- Mathematics Department, Işık University, İstanbul, Turkey
Thermoacoustic imaging is a crossbred approach taking advantages of electromagnetic and ultrasound disciplines, together. A significant number of current medical imaging strategies are based on reconstruction of source distribution from information collected by sensors over a surface covering the region to be imaged. Reconstruction in thermoacoustic imaging depends on the inverse solution of thermoacoustic wave equation. Homogeneous assumption of tissue to be imaged results in degradation of image quality. In our paper, inverse solution of the thermoacoustic wave equation using layered tissue model consisting of concentric annular layers on a cylindrical cross-section is investigated for cross-sectional thermoacustic imaging of breast and brain. By using Green’s functions and surface integral methods we derive an exact analytic inverse solution of thermoacoustic wave equation in frequency domain. Our inverse solution is an extension of conventional solution to layered cylindrical structures. By carrying out simulations, using numerical test phantoms consisting of thermoacoustic sources distributed in the layered model, our layered medium assumption solution was tested and benchmarked with conventional solutions based on homogeneous medium assumption in frequency domain. In thermoacoustic image reconstruction, where the medium is assumed as homogeneous medium, the solution of nonhomogeneous thermoacoustic wave equation results in geometrical distortions, artifacts and reduced image resolution due to inconvenient medium assumptions.
1 Introduction
In thermoacoustic imaging, non-ionizing radio frequency or microwave pulses are delivered in to biological tissues. Some of the delivered energy is absorbed and converted into heat, leading to thermoelastic expansion, which in turn leads to ultrasonic emission. The generated ultrasonic waves are then detected by ultrasonic transducers located on the boundary of the object in order to form images of absorption properties of the object. Reconstruction of source distribution (absorption function) is based on the inverse solution of thermoacoustic wave equation. Most of the research studies reported in the literature were based on inverse solution of thermoacoustic wave equation using homogeneous medium assumption [1], [2]. The boundary conditions for thermoacoustic imaging have been investigated by Wang and Yang [3]. In a more recent study, Schoonover and Anastasio have presented an inverse solution based on piecewise homogeneous planar layers structure consisting source distribution only in one certain layer [4]. Also, there are other studies taking acoustic heterogeneties into account combining conventional methods and acoustic speed distribution as apriori information so that reducing effect of inhomogeneity and improving image quality [5–8]. Here our approach is based on the facts that two-dimensional cross-sections of many organs and tissue structures such as breast and brain can efficiently be modelled by piecewise homogeneous cylindrical layers and thermoacoustic sources are distributed over all layer structure. Thus, in this study, for cross-sectional two-dimensional thermoacustic imaging of breast and brain, we explore solution of the wave equation using layered tissue model consisting of concentric annular layers on a cylindrical cross-section. At first, we state the inverse source problem concerning thermoacoustic wave equation on a nonhomogeneous medium and describe the layered cylindrical medium; and then we derive the Green’s function of described media involving Bessel and Hankel functions to obtain the forward and inverse solutions of the nonhomogeneous thermoacoustic wave equation. The geometrical and acoustic parameters (densities and velocities) of layered media are utilized together with temporal initial condition, radiation conditions and continuity conditions on the layers’ boundaries. After, we give a forward solution and represent a key property of Green’s function resulting from initial condition of thermoacoustic wave propagation, and using this, we derive an exact analytic inverse solution of thermoacoustic wave equation for N-layered media. Lastly, to test and compare our layered solution with conventional solution based on homogeneous medium assumption, we performed simulations using numerical test phantoms consisting of sources distributed in the layered structure. The image reconstruction based on this approach involves the layer parameters as the apriori information which can be estimated from the acquired thermoacoustic data or additional transmission ultrasound scan.
2 Problem Statement
Consider a region having (N − 1) concentric annular cylindrical layers with different acoustic properties in the space
The acoustic waves are measured by the transducer for a sufficiently long time interval so that the waves emitted from every source location reach to the transducer. When the regions are different, there will be reflections and transmissions at the boundaries, Sm, for 1 ≤ m ≤ N − 1. Hence, the thermoacoustic wave propagation is governed by the non-homogeneous wave equation for the pressure
with 2(N − 1) boundary conditions
and
on each boundary Sm. Here, pm and pm+1 are the pressure representing acoustic waves, ρm and ρm+1 are the densities for Region m and Region m + 1, respectively, and − p0(r).δ′(t) is the source term emitting the thermoacoustic wave [9–13]. Also, the pressure function assures radiation condition. As the nature of the problem, thermoacoustic pressure function must satisfy the following initial conditions
In an inverse source problem, p0(r) is to be reconstructed given that acoustic field is measured by the transducer and is known on the surface SN.
To solve inverse problem, we firstly derive Green’s functions and state the forward solution of the problem by using Green’s function method. The Equation 1 in frequency domain corresponds to the nonhomogeneous three dimensional Helmholtz equation
where P(r, w) is the temporal Fourier Transform of p(r, t) and
The outgoing and incoming waves are represented by superscripts ‘out’ and ‘in’ for pressure function and we use the fact that
3 Green’s Function of Medium
The Green’s function is the solution of homogeneous wave equation except the point r′ where the point source located:
where δ(.) is the Dirac delta function.
It is convenient to study in cylindrical coordinates for the N-layered cylindrical configuration. Before transforming the Helmholtz equation Eq. 7 to cylindrical system, we take the spatial Fourier transform in z-direction to derive forward solution. We represent the spatial transform with a tilde symbol above of a function name, that is
and from Eq. 7, we obtain two dimensional Helmholtz equation
where k = w/c is the wave number, kz is the spatial frequency. The wave equation given in Eq. 9 is expressed in cylindrical coordinates (r, ϕ, z) as
It is known that the solution of above equation has a series form consisting of Bessel’s functions and exponential functions [14]. When
in which r = (r, ϕ, z) and r′ = (r′, ϕ′, z′).The Bessel functions Jn(.), Yn(.), In(.) and Kn(.) are real valued functions for positive real arguments. Hence all the terms in summations except unknown coefficients in Eq. 11 are all real. When ∥ k ∥≥∥ kz ∥ we apply Sommerfeld radiation condition and when ∥ k ∥ ≤ ∥ kz ∥ we choose evanescent waves for outer most layer, so that the waves will not grow to infinity. In light of these, the derivations made for the argument
When the point source r′ locates in Layer m, we denote Green’s function as Gm (1 ≤ m ≤ N). Each Green’s function Gm represents the unit impulse response of the layered medium and is partially defined with respect to observation point r:
We call the parts of Gm as Gmj for 1 ≤ j ≤ N when observation point r in Layer j. In the derivations of inverse problem, the observation points are on the transducer in Layer N so we need to calculate only the last parts GmN of Green’s function Gm wherever the source location m is
The coefficients in each Green’s function Gm are obtained by (2N + 2) equalities coming from the boundary conditions, Green’s functions conditions and radiation conditions: The given boundary conditions Eqs 2, 3 state that acoustic pressure function is continuous and its normal derivative is continuous with a scaling factor on the layer boundaries r = ri:
for 1 ≤ i ≤ m − 1,
for m ≤ i ≤ N − 1.
On the other hand, Green’s function is continuous and its normal derivative has jump discontinuity on a cylinder r = r′ where the point source locates [16]. Hence, these conditions give us
Additionally, second kind of Bessel function Yn is undefined when r = 0. Therefore in Layer 1, Green’s function cannot include Yn implying
Lastly, the pressure function must satisfy Sommerfeld radiation condition
By writing Bessel functions Jn and Yn in a linear combination of Hankel functions
We can write the system of equations in the matrix form as follows:
in which Mn is the coefficient matrix (A.1) given in Supplementary Appendix. Here, when the point source location changes, the coefficient matrix of a system will differ. But, we show that for all possible matrix systems results from location of source point, the determinant of coefficient matrices are the same. (See Supplementary Appendix to find derivation of this fact). As a result, determinant of coefficient matrix Mn, say Mn, is independent of the source location and the same for all n (index set for Bessel functions’ order) and for all m (location of source point r′).
In the derivation of inverse solution, we need the last part of Green’s function Gm Eq. 12, that is GmN. The coefficient Dn(N) in GmN is obtained by Kramer’s rule as follows:
in which
4 Forward Solution and Initial Condition Yields
If the source distribution in the medium is p0(r) and Green’s function of the medium is G(r, r′, w), then the forward solution of thermoacoustic wave equation in frequency domain can be written as [15], [16]
by make use of Green’s theorem and radiation conditions. Here, V is the support of source function. When we take the inverse Fourier transform of both sides of above forward solution at the discontinuity t = 0, we get
By the initial condition of the thermoacoustic wave equation, we obtain
Now, if we choose the source distribution p0(r) in the medium as point source function namely Dirac delta function δ(r − r∗) and substitute in above equation, we obtain
which implies
for any r and r∗. We use this result in the proof of the inverse solution.
5 Inverse Solution
Inverse source problem has been studied for homogeneous medium by Xu and Wang [2] for specific measurement geometries: two parallel planes, an infinitely long circular cylinder and a sphere, and this solution extended to the arbitrary measurement geometry by İdemen and Alkumru [1]. In these studies, in frequency domain, the source distribution inside the medium is determined by the following integral equation:
where S is a measurement surface and Gh is a free space Green’s function. In our study, for three dimensional N-layered configuration (see Figure 1) we extend the conventional solution and prove that the source distribution in each layer can be determined, by using corresponding Green’s function which describe the unit source distribution on the measurement surface:
where P(rs, w) is the acoustic pressure measured on the surface SN, GiN is the corresponding Green’s function for 1 ≤ i ≤ N and ρ(r) is a density function such that
For the proof of Eq. 28, let us write that equation as independent of index set and call the integral expression as q(r):
The acoustic pressure measured on the surface SN is given by forward solution of the wave equation Eq. 22:
where V′ is the volume covered by measurement surface. By substituting P(rs, w) in q(r), we get
in which ∇s means the gradient operator is applied with respect to variable rs. Now, we change the order of integration and use inner product properties, hence we obtain
Let call the term in outer integral as follows:
We know that the Green’s function is continuous on whole space. Also the normal derivative of Green’s function with a scaling factor (the density function) is continuous, too. Hence, the expression in the above integral is continuous which makes possible to apply the Divergence theorem as follows:
Since each layer is homogeneous in itself, the density function ρ(rs) is constant on each volume Vi for 1 ≤ i ≤ N. Therefore
The solutions Gin and Gout satisfy the Helmholtz equation:
in which ks = w/cs and cs is the acoustic speed in the region where rs in. If we multiply Eq. 37 by Gout(r′,rs, w) and Eq. 38 by Gin(r,rs, w) and subtract each other, we get
By adding the term
When we substitute the last equality Eq. 40 instead of the integrand seen in the integral Eq. 36, P(r,r′) can be written as
By utilizing the Dirac delta function properties and using the result Eq. 26 obtained by the initial condition, we obtain
To explore the first term of P(r, r′), say P1(r, r′), we substitute the Green’s functions in the expression for all location combinations of r, r′ and rs in
since the normal derivative on a cylinder is equal to the derivative with respect to the variable rs in cylindrical coordinates and the partial derivative operator is independent of integral variable w. In the light of the definition of wave function P at negative frequency, that is P(−w) = P(w)∗, we induce the integral expression in P1(r, r′) as follows:
This result shows that the real part of the integrand appearing in Eq. 44 has no contribution to the integral. To examine this integrand term, we substitute Green’s functions Eq. 21 of layered media: On SN, the second variable rs in Gout and Gin is an element of Region N. But r and r′ are free to be in any region. Thus, depending on locations of points r and r′, we have N2 cases for the combination of product terms
The exponential functions
by the properties of Dirac delta distribution. In the derivation of Green’s function, we prove that the functions
by making substitutions in summation and integral operators. Hence the integrand term multiplied by iw of outer integral in Eq. 44 is purely real implying
by Eq. 45.
Consequently, we obtain the function P(r, r′) as follows
Hence,
where ρ(r) is a density function.
At the beginning in Eq. 30, we suppose
and therefore the source distribution p0(r) is given by
6 Numerical Simulations
To test and compare our layered solution with conventional solution based on homogeneous medium assumption, we perform simulations using numerical test phantom a cross section of three layered cylindrical region depicted in Figures 2, 3. In Figure 2, first layer is the region 0 mm ≤ r ≤ 2.5 mm, second layer is the region 2.5 mm ≤ r ≤ 5 mm and third layer is the region r ≥ 5 mm. Densities and acoustic speeds for layers are choosen as 1.06 g/m3, 0.95 g/m3, 1 g/m3 and 1,000 m/s, 1,500 m/s, 2000 m/s from inner to outer. This phantom consists of thermoacoustic point sources at each layer, their polar coordinates are (1.25 mm, 0) (3.75 mm, 5π/4) and (6.25 mm, 2π/3). In Figure 3, we model breast as three main layers: Glandular tissue is the region 0 mm ≤ r ≤ 6.75 mm, fat tissue is the region 6.75 mm ≤ r ≤ 7.35 mm and skin is the region 7.35 mm ≤ r ≤ 7.5 mm considering ratios of actual thicknesses of these breast layers. Densities and acoustic speeds for breast layers are taken as 1 g/m3, 0.95 g/m3, 1.15 g/m3 and 1,480 m/s, 1,450 m/s, 1730 m/s respectively. This phantom consists of thermoacoustic point sources at glandular tissue layer, their polar coordinates are (3.4 mm, 0) and (5.4 mm, 0). In the simulations, we generate synthetic data by using layered medium Green’s function in forward solution Eq. 22 of thermoacoustic wave equation. For this purpose, we choose 3 MHz temporal frequency band between 1.5 and 4.5 MHz, and collect data by 512 element transducer located on a circle r = 7.5 mm in third layer. Then we reconstruct the thermoacoustic source distribution from this data using the existing homogeneous inverse solution Eq. 27 including free space Green’s functions and our layered inverse solution Eq. 28 including layered medium Green’s functions. Here we present an illustrative test result in the Figures 2, 3, where the numerical phantom and the reconstructed inverse source distributions (thermoacoustic images of point targets) are displayed. The middle panel show us that reflections and refractions on layer boundaries causes smearing and morphologically deformation of image because of homogeneous assumption in inversion algorithm. Hence, the test results ensure that the homogeneous medium assumption, as expected, produces incorrect source locations and poor point-spread-functions with severe side-lobes associated with the original point sources. Our layered solution produces source locations correctly and point-spread-functions with relatively narrower main-lobe and lower side-lobes.
FIGURE 2. (A): Test phantom (B): Numerical simulation obtained under homogeneous medium assumption (C): Numerical simulation obtained for layered medium, showing correct source locations.
FIGURE 3. (A): Breast phantom (B): Numerical simulation obtained under homogeneous medium assumption (C): Numerical simulation obtained for layered medium, showing correct source locations.
We also test our inverse solution for the capability in measuring the strength of sources by using same layer properties, frequency band and sampling rates with above simulation. We locate two point sources at coordinates (1.25 mm, 0) and (1.25 mm, π) with amplitude values 1 and 10, respectively. In the simulations, again, we generate synthetic data by using layered medium Green’s function in forward solution Eq. 22 of thermoacoustic wave equation and reconstruct the thermoacoustic source distribution from this data using layered inverse solution Eq. 28 including layered medium Green’s functions. The test phantom and numerical simulation results are depicted in Figure 4. Simulation results shows the inverse solution is accurate in distinguishing different source strengths.
FIGURE 4. (A): Test phantom (B): Numerical simulation obtained by layered medium inverse solution (C): Numerical simulation obtained for layered medium, showing correct source locations.
The limitation of the proposed inverse solution for thermoacoustic imaging is need of tissue properties and structures as apriori information. But, these informations can be obtained from acquired thermoacoustic data or additional transmission ultrasound scan. The error in apriori information of tissue structures will reduce the image quality but this effect can be minimized by some iterative methods.
7 Conclusion
In this study, we have considered the inverse source problem for thermoacoustic wave equation in layered cylindrical models. We have derived an exact analytic inverse solution in frequency domain under boundary conditions. Also, the derived solution was tested in a three-layer numerical tissue models. The solution presented here is a suitable approach for cross-sectional imaging of cylindrical and spherical structures (such as the breast and brain) to get better image quality.
Data Availability Statement
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.
Author Contributions
DE and BÜU did mathematical analysis, derivations and computations of this work. DE did the numerical simulations. DE and BÜU contributed to manuscript revision, read, and approved the submitted version.
Funding
This work was supported by TUBITAK of Turkey through ARDEB-1003 Program under Grant 213E038.
Conflict of Interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Publisher’s Note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
Acknowledgments
We express deep gratitude to Prof. Dr. Mustafa Karaman, who unfortunately passed away, for his contribution to this research by his extensive knowledge and enlightening vision. Previously, outline of this work has been presented as conference paper [17].
Supplementary Material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphy.2022.736555/full#supplementary-material
References
1. Idemen M, Alkumru A. On an Inverse Source Problem Connected with Photo-Acoustic and Thermo-Acoustic Tomographies. Wave Motion (2012) 49:595–604. doi:10.1016/j.wavemoti.2012.03.004
2. Xu M, Wang LV. Universal Back-Projection Algorithm for Photoacoustic Computed Tomography. Phys Rev E (2005) 71:016706–10167067. doi:10.1103/PhysRevE.71.016706
3. Wang LV, Yang X. Boundary Conditions in Photoacoustic Tomography and Image Reconstruction. J Biomed Opt (2007) 12:014027–101402710. doi:10.1117/1.2709861
4. Schoonover RW, Anastasio MA. Image Reconstruction in Photoacoustic Tomography Involving Layered Acoustic media. J Opt Soc Am A (2011) 28:1114–20. doi:10.1364/JOSAA.28.001114
5. Yuan Xu Y, Wang LV. Effects of Acoustic Heterogeneity in Breast Thermoacoustic Tomography. IEEE Trans Ultrason Ferroelect., Freq Contr (2003) 50:1134–46. doi:10.1109/TUFFC.2003.1235325
6. Wang J, Zhao Z, Song J, Chen G, Nie Z, Liu Q-H. Reducing the Effects of Acoustic Heterogeneity with an Iterative Reconstruction Method from Experimental Data in Microwave Induced Thermoacoustic Tomography. Med Phys (2015) 42:2103–12. doi:10.1118/1.4916660
7. Wang B, Zhao Z, Liu S, Nie Z, Liu Q. Mitigating Acoustic Heterogeneous Effects in Microwave-Induced Breast Thermoacoustic Tomography Using Multi-Physical K-Means Clustering. Appl Phys Lett (2017) 111:223701. doi:10.1063/1.5008839
8. Liu S, Lu Y, Zhu X, Jin H. Measurement Matrix Uncertainty Model-Based Microwave Induced Thermoacoustic Sparse Reconstruction in Acoustically Heterogeneous media. Appl Phys Lett (2021) 119:263701. doi:10.1063/5.0076449
9. Ammari H. An Introduction to Mathematics of Emerging Biomedical Imaging. Berlin: Springer (2008).
10. Colton D, Kress R. Inverse Acoustic and Electromagnetic Scattering Theory. New York: Springer (2013).
11. Kruger RA, Kiser WL, Miller KD, Reynolds HE, Reinecke DR, Kruger GA, et al. Thermoacoustic Ct: Imaging Principles. In: Proc. SPIE 3916, Biomedical Optoacoustics (2000), 150–159. doi:10.1117/12.386316
12. Kuchment P, Kunyansky L. Mathematics of Thermoacoustic Tomography. Eur J Appl Math (2008) 19:191–224. doi:10.1017/S0956792508007353
13. Xu M, Wang LV. Photoacoustic Imaging in Biomedicine. Rev Scientific Instr (2006) 77:041101–104110122. doi:10.1063/1.2195024
Keywords: inverse source problem, thermoacoustic imaging, green’s functions, integral equations, layered medium
Citation: Elmas D and Uzun BÜ (2022) Inverse Solution of Thermoacoustic Wave Equation for Cylindrical Layered Media. Front. Phys. 10:736555. doi: 10.3389/fphy.2022.736555
Received: 05 July 2021; Accepted: 07 February 2022;
Published: 30 March 2022.
Edited by:
Andre P. Vieira, University of São Paulo, BrazilReviewed by:
Pragya Shukla, Indian Institute of Technology Kharagpur, IndiaYuan Zhao, Chongqing Medical University, China
Copyright © 2022 Elmas and Uzun. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Demet Elmas, ZHZ1cmd1bmVsbWFzQGdtYWlsLmNvbQ==