frequency domain. Specifically, the researchers transform the nonlinear wave equation in the time domain to the frequency domain, thereby obtaining the Helmholtz equations with quadratic nonlinearity. For the case where the nonlinear coefficient to be determined is the characteristic function of the region D which is unknown and not necessarily connected, they design and test a reconstruction algorithm based on the weighted point - source approximation combined with the Newton method.

In a more abstract setting, the convergence of the regularized Newton - type method for such inverse problems is proved by verifying the range invariance condition of the forward operator and establishing the injectivity of its linearization.

### Formula Representation

1. **Westervelt Equation**:
\[
p_{tt}-c^{2}\Delta p - b\Delta p_{t}=\eta(p^{2})_{tt}+h\quad\text{in }(0,T)\times\Omega
\]
where \(p\) is the acoustic pressure, \(c\) is the speed of sound, \(b\) is the acoustic diffusivity, and \(h\) is the excitation term.

2. **Observation Data**:
\[
y(x,t) = p(x,t),\quad(x,t)\in\Sigma\times(0,T)
\]

3. **Helmholtz Equations in the Frequency Domain**:
\[
m = 1:-\omega^{2}\hat{p}_{1}-(c^{2}+i\omega b)\Delta\hat{p}_{1}=\hat{h}-\eta\sum_{k = 3:2}\frac{\hat{p}_{k - 1}}{2}-\frac{\hat{p}_{k+1}}{2}
\]
\[
m\in\{2,\dots,M\}:-\omega^{2}m^{2}\hat{p}_{m}-(c^{2}+i\omega mb)\Delta\hat{p}_{m}=-\eta\left(\frac{1}{4}\sum_{\ell = 1}^{m - 1}\hat{p}_{\ell}\hat{p}_{m-\ell}+\sum_{k = m+2:2}\frac{\hat{p}_{k - m}}{2}-\frac{\hat{p}_{k+m}}{2}\right)
\]

4. **Reconstruction Problem**:
\[
\Delta\hat{p}_{1}+\kappa^{2}\hat{p}_{1}=\hat{h},\quad\kappa=\omega\sqrt{c^{2}+i\omega b}
\]
\[
\Delta\hat{p}_{m}+m^{2}\kappa^{2}\hat{p}_{m}=\eta_{0}\chi_{D}\left(\frac{1}{4}\sum_{\ell = 1}^{m - 1}\hat{p}_{\ell}\hat{p}_{m-\ell}\right),\quad m\in\{2,\dots,M\}
\]

5. **Boundary Conditions**:
\[
\partial_{\nu}\hat{p}_{m}+(im\omega\beta+\gamma)\hat{p}_{m}=0\quad\text{on }\partial\Omega
\]

### Summary

This paper aims to accurately identify the nonlinear coefficients in ultrasonic propagation models through nonlinear parametric tomography in the frequency domain by solving the Helmholtz equations and their inverse problems. This not only helps to improve the quality of ultrasonic imaging but also provides a theoretical basis and technical support for the application of nonlinear ultrasound.