ABSTRACT
We study the inverse problem in Optical Tomography of determining the optical properties of a medium , with , under the so-called diffusion approximation. We consider the time-harmonic case where Ω is probed with an input field that is modulated with a fixed harmonic frequency , where c is the speed of light and k is the wave number. We prove a result of Lipschitz stability of the absorption coefficient at the boundary in terms of the measurements in the case when the scattering coefficient is assumed to be known and k belongs to certain intervals depending on some a-priori bounds on , .
1. Introduction
Although Maxwell's equations provide a complete model for the light propagation in a scattering medium on a micro scale, on the scale suitable for medical diffuse Optical Tomography (OT) an appropriate model is given by the radiative transfer equation (or Boltzmann equation) [Citation1]. If Ω is a domain in , with with smooth boundary and radiation is considered in the body Ω, then it is well known that if the input field is modulated with a fixed harmonic frequency ω, the so-called diffusion approximation leads to the complex partial differential equation (see [Citation2]) for the energy current density u (1) (1) Here is the wave number, c is the speed of light and, in the anisotropic case, the so-called diffusion tensor K, is the complex matrix-valued function (2) (2) where is a real matrix-valued function, I is the identity matrix and I−B is positive definite [Citation2–4] on Ω. The spacially dependent real-valued coefficients and are called the absorption and the scattering coefficients of the medium Ω respectively and represent the optical properties of Ω. It is worth noticing that many tissues including parts of the brain, muscle and breast tissue have fibrous structure on a microscopic scale which results in anisotropic physical properties on a larger scale. Therefore the model considered in this manuscript seems appropriate for the case of medical applications of OT (see [Citation33]). Although it is common practise in OT to use the Robin-to-Robin map to describe the boundary measurements (see [Citation2]), the Dirichlet-to-Neumann (D–N) map will be employed here instead. This is justified by the fact that in OT, prescribing its inverse, the Neumnann-to-Dirichlet (N-D) map (on the appropriate spaces), is equivalent to prescribing the Robin-to-Robin boundary map. A rigorous definition of the D–N map for Equation (Equation1(1) (1) ) will be given in Section 2.
It is also well known that prescribing the N-D map is insufficient to recover both coefficients and uniquely [Citation5] unless a-priori smoothness assumptions are imposed [Citation6]. In this paper we consider the problem of determining the absorption coefficient in a medium , , that is probed with an input field which is modulated with a fixed harmonic frequency , with (time-harmonic case) and whose scattering coefficient is assumed to be known. More precisely, we show that , restricted to the boundary , depends upon the D–N map of (Equation1(1) (1) ), , in a Lipschitz way when k is chosen in certain intervals that depend on a-priori bounds on , and on the ellipticity constant of I−B (Theorem 2.4). The static case (k = 0), for which (Equation1(1) (1) ) is a single real elliptic equation, was studied in [Citation7], where the author proved Lipschitz stability of and Hölder stability of the derivatives of at the boundary in terms of . In the present paper we show that in the time-harmonic case, for which (Equation1(1) (1) ) is a complex elliptic equation, a Lipschitz stability estimate of at the boundary in terms of still holds true if k is chosen within certain ranges. The case where is assumed to be known and the scattering coefficient is to be determined, can be treated in a similar manner. The choice in this paper of focusing on the determination of rather than the one of is driven by the medical application of OT we have in mind. While varies from tissue to tissue, it is the absorption coefficient that carries the more interesting physiological information as it is related to the global concentrations of certain metabolites in their oxygenated and deoxygenated states.
Our main result (Theorem 2.4) is based on the construction of singular solutions to the complex elliptic equation (Equation1(1) (1) ), having an isolated singularity outside Ω. Such solutions were first constructed in [Citation8] for equations of type (3) (3) when K is a real matrix-valued function belonging to , with p>n and they were employed to prove stability results at the boundary in [Citation8], [Citation9], [Citation10] and [Citation11] in the case of Calderón's problem (see [Citation12]) with global, local data and on manifolds. The singular solutions introduced in [Citation8] were extended in [Citation13] to equations of type (4) (4) with real coefficients, where K is merely Hölder continuous. Singular solutions were also studied in [Citation14].
In this paper we extend the singular solutions introduced in [Citation8] to the case of elliptic equations of type (Equation1(1) (1) ) with complex coefficients. Such a construction is done by treating (Equation1(1) (1) ) as a strongly elliptic system with real coefficients, since , where is a positive constant depending on the a-priori information on , B and . We wish to stress out, however, that in [Citation8] the author constructed singular solutions to (Equation3(3) (3) ) which have an isolated singularity of arbitrary high order, where the current paper extends such construction to singular solutions to the complex Equation (Equation1(1) (1) ) having an isolated singularity of Green's type only. This is sufficient to prove the Lipschitz continuity of the boundary values of in terms of the D–N map. The more general construction of the singular solutions with an isolated singularity of arbitrary high order for elliptic complex partial differential equations will be material of future work.
This paper is stimulated by the work of Alessandrini and Vessella [Citation15], where the authors proved global Lipschitz stability of the conductivity in a medium Ω in terms of the D–N map for Calderón's problem, in the case when the conductivity is real, isotropic and piecewise constant on a given partition of Ω. This fundamental result was extended to the complex case in [Citation16] and in the context of various inverse problems for example in [Citation17], [Citation18], [Citation19] and [Citation20], [Citation21], [Citation22] in the isotropic and anisotropic settings, respectively. The machinery of the proof introduced in [Citation15] is based on an induction argument that combines quantitative estimates of unique continuation together with a careful asymptotic analysis of Green's functions. The initial step of their induction argument relies on Lipschitz (or Hölder) stability estimates at the boundary of the physical parameter that one wants to estimate in terms of the boundary measurements, which is the subject of the current manuscript. Our paper also provides a first step towards a reconstruction procedure of by boundary measurements based on a Landweber iterative method for nonlinear problems studied in [Citation23], where the authors provided an analysis of the convergence of such algorithm in terms of either a Hölder or Lipschitz global stability estimates (see also [Citation24]). We also refer to [Citation25] and [Citation32] for further reconstruction techniques of the optical properties of a medium.
The paper is organised as follows. Section 2 contains the formulation of the problem (Subsections 2.1 and 2.2) and our main result (Subsection 2.3, Theorem 2.4). Section 3 is devoted to the construction of singular solutions of Equation (Equation1(1) (1) ) having a Green's type isolated singularity outside Ω. The proof of our main result (Theorem 2.4) is given in Section 4.
2. Formulation of the problem and main result
2.1. Main assumptions
We rigorously formulate the problem by introducing the following notation, definitions and assumptions. For , a point will be denoted by , where and . Moreover, given a point , we will denote with the open balls in , centred at x and respectively with radius r and by the cylinder We will also denote , and .
Definition 2.1
Let Ω be a bounded domain in , with . We shall say that the boundary of Ω, , is of Lipschitz class with constants , if for any there exists a rigid transformation of coordinates under which we have P = 0 and where ϕ is a Lipschitz function on satisfying and
We consider, for a fixed k>0, (5) (5) where K is the complex matrix-valued function (6) (6) and q is the complex-valued function (7) (7) We recall that I denotes the identity matrix, where the matrix B is given by the OT physical experiment and it is such that , where denotes the class of real-valued symmetric matrices and such that I−B is a positive definite matrix [Citation2–4]. In this paper, we assume that the scattering coefficient is also known in Ω and it is the absorption coefficient that we seek to estimate from boundary measurements.
We assume that there are positive constants λ, E and and p>n such that the known quantities B, and the unknown quantity satisfy the two assumptions below respectively.
Assumption 2.1
Assumption on and B
(8) (8) (9) (9) and (10) (10)
Assumption 2.2
Assumption on
(11) (11) (12) (12)
We state below some facts needed in the sequel of the paper. Most of them are straightforward consequences of our assumptions.
The inverse of K (13) (13) has real and imaginary parts given by the symmetric, real matrix valued-functions on Ω (14) (14) (15) (15) respectively. As an immediate consequence of Assumptions 2.1 and 2.2 we have (16) (16) (17) (17) for a.e. and any . Moreover and commute, therefore the real and imaginary parts of K are the symmetric, real matrix valued-functions on Ω (18) (18) (19) (19) respectively. Assumptions 2.1 and 2.2 also imply that (20) (20) (21) (21) for a.e. , for every and the boundness condition (22) (22) for a.e. .
Moreover and q satisfy (23) (23) and (24) (24) respectively, where is a positive constant depending on λ, E, , k and n.
By denoting , the complex equation (25) (25) is equivalent to the system for the vector field (26) (26) which can be written in a more compact form as (27) (27) or, in components, as (28) (28) where is defined by (29) (29) and is a real matrix valued function on Ω defined by (30) (30) (Equation20(20) (20) ), together with (Equation22(22) (22) ) imply that system (Equation26(26) (26) ) is uniformly elliptic and bounded, therefore it satisfies the strong ellipticity condition (31) (31) where is a constant depending on λ, , k and n.
Remark 2.3
Matrix (32) (32) is uniformly positive definite on Ω and it satisfies (33) (33)
Definition 2.2
We will refer in the sequel to the set of positive numbers , L, λ, E, introduced above, along with the space dimension n, p>n, the wave number k and the diameter of Ω, , as to the a-priori data.
2.2. The Dirichlet-to-Neumann map
Let K be the complex matrix valued-function on Ω introduced in (Equation6(6) (6) ) and , satisfying Assumptions 2.1 and 2.2. B and are assumed to be known in Ω and satisfying Assumption 2.1, so that K is completely determined by , satisfying Assumption 2.2, on Ω. Denoting by the -pairing between and its dual , we will emphasise such dependence of K on by denoting K by For any , with , , we will denote throughout this paper by , the expression
Definition 2.3
The Dirichlet-to-Neumann (D–N) map corresponding to is the operator (34) (34) defined by (35) (35) for any f, , where is the weak solution to and is any function such that in the trace sense.
Given B, , , and the corresponding diffusion tensors , for i = 1, 2, satisfying Assumptions 2.1 and 2.2, the well known Alessandrini's identity (see [Citation8, (5.0.4), p.129]) (36) (36) holds true for any , where are the unique weak solutions to the Dirichlet problems and respectively.
We will denote in the sequel by the norm on the Banach space of bounded linear operators between and .
2.3. The main result
Theorem 2.4
Lipschitz stability of boundary values
Let and Ω be a bounded domain in with Lipschitz boundary with constants as in Definition 2.1. If B, and for satisfy Assumptions 2.1 and 2.2 and the wave number k satisfies either (37) (37) or (38) (38) where, λ and are the positive numbers introduced in Assumptions 2.1 and 2.2, then (39) (39) where C>0 is a constant depending on n, p, L, λ, E, and k.
3. Singular solutions
We consider (40) (40) where and q are the complex matrix valued-function and the complex function respectively introduced in Section 1 and satisfying Assumptions 2.1 and 2.2 on .
Theorem 3.1
Singular solutions for
Given L on as in (Equation40(40) (40) ), there exists such that (41) (41) and furthermore (42) (42) where w satisfies (43) (43) (44) (44) Here α is such that and C is a positive constant depending only on λ, E, and k.
Remark 3.2
Since is a complex matrix, the expression (45) (45) appearing in the leading term in (Equation42(42) (42) ) is defined as the principal branch of (Equation45(45) (45) ), where a branch cut along the negative real axis of the complex plane has been defined for , . Expressions like (Equation45(45) (45) ) will appear in the sequel of the paper and they will be understood in the same way.
Next we consider two technical lemmas that are needed for the proof of Theorem 3.1. The proofs of these results for the case where , with K a real matrix valued-function, are treated in detail in [Citation8] and their extension to the more general case , with K, q a real matrix valued-function and a real function respectively, was extended in [Citation7], therefore only the key points of their proof will be highlighted in the complex case below.
Lemma 3.3
Let p>n and be such that, for some positive s, (46) (46) (47) (47) Then we have (48) (48) (49) (49) where C is a positive constant depending only on A, n, p, λ, E, and k.
Proof of Lemma 3.3.
Proof of Lemma 3.3
The proof of (Equation49(49) (49) ) is based on the interior - Schauder estimate for uniformly elliptic systems (50) (50) for every r, , which, combined with interpolation inequality (51) (51) leads to (Equation48(48) (48) ). The positive constant C appearing in (Equation50(50) (50) ) depends on n, p, λ, E, and k only, whereas the positive constant C in (Equation51(51) (51) ) depends on n and p only. For (Equation50(50) (50) ) we refer to [Citation26, Lemma 6.2.6]) and for a detailed proof of it, in the case of a single real equation in divergence form, we refer to [Citation8, Proof of Lemma 2.1]. We refer to [Citation27, Theorem 5.12] for a detailed proof of (Equation51(51) (51) ) in the real case. For the complex case, (Equation51(51) (51) ) can be derived by denoting and combining (52) (52) for i = 1, 2 together with (53) (53)
Lemma 3.4
Let satisfy (54) (54) with 2<s<n<p. Then there exists satisfying (55) (55) and (56) (56) where C is a positive constant depending only on A, s, n, p, R, λ, E, and k.
Proof of Lemma 3.4.
Proof of Lemma 3.4
If then there exists a unique Green matrix defined in such that (57) (57) in the sense that for every we have (58) (58) Moreover (59) (59) where C is a positive constant depending on n, λ, E, and k and the vector valued-function defined by (60) (60) satisfies Lu = f with (61) (61) where and (62) (62) (63) (63) For the existence, uniqueness and asymptotic behaviour of the Green's matrix G on as in (Equation57(57) (57) )–(Equation59(59) (59) ) we refer to [Citation28]. We also refer to [Citation29], [Citation30] and the more recent result [Citation31] for further reading on the issue of the Green's matrix for elliptic systems of the second order. By an argument based on the monotone convergence theorem, one can show that and are both bounded from above by , where C is a positive constant depending on , E, and k.
If , we introduce a sequence , with , for , defined by for j = 1, 2. , for any and pointwise on . For any , let be the solution to (64) (64) such that (65) (65) on , therefore , for any , , for any . By applying interior - Schauder estimates to and using the fact that we obtain that (66) (66) where C is a positive constant that depends on . By applying a diagonal process we can find a subsequence weakly converging in to some function . This limit satisfies both (Equation55(55) (55) ) and (Equation56(56) (56) ).
We proceed next with the proof of Theorem 3.1.
Proof of Theorem 3.1.
Proof of Theorem 3.1.
We start by considering solution to (67) (67) where on . We want to find w such that (68) (68) satisfying (Equation43(43) (43) ), (Equation44(44) (44) ), where L is defined by (Equation5(5) (5) ). We have (69) (69) Therefore for any r, we have (70) (70) where and C is a positive constant depending on λ, E, , R and k only. If we take to be the solution to Lw = f given by Lemma 3.4, with f = −LH and , then (71) (71) and, by Lemma 3.3, properties (Equation43(43) (43) ), (Equation44(44) (44) ) are satisfied.
4. Proof of the main result
Since the boundary is Lipschitz, the normal unit vector field might not be defined on . We shall therefore introduce a unitary vector field locally defined near such that: (i) is smooth, (ii) is non-tangential to and it points to the exterior of Ω (see [Citation9, Lemmas 3.1–3.3] for a precised construction of ). Here we simply recall that any point , where , satisfies (72) (72) where and C depend on L, only.
Remark 4.1
Several constants depending on the a-priori data introduced in Definition 2.2 will appear in the proof of the main result below. In order to simplify our notation, we shall denote by C any of these constants, avoiding in most cases to point out their specific dependence on the a-priori data which may vary from case to case.
Proof of Theorem 2.4.
Proof of Theorem 2.4.
We start by recalling that by (Equation36(36) (36) ) we have for any that solve (73) (73) (74) (74) We set such that and , with , where is the number fixed in (Equation72(72) (72) ). Let be the singular solutions of Theorem 3.1 to (Equation73(73) (73) ), (Equation74(74) (74) ), respectively, having a singularity at (75) (75) By setting we have that and from (Equation36(36) (36) ) we obtain (76) (76) By (Equation75(75) (75) ) and Theorem 3.1 we have (77) (77) Recalling that for i = 1, 2 the real and imaginary parts of satisfy (Equation16(16) (16) ) and (Equation17(17) (17) ), respectively, we have (78) (78) and combining (Equation76(76) (76) ) together with (Equation75(75) (75) ), (Equation77(77) (77) ) and (Equation78(78) (78) ) we obtain (79) (79) The left-hand side of (Equation79(79) (79) ) can be estimated from below by recalling that is Hölder continuous on with exponent , for i = 1, 2 and by recalling again (Equation75(75) (75) ), which leads to (80) (80) and combining (Equation80(80) (80) ) together with (Equation79(79) (79) ) we obtain (81) (81) Recalling (Equation78(78) (78) ) and combining it together with (Equation77(77) (77) ), we can estimate the left-hand side of (Equation81(81) (81) ) from below as (82) (82) (Equation82(82) (82) ) together with (Equation81(81) (81) ) leads to (83) (83) is Hölder continuous on , with , for i = 1, 2 and, recalling that , we have (84) (84) Hence (Equation83(83) (83) ), combined with (Equation84(84) (84) ) and again with (Equation78(78) (78) ), leads to (85) (85) The integrand appearing on the left-hand side of (Equation85(85) (85) ) can be expressed as (86) (86) where the complex-valued function F is defined by (87) (87) The choices of k in either (Equation37(37) (37) ) or (Equation38(38) (38) ) imply (88) (88) where and denote the real and imaginary parts of a complex number z respectively. By combining (Equation88(88) (88) ) together with (Equation78(78) (78) ), the left-hand side of inequality (Equation85(85) (85) ) can be estimated from below as (89) (89) Combing (Equation89(89) (89) ) together with (Equation78(78) (78) ), we obtain (90) (90)
(Equation90(90) (90) ) combined with (Equation85(85) (85) ) and (Equation86(86) (86) ) then leads to (91) (91) By recalling (Equation72(72) (72) ), the first integral appearing on the right-hand side of (Equation91(91) (91) ) can be estimated from above by observing that , therefore (92) (92)
(see also [Citation8], [Citation9]), where denotes the surface measure on the unit sphere. Similarly to (Equation92(92) (92) ), the second, third and forth integrals on the right-hand side of inequality (Equation91(91) (91) ) are estimated from above as (93) (93) By observing that , where R depends on , the last two integrals appearing on the right-hand side of (Equation91(91) (91) ) can be estimated from above as (94) (94) The integral appearing on the left-hand side of (Equation91(91) (91) ) can be estimated from below as (95) (95) and we refer to [Citation13, p.66] for a detailed calculation of estimate (Equation95(95) (95) ). By combining (Equation91(91) (91) ) together with (Equation92(92) (92) )–(Equation95(95) (95) ) and the norms of u, v (see [Citation8], [Citation9]), we obtain (96) (96) By multiplying (Equation96(96) (96) ) by we obtain (97) (97) where as , which concludes the proof.
Remark 4.2
When n = 3 the ranges for k, (Equation37(37) (37) ) and (Equation38(38) (38) ), simplify to (98) (98) and (99) (99)
Acknowledgments
R. Gaburro would like to express her gratitude to Giovanni Alessandrini for the fruitful conversations regarding the construction of the singular solutions in this manuscript and to Simon Arridge who kindly offered insights on the physical problem of diffuse Optical Tomography some time ago.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Additional information
Funding
References
- Arridge SR, Schotland JC. Optical tomography: forward and inverse problems. Inverse Probl. 2009;25(12):123010. doi: https://doi.org/10.1088/0266-5611/25/12/123010
- Arridge SR. Optical tomography in medical imaging. Inverse Probl. 1999;15(2):R41. doi: https://doi.org/10.1088/0266-5611/15/2/022
- Heino J, Somersalo E. Estimation of optical absorption in anisotropic background. Inverse Probl. 2002;18(3):559–573. doi: https://doi.org/10.1088/0266-5611/18/3/304
- Hyvönen N. Characterizing inclusions in optical tomography. Inverse Probl. 2004;20(3):737. doi: https://doi.org/10.1088/0266-5611/20/3/006
- Arridge SR, Lionheart WRB. Nonuniqueness in diffusion-based optical tomography. Opt Lett. 1998;23(11):882–884. doi: https://doi.org/10.1364/OL.23.000882
- Harrach B. On uniqueness in diffuse optical tomography. Inverse Probl. 2009;25(5):055010. doi: https://doi.org/10.1088/0266-5611/25/5/055010
- Gaburro R. Stable determination at the boundary of the optical properties of a medium: the static case. Rend Istit Mat Univ Trieste. 2016;48:407–431.
- Alessandrini G. Singular solutions of elliptic equations and the determination of conductivity by boundary measurements. J Differ Equ. 1990;84(2):252–272. doi: https://doi.org/10.1016/0022-0396(90)90078-4
- Alessandrini G, Gaburro R. Determining conductivity with special anisotropy by boundary measurements. SIAM J Math Anal. 2001;33:153–171. doi: https://doi.org/10.1137/S0036141000369563
- Alessandrini G, Gaburro R. The local Calderón problem and the determination at the boundary of the conductivity. Comm Partial Differ Equ. 2009;34:918–936. doi: https://doi.org/10.1080/03605300903017397
- Gaburro R, Lionheart WRB. Recovering riemannian metrics in monotone families from boundary data. Inverse Probl. 2009;25(4):045004, (14pp). doi: https://doi.org/10.1088/0266-5611/25/4/045004
- Calderón AP. On an inverse boundary value problem, seminar on numerical analysis and its applications to continuum physics (Rio de Janeiro, 1980), 65-73, Soc. Brasil. Mat., Rio de Janeiro, 1980. Reprinted. Comput Appl Math. 2006;25(2–3):133–138. doi: https://doi.org/10.1590/S0101-82052006000200002
- Salo M. Inverse problems for nonsmooth first order perturbations of the Laplacian [PhD Thesis] Helsinki: University of Helsinki; 2004.
- Isakov V. On the uniqueness in the inverse conductivity problem with local data. Inverse Prob Imaging. 2007;1(1):95–105. doi: https://doi.org/10.3934/ipi.2007.1.95
- Alessandrini G, Vessella S. Lipschitz stability for the inverse conductivity problem. Adv Appl Math. 2005;35:207–241. doi: https://doi.org/10.1016/j.aam.2004.12.002
- Beretta E, Francini E. Lipschitz stability for the electrical impedance tomography problem: the complex case. Commun Partial Differ Equ. 2011;36:1723–1749. doi: https://doi.org/10.1080/03605302.2011.552930
- Alessandrini G, de Hoop MV, Gaburro R, et al. Lipschitz stability for the electrostatic inverse boundary value problem with piecewise linear conductivities. J Math Pures Appl. 2017;107(5):638–664. doi: https://doi.org/10.1016/j.matpur.2016.10.001
- Alessandrini G, de Hoop MV, Gaburro R, et al. Lipschitz stability for a piecewise linear Schrödinger potential from local cauchy data. Asymptotic Anal. 2018;108(3):115–149. doi: https://doi.org/10.3233/ASY-171457
- Beretta E, de Hoop MV, Francini E, et al. Uniqueness and Lipschitz stability of an inverse boundary value problem for time-harmonic elastic waves. Inverse Probl. 2017;33(3):035013.
- Alessandrini G, de Hoop MV, Gaburro R. Uniqueness for the electrostatic inverse boundary value problem with piecewise constant anisotropic conductivities. Inverse Probl. 2017;33(12):125013. doi: https://doi.org/10.1088/1361-6420/aa982d
- Alessandrini G, de Hoop MV, Gaburro R, et al. EIT in a layered anisotropic medium. Inverse Prob Imaging. 2018;12(3):667–676. doi: https://doi.org/10.3934/ipi.2018028
- Gaburro R, Sincich E. Lipschitz stability for the inverse conductivity problem for a conformal class of anisotropic conductivities. Inverse Probl. 2015;31:015008. doi: https://doi.org/10.1088/0266-5611/31/1/015008
- de Hoop MV, Qiu L, Scherzer O. Local analysis of inverse problems: Hölder stability and iterative regularization. Inverse Probl. 2012;23:045001, (16pp).
- Alessandrini G, Faucher F, de Hoop MV, et al. Inverse problem for the Helmholtz equation with cauchy data: reconstruction with conditional well-posedness driven iterative regularization. ESAIM: Math Modell Numer Anal. 2019;53(3):1005–1030. doi: https://doi.org/10.1051/m2an/2019009
- Kolehmainen V, Vauhkonen M, Kaipio JP, et al. Recovery of piecewise constant coefficients in optical diffusion tomography. Optic Express. 2000;7(13):468–480. doi: https://doi.org/10.1364/OE.7.000468
- Morrey CB. Multiple integrals in the calculus of variations. Berlin: Springer; 1966.
- Adams R, Fournier J. Sobolev spaces. Amsterdam: Academic Press; 2003.
- Dolzmann G, Müller S. Estimates for Green's matrices of elliptic systems by Lp theory. Manuscripta Math. 1995;88:261–273. doi: https://doi.org/10.1007/BF02567822
- Fuchs M. The green-matrix for elliptic systems which satisfy the legendre-hadamard condition. Manuscripta Math. 1984;46:97–115. doi: https://doi.org/10.1007/BF01185197
- Fuchs M. The green matrix for strongly elliptic systems of second order with continuous coefficients. Zeitschrift Für Analysis und ihre anwendungen. 1986;5(6):507–531. doi: https://doi.org/10.4171/ZAA/219
- Davey B, Hill J, Mayboroda S. Fundamental matrices and Green matrices for non-homogeneous elliptic systems. Publ Math. 2018;62(2):537–614. doi: https://doi.org/10.5565/PUBLMAT6221807
- Arridge SR, Hebden JC. Optical imaging in medicine II: modelling and reconstruction. Phys Med Biol. 1997;42(5):841. doi: https://doi.org/10.1088/0031-9155/42/5/008
- Heino J, Arridge S, Sikora J, et al. Anisotropic effects in highly scattering media. Phys Rev E. 2003;68(3):031908. doi: https://doi.org/10.1103/PhysRevE.68.031908