-
Full Field Inversion of the Attenuated Wave Equation: Theory and Numerical Inversion
Authors:
Ngoc Do,
Markus Haltmeier,
Richard Kowar,
Linh V. Nguyen,
Robert Nuster
Abstract:
Standard photoacoustic tomography (PAT) provides data that consist of time-dependent signals governed by the wave equation, which are measured on an observation surface. In contrast, the measured data from the recently invented full-field PAT is the Radon transform of the solution of the wave equation on a spatial domain at a single instant in time. While reconstruction using classical PAT data ha…
▽ More
Standard photoacoustic tomography (PAT) provides data that consist of time-dependent signals governed by the wave equation, which are measured on an observation surface. In contrast, the measured data from the recently invented full-field PAT is the Radon transform of the solution of the wave equation on a spatial domain at a single instant in time. While reconstruction using classical PAT data has been extensively studied, not much is known about the full-field PAT problem. In this paper, we study full-field photoacoustic tomography with spatially variable sound speed and spatially variable dam**. In particular, we prove the uniqueness and stability of the associated single-time full-field wave inversion problem and develop algorithms for its numerical inversion using iterative and variational regularization methods. Numerical simulations are presented for both full-angle and limited-angle data cases
△ Less
Submitted 7 June, 2024; v1 submitted 28 May, 2024;
originally announced May 2024.
-
Data-driven Morozov regularization of inverse problems
Authors:
Markus Haltmeier,
Richard Kowar,
Markus Tiefentaler
Abstract:
The solution of inverse problems is central to a wide range of applications including medicine, biology, and engineering. These problems require finding a desired solution in the presence of noisy observations. A key feature of inverse problems is their ill-posedness, which leads to unstable behavior under noise when standard solution methods are used. For this reason, regularization methods have…
▽ More
The solution of inverse problems is central to a wide range of applications including medicine, biology, and engineering. These problems require finding a desired solution in the presence of noisy observations. A key feature of inverse problems is their ill-posedness, which leads to unstable behavior under noise when standard solution methods are used. For this reason, regularization methods have been developed that compromise between data fitting and prior structure. Recently, data-driven variational regularization methods have been introduced, where the prior in the form of a regularizer is derived from provided ground truth data. However, these methods have mainly been analyzed for Tikhonov regularization, referred to as Network Tikhonov Regularization (NETT). In this paper, we propose and analyze Morozov regularization in combination with a learned regularizer. The regularizers, which can be adapted to the training data, are defined by neural networks and are therefore non-convex. We give a convergence analysis in the non-convex setting allowing noise-dependent regularizers, and propose a possible training strategy. We present numerical results for attenuation correction in the context of photoacoustic tomography.
△ Less
Submitted 22 October, 2023;
originally announced October 2023.
-
Analysis for Full Field Photoacoustic Tomography with Variable Sound Speed
Authors:
Linh V. Nguyen,
Markus Haltmeier,
Richard Kowar,
Ngoc Do
Abstract:
Photoacoustic tomography (PAT) is a non-invasive imaging modality that requires recovering the initial data of the wave equation from certain measurements of the solution outside the object. In the standard PAT measurement setup, the used data consist of time-dependent signals measured on an observation surface. In contrast, the measured data from the recently invented full-field detection techniq…
▽ More
Photoacoustic tomography (PAT) is a non-invasive imaging modality that requires recovering the initial data of the wave equation from certain measurements of the solution outside the object. In the standard PAT measurement setup, the used data consist of time-dependent signals measured on an observation surface. In contrast, the measured data from the recently invented full-field detection technique provide the solution of the wave equation on a spatial domain at a single instant in time. While reconstruction using classical PAT data has been extensively studied, not much is known for the full field PAT problem. In this paper, we build mathematical foundations of the latter problem for variable sound speed and settle its uniqueness and stability. Moreover, we introduce an exact inversion method using time-reversal and study its convergence. Our results demonstrate the suitability of both the full field approach and the proposed time-reversal technique for high resolution photoacoustic imaging.
△ Less
Submitted 4 December, 2021;
originally announced December 2021.
-
Extension of the principle of least action with focus on dissipative equations
Authors:
Richard Kowar
Abstract:
In this paper, we extend the \emph{principle of least action} and show that a \emph{Lagrange density} always exists for the usual linear pde or linear fractional problems $\oA\,u=f$ in physics, if the usual causality conditions $u|_{t<0}=0$ and $f|_{t<0}=0$ are assumed. (The approach is actually applicable to uniquely solvable linear operator equations for which an adjoint exist.) The set of Lagra…
▽ More
In this paper, we extend the \emph{principle of least action} and show that a \emph{Lagrange density} always exists for the usual linear pde or linear fractional problems $\oA\,u=f$ in physics, if the usual causality conditions $u|_{t<0}=0$ and $f|_{t<0}=0$ are assumed. (The approach is actually applicable to uniquely solvable linear operator equations for which an adjoint exist.) The set of Lagrange densities together with the zero vector form a non-trivial vector space and for each different set of variables, e.g. $\{u_t,f\}$, $\{u_{xt},f\}$ or $\{u_t,u_x,u_y,u_z,f\}$, there exists a Lagrange density that implies a Lagrange equation, which is equivalent to the considered problem. The usual Lagrange density is such that it implies the 'original equation'. But there are pde's for which the standard theory does not imply a Lagrange density. We show that for each of these equations a (covariant) Lagrange density exists that leads to an equivalent \emph{higher order pde} (if it is formulated with the above causality conditions). For each of these equations, there exists a Lagrange density that implies a Lagrange equation that equals the original equation, but this Lagrange density contains at least one \emph{linear integral operator}. A new point of view is that each of these equivalent Lagrange densities for a given set of variables implies a (usually different) \emph{generalized Hamiltonian density}, where the respective 'Hamiltonian' is conserved if $\oA$ and $f$ are appropriate. The standard Lagrange density implies an Hamiltonian that (frequently) models the energy. Morever, each conserved Hamiltonian implies countable many higher order Hamiltonians that are conserved (if the solution of the considered problem is sufficiently smooth.)
△ Less
Submitted 10 December, 2020;
originally announced December 2020.
-
Regularization of systems of nonlinear ill-posed equations: II. Applications
Authors:
M. Haltmeier,
R. Kowar,
A. Leitao,
O. Scherzer
Abstract:
In part I we introduced modified Landweber-Kaczmarz methods and have established a convergence analysis. In the present work we investigate three applications: an inverse problem related to thermoacoustic tomography, a nonlinear inverse problem for semiconductor equations, and a nonlinear problem in Schlieren tomography. Each application is considered in the framework established in the previous p…
▽ More
In part I we introduced modified Landweber-Kaczmarz methods and have established a convergence analysis. In the present work we investigate three applications: an inverse problem related to thermoacoustic tomography, a nonlinear inverse problem for semiconductor equations, and a nonlinear problem in Schlieren tomography. Each application is considered in the framework established in the previous part. The novel algorithms show robustness, stability, computational efficiency and high accuracy.
△ Less
Submitted 19 November, 2020;
originally announced November 2020.
-
A representation formula for solutions of second order ode's with time dependent coefficients and its application to model dissipative oscillations and waves
Authors:
Richard Kowar
Abstract:
In this paper, we model, classify and investigate the solutions of (normalized) second order ode's with \emph{nonconstant continuous coefficients}. We introduce a generalized \emph{frequency function} as the solution of a \emph{nonlinear integro-differential equation}, show its existence and then derive a representation formula for (all) solutions of (normalized) second order ode's with \emph{nonc…
▽ More
In this paper, we model, classify and investigate the solutions of (normalized) second order ode's with \emph{nonconstant continuous coefficients}. We introduce a generalized \emph{frequency function} as the solution of a \emph{nonlinear integro-differential equation}, show its existence and then derive a representation formula for (all) solutions of (normalized) second order ode's with \emph{nonconstant continuous coefficients}. Because this formula specifies the interplay between the coefficients of the ode, the \emph{relaxation function} ("strongly" decreasing positive function) and the frequency function of the oscillation, it can be applied to design models of dissipative oscillations. As an application, we present and discuss some oscillation models that stop within a finite time period. Moreover, we demonstrate that a large class of oscillations can be used to design and analyze dissipative waves. In particular, it is easy to model dissipative waves that cause in each point of space an oscillation that stops after a finite time period.
△ Less
Submitted 4 July, 2019; v1 submitted 15 March, 2019;
originally announced March 2019.
-
Dynamical systems with finite stop** times. Part 1: Relaxation, oscillation and their application to diffusion and wave dissipation
Authors:
Richard Kowar
Abstract:
In this paper, we derive general theorems for controlling (vector-valued) first order ordinary differential equations such that its solutions stop at a finite time $T>0$ and apply them to relaxation and dissipative oscillation processes. We discuss several interesting examples for relaxation processes with finite stop** time and their energy behaviour. Our results on relaxation and dissipative o…
▽ More
In this paper, we derive general theorems for controlling (vector-valued) first order ordinary differential equations such that its solutions stop at a finite time $T>0$ and apply them to relaxation and dissipative oscillation processes. We discuss several interesting examples for relaxation processes with finite stop** time and their energy behaviour. Our results on relaxation and dissipative oscillations enable us to model diffusion processes with finite front speeds and dissipative waves that cause in each space point $x$ an oscillation with a finite stop** time $T(x)$. In the latter case, we derive the relation between $T(0)$ and $T(x)$. Moreover, the relations beteween the control functions in the ode model and the respective pde model are derived.In particular, we present an application of the Paley-Wiener-Schwartz Theorem that is used in our analysis. A complementary approach for dissipative oscillations and its application to dissipative waves is presented in [Ko19b], where the finite stop** time is achieved due to nonconstant coefficients in second order odes.
△ Less
Submitted 15 March, 2019;
originally announced March 2019.
-
Iterative Methods for Photoacoustic Tomography in Attenuating Acoustic Media
Authors:
Markus Haltmeier,
Richard Kowar,
Linh V. Nguyen
Abstract:
The development of efficient and accurate reconstruction methods is an important aspect of tomographic imaging. In this article, we address this issue for photoacoustic tomography. To this aim, we use models for acoustic wave propagation accounting for frequency dependent attenuation according to a wide class of attenuation laws that may include memory. We formulate the inverse problem of photoaco…
▽ More
The development of efficient and accurate reconstruction methods is an important aspect of tomographic imaging. In this article, we address this issue for photoacoustic tomography. To this aim, we use models for acoustic wave propagation accounting for frequency dependent attenuation according to a wide class of attenuation laws that may include memory. We formulate the inverse problem of photoacoustic tomography in attenuating medium as an ill-posed operator equation in a Hilbert space framework that is tackled by iterative regularization methods. Our approach comes with a clear convergence analysis. For that purpose we derive explicit expressions for the adjoint problem that can efficiently be implemented. In contrast to time reversal, the employed adjoint wave equation is again dam** and, thus has a stable solution. This stability property can be clearly seen in our numerical results. Moreover, the presented numerical results clearly demonstrate the Efficiency and accuracy of the derived iterative reconstruction algorithms in various situations including the limited view case.
△ Less
Submitted 9 August, 2017; v1 submitted 24 April, 2017;
originally announced April 2017.
-
On the nonlocality of the state and wave equation of Treeby and Cox
Authors:
Richard Kowar
Abstract:
In this paper it is shown that the state equation of Treeby and Cox [B. E. Treeby and B. T. Cox, J. Acoust. Soc. Am. \textbf{127} 5, (2010)] is \emph{nonlocal}, more precisely, a \emph{local} density variation causes an \emph{instant global} pressure variation and a \emph{local} pressure variation can only be caused by an \emph{instant global} density variation. This is in contrast to all frequenc…
▽ More
In this paper it is shown that the state equation of Treeby and Cox [B. E. Treeby and B. T. Cox, J. Acoust. Soc. Am. \textbf{127} 5, (2010)] is \emph{nonlocal}, more precisely, a \emph{local} density variation causes an \emph{instant global} pressure variation and a \emph{local} pressure variation can only be caused by an \emph{instant global} density variation. This is in contrast to all frequency dependent dissipative state equations known to the author. Moreover, it is shown that the Green function $G$ of the wave equation of Treeby and Cox cannot have a \emph{finite} wave front speed, i.e. there exists no finite $c_F>0$ such that $$
G(\mathbf{x},t) = 0 \qquad\mbox{for}\qquad |\mathbf{x}|/c_F > t $$ holds, where $|\mathbf{x}|/c_F$ corresponds to the \emph{travel time} of a wave propagating with speed $c_F$ from point $\mathbf{0}$ to point $\mathbf{x}$. As a consequence, the density and pressure waves satisfying (i) the state equation of Treeby and Cox, (ii) the equation of motion and (iii) the equation of continuity do not have a \emph{finite} wave front speed.
△ Less
Submitted 5 February, 2014;
originally announced February 2014.
-
Time reversal for photoacoustic tomography based on the wave equation of Nachman, Smith and Waag
Authors:
Richard Kowar
Abstract:
The goal of \emph{photoacoustic tomography} (PAT) is to estimate an \emph{initial pressure function} $\varphi$ from pressure data measured at a boundary surrounding the object of interest. This paper is concerned with a time reversal method for PAT that is based on the dissipative wave equation of Nachman, Smith and Waag\cite{NaSmWa90}. This equation has the advantage that it is more accurate than…
▽ More
The goal of \emph{photoacoustic tomography} (PAT) is to estimate an \emph{initial pressure function} $\varphi$ from pressure data measured at a boundary surrounding the object of interest. This paper is concerned with a time reversal method for PAT that is based on the dissipative wave equation of Nachman, Smith and Waag\cite{NaSmWa90}. This equation has the advantage that it is more accurate than the \emph{thermo-viscous} wave equation. For simplicity, we focus on the case of one \emph{relaxation process}. We derive an exact formula for the \emph{time reversal image} $\I$, which depends on the \emph{relaxation time} $τ_1$ and the \emph{compressibility} $κ_1$ of the dissipative medium, and show $\I(τ_1,κ_1)\to\varphi$ for $κ_1\to 0$. This implies that $\I=\varphi$ holds in the dissipation-free case and that $\I$ is similar to $\varphi$ for sufficiently small compressibility $κ_1$. Moreover, we show for tissue similar to water that the \emph{small wave number approximation} $\I_0$ of the time reversal image satisfies $\I_0 = η_0 *_\x \varphi$ with $\hat η_0(|\k|)\approx const.$ for $|\k|<< \frac{1}{c_0\,τ_1}$. For such tissue, our theoretical analysis and numerical simulations show that the time reversal image $\I$ is very similar to the initial pressure function $\varphi$ and that a resolution of $σ\approx 0.036\cdot mm$ is feasible (in the noise-free case).
△ Less
Submitted 18 September, 2013;
originally announced September 2013.
-
On time reversal in photoacoustic tomography for tissue similar to water
Authors:
Richard Kowar
Abstract:
This paper is concerned with time reversal in \emph{photoacoustic tomography} (PAT) of dissipative media that are similar to water. Under an appropriate condition, it is shown that the time reversal method in \cite{Wa11,AmBrGaWa11} based on the non-causal thermo-viscous wave equation can be used if the non-causal data is replaced by a \emph{time shifted} set of causal data. We investigate a simila…
▽ More
This paper is concerned with time reversal in \emph{photoacoustic tomography} (PAT) of dissipative media that are similar to water. Under an appropriate condition, it is shown that the time reversal method in \cite{Wa11,AmBrGaWa11} based on the non-causal thermo-viscous wave equation can be used if the non-causal data is replaced by a \emph{time shifted} set of causal data. We investigate a similar imaging functional for time reversal and an operator equation with the time reversal image as right hand side. If required, an enhanced image can be obtained by solving this operator equation. Although time reversal (for noise-free data) does not lead to the exact initial pressure function, the theoretical and numerical results of this paper show that regularized time reversal in dissipative media similar to water is a valuable method. We note that the presented time reversal method can be considered as an alternative to the causal approach in \cite{KaSc13} and a similar operator equation may hold for their approach.
△ Less
Submitted 2 August, 2013;
originally announced August 2013.
-
Small frequency approximation of (causal) dissipative pressure waves
Authors:
Richard Kowar
Abstract:
In this paper we discuss the problem of small frequency approximation of the causal dissipative pressure wave model proposed in \cite{KoScBo:11}. We show that for appropriate situations the Green function $G^c$ of the causal wave model can be approximated by a noncausal Green function $G_M^{pl}$ that has frequencies only in the small frequency range $[-M,M]$ ($M\leq 1/τ_0$, $τ_0$ relaxation time)…
▽ More
In this paper we discuss the problem of small frequency approximation of the causal dissipative pressure wave model proposed in \cite{KoScBo:11}. We show that for appropriate situations the Green function $G^c$ of the causal wave model can be approximated by a noncausal Green function $G_M^{pl}$ that has frequencies only in the small frequency range $[-M,M]$ ($M\leq 1/τ_0$, $τ_0$ relaxation time) and obeys a power law. For such cases, the noncausal wave $G^{pl}_M$ contains partial waves propagating arbitrarily fast but the sum of the noncausal waves is small in the $L^2-$sense.
△ Less
Submitted 31 January, 2012; v1 submitted 27 January, 2012;
originally announced January 2012.
-
On the causality of electrodynamics and the Aharonov-Bohm effect
Authors:
Richard Kowar
Abstract:
This paper presents a \emph{non-instant field model} for electrodynamics that permits a causal explanation of the \emph{Aharonov-Bohm effect} and a \emph{covariant quantization} of the respective Maxwell equations via the \emph{Gupta-Bleuler method}. Our model satisfies the following \emph{correspondence principle}: if $A^μ$, $\vE$, $\vB$ denote the four potential, the electric field and the magne…
▽ More
This paper presents a \emph{non-instant field model} for electrodynamics that permits a causal explanation of the \emph{Aharonov-Bohm effect} and a \emph{covariant quantization} of the respective Maxwell equations via the \emph{Gupta-Bleuler method}. Our model satisfies the following \emph{correspondence principle}: if $A^μ$, $\vE$, $\vB$ denote the four potential, the electric field and the magnetic field of the non-instant field model, then the respective classical quantities are $\A[A^μ]$, $\A[\vE]$, $\A[\vB]$, where $\A$ is a covariant time averaging operator. Here $\A[A^μ]$ is interpreted as the best possible measurement of the four potential $A^μ$. Although the Lorentz condition is not satisfied for $A^μ$, it is satisfied for $\A[A^μ]$. The latter fact means that the Lorentz condition does not hold for the quantized field but for its expectation value (cf. \emph{Gupta-Bleuler method} of quantization). Finally, we derive the energy conservation law of our field model and show that the field energy is quantized.
△ Less
Submitted 24 November, 2011;
originally announced November 2011.
-
Causal diffusion and its backwards diffusion problem
Authors:
Richard Kowar
Abstract:
This article starts over the backwards diffusion problem by replacing the \emph{noncausal} diffusion equation, the direct problem, by the \emph{causal} diffusion model developed in \cite{Kow11} for the case of constant diffusion speed. For this purpose we derive an analytic representation of the Green function of causal diffusion in the wave vector-time space for arbitrary (wave vector) dimension…
▽ More
This article starts over the backwards diffusion problem by replacing the \emph{noncausal} diffusion equation, the direct problem, by the \emph{causal} diffusion model developed in \cite{Kow11} for the case of constant diffusion speed. For this purpose we derive an analytic representation of the Green function of causal diffusion in the wave vector-time space for arbitrary (wave vector) dimension $N$. We prove that the respective backwards diffusion problem is ill-posed, but not exponentially ill-posed, if the data acquisition time is larger than a characteristic time period $τ$ ($2\,τ$) for space dimension $N\geq 3$ (N=2). In contrast to the noncausal case, the inverse problem is well-posed for N=1. Moreover, we perform a theoretical and numerical comparison between causal and noncausal diffusion in the \emph{space-time domain} and the \emph{wave vector-time domain}. The paper is concluded with numerical simulations of the backwards diffusion problem via the Landweber method.
△ Less
Submitted 1 March, 2012; v1 submitted 20 July, 2011;
originally announced July 2011.
-
On the causality of real-valued semigroups and diffusion
Authors:
Richard Kowar
Abstract:
In this paper we show that a process modeled by a strongly continuous real-valued semigroup (that has a space convolution operator as infinitesimal generator) cannot satisfy causality. We present and analyze a causal model of diffusion that satisfies the semigroup property at a discrete set of time points $M:=\{τ_m\,|\,m\in\N_0\}$ and that is in contrast to the classical diffusion model not smooth…
▽ More
In this paper we show that a process modeled by a strongly continuous real-valued semigroup (that has a space convolution operator as infinitesimal generator) cannot satisfy causality. We present and analyze a causal model of diffusion that satisfies the semigroup property at a discrete set of time points $M:=\{τ_m\,|\,m\in\N_0\}$ and that is in contrast to the classical diffusion model not smooth. More precisely, if $v$ denotes the concentration of a substance diffusing with constant speed, then $v$ is continuous but its time derivative is discontinuous at the discrete set $M$ of time points. It is this property of diffusion that forbids the classical limit procedure that leads to the noncausal diffusion model in Stochastics. Furthermore, we show that diffusion with constant speed satisfies an inhomogeneous wave equation with a time dependent coefficient.
△ Less
Submitted 16 February, 2011;
originally announced February 2011.
-
Photoacoustic Imaging Taking into Account Attenuation
Authors:
Richard Kowar,
Otmar Scherzer
Abstract:
First, we review existing attenuation models and discuss their causality properties, which we believe to be essential for algorithms for inversion with attenuated data. Then, we survey causality properties of common attenuation models. We also derive integro-differential equations which the attenuated waves are satisfying. In addition we discuss the ill--conditionness of the inverse problem for ca…
▽ More
First, we review existing attenuation models and discuss their causality properties, which we believe to be essential for algorithms for inversion with attenuated data. Then, we survey causality properties of common attenuation models. We also derive integro-differential equations which the attenuated waves are satisfying. In addition we discuss the ill--conditionness of the inverse problem for calculating the unattenuated wave from the attenuated one.
△ Less
Submitted 22 September, 2010;
originally announced September 2010.
-
Integral equation models for thermoacoustic imaging of dissipative tissue
Authors:
Richard Kowar
Abstract:
In case of non-dissipative tissue the inverse problem of thermoacoustic imaging basically consists of two inverse problems. First, a function $φ$ depending on the \emph{electromagnetic absorption function}, is estimated from one of three types of projections (spherical, circular or planar) and secondly, the \emph{electromagnetic absorption function} is estimated from $φ$. In case of dissipative…
▽ More
In case of non-dissipative tissue the inverse problem of thermoacoustic imaging basically consists of two inverse problems. First, a function $φ$ depending on the \emph{electromagnetic absorption function}, is estimated from one of three types of projections (spherical, circular or planar) and secondly, the \emph{electromagnetic absorption function} is estimated from $φ$. In case of dissipative tissue, it is no longer possible to calculate explicitly the projection of $φ$ from the respective pressure data (measured by point, planar or line detectors). The goal of this paper is to derive for each of the three types of pressure data, an integral equation that allows estimating the respective projection of $φ$. The advantage of this approach is that all known reconstruction formulas for $φ$ from the respective projection can be exploited.
△ Less
Submitted 25 February, 2010;
originally announced February 2010.
-
Causality analysis of waves and wave equations obeying attenuation
Authors:
Richard Kowar
Abstract:
In this paper we show that the standard causality condition for attenuated waves, i.e. the Kramers-Kronig relation that relates the attenuation law and the phase speed of the wave, is necessary but not sufficient for causality of a wave. By causality of a wave we understand the property that its wave front speed is bounded. Although this condition is not new, the consequences for wave attenuatio…
▽ More
In this paper we show that the standard causality condition for attenuated waves, i.e. the Kramers-Kronig relation that relates the attenuation law and the phase speed of the wave, is necessary but not sufficient for causality of a wave. By causality of a wave we understand the property that its wave front speed is bounded. Although this condition is not new, the consequences for wave attenuation have not been analysed sufficiently well. We derive the wave equation (for a homogeneous and isotropic medium) obeying attenuation and causality and with a generalization of the Paley-Wiener-Schwartz Theorem (cf. Theorem 7.4.3. in \cite{Ho03}), we perform a causality analysis of waves obeying the frequency power attenuation law. Afterwards the causality behaviour of Szabo's wave equation (cf. \cite{Szabo94}) and the thermo-viscous wave equation are investigated. Finally, we present a generalization of the thermo-viscous wave equation that obeys causality and the frequency power law (for powers in $(1,2]$ and) for small frequencies, which we propose for Thermoacoustic Tomography.
△ Less
Submitted 9 January, 2009;
originally announced January 2009.
