research article recent progress on the …downloads.hindawi.com/journals/cmmm/2013/425184.pdfwe...
TRANSCRIPT
Hindawi Publishing CorporationComputational and Mathematical Methods in MedicineVolume 2013 Article ID 425184 8 pageshttpdxdoiorg1011552013425184
Research ArticleRecent Progress on the Factorization Method for ElectricalImpedance Tomography
Bastian Harrach
Department of Mathematics University of Wurzburg 97074 Wurzburg Germany
Correspondence should be addressed to Bastian Harrach bastianharrachuni-wuerzburgde
Received 12 November 2012 Accepted 15 March 2013
Academic Editor Bill Lionheart
Copyright copy 2013 Bastian Harrach This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
The FactorizationMethod is a noniterative method to detect the shape and position of conductivity anomalies inside an objectThemethod was introduced by Kirsch for inverse scattering problems and extended to electrical impedance tomography (EIT) by Bruhland Hanke Since these pioneering works substantial progress has been made on the theoretical foundations of the method Thenecessary assumptions have been weakened and the proofs have been considerably simplified In this work we aim to summarizethis progress and present a state-of-the-art formulation of the Factorization Method for EIT with continuous data In particularwe formulate the method for general piecewise analytic conductivities and give short and self-contained proofs
1 Introduction
Electrical impedance tomography (EIT) aims to reconstructthe spatial conductivity distribution inside an imaging sub-ject Ω sube R119899 from current-voltage measurements on a partof its surface Σ sube 120597Ω Mathematically this leads to theproblem of recovering the coefficient 120590(119909) in the ellipticpartial differential equation
nabla sdot 120590nabla119906119892
120590= 0 in Ω 120590120597]119906
119892
120590|120597Ω
= 119892 on Σ
0 else(1)
fromknowledge of the correspondingNeumann-to-Dirichletoperator (NtD)
Λ (120590) 119892 997891997888rarr 119906119892
120590|Σ (2)
where 119906119892
120590is the solution of (1) We describe the precise
mathematical setting in Section 21In several applications EIT is used to determine the posi-
tion of conductivity changesThis includes anomaly detectionproblems where Λ(120590) is compared to a reference NtD Λ(120590
0)
in order to determine if and where 120590 differs from a knownbackground conductivity 120590
0 This problem also appears in
time-difference EIT where measurements at different timesare compared to monitor temporal conductivity changes
These applications lead to the shape reconstruction problemof determining the support of 120590 minus 120590
0from Λ(120590) and Λ(120590
0)
A prominent noniterative shape reconstruction methodis the Factorization Method It was introduced by Kirsch [1]for inverse scattering problems and extended to EIT by Bruhland Hanke [2ndash4] In its original form (cf [4]) the methodassumes that
1205900(119909) = 1 120590 (119909) = 1 + 120581 (119909) 120594
119863(119909) (3)
where119863 sube Ω is a union of separated smoothly bounded andsimply connected domains on which there is a conductivityjump of at least 120598 gt 0 that is
120581 (119909) ge 120598 forall119909 isin 119863 or 120581 (119909) le minus120598 forall119909 isin 119863 (4)
The method then characterizes the unknown shape 119863 by arange criterion For all unit vectors 119889 isin R119899 119889 = 1
119911 isin 119863 iff Φ119911119889|Σisin R (|Λ (120590) minus Λ (1)|
12
) (5)
whereΦ119911119889
is the so-called dipole function that is the solutionof
ΔΦ119911119889
= 119889 sdot nabla120575119911
in Ω 120597]Φ119911119889|120597Ω = 0 (6)
The range criterion (5) can be implemented numerically sothat each point 119911 isin Ω can be tested whether it belongs to theunknown inclusion or not
2 Computational and Mathematical Methods in Medicine
Substantial progress has been made on the FactorizationMethod since the original works of Kirsch Bruhl andHankeIn the following we restrict ourselves to progress in thecontext of EIT Overviews on the FM for EIT have beengiven by Hanke and Bruhl [5] in the book of Kirsch andGrinberg [6] and in a recent chapter of Hanke and Kirsch inScherzerrsquos Handbook of Mathematical Methods in Imaging[7]The FM for EIT has been treated as a special case of moregeneral elliptic problems by Kirsch [8] the author [9] andby Nachman Paivarinta and Teirila [10] A half-space settinghas been considered by Hanke and Schappel [11] Electrodemodels have been covered in the works of Bruhl LechleiterHakula Hanke Hyvonen and Pursiainen [5 12ndash15] The FMhas been extended to the complex conductivity case arisingin frequency-difference EIT by Seo Woo and the author[16 17] A priori separated indefinite inclusions have beentreated by Schmitt [18] and Schmitt and Kirsch discussedthe determination of the contrast level in the context of theFM for EIT in [19] Hyvonen and the author have removedthe assumptions on the inclusion contrast and boundaryregularity in [20] and [21] discusses the relation of the FMto localized potentials
In this work we aim to summarize the theoreticalprogress and present a state-of-the-art formulation of theFactorization Method for EIT with continuous data Inparticular wewill formulate themethod for general piecewiseanalytic conductivities and give short and self-containedproofs
2 Setting and Auxiliary Results
21The Setting We start by making the mathematical settingprecise Let Ω sube R119899 119899 ge 2 denote a bounded domainwith smooth boundary 120597Ω and outer normal vector ] LetΣ sube 120597Ω be an open part of the boundary 119871infin
+(Ω) denotes the
subspace of 119871infin(Ω)-functions with positive essential infima1198671
(Ω) and 1198712
(Σ) denote the spaces of1198671- and 1198712-functions
with vanishing integral mean on 120597Ω (resp Σ)For 120590 isin 119871
infin
+(Ω) and 119892 isin 119871
2
(Σ) there exists a unique
solution 119906119892
120590isin 119867
1
(Ω) of the elliptic partial differential
equation
nabla sdot 120590nabla119906119892
120590= 0 in Ω 120590120597]119906
119892
120590|120597Ω
= 119892 on Σ
0 else(7)
so that we can define the Neumann-to-Dirichlet operator(NtD)
Λ (120590) 1198712
(Σ) 997888rarr 119871
2
(Σ) 119892 997891997888rarr 119906
119892
120590|Σ (8)
where 119906119892120590isin 119867
1
(Ω) solves (7) Λ(120590) is a self-adjoint compact
linear operatorLet 120590
0isin 119871
infin
+(Ω) be piecewise analytic For each point 119911 isin
Ω that has a neighborhood in which 1205900is analytic and each
unit vector 119889 isin R119899 119889 = 1 let Φ119911119889
be the solution of
nabla sdot 1205900nablaΦ
119911119889= 119889 sdot nabla120575
119911in Ω 120590
0120597]Φ119911119889|120597Ω = 0 (9)
Φ119911119889
is called a dipole function
22 Auxiliary Results Our presentation of the FactorizationMethod in the next section relies on the following fourlemmas The first lemma is frequently called a monotonylemma since it shows that a larger conductivity leads to asmaller NtD More precisely it shows a relation betweenthe difference of two NtDs and the difference of the corre-sponding conductivities and the interior energy of an electricpotential The second lemma shows that this energy term isthe image of the adjoint of an auxiliary virtual measurementoperator that is defined on a subregion ofΩThe third lemmais a functional analytic relation between the norm of animage of an operator and the range of its adjoint Togetherwith the first two lemmas it implies that the range of theauxiliary virtual measurement operator can be calculatedfrom the NtDs Finally using the previous dipole functionsthe last lemma shows that the range of the auxiliary virtualmeasurement operator determines the region on which theyare defined
We start with the monotony lemma
Lemma 1 Let 1205901 1205900isin 119871
infin
+(Ω) Then for all 119892 isin 119871
2
(120597Ω)
intΩ(1205900minus 120590
1)1003816100381610038161003816nabla119906
0
1003816100381610038161003816
2d119909 leintΣ119892 (Λ
1minus Λ
0) 119892d119904 leint
Ω
1205900
1205901
(1205900minus 120590
1)1003816100381610038161003816nabla119906
0
1003816100381610038161003816
2d119909
(10)
where we abbreviated Λ119895= Λ(120590
119895) 119895 = 0 1 and 119906
0= 119906
119892
1205900
Proof The lemma seems to go back to Ikehata Kang Seo andSheen [22 23] cf also the similar arguments inKirsch [8] Ideet al [24] and in the works of Seo and the author [16 25] Forthe sake of completeness we copy the short proof from [25]For all 119892 isin 119871
2
(120597Ω) we have that
intΩ
1205901nabla119906
1sdot nabla119906
0d119909 = int
Σ
1198921199060d119904
= intΩ
1205900nabla119906
0sdot nabla119906
0d119909 = int
Σ
119892Λ0119892d119904
(11)
Hence from
intΩ
1205901nabla (119906
1minus 119906
0) sdot nabla (119906
1minus 119906
0) d119909
= intΩ
1205901
1003816100381610038161003816nabla11990611003816100381610038161003816
2d119909 minus intΩ
1205900
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
+ intΩ
(1205901minus 120590
0)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909
(12)
we obtain that
intΣ
119892 (Λ1minus Λ
0) 119892d119904 = int
Ω
(1205900minus 120590
1)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909
+ intΩ
1205901
1003816100381610038161003816nabla (1199061minus 119906
0)1003816100381610038161003816
2d119909(13)
which already yields the first asserted inequality
Computational and Mathematical Methods in Medicine 3
By interchanging 1205901and 120590
0 we conclude that
intΣ
119892 (Λ0minus Λ
1) 119892d119904
= intΩ
(1205901minus 120590
0)1003816100381610038161003816nabla1199061
1003816100381610038161003816
2d119909 + intΩ
1205900
1003816100381610038161003816nabla (1199060minus 119906
1)1003816100381610038161003816
2d119909
= intΩ
(1205901
10038161003816100381610038161003816100381610038161003816
nabla1199061minus1205900
1205901
nabla1199060
10038161003816100381610038161003816100381610038161003816
2
+ (1205900minus1205902
0
1205901
)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2
) d119909
(14)
and hence obtain the second inequality
Given a reference conductivity 1205900
isin 119871infin
+(Ω) and a
measurable subset119863 sube Ω we define the virtual measurementoperator 119871
119863by
119871119863 1198712
(119863)119899
997888rarr 1198712
(Σ) 119865 997891997888rarr V|
Σ (15)
where V isin 1198671
(Ω) solves
intΩ
1205900nablaV sdot nabla119908d119909 = int
119863
119865 sdot nabla119908d119909 forall119908 isin 1198671
(Ω) (16)
The energy term |nabla1199060|2 in Lemma 1 can be identified with the
norm of the adjoint of this virtual measurement operator
Lemma 2 The adjoint operator of 119871119863is given by
119871lowast
119863 1198712
(Σ) 997888rarr 119871
2
(119863)119899
119892 997891997888rarr nabla1199060|119863 (17)
where 1199060isin 119867
1
(Ω) solves
nabla sdot 1205900nabla119906
0= 0 119894119899 Ω 120590
0120597]1199060|120597Ω =
119892 119900119899 Σ
0 119890119897119904119890(18)
Proof For all 119892 isin 1198712
(Σ) and 119865 isin 119871
2
(119863)119899 we have that
int119863
(119871lowast
119863119892) sdot 119865 d119909 = int
Σ
119892 (119871119863119865) d119909 = int
Σ
119892V|Σd119909
= intΩ
1205900nabla119906
0sdot nablaV d119909 = int
119863
nabla1199060sdot 119865 d119909
(19)
which shows the assertion
The following functional analytic lemma uses bounds onthe image of an operator to characterize the range of its dualoperator
Lemma 3 Let 119883 and 119884 be real Hilbert spaces with innerproducts (sdot sdot)
119883and (sdot sdot)
119884 respectively Let 119860 isin L(119883 119884) and
1199091015840
isin 119883 Then
1199091015840
isin R (119860lowast
) iff exist119862 gt 0 10038161003816100381610038161003816(1199091015840
119909)119883
10038161003816100381610038161003816le 119862 119860119909 forall119909 isin 119883
(20)
In particular if1198831198841 and119884
2are three real Hilbert spaces119860
119894isin
L(119884119894 119883) 119894 = 1 2 and if there exists 119862 gt 0 with
1003817100381710038171003817119860lowast
11199091003817100381710038171003817 le 119862
1003817100381710038171003817119860lowast
21199091003817100381710038171003817 forall119909 isin 119883 (21)
thenR(1198601) sube R(119860
2)
Proof The assertion can be generalized to Banach spacesand in that context it is called the ldquo14th important propertyof Banach spacesrdquo in Bourbaki [26] For the sake of complete-ness we rewrite the proof from [27] to Hilbert spaces
If 1199091015840 isin R(119860lowast
) then there exists 1199101015840 isin 119884 such that 1199091015840 =119860lowast
1199101015840 Hence10038161003816100381610038161003816(1199091015840
119909)119883
10038161003816100381610038161003816=10038161003816100381610038161003816(119860
lowast
1199101015840
119909)119883
10038161003816100381610038161003816
=10038161003816100381610038161003816(1199101015840
119860119909)119884
10038161003816100381610038161003816le10038171003817100381710038171003817119910101584010038171003817100381710038171003817119860119909 forall119909 isin 119883
(22)
so that the assertion holds with 119862 = 1199101015840
Now let 1199091015840 isin 119883 be such that there exists 119862 gt 0 with
|(1199091015840
119909)119883| le 119862119860119909 for all 119909 isin 119883 We define
119891 (119910) = (1199091015840
119909)119883
for every 119910 = 119860119909 isin R (119860) (23)
Then 119891 is a well-defined continuous linear functional onR(119860) By setting it to zero on R(119860)
perp we can extend 119891 to acontinuous linear functional on 119884 Using the Riesz theoremit follows that there exists 1199101015840 isin 119884 with
(1199101015840
119910)119884
= 119891 (119910) forall119910 isin R (119860) (24)
Hence for all 119909 isin 119883 we have
(119860lowast
1199101015840
119909)119883
= (1199101015840
119860119909)119884
= 119891 (119860119909) = (1199091015840
119909)119883
(25)
so that 1199091015840 = 119860lowast
1199101015840
isin R(119860lowast
)
The last lemma shows that the range of the virtualmeasurement operator 119871
119863determines the region119863 onwhich
it is defined We state the lemma for a simple special casea generalized version of the lemma will be formulated inSection 32
Lemma 4 Let 1205900= 1 119863 sube Ω be open and 119863 sube Ω have a
connected complement Ω 119863Then for all unit vectors 119889 isin R119899 119889 = 1 and every point
119911 isin Ω 120597119863 it holds that
119911 isin 119863 iff Φ119911119889|Σisin R (119871
119863) (26)
Proof The proof is similar to the one of [21 Lemma 29]First let 119911 isin 119863 and 120598 gt 0 be such that 119861
120598(119911) sube 119863 We
choose
1198911isin 119867
1
(119861120598(119911)) with 119891
1|120597119861120598(119911)
= Φ119911119889|120597119861120598(119911)
1198912isin 119867
1
(119861120598(119911)) with Δ119891
2= 0
120597]1198912|120597119861120598(119911)= 120597]Φ119911119889|120597119861120598(119911)
(27)
and let 119865 isin 1198712
(119863)119899 be the zero continuation of nabla(119891
1minus 119891
2) to
119863Then the function
V = Φ119911119889
in Ω 119861120598(119911)
1198911
in 119861120598(119911)
(28)
4 Computational and Mathematical Methods in Medicine
fulfills V isin 1198671
(Ω) and for all 119908 isin 119867
1
(Ω)
intΩ
nablaV sdot nabla119908d119909
= intΩ119861120598(119911)
nablaΦ119911119889
sdot nabla119908d119909 + int119861120598(119911)
nabla1198911sdot nabla119908d119909
= minusint120597119861120598(119911)
120597]Φ119911119889 119908|120597119861120598(119911)d119904 + int
119861120598(119911)
nabla1198911sdot nabla119908d119909
= int119861120598(119911)
nabla (1198911minus 119891
2) sdot nabla119908d119909 = int
119863
119865 sdot nabla119908d119909
(29)
This shows thatΦ119911119889|Σ= V|
Σ= 119871
119863(119865) isin R(119871
119863)
Now let Φ119911119889|Σisin R(119871
119863) Let V isin 119867
1
(Ω) be the function
from the definition of 119871119863 Then
V|Σ= Φ
119911119889|Σ 120597]V|Σ = 0 = 120597]Φ119911119889|Σ (30)
so that it follows by unique continuation that V = Φ119911119889
in theconnected setΩ (119863 cup 119911)
If 119911 notin 119863 then 119889 sdot nabla120575119911notin 119867
minus2
(Ω 119863) and thus Φ119911119889
notin
1198712
(Ω 119863) which contradicts that V = Φ119911119889
in Ω (119863 cup 119911)Hence 119911 isin 119863
3 The Factorization Method
Nowwe will formulate the FactorizationMethod and charac-terize a region where a conductivity 120590 differs from a referenceconductivity 120590
0by a range criterion Before we turn to a
new general formulation of the method we first state it fora special case that is similar to the one that was treated in theoriginal works of Bruhl and Hanke [3 4]
31 The Factorization Method for a Simple Special Case
Theorem 5 Let 1205900= 1 and 120590 = 1 + 120594
119863 where 119863 sube Ω is an
open set so that 119863 sube Ω has a connected complement Ω 119863Then for all 119911 isin Ω 119911 notin 120597119863 and all dipole directions 119889 isin R119899119889 = 1
119911 isin 119863 iff Φ119911119889
1003816100381610038161003816Σisin R (|Λ (120590) minus Λ (1)|
12
) (31)
Proof Themonotony Lemma 1 yields that for all119892 isin 1198712
(120597Ω)
int119863
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 ge intΣ
119892 (Λ (1) minus Λ (120590)) 119892d119904 ge int119863
1
2
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(32)
Hence |Λ(120590) minusΛ(1)| = Λ(1) minusΛ(120590) and using Lemma 2 wecan restate this in the form
1003817100381710038171003817119871lowast
1198631198921003817100381710038171003817
2
ge10038171003817100381710038171003817|Λ (120590) minus Λ (1)|
12
11989210038171003817100381710038171003817
2
ge1
2
1003817100381710038171003817119871lowast
1198631198921003817100381710038171003817
2
(33)
Using the functional analytic Lemma 3 this implies that
R (119871119863) = R (|Λ (120590) minus Λ (1)|
12
) (34)
and thus the assertion follows from the relation between 119863
andR(119871119863) in Lemma 4
Obviously the same arguments can be used to treat thecase 120590(119909) = 1 + 120581(119909)120594
119863(119909) when there exists a conductivity
jump 120598 gt 0 so that either
120581 (119909) ge 120598 forall119909 isin 119863 or 120581 (119909) le minus120598 forall119909 isin 119863 (35)
32 The Factorization Method for the General PiecewiseAnalytic Case Now we drop the assumptions that the back-ground is constant that there is a clear conductivity jumpand that the complement of the inclusions is connected Wewill merely assume that the reference conductivity 120590
0is a
piecewise analytic function and that either 120590 minus 1205900ge 0 or
120590 minus 1205900le 0 Roughly speaking under this general assumption
the Factorization Method then characterizes the support of120590 minus 120590
0up to holes in the support that have no connections to
Σ For a precise formulation we use the concept of the innerand outer support from [28] that has been inspired by the useof the infinity support of Kusiak and Sylvester [29] see also[25 30]
Definition 6 A relatively open set 119880 sube Ω is called connectedto Σ if 119880 cap Ω is connected and 119880 cap Σ = 0
For a measurable function 120581 Ω rarr R we define
(a) the support supp(120581) as the complement (in Ω) of theunion of those relatively open119880 sube Ω for which 120581|
119880equiv
0(b) the inner support inn supp 120581 as the union of those
open sets 119880 sube Ω for which ess inf119909isin119880
|120581(119909)| gt 0(c) the outer support out
Σsupp 120581 as the complement (in
Ω) of the union of those relatively open 119880 sube Ω thatare connected to Σ and for which 120581|
119880equiv 0
The interior of a set119872 sube Ω is denoted by int119872 and its closure(with respect toR119899) by119872 If119872 is measurable we also define
(d) outΣ119872 = out
Σsupp120594
119872
It is easily checked that outΣ(supp 120581) = out
Σsupp 120581
With this concept we can extend the range characteriza-tion in Lemma 4 to a general setting (see also Remark 9 later)
Lemma 7 Let 1205900isin 119871
infin
+(Ω) be piecewise analytic Let 119863 sube Ω
be measurableThen for all unit vectors 119889 isin R119899 119889 = 1 and every point
119911 isin Ω that has a neighborhood in which 1205900is analytic
119911 isin int119863 implies Φ119911119889
1003816100381610038161003816Σisin R (119871
119863) (36)
and
Φ119911119889
1003816100381610038161003816Σisin R (119871
119863) implies 119911 isin out
Σ119863 (37)
Proof If 119911 isin int119863 then there exists a small ball 119861120598(119911) sube 119863
and the first assertion follows as in the proof of Lemma 4To show the second assertion letΦ
119911119889|Σisin R(119871
119863) and let
V isin 1198671
(Ω) be the function from the definition of 119871
119863 so that
(as in the proof of Lemma 4)
V|Σ= Φ
119911119889|Σ 120597]V|Σ = 0 = 120597]Φ119911119889|Σ (38)
Computational and Mathematical Methods in Medicine 5
Assume that 119911 notin outΣ119863 Then there exists a relatively
open 119880 sube Ω that is connected to Σ and contains 119911Hence by unique continuation it follows that V|
119880= Φ
119911119889|119880
and we obtain the same contradiction as in the proof ofLemma 4
Now we can formulate and prove the FactorizationMethod for general piecewise analytic conductivities
Theorem 8 Let 120590 isin 119871infin
+(Ω) and let 120590
0isin 119871
infin
+(Ω) be a
piecewise analytic function Let either
120590 (119909) ge 1205900(119909) forall119909 isin Ω or 120590 (119909) le 120590
0(119909) forall119909 isin Ω
(39)
Then for all 119911 isin Ω that have a neighborhood in which 1205900is
analytic as well as all unit vectors 119889 isin R119899 119889 = 1
119911isin innsupp (120590 minus 1205900) implies Φ
119911119889
1003816100381610038161003816Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
)
(40)
and
Φ119911119889
1003816100381610038161003816ΣisinR (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) implies 119911isin outΣsupp (120590 minus 120590
0)
(41)
Proof Let 119911 isin Ω have a neighborhood in which 1205900is analytic
and let 119889 isin R119899 be a unit vector with 119889 = 1 We onlyprove the assertions for 120590 ge 120590
0 The other case is completely
analogousFirst let 119911 isin inn supp (120590 minus 120590
0) Then there exists a small
ball 119861120598(119911) and 120575 gt 0 so that 120590 minus 120590
0ge 120575 on 119861
120598(119911) Using the
monotony Lemma 1 it follows that for all 119892 isin 1198712
(120597Ω)
intΣ
119892 (Λ (1205900) minus Λ (120590)) 119892d119904
ge intΩ
1205900
120590(120590 minus 120590
0)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909
ge 12057510038171003817100381710038171205900
1003817100381710038171003817119871infin(Ω)
1003817100381710038171003817100381710038171003817
1
120590
1003817100381710038171003817100381710038171003817119871infin(Ω)int119861120598(119911)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
ge 12057510038171003817100381710038171205900
1003817100381710038171003817119871infin(Ω)
1003817100381710038171003817100381710038171003817
1
120590
1003817100381710038171003817100381710038171003817119871infin(Ω)
10038171003817100381710038171003817119871lowast
119861120598(119911)11989210038171003817100381710038171003817
2
(42)
Using the functional analytic Lemma 3 we obtain that
R (119871119861120598(119911)
) sube R (1003816100381610038161003816Λ (120590) minus Λ (120590
0)1003816100381610038161003816
12
) (43)
and Lemma 7 yields that
Φ119911119889|Σisin R (119871
119861120598(119911)) sube R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) (44)
On the other hand with 119863 = supp(120590 minus 1205900) the monotony
Lemma 1 shows that for all 119892 isin 1198712
(120597Ω)
intΣ
119892 (Λ (1205900) minus Λ (120590
1)) 119892d119904 le int
Ω
(1205901minus 120590
0)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909
le10038171003817100381710038171205901 minus 120590
0
1003817100381710038171003817119871infin(Ω)int119863
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
le10038171003817100381710038171205901 minus 120590
0
1003817100381710038171003817119871infin(Ω)
1003817100381710038171003817119871lowast
1198631198921003817100381710038171003817
2
(45)
so that we obtain from the functional analytic Lemma 3
R (1003816100381610038161003816Λ (120590) minus Λ (120590
0)1003816100381610038161003816
12
) sube R (119871119863) (46)
Hence Lemma 7 yields that
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) sube R (119871119863) implies
119911 isin outΣ(supp (120590 minus 120590
0)) = out
Σsupp (120590 minus 120590
0)
(47)
Remark 9 Theorem 8 shows that the Factorization Methodis able to detect the support of a conductivity difference up tothe difference between the outer and the inner support thatis roughly speaking up to holes in the support that have noconnections to the boundary It leaves openwhether points insuch holes will fulfill the range criterion of the FactorizationMethod or not
A result of Hyvonen and the author [30 Lemma 25]shows that for every smooth domain119863with119863 sub Ω and everyunit vector 119889 isin R119899 119889 = 1
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) forall119911 isin 120597119863 (48)
implies
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) forall119911 isin 119863 (49)
In that sense we can expect that holes in the support willbe filled up and that the set detected by the FactorizationMethod is essentially the outer support of the conductivitydifference
33 The Factorization Method for the Indefinite Case It is along standing open theoretical problem whether the rangecriterion of the Factorization Method holds true withoutthe definiteness assumption that 120590 ge 120590
0on Ω or 120590 le
1205900on Ω However Grinberg Kirsch and Schmitt [18 31]
showed how to exclude a region 119864 sube Ω from Ω insuch a way that the Factorization Method only requires thedefiniteness assumption onΩ119864 In this subsection we showhow their idea can be incorporated into our formulation ofthe method
To point out the main idea we first formulate the resultfor a simple special case Let us stress that for 120590 = 1 + 120594
119863+ minus
(12)120594119863minus it is not known whether
119911 isin 119863+
cup 119863minus iff Φ
119911119889|Σisin R (|Λ (120590) minus Λ (1)|
12
) (50)
However we can still use the FactorizationMethod if we havesome a priori knowledge that separates 119863+ and 119863
minus Moreprecisely if we know a subset 119864 that contains 119863minus withoutintersecting 119863
+ then we can use the Factorization Methodto find119863+ cup 119864 (and thus119863+)
6 Computational and Mathematical Methods in Medicine
Theorem 10 Let 1205900= 1 and 120590 = 1 + 120594
119863+ minus (12)120594
119863minus where
119863+
119863minus
sube Ω are open Let 119864 sube Ω be an open set
(a) If 119863+ sube 119864 and 119863minus cup 119864 sube Ω has a connectedcomplement then for all 119911 isin Ω 119911 notin 120597(119863
minus
cup119864) and alldipole directions 119889 isin R119899 119889 = 1
119911 isin 119863minus
cup 119864 iff Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1) + 119871119864119871lowast
119864
1003816100381610038161003816
12
)
(51)
(b) If 119863minus sube 119864 and 119863+ cup 119864 sube Ω has a connectedcomplement then for all 119911 isin Ω 119911 notin 120597(119863
+
cup119864) and alldipole directions 119889 isin R119899 119889 = 1
119911 isin 119863+
cup 119864 iff Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1) minus 119871119864119871lowast
119864
1003816100381610038161003816
12
)
(52)
Proof Themonotony Lemma 1 yields that for all119892 isin 1198712
(120597Ω)
1
2int119863minus
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 minus int119863+
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
le intΣ
119892 (Λ (120590) minus Λ (1)) 119892d119904
le int119863minus
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 minus1
2int119863+
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(53)
Since (cf Lemma 2)
intΣ
119892 (119871119864119871lowast
119864) 119892d119904 = 1003817100381710038171003817119871
lowast
1198641198921003817100381710038171003817
2
= int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 (54)
it follows for case (a) that
1
2int119863minuscup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 le intΣ
119892 (Λ (120590) minus Λ (1) + 2119871119864119871lowast
119864) 119892d119904
le int119863minuscup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909(55)
Using the functional analytic Lemma 3 this implies that
R (119871119863minuscup119864) = R (
1003816100381610038161003816Λ (120590) minus Λ (1) + 2119871119864119871lowast
119864
1003816100381610038161003816
12
) (56)
so that the assertion (a) follows from Lemma 4In case (b) we obtain that
3
2int119863+cup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 ge intΣ
119892 (Λ (1) minus Λ (120590) + 2119871119864119871lowast
119864) 119892d119904
ge1
2int119863+cup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909(57)
and the same arguments as above yield the assertion
We can also extend these ideas to the general setting ofSection 32
Theorem 11 Let 120590 isin 119871infin
+(Ω) and let 120590
0isin 119871
infin
+(Ω) be a
piecewise analytic function Let 119864 sube Ω be a measurable setChoose 120572 120573 isin R such that
120572 gt1003817100381710038171003817120590 minus 120590
0
1003817100381710038171003817119871infin(Ω) 120573 gt
1003817100381710038171003817100381710038171003817
1205900
120590(1205900minus 120590)
1003817100381710038171003817100381710038171003817119871infin(Ω) (58)
(a) If 120590 le 1205900on Ω 119864 then for all 119911 isin Ω that have a
neighborhood in which 1205900is analytic as well as all unit
vectors 119889 isin R119899 119889 = 1
119911 isin inn supp (120590 minus 1205900) cup 119864 implies
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590)minusΛ (1205900)+120572119871
119864119871lowast
119864
1003816100381610038161003816
12
)
(59)
and
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864
1003816100381610038161003816
12
) implies
119911 isin outΣ(supp (120590 minus 120590
0) cup 119864)
(60)
(b) If 120590 ge 1205900on Ω 119864 then for all 119911 isin Ω that have a
neighborhood in which 1205900is analytic as well as all unit
vectors 119889 isin R119899 119889 = 1
119911 isin inn supp (120590 minus 1205900) cup 119864 implies
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900) minus 120573119871
119864119871lowast
119864
1003816100381610038161003816
12
)
(61)
and
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900) minus 120573119871
119864119871lowast
119864
1003816100381610038161003816
12
) implies
119911 isin outΣ(supp (120590 minus 120590
0) cup 119864)
(62)
Proof For every 119911 isin inn supp(120590minus1205900)with 119911 notin 119864 there exists
a small ball 119861120598(119911) and 120575 gt 0 so that 120590
0minus120590 ge 120575 on 119861
120598(119911) Using
the monotony Lemma 1 it follows that for all 119892 isin 1198712
(120597Ω)
intΣ
119892 (Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864) 119892d119904
ge intΩ
(1205900minus 120590
1)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909 + 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
ge 120575int119861120598(119911)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909+(120572 minus1003817100381710038171003817120590 minus 120590
0
1003817100381710038171003817119871infin(Ω))int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(63)
Computational and Mathematical Methods in Medicine 7
As in the previous proofs we obtain from Lemmas 2 and 3that
R (1003816100381610038161003816Λ (120590) minus Λ (120590
0) + 120572119871
119864119871lowast
119864
1003816100381610038161003816
12
) supe R (119871119861120598(119911)cup119864
) (64)
so that the first implication of (a) follows from Lemma 7The monotony Lemma 1 also implies that
intΣ
119892 (Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864) 119892d119904
le intΩ
1205900
120590(1205900minus 120590)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 + 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
le
1003817100381710038171003817100381710038171003817
1205900
120590(1205900minus 120590)
1003817100381710038171003817100381710038171003817119871infin(Ω)intsupp(120590minus1205900)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
+ 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(65)
so that the second implication of (a) follows from Lemmas 23 and 7 Assertion (b) can be proven analogously
Remark 9 also applies to this case
4 Conclusions and Remarks
The Factorization Method can be used to detect regionsin which a conductivity differs from a known referenceconductivity In this work we summarized the progresson the methodrsquos theoretical foundation We formulated themethod for general piecewise analytic conductivities andgave comparatively simple and self-contained proofsWe alsoshowed how the idea of excluding a part of the imaging regioncan be incorporated into this formulation
The regularity assumptions can beweakened even furtherOur proofs only require unique continuation arguments forthe reference conductivity 120590
0and the existence of the dipole
functionsTwo major open theoretical questions still exist in the
context of the FactorizationMethodThe theoretical justifica-tion of the method requires a definiteness condition (on thewhole domain or after excluding an a priori known part of thedomain) It is unknownwhether themethodrsquos range criterionholds without such a definiteness conditionThe second openquestion concerns the numerical stability of the methodrsquosrange criterion So far there are no rigorous convergenceresults for numerical implementations of this range criterion(see however Lechleiter [32] for a first step in this direction)As a promising approach to overcome both problems wewould like to point out the recent work on monotony-basedmethods [28]
Acknowledgments
This paper was funded by the German Research Foundation(DFG) and the University of Wurzburg in the fundingprogramme Open Access Publishing
References
[1] A Kirsch ldquoCharacterization of the shape of a scattering obstacleusing the spectral data of the far field operatorrdquo Inverse Prob-lems vol 14 no 6 pp 1489ndash1512 1998
[2] M Bruhl Gebietserkennung in der elektrischen Impedanztomo-graphie [PhD thesis] Universitat Karlsruhe 1999
[3] M Bruhl and M Hanke ldquoNumerical implementation of twononiterative methods for locating inclusions by impedancetomographyrdquo Inverse Problems vol 16 no 4 pp 1029ndash10422000
[4] M Bruhl ldquoExplicit characterization of inclusions in electricalimpedance tomographyrdquo SIAM Journal on Mathematical Anal-ysis vol 32 no 6 pp 1327ndash1341 2001
[5] M Hanke and M Bruhl ldquoRecent progress in electricalimpedance tomographyrdquo Inverse Problems vol 19 no 6 ppS65ndashS90 2003
[6] A Kirsch andN GrinbergTheFactorizationMethod for InverseProblems vol 36 ofOxford Lecture Series inMathematics and ItsApplications Oxford University Press Oxford UK 2007
[7] M Hanke and A Kirsch ldquoSampling methodsrdquo in Handbook ofMathematical Models in Imaging O Scherzer Ed pp 501ndash550Springer 2011
[8] A Kirsch ldquoThe factorization method for a class of inverseelliptic problemsrdquo Mathematische Nachrichten vol 278 pp258ndash277 2005
[9] B Gebauer ldquoThe Factorization Method for real elliptic prob-lemsrdquo Zeitschrift fur Analysis und ihre Anwendung vol 25 no1 pp 81ndash102 2006
[10] A I Nachman L Paivarinta and A Teirila ldquoOn imagingobstacles inside inhomogeneous mediardquo Journal of FunctionalAnalysis vol 252 no 2 pp 490ndash516 2007
[11] M Hanke and B Schappel ldquoThe factorization method forelectrical impedance tomography in the half-spacerdquo SIAMJournal on Applied Mathematics vol 68 no 4 pp 907ndash9242008
[12] N Hyvonen ldquoComplete electrode model of electricalimpedance tomography approximation properties andcharacterization of inclusionsrdquo SIAM Journal on AppliedMathematics vol 64 pp 902ndash931 2004
[13] A Lechleiter N Hyvonen and H Hakula ldquoThe factorizationmethod applied to the complete electrode model of impedancetomographyrdquo SIAM Journal onAppliedMathematics vol 68 no4 pp 1097ndash1121 2008
[14] N Hyvonen H Hakula and S Pursiainen ldquoNumerical imple-mentation of the factorization method within the completeelectrode model of electrical impedance tomographyrdquo InverseProblems and Imaging vol 1 pp 299ndash317 2007
[15] HHakula andNHyvonen ldquoOn computation of test dipoles forfactorization methodrdquo BIT Numerical Mathematics vol 49 no1 pp 75ndash91 2009
[16] B Harrach and J K Seo ldquoDetecting inclusions in electri-cal impedance tomography without reference measurementsrdquoSIAM Journal on Applied Mathematics vol 69 no 6 pp 1662ndash1681 2009
[17] B Harrach J K Seo and E J Woo ldquoFactorization methodand its physical justification in frequency-difference electricalimpedance tomographyrdquo IEEE Transactions on Medical Imag-ing vol 29 no 11 pp 1918ndash1926 2010
[18] S Schmitt ldquoThe factorization method for EIT in the case ofmixed inclusionsrdquo Inverse Problems vol 25 no 6 Article ID065012 2009
8 Computational and Mathematical Methods in Medicine
[19] S Schmitt and A Kirsch ldquoA factorization scheme for determin-ing conductivity contrasts in impedance tomographyrdquo InverseProblems vol 27 Article ID 095005 2011
[20] B Gebauer and N Hyvonen ldquoFactorization method and irreg-ular inclusions in electrical impedance tomographyrdquo InverseProblems vol 23 no 5 pp 2159ndash2170 2007
[21] B Gebauer ldquoLocalized potentials in electrical impedancetomographyrdquo Inverse Problems and Imaging vol 2 pp 251ndash2692008
[22] H Kang J K Seo and D Sheen ldquoThe inverse conductivityproblem with one measurement Stability and estimation ofsizerdquo SIAM Journal on Mathematical Analysis vol 28 no 6 pp1389ndash1405 1997
[23] M Ikehata ldquoSize estimation of inclusionrdquo Journal of Inverse andIll-Posed Problems vol 6 no 2 pp 127ndash140 1998
[24] T Ide H Isozaki S Nakata S Siltanen and G UhlmannldquoProbing for electrical inclusions with complex sphericalwavesrdquo Communications on Pure and Applied Mathematics vol60 no 10 pp 1415ndash1442 2007
[25] B Harrach and J K Seo ldquoExact shape-reconstruction by one-step linearization in electrical impedance tomographyrdquo TheSIAM Journal onMathematical Analysis vol 42 no 4 pp 1505ndash1518 2010
[26] N Bourbaki Elements of Mathematics Topological VectorSpaces chapters 1ndash5 Springer Berlin Germany 2003
[27] F Fruhauf B Gebauer andO Scherzer ldquoDetecting interfaces ina parabolic-elliptic problem from surfacemeasurementsrdquo SIAMJournal on Numerical Analysis vol 45 no 2 pp 810ndash836 2007
[28] B Harrach and M Ullrich ldquoMonotony based shape recon-struction in electrical impedance tomographyrdquo submitted forpublication
[29] S Kusiak and J Sylvester ldquoThe scattering supportrdquoCommunica-tions on Pure and Applied Mathematics vol 56 no 11 pp 1525ndash1548 2003
[30] B Gebauer and N Hyvonen ldquoFactorization method andinclusions of mixed type in an inverse elliptic boundary valueproblemrdquo Inverse Problems and Imaging vol 2 pp 355ndash3722008
[31] N I Grinberg and A Kirsch ldquoThe factorization method forobstacles with a-priori separated sound-soft and sound-hardpartsrdquo Mathematics and Computers in Simulation vol 66 no4-5 pp 267ndash279 2004
[32] A Lechleiter ldquoA regularization technique for the factorizationmethodrdquo Inverse Problems vol 22 no 5 pp 1605ndash1625 2006
Submit your manuscripts athttpwwwhindawicom
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MEDIATORSINFLAMMATION
of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Behavioural Neurology
EndocrinologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Disease Markers
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
OncologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Oxidative Medicine and Cellular Longevity
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
PPAR Research
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Immunology ResearchHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
ObesityJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational and Mathematical Methods in Medicine
OphthalmologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Diabetes ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentAIDS
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Gastroenterology Research and Practice
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Parkinsonrsquos Disease
Evidence-Based Complementary and Alternative Medicine
Volume 2014Hindawi Publishing Corporationhttpwwwhindawicom
2 Computational and Mathematical Methods in Medicine
Substantial progress has been made on the FactorizationMethod since the original works of Kirsch Bruhl andHankeIn the following we restrict ourselves to progress in thecontext of EIT Overviews on the FM for EIT have beengiven by Hanke and Bruhl [5] in the book of Kirsch andGrinberg [6] and in a recent chapter of Hanke and Kirsch inScherzerrsquos Handbook of Mathematical Methods in Imaging[7]The FM for EIT has been treated as a special case of moregeneral elliptic problems by Kirsch [8] the author [9] andby Nachman Paivarinta and Teirila [10] A half-space settinghas been considered by Hanke and Schappel [11] Electrodemodels have been covered in the works of Bruhl LechleiterHakula Hanke Hyvonen and Pursiainen [5 12ndash15] The FMhas been extended to the complex conductivity case arisingin frequency-difference EIT by Seo Woo and the author[16 17] A priori separated indefinite inclusions have beentreated by Schmitt [18] and Schmitt and Kirsch discussedthe determination of the contrast level in the context of theFM for EIT in [19] Hyvonen and the author have removedthe assumptions on the inclusion contrast and boundaryregularity in [20] and [21] discusses the relation of the FMto localized potentials
In this work we aim to summarize the theoreticalprogress and present a state-of-the-art formulation of theFactorization Method for EIT with continuous data Inparticular wewill formulate themethod for general piecewiseanalytic conductivities and give short and self-containedproofs
2 Setting and Auxiliary Results
21The Setting We start by making the mathematical settingprecise Let Ω sube R119899 119899 ge 2 denote a bounded domainwith smooth boundary 120597Ω and outer normal vector ] LetΣ sube 120597Ω be an open part of the boundary 119871infin
+(Ω) denotes the
subspace of 119871infin(Ω)-functions with positive essential infima1198671
(Ω) and 1198712
(Σ) denote the spaces of1198671- and 1198712-functions
with vanishing integral mean on 120597Ω (resp Σ)For 120590 isin 119871
infin
+(Ω) and 119892 isin 119871
2
(Σ) there exists a unique
solution 119906119892
120590isin 119867
1
(Ω) of the elliptic partial differential
equation
nabla sdot 120590nabla119906119892
120590= 0 in Ω 120590120597]119906
119892
120590|120597Ω
= 119892 on Σ
0 else(7)
so that we can define the Neumann-to-Dirichlet operator(NtD)
Λ (120590) 1198712
(Σ) 997888rarr 119871
2
(Σ) 119892 997891997888rarr 119906
119892
120590|Σ (8)
where 119906119892120590isin 119867
1
(Ω) solves (7) Λ(120590) is a self-adjoint compact
linear operatorLet 120590
0isin 119871
infin
+(Ω) be piecewise analytic For each point 119911 isin
Ω that has a neighborhood in which 1205900is analytic and each
unit vector 119889 isin R119899 119889 = 1 let Φ119911119889
be the solution of
nabla sdot 1205900nablaΦ
119911119889= 119889 sdot nabla120575
119911in Ω 120590
0120597]Φ119911119889|120597Ω = 0 (9)
Φ119911119889
is called a dipole function
22 Auxiliary Results Our presentation of the FactorizationMethod in the next section relies on the following fourlemmas The first lemma is frequently called a monotonylemma since it shows that a larger conductivity leads to asmaller NtD More precisely it shows a relation betweenthe difference of two NtDs and the difference of the corre-sponding conductivities and the interior energy of an electricpotential The second lemma shows that this energy term isthe image of the adjoint of an auxiliary virtual measurementoperator that is defined on a subregion ofΩThe third lemmais a functional analytic relation between the norm of animage of an operator and the range of its adjoint Togetherwith the first two lemmas it implies that the range of theauxiliary virtual measurement operator can be calculatedfrom the NtDs Finally using the previous dipole functionsthe last lemma shows that the range of the auxiliary virtualmeasurement operator determines the region on which theyare defined
We start with the monotony lemma
Lemma 1 Let 1205901 1205900isin 119871
infin
+(Ω) Then for all 119892 isin 119871
2
(120597Ω)
intΩ(1205900minus 120590
1)1003816100381610038161003816nabla119906
0
1003816100381610038161003816
2d119909 leintΣ119892 (Λ
1minus Λ
0) 119892d119904 leint
Ω
1205900
1205901
(1205900minus 120590
1)1003816100381610038161003816nabla119906
0
1003816100381610038161003816
2d119909
(10)
where we abbreviated Λ119895= Λ(120590
119895) 119895 = 0 1 and 119906
0= 119906
119892
1205900
Proof The lemma seems to go back to Ikehata Kang Seo andSheen [22 23] cf also the similar arguments inKirsch [8] Ideet al [24] and in the works of Seo and the author [16 25] Forthe sake of completeness we copy the short proof from [25]For all 119892 isin 119871
2
(120597Ω) we have that
intΩ
1205901nabla119906
1sdot nabla119906
0d119909 = int
Σ
1198921199060d119904
= intΩ
1205900nabla119906
0sdot nabla119906
0d119909 = int
Σ
119892Λ0119892d119904
(11)
Hence from
intΩ
1205901nabla (119906
1minus 119906
0) sdot nabla (119906
1minus 119906
0) d119909
= intΩ
1205901
1003816100381610038161003816nabla11990611003816100381610038161003816
2d119909 minus intΩ
1205900
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
+ intΩ
(1205901minus 120590
0)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909
(12)
we obtain that
intΣ
119892 (Λ1minus Λ
0) 119892d119904 = int
Ω
(1205900minus 120590
1)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909
+ intΩ
1205901
1003816100381610038161003816nabla (1199061minus 119906
0)1003816100381610038161003816
2d119909(13)
which already yields the first asserted inequality
Computational and Mathematical Methods in Medicine 3
By interchanging 1205901and 120590
0 we conclude that
intΣ
119892 (Λ0minus Λ
1) 119892d119904
= intΩ
(1205901minus 120590
0)1003816100381610038161003816nabla1199061
1003816100381610038161003816
2d119909 + intΩ
1205900
1003816100381610038161003816nabla (1199060minus 119906
1)1003816100381610038161003816
2d119909
= intΩ
(1205901
10038161003816100381610038161003816100381610038161003816
nabla1199061minus1205900
1205901
nabla1199060
10038161003816100381610038161003816100381610038161003816
2
+ (1205900minus1205902
0
1205901
)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2
) d119909
(14)
and hence obtain the second inequality
Given a reference conductivity 1205900
isin 119871infin
+(Ω) and a
measurable subset119863 sube Ω we define the virtual measurementoperator 119871
119863by
119871119863 1198712
(119863)119899
997888rarr 1198712
(Σ) 119865 997891997888rarr V|
Σ (15)
where V isin 1198671
(Ω) solves
intΩ
1205900nablaV sdot nabla119908d119909 = int
119863
119865 sdot nabla119908d119909 forall119908 isin 1198671
(Ω) (16)
The energy term |nabla1199060|2 in Lemma 1 can be identified with the
norm of the adjoint of this virtual measurement operator
Lemma 2 The adjoint operator of 119871119863is given by
119871lowast
119863 1198712
(Σ) 997888rarr 119871
2
(119863)119899
119892 997891997888rarr nabla1199060|119863 (17)
where 1199060isin 119867
1
(Ω) solves
nabla sdot 1205900nabla119906
0= 0 119894119899 Ω 120590
0120597]1199060|120597Ω =
119892 119900119899 Σ
0 119890119897119904119890(18)
Proof For all 119892 isin 1198712
(Σ) and 119865 isin 119871
2
(119863)119899 we have that
int119863
(119871lowast
119863119892) sdot 119865 d119909 = int
Σ
119892 (119871119863119865) d119909 = int
Σ
119892V|Σd119909
= intΩ
1205900nabla119906
0sdot nablaV d119909 = int
119863
nabla1199060sdot 119865 d119909
(19)
which shows the assertion
The following functional analytic lemma uses bounds onthe image of an operator to characterize the range of its dualoperator
Lemma 3 Let 119883 and 119884 be real Hilbert spaces with innerproducts (sdot sdot)
119883and (sdot sdot)
119884 respectively Let 119860 isin L(119883 119884) and
1199091015840
isin 119883 Then
1199091015840
isin R (119860lowast
) iff exist119862 gt 0 10038161003816100381610038161003816(1199091015840
119909)119883
10038161003816100381610038161003816le 119862 119860119909 forall119909 isin 119883
(20)
In particular if1198831198841 and119884
2are three real Hilbert spaces119860
119894isin
L(119884119894 119883) 119894 = 1 2 and if there exists 119862 gt 0 with
1003817100381710038171003817119860lowast
11199091003817100381710038171003817 le 119862
1003817100381710038171003817119860lowast
21199091003817100381710038171003817 forall119909 isin 119883 (21)
thenR(1198601) sube R(119860
2)
Proof The assertion can be generalized to Banach spacesand in that context it is called the ldquo14th important propertyof Banach spacesrdquo in Bourbaki [26] For the sake of complete-ness we rewrite the proof from [27] to Hilbert spaces
If 1199091015840 isin R(119860lowast
) then there exists 1199101015840 isin 119884 such that 1199091015840 =119860lowast
1199101015840 Hence10038161003816100381610038161003816(1199091015840
119909)119883
10038161003816100381610038161003816=10038161003816100381610038161003816(119860
lowast
1199101015840
119909)119883
10038161003816100381610038161003816
=10038161003816100381610038161003816(1199101015840
119860119909)119884
10038161003816100381610038161003816le10038171003817100381710038171003817119910101584010038171003817100381710038171003817119860119909 forall119909 isin 119883
(22)
so that the assertion holds with 119862 = 1199101015840
Now let 1199091015840 isin 119883 be such that there exists 119862 gt 0 with
|(1199091015840
119909)119883| le 119862119860119909 for all 119909 isin 119883 We define
119891 (119910) = (1199091015840
119909)119883
for every 119910 = 119860119909 isin R (119860) (23)
Then 119891 is a well-defined continuous linear functional onR(119860) By setting it to zero on R(119860)
perp we can extend 119891 to acontinuous linear functional on 119884 Using the Riesz theoremit follows that there exists 1199101015840 isin 119884 with
(1199101015840
119910)119884
= 119891 (119910) forall119910 isin R (119860) (24)
Hence for all 119909 isin 119883 we have
(119860lowast
1199101015840
119909)119883
= (1199101015840
119860119909)119884
= 119891 (119860119909) = (1199091015840
119909)119883
(25)
so that 1199091015840 = 119860lowast
1199101015840
isin R(119860lowast
)
The last lemma shows that the range of the virtualmeasurement operator 119871
119863determines the region119863 onwhich
it is defined We state the lemma for a simple special casea generalized version of the lemma will be formulated inSection 32
Lemma 4 Let 1205900= 1 119863 sube Ω be open and 119863 sube Ω have a
connected complement Ω 119863Then for all unit vectors 119889 isin R119899 119889 = 1 and every point
119911 isin Ω 120597119863 it holds that
119911 isin 119863 iff Φ119911119889|Σisin R (119871
119863) (26)
Proof The proof is similar to the one of [21 Lemma 29]First let 119911 isin 119863 and 120598 gt 0 be such that 119861
120598(119911) sube 119863 We
choose
1198911isin 119867
1
(119861120598(119911)) with 119891
1|120597119861120598(119911)
= Φ119911119889|120597119861120598(119911)
1198912isin 119867
1
(119861120598(119911)) with Δ119891
2= 0
120597]1198912|120597119861120598(119911)= 120597]Φ119911119889|120597119861120598(119911)
(27)
and let 119865 isin 1198712
(119863)119899 be the zero continuation of nabla(119891
1minus 119891
2) to
119863Then the function
V = Φ119911119889
in Ω 119861120598(119911)
1198911
in 119861120598(119911)
(28)
4 Computational and Mathematical Methods in Medicine
fulfills V isin 1198671
(Ω) and for all 119908 isin 119867
1
(Ω)
intΩ
nablaV sdot nabla119908d119909
= intΩ119861120598(119911)
nablaΦ119911119889
sdot nabla119908d119909 + int119861120598(119911)
nabla1198911sdot nabla119908d119909
= minusint120597119861120598(119911)
120597]Φ119911119889 119908|120597119861120598(119911)d119904 + int
119861120598(119911)
nabla1198911sdot nabla119908d119909
= int119861120598(119911)
nabla (1198911minus 119891
2) sdot nabla119908d119909 = int
119863
119865 sdot nabla119908d119909
(29)
This shows thatΦ119911119889|Σ= V|
Σ= 119871
119863(119865) isin R(119871
119863)
Now let Φ119911119889|Σisin R(119871
119863) Let V isin 119867
1
(Ω) be the function
from the definition of 119871119863 Then
V|Σ= Φ
119911119889|Σ 120597]V|Σ = 0 = 120597]Φ119911119889|Σ (30)
so that it follows by unique continuation that V = Φ119911119889
in theconnected setΩ (119863 cup 119911)
If 119911 notin 119863 then 119889 sdot nabla120575119911notin 119867
minus2
(Ω 119863) and thus Φ119911119889
notin
1198712
(Ω 119863) which contradicts that V = Φ119911119889
in Ω (119863 cup 119911)Hence 119911 isin 119863
3 The Factorization Method
Nowwe will formulate the FactorizationMethod and charac-terize a region where a conductivity 120590 differs from a referenceconductivity 120590
0by a range criterion Before we turn to a
new general formulation of the method we first state it fora special case that is similar to the one that was treated in theoriginal works of Bruhl and Hanke [3 4]
31 The Factorization Method for a Simple Special Case
Theorem 5 Let 1205900= 1 and 120590 = 1 + 120594
119863 where 119863 sube Ω is an
open set so that 119863 sube Ω has a connected complement Ω 119863Then for all 119911 isin Ω 119911 notin 120597119863 and all dipole directions 119889 isin R119899119889 = 1
119911 isin 119863 iff Φ119911119889
1003816100381610038161003816Σisin R (|Λ (120590) minus Λ (1)|
12
) (31)
Proof Themonotony Lemma 1 yields that for all119892 isin 1198712
(120597Ω)
int119863
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 ge intΣ
119892 (Λ (1) minus Λ (120590)) 119892d119904 ge int119863
1
2
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(32)
Hence |Λ(120590) minusΛ(1)| = Λ(1) minusΛ(120590) and using Lemma 2 wecan restate this in the form
1003817100381710038171003817119871lowast
1198631198921003817100381710038171003817
2
ge10038171003817100381710038171003817|Λ (120590) minus Λ (1)|
12
11989210038171003817100381710038171003817
2
ge1
2
1003817100381710038171003817119871lowast
1198631198921003817100381710038171003817
2
(33)
Using the functional analytic Lemma 3 this implies that
R (119871119863) = R (|Λ (120590) minus Λ (1)|
12
) (34)
and thus the assertion follows from the relation between 119863
andR(119871119863) in Lemma 4
Obviously the same arguments can be used to treat thecase 120590(119909) = 1 + 120581(119909)120594
119863(119909) when there exists a conductivity
jump 120598 gt 0 so that either
120581 (119909) ge 120598 forall119909 isin 119863 or 120581 (119909) le minus120598 forall119909 isin 119863 (35)
32 The Factorization Method for the General PiecewiseAnalytic Case Now we drop the assumptions that the back-ground is constant that there is a clear conductivity jumpand that the complement of the inclusions is connected Wewill merely assume that the reference conductivity 120590
0is a
piecewise analytic function and that either 120590 minus 1205900ge 0 or
120590 minus 1205900le 0 Roughly speaking under this general assumption
the Factorization Method then characterizes the support of120590 minus 120590
0up to holes in the support that have no connections to
Σ For a precise formulation we use the concept of the innerand outer support from [28] that has been inspired by the useof the infinity support of Kusiak and Sylvester [29] see also[25 30]
Definition 6 A relatively open set 119880 sube Ω is called connectedto Σ if 119880 cap Ω is connected and 119880 cap Σ = 0
For a measurable function 120581 Ω rarr R we define
(a) the support supp(120581) as the complement (in Ω) of theunion of those relatively open119880 sube Ω for which 120581|
119880equiv
0(b) the inner support inn supp 120581 as the union of those
open sets 119880 sube Ω for which ess inf119909isin119880
|120581(119909)| gt 0(c) the outer support out
Σsupp 120581 as the complement (in
Ω) of the union of those relatively open 119880 sube Ω thatare connected to Σ and for which 120581|
119880equiv 0
The interior of a set119872 sube Ω is denoted by int119872 and its closure(with respect toR119899) by119872 If119872 is measurable we also define
(d) outΣ119872 = out
Σsupp120594
119872
It is easily checked that outΣ(supp 120581) = out
Σsupp 120581
With this concept we can extend the range characteriza-tion in Lemma 4 to a general setting (see also Remark 9 later)
Lemma 7 Let 1205900isin 119871
infin
+(Ω) be piecewise analytic Let 119863 sube Ω
be measurableThen for all unit vectors 119889 isin R119899 119889 = 1 and every point
119911 isin Ω that has a neighborhood in which 1205900is analytic
119911 isin int119863 implies Φ119911119889
1003816100381610038161003816Σisin R (119871
119863) (36)
and
Φ119911119889
1003816100381610038161003816Σisin R (119871
119863) implies 119911 isin out
Σ119863 (37)
Proof If 119911 isin int119863 then there exists a small ball 119861120598(119911) sube 119863
and the first assertion follows as in the proof of Lemma 4To show the second assertion letΦ
119911119889|Σisin R(119871
119863) and let
V isin 1198671
(Ω) be the function from the definition of 119871
119863 so that
(as in the proof of Lemma 4)
V|Σ= Φ
119911119889|Σ 120597]V|Σ = 0 = 120597]Φ119911119889|Σ (38)
Computational and Mathematical Methods in Medicine 5
Assume that 119911 notin outΣ119863 Then there exists a relatively
open 119880 sube Ω that is connected to Σ and contains 119911Hence by unique continuation it follows that V|
119880= Φ
119911119889|119880
and we obtain the same contradiction as in the proof ofLemma 4
Now we can formulate and prove the FactorizationMethod for general piecewise analytic conductivities
Theorem 8 Let 120590 isin 119871infin
+(Ω) and let 120590
0isin 119871
infin
+(Ω) be a
piecewise analytic function Let either
120590 (119909) ge 1205900(119909) forall119909 isin Ω or 120590 (119909) le 120590
0(119909) forall119909 isin Ω
(39)
Then for all 119911 isin Ω that have a neighborhood in which 1205900is
analytic as well as all unit vectors 119889 isin R119899 119889 = 1
119911isin innsupp (120590 minus 1205900) implies Φ
119911119889
1003816100381610038161003816Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
)
(40)
and
Φ119911119889
1003816100381610038161003816ΣisinR (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) implies 119911isin outΣsupp (120590 minus 120590
0)
(41)
Proof Let 119911 isin Ω have a neighborhood in which 1205900is analytic
and let 119889 isin R119899 be a unit vector with 119889 = 1 We onlyprove the assertions for 120590 ge 120590
0 The other case is completely
analogousFirst let 119911 isin inn supp (120590 minus 120590
0) Then there exists a small
ball 119861120598(119911) and 120575 gt 0 so that 120590 minus 120590
0ge 120575 on 119861
120598(119911) Using the
monotony Lemma 1 it follows that for all 119892 isin 1198712
(120597Ω)
intΣ
119892 (Λ (1205900) minus Λ (120590)) 119892d119904
ge intΩ
1205900
120590(120590 minus 120590
0)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909
ge 12057510038171003817100381710038171205900
1003817100381710038171003817119871infin(Ω)
1003817100381710038171003817100381710038171003817
1
120590
1003817100381710038171003817100381710038171003817119871infin(Ω)int119861120598(119911)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
ge 12057510038171003817100381710038171205900
1003817100381710038171003817119871infin(Ω)
1003817100381710038171003817100381710038171003817
1
120590
1003817100381710038171003817100381710038171003817119871infin(Ω)
10038171003817100381710038171003817119871lowast
119861120598(119911)11989210038171003817100381710038171003817
2
(42)
Using the functional analytic Lemma 3 we obtain that
R (119871119861120598(119911)
) sube R (1003816100381610038161003816Λ (120590) minus Λ (120590
0)1003816100381610038161003816
12
) (43)
and Lemma 7 yields that
Φ119911119889|Σisin R (119871
119861120598(119911)) sube R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) (44)
On the other hand with 119863 = supp(120590 minus 1205900) the monotony
Lemma 1 shows that for all 119892 isin 1198712
(120597Ω)
intΣ
119892 (Λ (1205900) minus Λ (120590
1)) 119892d119904 le int
Ω
(1205901minus 120590
0)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909
le10038171003817100381710038171205901 minus 120590
0
1003817100381710038171003817119871infin(Ω)int119863
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
le10038171003817100381710038171205901 minus 120590
0
1003817100381710038171003817119871infin(Ω)
1003817100381710038171003817119871lowast
1198631198921003817100381710038171003817
2
(45)
so that we obtain from the functional analytic Lemma 3
R (1003816100381610038161003816Λ (120590) minus Λ (120590
0)1003816100381610038161003816
12
) sube R (119871119863) (46)
Hence Lemma 7 yields that
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) sube R (119871119863) implies
119911 isin outΣ(supp (120590 minus 120590
0)) = out
Σsupp (120590 minus 120590
0)
(47)
Remark 9 Theorem 8 shows that the Factorization Methodis able to detect the support of a conductivity difference up tothe difference between the outer and the inner support thatis roughly speaking up to holes in the support that have noconnections to the boundary It leaves openwhether points insuch holes will fulfill the range criterion of the FactorizationMethod or not
A result of Hyvonen and the author [30 Lemma 25]shows that for every smooth domain119863with119863 sub Ω and everyunit vector 119889 isin R119899 119889 = 1
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) forall119911 isin 120597119863 (48)
implies
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) forall119911 isin 119863 (49)
In that sense we can expect that holes in the support willbe filled up and that the set detected by the FactorizationMethod is essentially the outer support of the conductivitydifference
33 The Factorization Method for the Indefinite Case It is along standing open theoretical problem whether the rangecriterion of the Factorization Method holds true withoutthe definiteness assumption that 120590 ge 120590
0on Ω or 120590 le
1205900on Ω However Grinberg Kirsch and Schmitt [18 31]
showed how to exclude a region 119864 sube Ω from Ω insuch a way that the Factorization Method only requires thedefiniteness assumption onΩ119864 In this subsection we showhow their idea can be incorporated into our formulation ofthe method
To point out the main idea we first formulate the resultfor a simple special case Let us stress that for 120590 = 1 + 120594
119863+ minus
(12)120594119863minus it is not known whether
119911 isin 119863+
cup 119863minus iff Φ
119911119889|Σisin R (|Λ (120590) minus Λ (1)|
12
) (50)
However we can still use the FactorizationMethod if we havesome a priori knowledge that separates 119863+ and 119863
minus Moreprecisely if we know a subset 119864 that contains 119863minus withoutintersecting 119863
+ then we can use the Factorization Methodto find119863+ cup 119864 (and thus119863+)
6 Computational and Mathematical Methods in Medicine
Theorem 10 Let 1205900= 1 and 120590 = 1 + 120594
119863+ minus (12)120594
119863minus where
119863+
119863minus
sube Ω are open Let 119864 sube Ω be an open set
(a) If 119863+ sube 119864 and 119863minus cup 119864 sube Ω has a connectedcomplement then for all 119911 isin Ω 119911 notin 120597(119863
minus
cup119864) and alldipole directions 119889 isin R119899 119889 = 1
119911 isin 119863minus
cup 119864 iff Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1) + 119871119864119871lowast
119864
1003816100381610038161003816
12
)
(51)
(b) If 119863minus sube 119864 and 119863+ cup 119864 sube Ω has a connectedcomplement then for all 119911 isin Ω 119911 notin 120597(119863
+
cup119864) and alldipole directions 119889 isin R119899 119889 = 1
119911 isin 119863+
cup 119864 iff Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1) minus 119871119864119871lowast
119864
1003816100381610038161003816
12
)
(52)
Proof Themonotony Lemma 1 yields that for all119892 isin 1198712
(120597Ω)
1
2int119863minus
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 minus int119863+
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
le intΣ
119892 (Λ (120590) minus Λ (1)) 119892d119904
le int119863minus
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 minus1
2int119863+
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(53)
Since (cf Lemma 2)
intΣ
119892 (119871119864119871lowast
119864) 119892d119904 = 1003817100381710038171003817119871
lowast
1198641198921003817100381710038171003817
2
= int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 (54)
it follows for case (a) that
1
2int119863minuscup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 le intΣ
119892 (Λ (120590) minus Λ (1) + 2119871119864119871lowast
119864) 119892d119904
le int119863minuscup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909(55)
Using the functional analytic Lemma 3 this implies that
R (119871119863minuscup119864) = R (
1003816100381610038161003816Λ (120590) minus Λ (1) + 2119871119864119871lowast
119864
1003816100381610038161003816
12
) (56)
so that the assertion (a) follows from Lemma 4In case (b) we obtain that
3
2int119863+cup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 ge intΣ
119892 (Λ (1) minus Λ (120590) + 2119871119864119871lowast
119864) 119892d119904
ge1
2int119863+cup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909(57)
and the same arguments as above yield the assertion
We can also extend these ideas to the general setting ofSection 32
Theorem 11 Let 120590 isin 119871infin
+(Ω) and let 120590
0isin 119871
infin
+(Ω) be a
piecewise analytic function Let 119864 sube Ω be a measurable setChoose 120572 120573 isin R such that
120572 gt1003817100381710038171003817120590 minus 120590
0
1003817100381710038171003817119871infin(Ω) 120573 gt
1003817100381710038171003817100381710038171003817
1205900
120590(1205900minus 120590)
1003817100381710038171003817100381710038171003817119871infin(Ω) (58)
(a) If 120590 le 1205900on Ω 119864 then for all 119911 isin Ω that have a
neighborhood in which 1205900is analytic as well as all unit
vectors 119889 isin R119899 119889 = 1
119911 isin inn supp (120590 minus 1205900) cup 119864 implies
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590)minusΛ (1205900)+120572119871
119864119871lowast
119864
1003816100381610038161003816
12
)
(59)
and
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864
1003816100381610038161003816
12
) implies
119911 isin outΣ(supp (120590 minus 120590
0) cup 119864)
(60)
(b) If 120590 ge 1205900on Ω 119864 then for all 119911 isin Ω that have a
neighborhood in which 1205900is analytic as well as all unit
vectors 119889 isin R119899 119889 = 1
119911 isin inn supp (120590 minus 1205900) cup 119864 implies
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900) minus 120573119871
119864119871lowast
119864
1003816100381610038161003816
12
)
(61)
and
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900) minus 120573119871
119864119871lowast
119864
1003816100381610038161003816
12
) implies
119911 isin outΣ(supp (120590 minus 120590
0) cup 119864)
(62)
Proof For every 119911 isin inn supp(120590minus1205900)with 119911 notin 119864 there exists
a small ball 119861120598(119911) and 120575 gt 0 so that 120590
0minus120590 ge 120575 on 119861
120598(119911) Using
the monotony Lemma 1 it follows that for all 119892 isin 1198712
(120597Ω)
intΣ
119892 (Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864) 119892d119904
ge intΩ
(1205900minus 120590
1)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909 + 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
ge 120575int119861120598(119911)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909+(120572 minus1003817100381710038171003817120590 minus 120590
0
1003817100381710038171003817119871infin(Ω))int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(63)
Computational and Mathematical Methods in Medicine 7
As in the previous proofs we obtain from Lemmas 2 and 3that
R (1003816100381610038161003816Λ (120590) minus Λ (120590
0) + 120572119871
119864119871lowast
119864
1003816100381610038161003816
12
) supe R (119871119861120598(119911)cup119864
) (64)
so that the first implication of (a) follows from Lemma 7The monotony Lemma 1 also implies that
intΣ
119892 (Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864) 119892d119904
le intΩ
1205900
120590(1205900minus 120590)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 + 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
le
1003817100381710038171003817100381710038171003817
1205900
120590(1205900minus 120590)
1003817100381710038171003817100381710038171003817119871infin(Ω)intsupp(120590minus1205900)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
+ 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(65)
so that the second implication of (a) follows from Lemmas 23 and 7 Assertion (b) can be proven analogously
Remark 9 also applies to this case
4 Conclusions and Remarks
The Factorization Method can be used to detect regionsin which a conductivity differs from a known referenceconductivity In this work we summarized the progresson the methodrsquos theoretical foundation We formulated themethod for general piecewise analytic conductivities andgave comparatively simple and self-contained proofsWe alsoshowed how the idea of excluding a part of the imaging regioncan be incorporated into this formulation
The regularity assumptions can beweakened even furtherOur proofs only require unique continuation arguments forthe reference conductivity 120590
0and the existence of the dipole
functionsTwo major open theoretical questions still exist in the
context of the FactorizationMethodThe theoretical justifica-tion of the method requires a definiteness condition (on thewhole domain or after excluding an a priori known part of thedomain) It is unknownwhether themethodrsquos range criterionholds without such a definiteness conditionThe second openquestion concerns the numerical stability of the methodrsquosrange criterion So far there are no rigorous convergenceresults for numerical implementations of this range criterion(see however Lechleiter [32] for a first step in this direction)As a promising approach to overcome both problems wewould like to point out the recent work on monotony-basedmethods [28]
Acknowledgments
This paper was funded by the German Research Foundation(DFG) and the University of Wurzburg in the fundingprogramme Open Access Publishing
References
[1] A Kirsch ldquoCharacterization of the shape of a scattering obstacleusing the spectral data of the far field operatorrdquo Inverse Prob-lems vol 14 no 6 pp 1489ndash1512 1998
[2] M Bruhl Gebietserkennung in der elektrischen Impedanztomo-graphie [PhD thesis] Universitat Karlsruhe 1999
[3] M Bruhl and M Hanke ldquoNumerical implementation of twononiterative methods for locating inclusions by impedancetomographyrdquo Inverse Problems vol 16 no 4 pp 1029ndash10422000
[4] M Bruhl ldquoExplicit characterization of inclusions in electricalimpedance tomographyrdquo SIAM Journal on Mathematical Anal-ysis vol 32 no 6 pp 1327ndash1341 2001
[5] M Hanke and M Bruhl ldquoRecent progress in electricalimpedance tomographyrdquo Inverse Problems vol 19 no 6 ppS65ndashS90 2003
[6] A Kirsch andN GrinbergTheFactorizationMethod for InverseProblems vol 36 ofOxford Lecture Series inMathematics and ItsApplications Oxford University Press Oxford UK 2007
[7] M Hanke and A Kirsch ldquoSampling methodsrdquo in Handbook ofMathematical Models in Imaging O Scherzer Ed pp 501ndash550Springer 2011
[8] A Kirsch ldquoThe factorization method for a class of inverseelliptic problemsrdquo Mathematische Nachrichten vol 278 pp258ndash277 2005
[9] B Gebauer ldquoThe Factorization Method for real elliptic prob-lemsrdquo Zeitschrift fur Analysis und ihre Anwendung vol 25 no1 pp 81ndash102 2006
[10] A I Nachman L Paivarinta and A Teirila ldquoOn imagingobstacles inside inhomogeneous mediardquo Journal of FunctionalAnalysis vol 252 no 2 pp 490ndash516 2007
[11] M Hanke and B Schappel ldquoThe factorization method forelectrical impedance tomography in the half-spacerdquo SIAMJournal on Applied Mathematics vol 68 no 4 pp 907ndash9242008
[12] N Hyvonen ldquoComplete electrode model of electricalimpedance tomography approximation properties andcharacterization of inclusionsrdquo SIAM Journal on AppliedMathematics vol 64 pp 902ndash931 2004
[13] A Lechleiter N Hyvonen and H Hakula ldquoThe factorizationmethod applied to the complete electrode model of impedancetomographyrdquo SIAM Journal onAppliedMathematics vol 68 no4 pp 1097ndash1121 2008
[14] N Hyvonen H Hakula and S Pursiainen ldquoNumerical imple-mentation of the factorization method within the completeelectrode model of electrical impedance tomographyrdquo InverseProblems and Imaging vol 1 pp 299ndash317 2007
[15] HHakula andNHyvonen ldquoOn computation of test dipoles forfactorization methodrdquo BIT Numerical Mathematics vol 49 no1 pp 75ndash91 2009
[16] B Harrach and J K Seo ldquoDetecting inclusions in electri-cal impedance tomography without reference measurementsrdquoSIAM Journal on Applied Mathematics vol 69 no 6 pp 1662ndash1681 2009
[17] B Harrach J K Seo and E J Woo ldquoFactorization methodand its physical justification in frequency-difference electricalimpedance tomographyrdquo IEEE Transactions on Medical Imag-ing vol 29 no 11 pp 1918ndash1926 2010
[18] S Schmitt ldquoThe factorization method for EIT in the case ofmixed inclusionsrdquo Inverse Problems vol 25 no 6 Article ID065012 2009
8 Computational and Mathematical Methods in Medicine
[19] S Schmitt and A Kirsch ldquoA factorization scheme for determin-ing conductivity contrasts in impedance tomographyrdquo InverseProblems vol 27 Article ID 095005 2011
[20] B Gebauer and N Hyvonen ldquoFactorization method and irreg-ular inclusions in electrical impedance tomographyrdquo InverseProblems vol 23 no 5 pp 2159ndash2170 2007
[21] B Gebauer ldquoLocalized potentials in electrical impedancetomographyrdquo Inverse Problems and Imaging vol 2 pp 251ndash2692008
[22] H Kang J K Seo and D Sheen ldquoThe inverse conductivityproblem with one measurement Stability and estimation ofsizerdquo SIAM Journal on Mathematical Analysis vol 28 no 6 pp1389ndash1405 1997
[23] M Ikehata ldquoSize estimation of inclusionrdquo Journal of Inverse andIll-Posed Problems vol 6 no 2 pp 127ndash140 1998
[24] T Ide H Isozaki S Nakata S Siltanen and G UhlmannldquoProbing for electrical inclusions with complex sphericalwavesrdquo Communications on Pure and Applied Mathematics vol60 no 10 pp 1415ndash1442 2007
[25] B Harrach and J K Seo ldquoExact shape-reconstruction by one-step linearization in electrical impedance tomographyrdquo TheSIAM Journal onMathematical Analysis vol 42 no 4 pp 1505ndash1518 2010
[26] N Bourbaki Elements of Mathematics Topological VectorSpaces chapters 1ndash5 Springer Berlin Germany 2003
[27] F Fruhauf B Gebauer andO Scherzer ldquoDetecting interfaces ina parabolic-elliptic problem from surfacemeasurementsrdquo SIAMJournal on Numerical Analysis vol 45 no 2 pp 810ndash836 2007
[28] B Harrach and M Ullrich ldquoMonotony based shape recon-struction in electrical impedance tomographyrdquo submitted forpublication
[29] S Kusiak and J Sylvester ldquoThe scattering supportrdquoCommunica-tions on Pure and Applied Mathematics vol 56 no 11 pp 1525ndash1548 2003
[30] B Gebauer and N Hyvonen ldquoFactorization method andinclusions of mixed type in an inverse elliptic boundary valueproblemrdquo Inverse Problems and Imaging vol 2 pp 355ndash3722008
[31] N I Grinberg and A Kirsch ldquoThe factorization method forobstacles with a-priori separated sound-soft and sound-hardpartsrdquo Mathematics and Computers in Simulation vol 66 no4-5 pp 267ndash279 2004
[32] A Lechleiter ldquoA regularization technique for the factorizationmethodrdquo Inverse Problems vol 22 no 5 pp 1605ndash1625 2006
Submit your manuscripts athttpwwwhindawicom
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MEDIATORSINFLAMMATION
of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Behavioural Neurology
EndocrinologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Disease Markers
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
OncologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Oxidative Medicine and Cellular Longevity
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
PPAR Research
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Immunology ResearchHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
ObesityJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational and Mathematical Methods in Medicine
OphthalmologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Diabetes ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentAIDS
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Gastroenterology Research and Practice
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Parkinsonrsquos Disease
Evidence-Based Complementary and Alternative Medicine
Volume 2014Hindawi Publishing Corporationhttpwwwhindawicom
Computational and Mathematical Methods in Medicine 3
By interchanging 1205901and 120590
0 we conclude that
intΣ
119892 (Λ0minus Λ
1) 119892d119904
= intΩ
(1205901minus 120590
0)1003816100381610038161003816nabla1199061
1003816100381610038161003816
2d119909 + intΩ
1205900
1003816100381610038161003816nabla (1199060minus 119906
1)1003816100381610038161003816
2d119909
= intΩ
(1205901
10038161003816100381610038161003816100381610038161003816
nabla1199061minus1205900
1205901
nabla1199060
10038161003816100381610038161003816100381610038161003816
2
+ (1205900minus1205902
0
1205901
)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2
) d119909
(14)
and hence obtain the second inequality
Given a reference conductivity 1205900
isin 119871infin
+(Ω) and a
measurable subset119863 sube Ω we define the virtual measurementoperator 119871
119863by
119871119863 1198712
(119863)119899
997888rarr 1198712
(Σ) 119865 997891997888rarr V|
Σ (15)
where V isin 1198671
(Ω) solves
intΩ
1205900nablaV sdot nabla119908d119909 = int
119863
119865 sdot nabla119908d119909 forall119908 isin 1198671
(Ω) (16)
The energy term |nabla1199060|2 in Lemma 1 can be identified with the
norm of the adjoint of this virtual measurement operator
Lemma 2 The adjoint operator of 119871119863is given by
119871lowast
119863 1198712
(Σ) 997888rarr 119871
2
(119863)119899
119892 997891997888rarr nabla1199060|119863 (17)
where 1199060isin 119867
1
(Ω) solves
nabla sdot 1205900nabla119906
0= 0 119894119899 Ω 120590
0120597]1199060|120597Ω =
119892 119900119899 Σ
0 119890119897119904119890(18)
Proof For all 119892 isin 1198712
(Σ) and 119865 isin 119871
2
(119863)119899 we have that
int119863
(119871lowast
119863119892) sdot 119865 d119909 = int
Σ
119892 (119871119863119865) d119909 = int
Σ
119892V|Σd119909
= intΩ
1205900nabla119906
0sdot nablaV d119909 = int
119863
nabla1199060sdot 119865 d119909
(19)
which shows the assertion
The following functional analytic lemma uses bounds onthe image of an operator to characterize the range of its dualoperator
Lemma 3 Let 119883 and 119884 be real Hilbert spaces with innerproducts (sdot sdot)
119883and (sdot sdot)
119884 respectively Let 119860 isin L(119883 119884) and
1199091015840
isin 119883 Then
1199091015840
isin R (119860lowast
) iff exist119862 gt 0 10038161003816100381610038161003816(1199091015840
119909)119883
10038161003816100381610038161003816le 119862 119860119909 forall119909 isin 119883
(20)
In particular if1198831198841 and119884
2are three real Hilbert spaces119860
119894isin
L(119884119894 119883) 119894 = 1 2 and if there exists 119862 gt 0 with
1003817100381710038171003817119860lowast
11199091003817100381710038171003817 le 119862
1003817100381710038171003817119860lowast
21199091003817100381710038171003817 forall119909 isin 119883 (21)
thenR(1198601) sube R(119860
2)
Proof The assertion can be generalized to Banach spacesand in that context it is called the ldquo14th important propertyof Banach spacesrdquo in Bourbaki [26] For the sake of complete-ness we rewrite the proof from [27] to Hilbert spaces
If 1199091015840 isin R(119860lowast
) then there exists 1199101015840 isin 119884 such that 1199091015840 =119860lowast
1199101015840 Hence10038161003816100381610038161003816(1199091015840
119909)119883
10038161003816100381610038161003816=10038161003816100381610038161003816(119860
lowast
1199101015840
119909)119883
10038161003816100381610038161003816
=10038161003816100381610038161003816(1199101015840
119860119909)119884
10038161003816100381610038161003816le10038171003817100381710038171003817119910101584010038171003817100381710038171003817119860119909 forall119909 isin 119883
(22)
so that the assertion holds with 119862 = 1199101015840
Now let 1199091015840 isin 119883 be such that there exists 119862 gt 0 with
|(1199091015840
119909)119883| le 119862119860119909 for all 119909 isin 119883 We define
119891 (119910) = (1199091015840
119909)119883
for every 119910 = 119860119909 isin R (119860) (23)
Then 119891 is a well-defined continuous linear functional onR(119860) By setting it to zero on R(119860)
perp we can extend 119891 to acontinuous linear functional on 119884 Using the Riesz theoremit follows that there exists 1199101015840 isin 119884 with
(1199101015840
119910)119884
= 119891 (119910) forall119910 isin R (119860) (24)
Hence for all 119909 isin 119883 we have
(119860lowast
1199101015840
119909)119883
= (1199101015840
119860119909)119884
= 119891 (119860119909) = (1199091015840
119909)119883
(25)
so that 1199091015840 = 119860lowast
1199101015840
isin R(119860lowast
)
The last lemma shows that the range of the virtualmeasurement operator 119871
119863determines the region119863 onwhich
it is defined We state the lemma for a simple special casea generalized version of the lemma will be formulated inSection 32
Lemma 4 Let 1205900= 1 119863 sube Ω be open and 119863 sube Ω have a
connected complement Ω 119863Then for all unit vectors 119889 isin R119899 119889 = 1 and every point
119911 isin Ω 120597119863 it holds that
119911 isin 119863 iff Φ119911119889|Σisin R (119871
119863) (26)
Proof The proof is similar to the one of [21 Lemma 29]First let 119911 isin 119863 and 120598 gt 0 be such that 119861
120598(119911) sube 119863 We
choose
1198911isin 119867
1
(119861120598(119911)) with 119891
1|120597119861120598(119911)
= Φ119911119889|120597119861120598(119911)
1198912isin 119867
1
(119861120598(119911)) with Δ119891
2= 0
120597]1198912|120597119861120598(119911)= 120597]Φ119911119889|120597119861120598(119911)
(27)
and let 119865 isin 1198712
(119863)119899 be the zero continuation of nabla(119891
1minus 119891
2) to
119863Then the function
V = Φ119911119889
in Ω 119861120598(119911)
1198911
in 119861120598(119911)
(28)
4 Computational and Mathematical Methods in Medicine
fulfills V isin 1198671
(Ω) and for all 119908 isin 119867
1
(Ω)
intΩ
nablaV sdot nabla119908d119909
= intΩ119861120598(119911)
nablaΦ119911119889
sdot nabla119908d119909 + int119861120598(119911)
nabla1198911sdot nabla119908d119909
= minusint120597119861120598(119911)
120597]Φ119911119889 119908|120597119861120598(119911)d119904 + int
119861120598(119911)
nabla1198911sdot nabla119908d119909
= int119861120598(119911)
nabla (1198911minus 119891
2) sdot nabla119908d119909 = int
119863
119865 sdot nabla119908d119909
(29)
This shows thatΦ119911119889|Σ= V|
Σ= 119871
119863(119865) isin R(119871
119863)
Now let Φ119911119889|Σisin R(119871
119863) Let V isin 119867
1
(Ω) be the function
from the definition of 119871119863 Then
V|Σ= Φ
119911119889|Σ 120597]V|Σ = 0 = 120597]Φ119911119889|Σ (30)
so that it follows by unique continuation that V = Φ119911119889
in theconnected setΩ (119863 cup 119911)
If 119911 notin 119863 then 119889 sdot nabla120575119911notin 119867
minus2
(Ω 119863) and thus Φ119911119889
notin
1198712
(Ω 119863) which contradicts that V = Φ119911119889
in Ω (119863 cup 119911)Hence 119911 isin 119863
3 The Factorization Method
Nowwe will formulate the FactorizationMethod and charac-terize a region where a conductivity 120590 differs from a referenceconductivity 120590
0by a range criterion Before we turn to a
new general formulation of the method we first state it fora special case that is similar to the one that was treated in theoriginal works of Bruhl and Hanke [3 4]
31 The Factorization Method for a Simple Special Case
Theorem 5 Let 1205900= 1 and 120590 = 1 + 120594
119863 where 119863 sube Ω is an
open set so that 119863 sube Ω has a connected complement Ω 119863Then for all 119911 isin Ω 119911 notin 120597119863 and all dipole directions 119889 isin R119899119889 = 1
119911 isin 119863 iff Φ119911119889
1003816100381610038161003816Σisin R (|Λ (120590) minus Λ (1)|
12
) (31)
Proof Themonotony Lemma 1 yields that for all119892 isin 1198712
(120597Ω)
int119863
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 ge intΣ
119892 (Λ (1) minus Λ (120590)) 119892d119904 ge int119863
1
2
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(32)
Hence |Λ(120590) minusΛ(1)| = Λ(1) minusΛ(120590) and using Lemma 2 wecan restate this in the form
1003817100381710038171003817119871lowast
1198631198921003817100381710038171003817
2
ge10038171003817100381710038171003817|Λ (120590) minus Λ (1)|
12
11989210038171003817100381710038171003817
2
ge1
2
1003817100381710038171003817119871lowast
1198631198921003817100381710038171003817
2
(33)
Using the functional analytic Lemma 3 this implies that
R (119871119863) = R (|Λ (120590) minus Λ (1)|
12
) (34)
and thus the assertion follows from the relation between 119863
andR(119871119863) in Lemma 4
Obviously the same arguments can be used to treat thecase 120590(119909) = 1 + 120581(119909)120594
119863(119909) when there exists a conductivity
jump 120598 gt 0 so that either
120581 (119909) ge 120598 forall119909 isin 119863 or 120581 (119909) le minus120598 forall119909 isin 119863 (35)
32 The Factorization Method for the General PiecewiseAnalytic Case Now we drop the assumptions that the back-ground is constant that there is a clear conductivity jumpand that the complement of the inclusions is connected Wewill merely assume that the reference conductivity 120590
0is a
piecewise analytic function and that either 120590 minus 1205900ge 0 or
120590 minus 1205900le 0 Roughly speaking under this general assumption
the Factorization Method then characterizes the support of120590 minus 120590
0up to holes in the support that have no connections to
Σ For a precise formulation we use the concept of the innerand outer support from [28] that has been inspired by the useof the infinity support of Kusiak and Sylvester [29] see also[25 30]
Definition 6 A relatively open set 119880 sube Ω is called connectedto Σ if 119880 cap Ω is connected and 119880 cap Σ = 0
For a measurable function 120581 Ω rarr R we define
(a) the support supp(120581) as the complement (in Ω) of theunion of those relatively open119880 sube Ω for which 120581|
119880equiv
0(b) the inner support inn supp 120581 as the union of those
open sets 119880 sube Ω for which ess inf119909isin119880
|120581(119909)| gt 0(c) the outer support out
Σsupp 120581 as the complement (in
Ω) of the union of those relatively open 119880 sube Ω thatare connected to Σ and for which 120581|
119880equiv 0
The interior of a set119872 sube Ω is denoted by int119872 and its closure(with respect toR119899) by119872 If119872 is measurable we also define
(d) outΣ119872 = out
Σsupp120594
119872
It is easily checked that outΣ(supp 120581) = out
Σsupp 120581
With this concept we can extend the range characteriza-tion in Lemma 4 to a general setting (see also Remark 9 later)
Lemma 7 Let 1205900isin 119871
infin
+(Ω) be piecewise analytic Let 119863 sube Ω
be measurableThen for all unit vectors 119889 isin R119899 119889 = 1 and every point
119911 isin Ω that has a neighborhood in which 1205900is analytic
119911 isin int119863 implies Φ119911119889
1003816100381610038161003816Σisin R (119871
119863) (36)
and
Φ119911119889
1003816100381610038161003816Σisin R (119871
119863) implies 119911 isin out
Σ119863 (37)
Proof If 119911 isin int119863 then there exists a small ball 119861120598(119911) sube 119863
and the first assertion follows as in the proof of Lemma 4To show the second assertion letΦ
119911119889|Σisin R(119871
119863) and let
V isin 1198671
(Ω) be the function from the definition of 119871
119863 so that
(as in the proof of Lemma 4)
V|Σ= Φ
119911119889|Σ 120597]V|Σ = 0 = 120597]Φ119911119889|Σ (38)
Computational and Mathematical Methods in Medicine 5
Assume that 119911 notin outΣ119863 Then there exists a relatively
open 119880 sube Ω that is connected to Σ and contains 119911Hence by unique continuation it follows that V|
119880= Φ
119911119889|119880
and we obtain the same contradiction as in the proof ofLemma 4
Now we can formulate and prove the FactorizationMethod for general piecewise analytic conductivities
Theorem 8 Let 120590 isin 119871infin
+(Ω) and let 120590
0isin 119871
infin
+(Ω) be a
piecewise analytic function Let either
120590 (119909) ge 1205900(119909) forall119909 isin Ω or 120590 (119909) le 120590
0(119909) forall119909 isin Ω
(39)
Then for all 119911 isin Ω that have a neighborhood in which 1205900is
analytic as well as all unit vectors 119889 isin R119899 119889 = 1
119911isin innsupp (120590 minus 1205900) implies Φ
119911119889
1003816100381610038161003816Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
)
(40)
and
Φ119911119889
1003816100381610038161003816ΣisinR (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) implies 119911isin outΣsupp (120590 minus 120590
0)
(41)
Proof Let 119911 isin Ω have a neighborhood in which 1205900is analytic
and let 119889 isin R119899 be a unit vector with 119889 = 1 We onlyprove the assertions for 120590 ge 120590
0 The other case is completely
analogousFirst let 119911 isin inn supp (120590 minus 120590
0) Then there exists a small
ball 119861120598(119911) and 120575 gt 0 so that 120590 minus 120590
0ge 120575 on 119861
120598(119911) Using the
monotony Lemma 1 it follows that for all 119892 isin 1198712
(120597Ω)
intΣ
119892 (Λ (1205900) minus Λ (120590)) 119892d119904
ge intΩ
1205900
120590(120590 minus 120590
0)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909
ge 12057510038171003817100381710038171205900
1003817100381710038171003817119871infin(Ω)
1003817100381710038171003817100381710038171003817
1
120590
1003817100381710038171003817100381710038171003817119871infin(Ω)int119861120598(119911)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
ge 12057510038171003817100381710038171205900
1003817100381710038171003817119871infin(Ω)
1003817100381710038171003817100381710038171003817
1
120590
1003817100381710038171003817100381710038171003817119871infin(Ω)
10038171003817100381710038171003817119871lowast
119861120598(119911)11989210038171003817100381710038171003817
2
(42)
Using the functional analytic Lemma 3 we obtain that
R (119871119861120598(119911)
) sube R (1003816100381610038161003816Λ (120590) minus Λ (120590
0)1003816100381610038161003816
12
) (43)
and Lemma 7 yields that
Φ119911119889|Σisin R (119871
119861120598(119911)) sube R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) (44)
On the other hand with 119863 = supp(120590 minus 1205900) the monotony
Lemma 1 shows that for all 119892 isin 1198712
(120597Ω)
intΣ
119892 (Λ (1205900) minus Λ (120590
1)) 119892d119904 le int
Ω
(1205901minus 120590
0)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909
le10038171003817100381710038171205901 minus 120590
0
1003817100381710038171003817119871infin(Ω)int119863
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
le10038171003817100381710038171205901 minus 120590
0
1003817100381710038171003817119871infin(Ω)
1003817100381710038171003817119871lowast
1198631198921003817100381710038171003817
2
(45)
so that we obtain from the functional analytic Lemma 3
R (1003816100381610038161003816Λ (120590) minus Λ (120590
0)1003816100381610038161003816
12
) sube R (119871119863) (46)
Hence Lemma 7 yields that
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) sube R (119871119863) implies
119911 isin outΣ(supp (120590 minus 120590
0)) = out
Σsupp (120590 minus 120590
0)
(47)
Remark 9 Theorem 8 shows that the Factorization Methodis able to detect the support of a conductivity difference up tothe difference between the outer and the inner support thatis roughly speaking up to holes in the support that have noconnections to the boundary It leaves openwhether points insuch holes will fulfill the range criterion of the FactorizationMethod or not
A result of Hyvonen and the author [30 Lemma 25]shows that for every smooth domain119863with119863 sub Ω and everyunit vector 119889 isin R119899 119889 = 1
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) forall119911 isin 120597119863 (48)
implies
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) forall119911 isin 119863 (49)
In that sense we can expect that holes in the support willbe filled up and that the set detected by the FactorizationMethod is essentially the outer support of the conductivitydifference
33 The Factorization Method for the Indefinite Case It is along standing open theoretical problem whether the rangecriterion of the Factorization Method holds true withoutthe definiteness assumption that 120590 ge 120590
0on Ω or 120590 le
1205900on Ω However Grinberg Kirsch and Schmitt [18 31]
showed how to exclude a region 119864 sube Ω from Ω insuch a way that the Factorization Method only requires thedefiniteness assumption onΩ119864 In this subsection we showhow their idea can be incorporated into our formulation ofthe method
To point out the main idea we first formulate the resultfor a simple special case Let us stress that for 120590 = 1 + 120594
119863+ minus
(12)120594119863minus it is not known whether
119911 isin 119863+
cup 119863minus iff Φ
119911119889|Σisin R (|Λ (120590) minus Λ (1)|
12
) (50)
However we can still use the FactorizationMethod if we havesome a priori knowledge that separates 119863+ and 119863
minus Moreprecisely if we know a subset 119864 that contains 119863minus withoutintersecting 119863
+ then we can use the Factorization Methodto find119863+ cup 119864 (and thus119863+)
6 Computational and Mathematical Methods in Medicine
Theorem 10 Let 1205900= 1 and 120590 = 1 + 120594
119863+ minus (12)120594
119863minus where
119863+
119863minus
sube Ω are open Let 119864 sube Ω be an open set
(a) If 119863+ sube 119864 and 119863minus cup 119864 sube Ω has a connectedcomplement then for all 119911 isin Ω 119911 notin 120597(119863
minus
cup119864) and alldipole directions 119889 isin R119899 119889 = 1
119911 isin 119863minus
cup 119864 iff Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1) + 119871119864119871lowast
119864
1003816100381610038161003816
12
)
(51)
(b) If 119863minus sube 119864 and 119863+ cup 119864 sube Ω has a connectedcomplement then for all 119911 isin Ω 119911 notin 120597(119863
+
cup119864) and alldipole directions 119889 isin R119899 119889 = 1
119911 isin 119863+
cup 119864 iff Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1) minus 119871119864119871lowast
119864
1003816100381610038161003816
12
)
(52)
Proof Themonotony Lemma 1 yields that for all119892 isin 1198712
(120597Ω)
1
2int119863minus
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 minus int119863+
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
le intΣ
119892 (Λ (120590) minus Λ (1)) 119892d119904
le int119863minus
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 minus1
2int119863+
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(53)
Since (cf Lemma 2)
intΣ
119892 (119871119864119871lowast
119864) 119892d119904 = 1003817100381710038171003817119871
lowast
1198641198921003817100381710038171003817
2
= int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 (54)
it follows for case (a) that
1
2int119863minuscup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 le intΣ
119892 (Λ (120590) minus Λ (1) + 2119871119864119871lowast
119864) 119892d119904
le int119863minuscup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909(55)
Using the functional analytic Lemma 3 this implies that
R (119871119863minuscup119864) = R (
1003816100381610038161003816Λ (120590) minus Λ (1) + 2119871119864119871lowast
119864
1003816100381610038161003816
12
) (56)
so that the assertion (a) follows from Lemma 4In case (b) we obtain that
3
2int119863+cup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 ge intΣ
119892 (Λ (1) minus Λ (120590) + 2119871119864119871lowast
119864) 119892d119904
ge1
2int119863+cup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909(57)
and the same arguments as above yield the assertion
We can also extend these ideas to the general setting ofSection 32
Theorem 11 Let 120590 isin 119871infin
+(Ω) and let 120590
0isin 119871
infin
+(Ω) be a
piecewise analytic function Let 119864 sube Ω be a measurable setChoose 120572 120573 isin R such that
120572 gt1003817100381710038171003817120590 minus 120590
0
1003817100381710038171003817119871infin(Ω) 120573 gt
1003817100381710038171003817100381710038171003817
1205900
120590(1205900minus 120590)
1003817100381710038171003817100381710038171003817119871infin(Ω) (58)
(a) If 120590 le 1205900on Ω 119864 then for all 119911 isin Ω that have a
neighborhood in which 1205900is analytic as well as all unit
vectors 119889 isin R119899 119889 = 1
119911 isin inn supp (120590 minus 1205900) cup 119864 implies
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590)minusΛ (1205900)+120572119871
119864119871lowast
119864
1003816100381610038161003816
12
)
(59)
and
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864
1003816100381610038161003816
12
) implies
119911 isin outΣ(supp (120590 minus 120590
0) cup 119864)
(60)
(b) If 120590 ge 1205900on Ω 119864 then for all 119911 isin Ω that have a
neighborhood in which 1205900is analytic as well as all unit
vectors 119889 isin R119899 119889 = 1
119911 isin inn supp (120590 minus 1205900) cup 119864 implies
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900) minus 120573119871
119864119871lowast
119864
1003816100381610038161003816
12
)
(61)
and
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900) minus 120573119871
119864119871lowast
119864
1003816100381610038161003816
12
) implies
119911 isin outΣ(supp (120590 minus 120590
0) cup 119864)
(62)
Proof For every 119911 isin inn supp(120590minus1205900)with 119911 notin 119864 there exists
a small ball 119861120598(119911) and 120575 gt 0 so that 120590
0minus120590 ge 120575 on 119861
120598(119911) Using
the monotony Lemma 1 it follows that for all 119892 isin 1198712
(120597Ω)
intΣ
119892 (Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864) 119892d119904
ge intΩ
(1205900minus 120590
1)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909 + 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
ge 120575int119861120598(119911)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909+(120572 minus1003817100381710038171003817120590 minus 120590
0
1003817100381710038171003817119871infin(Ω))int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(63)
Computational and Mathematical Methods in Medicine 7
As in the previous proofs we obtain from Lemmas 2 and 3that
R (1003816100381610038161003816Λ (120590) minus Λ (120590
0) + 120572119871
119864119871lowast
119864
1003816100381610038161003816
12
) supe R (119871119861120598(119911)cup119864
) (64)
so that the first implication of (a) follows from Lemma 7The monotony Lemma 1 also implies that
intΣ
119892 (Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864) 119892d119904
le intΩ
1205900
120590(1205900minus 120590)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 + 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
le
1003817100381710038171003817100381710038171003817
1205900
120590(1205900minus 120590)
1003817100381710038171003817100381710038171003817119871infin(Ω)intsupp(120590minus1205900)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
+ 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(65)
so that the second implication of (a) follows from Lemmas 23 and 7 Assertion (b) can be proven analogously
Remark 9 also applies to this case
4 Conclusions and Remarks
The Factorization Method can be used to detect regionsin which a conductivity differs from a known referenceconductivity In this work we summarized the progresson the methodrsquos theoretical foundation We formulated themethod for general piecewise analytic conductivities andgave comparatively simple and self-contained proofsWe alsoshowed how the idea of excluding a part of the imaging regioncan be incorporated into this formulation
The regularity assumptions can beweakened even furtherOur proofs only require unique continuation arguments forthe reference conductivity 120590
0and the existence of the dipole
functionsTwo major open theoretical questions still exist in the
context of the FactorizationMethodThe theoretical justifica-tion of the method requires a definiteness condition (on thewhole domain or after excluding an a priori known part of thedomain) It is unknownwhether themethodrsquos range criterionholds without such a definiteness conditionThe second openquestion concerns the numerical stability of the methodrsquosrange criterion So far there are no rigorous convergenceresults for numerical implementations of this range criterion(see however Lechleiter [32] for a first step in this direction)As a promising approach to overcome both problems wewould like to point out the recent work on monotony-basedmethods [28]
Acknowledgments
This paper was funded by the German Research Foundation(DFG) and the University of Wurzburg in the fundingprogramme Open Access Publishing
References
[1] A Kirsch ldquoCharacterization of the shape of a scattering obstacleusing the spectral data of the far field operatorrdquo Inverse Prob-lems vol 14 no 6 pp 1489ndash1512 1998
[2] M Bruhl Gebietserkennung in der elektrischen Impedanztomo-graphie [PhD thesis] Universitat Karlsruhe 1999
[3] M Bruhl and M Hanke ldquoNumerical implementation of twononiterative methods for locating inclusions by impedancetomographyrdquo Inverse Problems vol 16 no 4 pp 1029ndash10422000
[4] M Bruhl ldquoExplicit characterization of inclusions in electricalimpedance tomographyrdquo SIAM Journal on Mathematical Anal-ysis vol 32 no 6 pp 1327ndash1341 2001
[5] M Hanke and M Bruhl ldquoRecent progress in electricalimpedance tomographyrdquo Inverse Problems vol 19 no 6 ppS65ndashS90 2003
[6] A Kirsch andN GrinbergTheFactorizationMethod for InverseProblems vol 36 ofOxford Lecture Series inMathematics and ItsApplications Oxford University Press Oxford UK 2007
[7] M Hanke and A Kirsch ldquoSampling methodsrdquo in Handbook ofMathematical Models in Imaging O Scherzer Ed pp 501ndash550Springer 2011
[8] A Kirsch ldquoThe factorization method for a class of inverseelliptic problemsrdquo Mathematische Nachrichten vol 278 pp258ndash277 2005
[9] B Gebauer ldquoThe Factorization Method for real elliptic prob-lemsrdquo Zeitschrift fur Analysis und ihre Anwendung vol 25 no1 pp 81ndash102 2006
[10] A I Nachman L Paivarinta and A Teirila ldquoOn imagingobstacles inside inhomogeneous mediardquo Journal of FunctionalAnalysis vol 252 no 2 pp 490ndash516 2007
[11] M Hanke and B Schappel ldquoThe factorization method forelectrical impedance tomography in the half-spacerdquo SIAMJournal on Applied Mathematics vol 68 no 4 pp 907ndash9242008
[12] N Hyvonen ldquoComplete electrode model of electricalimpedance tomography approximation properties andcharacterization of inclusionsrdquo SIAM Journal on AppliedMathematics vol 64 pp 902ndash931 2004
[13] A Lechleiter N Hyvonen and H Hakula ldquoThe factorizationmethod applied to the complete electrode model of impedancetomographyrdquo SIAM Journal onAppliedMathematics vol 68 no4 pp 1097ndash1121 2008
[14] N Hyvonen H Hakula and S Pursiainen ldquoNumerical imple-mentation of the factorization method within the completeelectrode model of electrical impedance tomographyrdquo InverseProblems and Imaging vol 1 pp 299ndash317 2007
[15] HHakula andNHyvonen ldquoOn computation of test dipoles forfactorization methodrdquo BIT Numerical Mathematics vol 49 no1 pp 75ndash91 2009
[16] B Harrach and J K Seo ldquoDetecting inclusions in electri-cal impedance tomography without reference measurementsrdquoSIAM Journal on Applied Mathematics vol 69 no 6 pp 1662ndash1681 2009
[17] B Harrach J K Seo and E J Woo ldquoFactorization methodand its physical justification in frequency-difference electricalimpedance tomographyrdquo IEEE Transactions on Medical Imag-ing vol 29 no 11 pp 1918ndash1926 2010
[18] S Schmitt ldquoThe factorization method for EIT in the case ofmixed inclusionsrdquo Inverse Problems vol 25 no 6 Article ID065012 2009
8 Computational and Mathematical Methods in Medicine
[19] S Schmitt and A Kirsch ldquoA factorization scheme for determin-ing conductivity contrasts in impedance tomographyrdquo InverseProblems vol 27 Article ID 095005 2011
[20] B Gebauer and N Hyvonen ldquoFactorization method and irreg-ular inclusions in electrical impedance tomographyrdquo InverseProblems vol 23 no 5 pp 2159ndash2170 2007
[21] B Gebauer ldquoLocalized potentials in electrical impedancetomographyrdquo Inverse Problems and Imaging vol 2 pp 251ndash2692008
[22] H Kang J K Seo and D Sheen ldquoThe inverse conductivityproblem with one measurement Stability and estimation ofsizerdquo SIAM Journal on Mathematical Analysis vol 28 no 6 pp1389ndash1405 1997
[23] M Ikehata ldquoSize estimation of inclusionrdquo Journal of Inverse andIll-Posed Problems vol 6 no 2 pp 127ndash140 1998
[24] T Ide H Isozaki S Nakata S Siltanen and G UhlmannldquoProbing for electrical inclusions with complex sphericalwavesrdquo Communications on Pure and Applied Mathematics vol60 no 10 pp 1415ndash1442 2007
[25] B Harrach and J K Seo ldquoExact shape-reconstruction by one-step linearization in electrical impedance tomographyrdquo TheSIAM Journal onMathematical Analysis vol 42 no 4 pp 1505ndash1518 2010
[26] N Bourbaki Elements of Mathematics Topological VectorSpaces chapters 1ndash5 Springer Berlin Germany 2003
[27] F Fruhauf B Gebauer andO Scherzer ldquoDetecting interfaces ina parabolic-elliptic problem from surfacemeasurementsrdquo SIAMJournal on Numerical Analysis vol 45 no 2 pp 810ndash836 2007
[28] B Harrach and M Ullrich ldquoMonotony based shape recon-struction in electrical impedance tomographyrdquo submitted forpublication
[29] S Kusiak and J Sylvester ldquoThe scattering supportrdquoCommunica-tions on Pure and Applied Mathematics vol 56 no 11 pp 1525ndash1548 2003
[30] B Gebauer and N Hyvonen ldquoFactorization method andinclusions of mixed type in an inverse elliptic boundary valueproblemrdquo Inverse Problems and Imaging vol 2 pp 355ndash3722008
[31] N I Grinberg and A Kirsch ldquoThe factorization method forobstacles with a-priori separated sound-soft and sound-hardpartsrdquo Mathematics and Computers in Simulation vol 66 no4-5 pp 267ndash279 2004
[32] A Lechleiter ldquoA regularization technique for the factorizationmethodrdquo Inverse Problems vol 22 no 5 pp 1605ndash1625 2006
Submit your manuscripts athttpwwwhindawicom
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MEDIATORSINFLAMMATION
of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Behavioural Neurology
EndocrinologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Disease Markers
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
OncologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Oxidative Medicine and Cellular Longevity
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
PPAR Research
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Immunology ResearchHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
ObesityJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational and Mathematical Methods in Medicine
OphthalmologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Diabetes ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentAIDS
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Gastroenterology Research and Practice
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Parkinsonrsquos Disease
Evidence-Based Complementary and Alternative Medicine
Volume 2014Hindawi Publishing Corporationhttpwwwhindawicom
4 Computational and Mathematical Methods in Medicine
fulfills V isin 1198671
(Ω) and for all 119908 isin 119867
1
(Ω)
intΩ
nablaV sdot nabla119908d119909
= intΩ119861120598(119911)
nablaΦ119911119889
sdot nabla119908d119909 + int119861120598(119911)
nabla1198911sdot nabla119908d119909
= minusint120597119861120598(119911)
120597]Φ119911119889 119908|120597119861120598(119911)d119904 + int
119861120598(119911)
nabla1198911sdot nabla119908d119909
= int119861120598(119911)
nabla (1198911minus 119891
2) sdot nabla119908d119909 = int
119863
119865 sdot nabla119908d119909
(29)
This shows thatΦ119911119889|Σ= V|
Σ= 119871
119863(119865) isin R(119871
119863)
Now let Φ119911119889|Σisin R(119871
119863) Let V isin 119867
1
(Ω) be the function
from the definition of 119871119863 Then
V|Σ= Φ
119911119889|Σ 120597]V|Σ = 0 = 120597]Φ119911119889|Σ (30)
so that it follows by unique continuation that V = Φ119911119889
in theconnected setΩ (119863 cup 119911)
If 119911 notin 119863 then 119889 sdot nabla120575119911notin 119867
minus2
(Ω 119863) and thus Φ119911119889
notin
1198712
(Ω 119863) which contradicts that V = Φ119911119889
in Ω (119863 cup 119911)Hence 119911 isin 119863
3 The Factorization Method
Nowwe will formulate the FactorizationMethod and charac-terize a region where a conductivity 120590 differs from a referenceconductivity 120590
0by a range criterion Before we turn to a
new general formulation of the method we first state it fora special case that is similar to the one that was treated in theoriginal works of Bruhl and Hanke [3 4]
31 The Factorization Method for a Simple Special Case
Theorem 5 Let 1205900= 1 and 120590 = 1 + 120594
119863 where 119863 sube Ω is an
open set so that 119863 sube Ω has a connected complement Ω 119863Then for all 119911 isin Ω 119911 notin 120597119863 and all dipole directions 119889 isin R119899119889 = 1
119911 isin 119863 iff Φ119911119889
1003816100381610038161003816Σisin R (|Λ (120590) minus Λ (1)|
12
) (31)
Proof Themonotony Lemma 1 yields that for all119892 isin 1198712
(120597Ω)
int119863
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 ge intΣ
119892 (Λ (1) minus Λ (120590)) 119892d119904 ge int119863
1
2
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(32)
Hence |Λ(120590) minusΛ(1)| = Λ(1) minusΛ(120590) and using Lemma 2 wecan restate this in the form
1003817100381710038171003817119871lowast
1198631198921003817100381710038171003817
2
ge10038171003817100381710038171003817|Λ (120590) minus Λ (1)|
12
11989210038171003817100381710038171003817
2
ge1
2
1003817100381710038171003817119871lowast
1198631198921003817100381710038171003817
2
(33)
Using the functional analytic Lemma 3 this implies that
R (119871119863) = R (|Λ (120590) minus Λ (1)|
12
) (34)
and thus the assertion follows from the relation between 119863
andR(119871119863) in Lemma 4
Obviously the same arguments can be used to treat thecase 120590(119909) = 1 + 120581(119909)120594
119863(119909) when there exists a conductivity
jump 120598 gt 0 so that either
120581 (119909) ge 120598 forall119909 isin 119863 or 120581 (119909) le minus120598 forall119909 isin 119863 (35)
32 The Factorization Method for the General PiecewiseAnalytic Case Now we drop the assumptions that the back-ground is constant that there is a clear conductivity jumpand that the complement of the inclusions is connected Wewill merely assume that the reference conductivity 120590
0is a
piecewise analytic function and that either 120590 minus 1205900ge 0 or
120590 minus 1205900le 0 Roughly speaking under this general assumption
the Factorization Method then characterizes the support of120590 minus 120590
0up to holes in the support that have no connections to
Σ For a precise formulation we use the concept of the innerand outer support from [28] that has been inspired by the useof the infinity support of Kusiak and Sylvester [29] see also[25 30]
Definition 6 A relatively open set 119880 sube Ω is called connectedto Σ if 119880 cap Ω is connected and 119880 cap Σ = 0
For a measurable function 120581 Ω rarr R we define
(a) the support supp(120581) as the complement (in Ω) of theunion of those relatively open119880 sube Ω for which 120581|
119880equiv
0(b) the inner support inn supp 120581 as the union of those
open sets 119880 sube Ω for which ess inf119909isin119880
|120581(119909)| gt 0(c) the outer support out
Σsupp 120581 as the complement (in
Ω) of the union of those relatively open 119880 sube Ω thatare connected to Σ and for which 120581|
119880equiv 0
The interior of a set119872 sube Ω is denoted by int119872 and its closure(with respect toR119899) by119872 If119872 is measurable we also define
(d) outΣ119872 = out
Σsupp120594
119872
It is easily checked that outΣ(supp 120581) = out
Σsupp 120581
With this concept we can extend the range characteriza-tion in Lemma 4 to a general setting (see also Remark 9 later)
Lemma 7 Let 1205900isin 119871
infin
+(Ω) be piecewise analytic Let 119863 sube Ω
be measurableThen for all unit vectors 119889 isin R119899 119889 = 1 and every point
119911 isin Ω that has a neighborhood in which 1205900is analytic
119911 isin int119863 implies Φ119911119889
1003816100381610038161003816Σisin R (119871
119863) (36)
and
Φ119911119889
1003816100381610038161003816Σisin R (119871
119863) implies 119911 isin out
Σ119863 (37)
Proof If 119911 isin int119863 then there exists a small ball 119861120598(119911) sube 119863
and the first assertion follows as in the proof of Lemma 4To show the second assertion letΦ
119911119889|Σisin R(119871
119863) and let
V isin 1198671
(Ω) be the function from the definition of 119871
119863 so that
(as in the proof of Lemma 4)
V|Σ= Φ
119911119889|Σ 120597]V|Σ = 0 = 120597]Φ119911119889|Σ (38)
Computational and Mathematical Methods in Medicine 5
Assume that 119911 notin outΣ119863 Then there exists a relatively
open 119880 sube Ω that is connected to Σ and contains 119911Hence by unique continuation it follows that V|
119880= Φ
119911119889|119880
and we obtain the same contradiction as in the proof ofLemma 4
Now we can formulate and prove the FactorizationMethod for general piecewise analytic conductivities
Theorem 8 Let 120590 isin 119871infin
+(Ω) and let 120590
0isin 119871
infin
+(Ω) be a
piecewise analytic function Let either
120590 (119909) ge 1205900(119909) forall119909 isin Ω or 120590 (119909) le 120590
0(119909) forall119909 isin Ω
(39)
Then for all 119911 isin Ω that have a neighborhood in which 1205900is
analytic as well as all unit vectors 119889 isin R119899 119889 = 1
119911isin innsupp (120590 minus 1205900) implies Φ
119911119889
1003816100381610038161003816Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
)
(40)
and
Φ119911119889
1003816100381610038161003816ΣisinR (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) implies 119911isin outΣsupp (120590 minus 120590
0)
(41)
Proof Let 119911 isin Ω have a neighborhood in which 1205900is analytic
and let 119889 isin R119899 be a unit vector with 119889 = 1 We onlyprove the assertions for 120590 ge 120590
0 The other case is completely
analogousFirst let 119911 isin inn supp (120590 minus 120590
0) Then there exists a small
ball 119861120598(119911) and 120575 gt 0 so that 120590 minus 120590
0ge 120575 on 119861
120598(119911) Using the
monotony Lemma 1 it follows that for all 119892 isin 1198712
(120597Ω)
intΣ
119892 (Λ (1205900) minus Λ (120590)) 119892d119904
ge intΩ
1205900
120590(120590 minus 120590
0)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909
ge 12057510038171003817100381710038171205900
1003817100381710038171003817119871infin(Ω)
1003817100381710038171003817100381710038171003817
1
120590
1003817100381710038171003817100381710038171003817119871infin(Ω)int119861120598(119911)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
ge 12057510038171003817100381710038171205900
1003817100381710038171003817119871infin(Ω)
1003817100381710038171003817100381710038171003817
1
120590
1003817100381710038171003817100381710038171003817119871infin(Ω)
10038171003817100381710038171003817119871lowast
119861120598(119911)11989210038171003817100381710038171003817
2
(42)
Using the functional analytic Lemma 3 we obtain that
R (119871119861120598(119911)
) sube R (1003816100381610038161003816Λ (120590) minus Λ (120590
0)1003816100381610038161003816
12
) (43)
and Lemma 7 yields that
Φ119911119889|Σisin R (119871
119861120598(119911)) sube R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) (44)
On the other hand with 119863 = supp(120590 minus 1205900) the monotony
Lemma 1 shows that for all 119892 isin 1198712
(120597Ω)
intΣ
119892 (Λ (1205900) minus Λ (120590
1)) 119892d119904 le int
Ω
(1205901minus 120590
0)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909
le10038171003817100381710038171205901 minus 120590
0
1003817100381710038171003817119871infin(Ω)int119863
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
le10038171003817100381710038171205901 minus 120590
0
1003817100381710038171003817119871infin(Ω)
1003817100381710038171003817119871lowast
1198631198921003817100381710038171003817
2
(45)
so that we obtain from the functional analytic Lemma 3
R (1003816100381610038161003816Λ (120590) minus Λ (120590
0)1003816100381610038161003816
12
) sube R (119871119863) (46)
Hence Lemma 7 yields that
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) sube R (119871119863) implies
119911 isin outΣ(supp (120590 minus 120590
0)) = out
Σsupp (120590 minus 120590
0)
(47)
Remark 9 Theorem 8 shows that the Factorization Methodis able to detect the support of a conductivity difference up tothe difference between the outer and the inner support thatis roughly speaking up to holes in the support that have noconnections to the boundary It leaves openwhether points insuch holes will fulfill the range criterion of the FactorizationMethod or not
A result of Hyvonen and the author [30 Lemma 25]shows that for every smooth domain119863with119863 sub Ω and everyunit vector 119889 isin R119899 119889 = 1
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) forall119911 isin 120597119863 (48)
implies
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) forall119911 isin 119863 (49)
In that sense we can expect that holes in the support willbe filled up and that the set detected by the FactorizationMethod is essentially the outer support of the conductivitydifference
33 The Factorization Method for the Indefinite Case It is along standing open theoretical problem whether the rangecriterion of the Factorization Method holds true withoutthe definiteness assumption that 120590 ge 120590
0on Ω or 120590 le
1205900on Ω However Grinberg Kirsch and Schmitt [18 31]
showed how to exclude a region 119864 sube Ω from Ω insuch a way that the Factorization Method only requires thedefiniteness assumption onΩ119864 In this subsection we showhow their idea can be incorporated into our formulation ofthe method
To point out the main idea we first formulate the resultfor a simple special case Let us stress that for 120590 = 1 + 120594
119863+ minus
(12)120594119863minus it is not known whether
119911 isin 119863+
cup 119863minus iff Φ
119911119889|Σisin R (|Λ (120590) minus Λ (1)|
12
) (50)
However we can still use the FactorizationMethod if we havesome a priori knowledge that separates 119863+ and 119863
minus Moreprecisely if we know a subset 119864 that contains 119863minus withoutintersecting 119863
+ then we can use the Factorization Methodto find119863+ cup 119864 (and thus119863+)
6 Computational and Mathematical Methods in Medicine
Theorem 10 Let 1205900= 1 and 120590 = 1 + 120594
119863+ minus (12)120594
119863minus where
119863+
119863minus
sube Ω are open Let 119864 sube Ω be an open set
(a) If 119863+ sube 119864 and 119863minus cup 119864 sube Ω has a connectedcomplement then for all 119911 isin Ω 119911 notin 120597(119863
minus
cup119864) and alldipole directions 119889 isin R119899 119889 = 1
119911 isin 119863minus
cup 119864 iff Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1) + 119871119864119871lowast
119864
1003816100381610038161003816
12
)
(51)
(b) If 119863minus sube 119864 and 119863+ cup 119864 sube Ω has a connectedcomplement then for all 119911 isin Ω 119911 notin 120597(119863
+
cup119864) and alldipole directions 119889 isin R119899 119889 = 1
119911 isin 119863+
cup 119864 iff Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1) minus 119871119864119871lowast
119864
1003816100381610038161003816
12
)
(52)
Proof Themonotony Lemma 1 yields that for all119892 isin 1198712
(120597Ω)
1
2int119863minus
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 minus int119863+
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
le intΣ
119892 (Λ (120590) minus Λ (1)) 119892d119904
le int119863minus
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 minus1
2int119863+
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(53)
Since (cf Lemma 2)
intΣ
119892 (119871119864119871lowast
119864) 119892d119904 = 1003817100381710038171003817119871
lowast
1198641198921003817100381710038171003817
2
= int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 (54)
it follows for case (a) that
1
2int119863minuscup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 le intΣ
119892 (Λ (120590) minus Λ (1) + 2119871119864119871lowast
119864) 119892d119904
le int119863minuscup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909(55)
Using the functional analytic Lemma 3 this implies that
R (119871119863minuscup119864) = R (
1003816100381610038161003816Λ (120590) minus Λ (1) + 2119871119864119871lowast
119864
1003816100381610038161003816
12
) (56)
so that the assertion (a) follows from Lemma 4In case (b) we obtain that
3
2int119863+cup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 ge intΣ
119892 (Λ (1) minus Λ (120590) + 2119871119864119871lowast
119864) 119892d119904
ge1
2int119863+cup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909(57)
and the same arguments as above yield the assertion
We can also extend these ideas to the general setting ofSection 32
Theorem 11 Let 120590 isin 119871infin
+(Ω) and let 120590
0isin 119871
infin
+(Ω) be a
piecewise analytic function Let 119864 sube Ω be a measurable setChoose 120572 120573 isin R such that
120572 gt1003817100381710038171003817120590 minus 120590
0
1003817100381710038171003817119871infin(Ω) 120573 gt
1003817100381710038171003817100381710038171003817
1205900
120590(1205900minus 120590)
1003817100381710038171003817100381710038171003817119871infin(Ω) (58)
(a) If 120590 le 1205900on Ω 119864 then for all 119911 isin Ω that have a
neighborhood in which 1205900is analytic as well as all unit
vectors 119889 isin R119899 119889 = 1
119911 isin inn supp (120590 minus 1205900) cup 119864 implies
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590)minusΛ (1205900)+120572119871
119864119871lowast
119864
1003816100381610038161003816
12
)
(59)
and
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864
1003816100381610038161003816
12
) implies
119911 isin outΣ(supp (120590 minus 120590
0) cup 119864)
(60)
(b) If 120590 ge 1205900on Ω 119864 then for all 119911 isin Ω that have a
neighborhood in which 1205900is analytic as well as all unit
vectors 119889 isin R119899 119889 = 1
119911 isin inn supp (120590 minus 1205900) cup 119864 implies
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900) minus 120573119871
119864119871lowast
119864
1003816100381610038161003816
12
)
(61)
and
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900) minus 120573119871
119864119871lowast
119864
1003816100381610038161003816
12
) implies
119911 isin outΣ(supp (120590 minus 120590
0) cup 119864)
(62)
Proof For every 119911 isin inn supp(120590minus1205900)with 119911 notin 119864 there exists
a small ball 119861120598(119911) and 120575 gt 0 so that 120590
0minus120590 ge 120575 on 119861
120598(119911) Using
the monotony Lemma 1 it follows that for all 119892 isin 1198712
(120597Ω)
intΣ
119892 (Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864) 119892d119904
ge intΩ
(1205900minus 120590
1)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909 + 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
ge 120575int119861120598(119911)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909+(120572 minus1003817100381710038171003817120590 minus 120590
0
1003817100381710038171003817119871infin(Ω))int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(63)
Computational and Mathematical Methods in Medicine 7
As in the previous proofs we obtain from Lemmas 2 and 3that
R (1003816100381610038161003816Λ (120590) minus Λ (120590
0) + 120572119871
119864119871lowast
119864
1003816100381610038161003816
12
) supe R (119871119861120598(119911)cup119864
) (64)
so that the first implication of (a) follows from Lemma 7The monotony Lemma 1 also implies that
intΣ
119892 (Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864) 119892d119904
le intΩ
1205900
120590(1205900minus 120590)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 + 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
le
1003817100381710038171003817100381710038171003817
1205900
120590(1205900minus 120590)
1003817100381710038171003817100381710038171003817119871infin(Ω)intsupp(120590minus1205900)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
+ 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(65)
so that the second implication of (a) follows from Lemmas 23 and 7 Assertion (b) can be proven analogously
Remark 9 also applies to this case
4 Conclusions and Remarks
The Factorization Method can be used to detect regionsin which a conductivity differs from a known referenceconductivity In this work we summarized the progresson the methodrsquos theoretical foundation We formulated themethod for general piecewise analytic conductivities andgave comparatively simple and self-contained proofsWe alsoshowed how the idea of excluding a part of the imaging regioncan be incorporated into this formulation
The regularity assumptions can beweakened even furtherOur proofs only require unique continuation arguments forthe reference conductivity 120590
0and the existence of the dipole
functionsTwo major open theoretical questions still exist in the
context of the FactorizationMethodThe theoretical justifica-tion of the method requires a definiteness condition (on thewhole domain or after excluding an a priori known part of thedomain) It is unknownwhether themethodrsquos range criterionholds without such a definiteness conditionThe second openquestion concerns the numerical stability of the methodrsquosrange criterion So far there are no rigorous convergenceresults for numerical implementations of this range criterion(see however Lechleiter [32] for a first step in this direction)As a promising approach to overcome both problems wewould like to point out the recent work on monotony-basedmethods [28]
Acknowledgments
This paper was funded by the German Research Foundation(DFG) and the University of Wurzburg in the fundingprogramme Open Access Publishing
References
[1] A Kirsch ldquoCharacterization of the shape of a scattering obstacleusing the spectral data of the far field operatorrdquo Inverse Prob-lems vol 14 no 6 pp 1489ndash1512 1998
[2] M Bruhl Gebietserkennung in der elektrischen Impedanztomo-graphie [PhD thesis] Universitat Karlsruhe 1999
[3] M Bruhl and M Hanke ldquoNumerical implementation of twononiterative methods for locating inclusions by impedancetomographyrdquo Inverse Problems vol 16 no 4 pp 1029ndash10422000
[4] M Bruhl ldquoExplicit characterization of inclusions in electricalimpedance tomographyrdquo SIAM Journal on Mathematical Anal-ysis vol 32 no 6 pp 1327ndash1341 2001
[5] M Hanke and M Bruhl ldquoRecent progress in electricalimpedance tomographyrdquo Inverse Problems vol 19 no 6 ppS65ndashS90 2003
[6] A Kirsch andN GrinbergTheFactorizationMethod for InverseProblems vol 36 ofOxford Lecture Series inMathematics and ItsApplications Oxford University Press Oxford UK 2007
[7] M Hanke and A Kirsch ldquoSampling methodsrdquo in Handbook ofMathematical Models in Imaging O Scherzer Ed pp 501ndash550Springer 2011
[8] A Kirsch ldquoThe factorization method for a class of inverseelliptic problemsrdquo Mathematische Nachrichten vol 278 pp258ndash277 2005
[9] B Gebauer ldquoThe Factorization Method for real elliptic prob-lemsrdquo Zeitschrift fur Analysis und ihre Anwendung vol 25 no1 pp 81ndash102 2006
[10] A I Nachman L Paivarinta and A Teirila ldquoOn imagingobstacles inside inhomogeneous mediardquo Journal of FunctionalAnalysis vol 252 no 2 pp 490ndash516 2007
[11] M Hanke and B Schappel ldquoThe factorization method forelectrical impedance tomography in the half-spacerdquo SIAMJournal on Applied Mathematics vol 68 no 4 pp 907ndash9242008
[12] N Hyvonen ldquoComplete electrode model of electricalimpedance tomography approximation properties andcharacterization of inclusionsrdquo SIAM Journal on AppliedMathematics vol 64 pp 902ndash931 2004
[13] A Lechleiter N Hyvonen and H Hakula ldquoThe factorizationmethod applied to the complete electrode model of impedancetomographyrdquo SIAM Journal onAppliedMathematics vol 68 no4 pp 1097ndash1121 2008
[14] N Hyvonen H Hakula and S Pursiainen ldquoNumerical imple-mentation of the factorization method within the completeelectrode model of electrical impedance tomographyrdquo InverseProblems and Imaging vol 1 pp 299ndash317 2007
[15] HHakula andNHyvonen ldquoOn computation of test dipoles forfactorization methodrdquo BIT Numerical Mathematics vol 49 no1 pp 75ndash91 2009
[16] B Harrach and J K Seo ldquoDetecting inclusions in electri-cal impedance tomography without reference measurementsrdquoSIAM Journal on Applied Mathematics vol 69 no 6 pp 1662ndash1681 2009
[17] B Harrach J K Seo and E J Woo ldquoFactorization methodand its physical justification in frequency-difference electricalimpedance tomographyrdquo IEEE Transactions on Medical Imag-ing vol 29 no 11 pp 1918ndash1926 2010
[18] S Schmitt ldquoThe factorization method for EIT in the case ofmixed inclusionsrdquo Inverse Problems vol 25 no 6 Article ID065012 2009
8 Computational and Mathematical Methods in Medicine
[19] S Schmitt and A Kirsch ldquoA factorization scheme for determin-ing conductivity contrasts in impedance tomographyrdquo InverseProblems vol 27 Article ID 095005 2011
[20] B Gebauer and N Hyvonen ldquoFactorization method and irreg-ular inclusions in electrical impedance tomographyrdquo InverseProblems vol 23 no 5 pp 2159ndash2170 2007
[21] B Gebauer ldquoLocalized potentials in electrical impedancetomographyrdquo Inverse Problems and Imaging vol 2 pp 251ndash2692008
[22] H Kang J K Seo and D Sheen ldquoThe inverse conductivityproblem with one measurement Stability and estimation ofsizerdquo SIAM Journal on Mathematical Analysis vol 28 no 6 pp1389ndash1405 1997
[23] M Ikehata ldquoSize estimation of inclusionrdquo Journal of Inverse andIll-Posed Problems vol 6 no 2 pp 127ndash140 1998
[24] T Ide H Isozaki S Nakata S Siltanen and G UhlmannldquoProbing for electrical inclusions with complex sphericalwavesrdquo Communications on Pure and Applied Mathematics vol60 no 10 pp 1415ndash1442 2007
[25] B Harrach and J K Seo ldquoExact shape-reconstruction by one-step linearization in electrical impedance tomographyrdquo TheSIAM Journal onMathematical Analysis vol 42 no 4 pp 1505ndash1518 2010
[26] N Bourbaki Elements of Mathematics Topological VectorSpaces chapters 1ndash5 Springer Berlin Germany 2003
[27] F Fruhauf B Gebauer andO Scherzer ldquoDetecting interfaces ina parabolic-elliptic problem from surfacemeasurementsrdquo SIAMJournal on Numerical Analysis vol 45 no 2 pp 810ndash836 2007
[28] B Harrach and M Ullrich ldquoMonotony based shape recon-struction in electrical impedance tomographyrdquo submitted forpublication
[29] S Kusiak and J Sylvester ldquoThe scattering supportrdquoCommunica-tions on Pure and Applied Mathematics vol 56 no 11 pp 1525ndash1548 2003
[30] B Gebauer and N Hyvonen ldquoFactorization method andinclusions of mixed type in an inverse elliptic boundary valueproblemrdquo Inverse Problems and Imaging vol 2 pp 355ndash3722008
[31] N I Grinberg and A Kirsch ldquoThe factorization method forobstacles with a-priori separated sound-soft and sound-hardpartsrdquo Mathematics and Computers in Simulation vol 66 no4-5 pp 267ndash279 2004
[32] A Lechleiter ldquoA regularization technique for the factorizationmethodrdquo Inverse Problems vol 22 no 5 pp 1605ndash1625 2006
Submit your manuscripts athttpwwwhindawicom
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MEDIATORSINFLAMMATION
of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Behavioural Neurology
EndocrinologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Disease Markers
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
OncologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Oxidative Medicine and Cellular Longevity
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
PPAR Research
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Immunology ResearchHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
ObesityJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational and Mathematical Methods in Medicine
OphthalmologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Diabetes ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentAIDS
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Gastroenterology Research and Practice
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Parkinsonrsquos Disease
Evidence-Based Complementary and Alternative Medicine
Volume 2014Hindawi Publishing Corporationhttpwwwhindawicom
Computational and Mathematical Methods in Medicine 5
Assume that 119911 notin outΣ119863 Then there exists a relatively
open 119880 sube Ω that is connected to Σ and contains 119911Hence by unique continuation it follows that V|
119880= Φ
119911119889|119880
and we obtain the same contradiction as in the proof ofLemma 4
Now we can formulate and prove the FactorizationMethod for general piecewise analytic conductivities
Theorem 8 Let 120590 isin 119871infin
+(Ω) and let 120590
0isin 119871
infin
+(Ω) be a
piecewise analytic function Let either
120590 (119909) ge 1205900(119909) forall119909 isin Ω or 120590 (119909) le 120590
0(119909) forall119909 isin Ω
(39)
Then for all 119911 isin Ω that have a neighborhood in which 1205900is
analytic as well as all unit vectors 119889 isin R119899 119889 = 1
119911isin innsupp (120590 minus 1205900) implies Φ
119911119889
1003816100381610038161003816Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
)
(40)
and
Φ119911119889
1003816100381610038161003816ΣisinR (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) implies 119911isin outΣsupp (120590 minus 120590
0)
(41)
Proof Let 119911 isin Ω have a neighborhood in which 1205900is analytic
and let 119889 isin R119899 be a unit vector with 119889 = 1 We onlyprove the assertions for 120590 ge 120590
0 The other case is completely
analogousFirst let 119911 isin inn supp (120590 minus 120590
0) Then there exists a small
ball 119861120598(119911) and 120575 gt 0 so that 120590 minus 120590
0ge 120575 on 119861
120598(119911) Using the
monotony Lemma 1 it follows that for all 119892 isin 1198712
(120597Ω)
intΣ
119892 (Λ (1205900) minus Λ (120590)) 119892d119904
ge intΩ
1205900
120590(120590 minus 120590
0)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909
ge 12057510038171003817100381710038171205900
1003817100381710038171003817119871infin(Ω)
1003817100381710038171003817100381710038171003817
1
120590
1003817100381710038171003817100381710038171003817119871infin(Ω)int119861120598(119911)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
ge 12057510038171003817100381710038171205900
1003817100381710038171003817119871infin(Ω)
1003817100381710038171003817100381710038171003817
1
120590
1003817100381710038171003817100381710038171003817119871infin(Ω)
10038171003817100381710038171003817119871lowast
119861120598(119911)11989210038171003817100381710038171003817
2
(42)
Using the functional analytic Lemma 3 we obtain that
R (119871119861120598(119911)
) sube R (1003816100381610038161003816Λ (120590) minus Λ (120590
0)1003816100381610038161003816
12
) (43)
and Lemma 7 yields that
Φ119911119889|Σisin R (119871
119861120598(119911)) sube R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) (44)
On the other hand with 119863 = supp(120590 minus 1205900) the monotony
Lemma 1 shows that for all 119892 isin 1198712
(120597Ω)
intΣ
119892 (Λ (1205900) minus Λ (120590
1)) 119892d119904 le int
Ω
(1205901minus 120590
0)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909
le10038171003817100381710038171205901 minus 120590
0
1003817100381710038171003817119871infin(Ω)int119863
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
le10038171003817100381710038171205901 minus 120590
0
1003817100381710038171003817119871infin(Ω)
1003817100381710038171003817119871lowast
1198631198921003817100381710038171003817
2
(45)
so that we obtain from the functional analytic Lemma 3
R (1003816100381610038161003816Λ (120590) minus Λ (120590
0)1003816100381610038161003816
12
) sube R (119871119863) (46)
Hence Lemma 7 yields that
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) sube R (119871119863) implies
119911 isin outΣ(supp (120590 minus 120590
0)) = out
Σsupp (120590 minus 120590
0)
(47)
Remark 9 Theorem 8 shows that the Factorization Methodis able to detect the support of a conductivity difference up tothe difference between the outer and the inner support thatis roughly speaking up to holes in the support that have noconnections to the boundary It leaves openwhether points insuch holes will fulfill the range criterion of the FactorizationMethod or not
A result of Hyvonen and the author [30 Lemma 25]shows that for every smooth domain119863with119863 sub Ω and everyunit vector 119889 isin R119899 119889 = 1
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) forall119911 isin 120597119863 (48)
implies
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900)1003816100381610038161003816
12
) forall119911 isin 119863 (49)
In that sense we can expect that holes in the support willbe filled up and that the set detected by the FactorizationMethod is essentially the outer support of the conductivitydifference
33 The Factorization Method for the Indefinite Case It is along standing open theoretical problem whether the rangecriterion of the Factorization Method holds true withoutthe definiteness assumption that 120590 ge 120590
0on Ω or 120590 le
1205900on Ω However Grinberg Kirsch and Schmitt [18 31]
showed how to exclude a region 119864 sube Ω from Ω insuch a way that the Factorization Method only requires thedefiniteness assumption onΩ119864 In this subsection we showhow their idea can be incorporated into our formulation ofthe method
To point out the main idea we first formulate the resultfor a simple special case Let us stress that for 120590 = 1 + 120594
119863+ minus
(12)120594119863minus it is not known whether
119911 isin 119863+
cup 119863minus iff Φ
119911119889|Σisin R (|Λ (120590) minus Λ (1)|
12
) (50)
However we can still use the FactorizationMethod if we havesome a priori knowledge that separates 119863+ and 119863
minus Moreprecisely if we know a subset 119864 that contains 119863minus withoutintersecting 119863
+ then we can use the Factorization Methodto find119863+ cup 119864 (and thus119863+)
6 Computational and Mathematical Methods in Medicine
Theorem 10 Let 1205900= 1 and 120590 = 1 + 120594
119863+ minus (12)120594
119863minus where
119863+
119863minus
sube Ω are open Let 119864 sube Ω be an open set
(a) If 119863+ sube 119864 and 119863minus cup 119864 sube Ω has a connectedcomplement then for all 119911 isin Ω 119911 notin 120597(119863
minus
cup119864) and alldipole directions 119889 isin R119899 119889 = 1
119911 isin 119863minus
cup 119864 iff Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1) + 119871119864119871lowast
119864
1003816100381610038161003816
12
)
(51)
(b) If 119863minus sube 119864 and 119863+ cup 119864 sube Ω has a connectedcomplement then for all 119911 isin Ω 119911 notin 120597(119863
+
cup119864) and alldipole directions 119889 isin R119899 119889 = 1
119911 isin 119863+
cup 119864 iff Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1) minus 119871119864119871lowast
119864
1003816100381610038161003816
12
)
(52)
Proof Themonotony Lemma 1 yields that for all119892 isin 1198712
(120597Ω)
1
2int119863minus
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 minus int119863+
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
le intΣ
119892 (Λ (120590) minus Λ (1)) 119892d119904
le int119863minus
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 minus1
2int119863+
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(53)
Since (cf Lemma 2)
intΣ
119892 (119871119864119871lowast
119864) 119892d119904 = 1003817100381710038171003817119871
lowast
1198641198921003817100381710038171003817
2
= int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 (54)
it follows for case (a) that
1
2int119863minuscup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 le intΣ
119892 (Λ (120590) minus Λ (1) + 2119871119864119871lowast
119864) 119892d119904
le int119863minuscup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909(55)
Using the functional analytic Lemma 3 this implies that
R (119871119863minuscup119864) = R (
1003816100381610038161003816Λ (120590) minus Λ (1) + 2119871119864119871lowast
119864
1003816100381610038161003816
12
) (56)
so that the assertion (a) follows from Lemma 4In case (b) we obtain that
3
2int119863+cup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 ge intΣ
119892 (Λ (1) minus Λ (120590) + 2119871119864119871lowast
119864) 119892d119904
ge1
2int119863+cup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909(57)
and the same arguments as above yield the assertion
We can also extend these ideas to the general setting ofSection 32
Theorem 11 Let 120590 isin 119871infin
+(Ω) and let 120590
0isin 119871
infin
+(Ω) be a
piecewise analytic function Let 119864 sube Ω be a measurable setChoose 120572 120573 isin R such that
120572 gt1003817100381710038171003817120590 minus 120590
0
1003817100381710038171003817119871infin(Ω) 120573 gt
1003817100381710038171003817100381710038171003817
1205900
120590(1205900minus 120590)
1003817100381710038171003817100381710038171003817119871infin(Ω) (58)
(a) If 120590 le 1205900on Ω 119864 then for all 119911 isin Ω that have a
neighborhood in which 1205900is analytic as well as all unit
vectors 119889 isin R119899 119889 = 1
119911 isin inn supp (120590 minus 1205900) cup 119864 implies
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590)minusΛ (1205900)+120572119871
119864119871lowast
119864
1003816100381610038161003816
12
)
(59)
and
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864
1003816100381610038161003816
12
) implies
119911 isin outΣ(supp (120590 minus 120590
0) cup 119864)
(60)
(b) If 120590 ge 1205900on Ω 119864 then for all 119911 isin Ω that have a
neighborhood in which 1205900is analytic as well as all unit
vectors 119889 isin R119899 119889 = 1
119911 isin inn supp (120590 minus 1205900) cup 119864 implies
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900) minus 120573119871
119864119871lowast
119864
1003816100381610038161003816
12
)
(61)
and
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900) minus 120573119871
119864119871lowast
119864
1003816100381610038161003816
12
) implies
119911 isin outΣ(supp (120590 minus 120590
0) cup 119864)
(62)
Proof For every 119911 isin inn supp(120590minus1205900)with 119911 notin 119864 there exists
a small ball 119861120598(119911) and 120575 gt 0 so that 120590
0minus120590 ge 120575 on 119861
120598(119911) Using
the monotony Lemma 1 it follows that for all 119892 isin 1198712
(120597Ω)
intΣ
119892 (Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864) 119892d119904
ge intΩ
(1205900minus 120590
1)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909 + 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
ge 120575int119861120598(119911)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909+(120572 minus1003817100381710038171003817120590 minus 120590
0
1003817100381710038171003817119871infin(Ω))int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(63)
Computational and Mathematical Methods in Medicine 7
As in the previous proofs we obtain from Lemmas 2 and 3that
R (1003816100381610038161003816Λ (120590) minus Λ (120590
0) + 120572119871
119864119871lowast
119864
1003816100381610038161003816
12
) supe R (119871119861120598(119911)cup119864
) (64)
so that the first implication of (a) follows from Lemma 7The monotony Lemma 1 also implies that
intΣ
119892 (Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864) 119892d119904
le intΩ
1205900
120590(1205900minus 120590)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 + 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
le
1003817100381710038171003817100381710038171003817
1205900
120590(1205900minus 120590)
1003817100381710038171003817100381710038171003817119871infin(Ω)intsupp(120590minus1205900)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
+ 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(65)
so that the second implication of (a) follows from Lemmas 23 and 7 Assertion (b) can be proven analogously
Remark 9 also applies to this case
4 Conclusions and Remarks
The Factorization Method can be used to detect regionsin which a conductivity differs from a known referenceconductivity In this work we summarized the progresson the methodrsquos theoretical foundation We formulated themethod for general piecewise analytic conductivities andgave comparatively simple and self-contained proofsWe alsoshowed how the idea of excluding a part of the imaging regioncan be incorporated into this formulation
The regularity assumptions can beweakened even furtherOur proofs only require unique continuation arguments forthe reference conductivity 120590
0and the existence of the dipole
functionsTwo major open theoretical questions still exist in the
context of the FactorizationMethodThe theoretical justifica-tion of the method requires a definiteness condition (on thewhole domain or after excluding an a priori known part of thedomain) It is unknownwhether themethodrsquos range criterionholds without such a definiteness conditionThe second openquestion concerns the numerical stability of the methodrsquosrange criterion So far there are no rigorous convergenceresults for numerical implementations of this range criterion(see however Lechleiter [32] for a first step in this direction)As a promising approach to overcome both problems wewould like to point out the recent work on monotony-basedmethods [28]
Acknowledgments
This paper was funded by the German Research Foundation(DFG) and the University of Wurzburg in the fundingprogramme Open Access Publishing
References
[1] A Kirsch ldquoCharacterization of the shape of a scattering obstacleusing the spectral data of the far field operatorrdquo Inverse Prob-lems vol 14 no 6 pp 1489ndash1512 1998
[2] M Bruhl Gebietserkennung in der elektrischen Impedanztomo-graphie [PhD thesis] Universitat Karlsruhe 1999
[3] M Bruhl and M Hanke ldquoNumerical implementation of twononiterative methods for locating inclusions by impedancetomographyrdquo Inverse Problems vol 16 no 4 pp 1029ndash10422000
[4] M Bruhl ldquoExplicit characterization of inclusions in electricalimpedance tomographyrdquo SIAM Journal on Mathematical Anal-ysis vol 32 no 6 pp 1327ndash1341 2001
[5] M Hanke and M Bruhl ldquoRecent progress in electricalimpedance tomographyrdquo Inverse Problems vol 19 no 6 ppS65ndashS90 2003
[6] A Kirsch andN GrinbergTheFactorizationMethod for InverseProblems vol 36 ofOxford Lecture Series inMathematics and ItsApplications Oxford University Press Oxford UK 2007
[7] M Hanke and A Kirsch ldquoSampling methodsrdquo in Handbook ofMathematical Models in Imaging O Scherzer Ed pp 501ndash550Springer 2011
[8] A Kirsch ldquoThe factorization method for a class of inverseelliptic problemsrdquo Mathematische Nachrichten vol 278 pp258ndash277 2005
[9] B Gebauer ldquoThe Factorization Method for real elliptic prob-lemsrdquo Zeitschrift fur Analysis und ihre Anwendung vol 25 no1 pp 81ndash102 2006
[10] A I Nachman L Paivarinta and A Teirila ldquoOn imagingobstacles inside inhomogeneous mediardquo Journal of FunctionalAnalysis vol 252 no 2 pp 490ndash516 2007
[11] M Hanke and B Schappel ldquoThe factorization method forelectrical impedance tomography in the half-spacerdquo SIAMJournal on Applied Mathematics vol 68 no 4 pp 907ndash9242008
[12] N Hyvonen ldquoComplete electrode model of electricalimpedance tomography approximation properties andcharacterization of inclusionsrdquo SIAM Journal on AppliedMathematics vol 64 pp 902ndash931 2004
[13] A Lechleiter N Hyvonen and H Hakula ldquoThe factorizationmethod applied to the complete electrode model of impedancetomographyrdquo SIAM Journal onAppliedMathematics vol 68 no4 pp 1097ndash1121 2008
[14] N Hyvonen H Hakula and S Pursiainen ldquoNumerical imple-mentation of the factorization method within the completeelectrode model of electrical impedance tomographyrdquo InverseProblems and Imaging vol 1 pp 299ndash317 2007
[15] HHakula andNHyvonen ldquoOn computation of test dipoles forfactorization methodrdquo BIT Numerical Mathematics vol 49 no1 pp 75ndash91 2009
[16] B Harrach and J K Seo ldquoDetecting inclusions in electri-cal impedance tomography without reference measurementsrdquoSIAM Journal on Applied Mathematics vol 69 no 6 pp 1662ndash1681 2009
[17] B Harrach J K Seo and E J Woo ldquoFactorization methodand its physical justification in frequency-difference electricalimpedance tomographyrdquo IEEE Transactions on Medical Imag-ing vol 29 no 11 pp 1918ndash1926 2010
[18] S Schmitt ldquoThe factorization method for EIT in the case ofmixed inclusionsrdquo Inverse Problems vol 25 no 6 Article ID065012 2009
8 Computational and Mathematical Methods in Medicine
[19] S Schmitt and A Kirsch ldquoA factorization scheme for determin-ing conductivity contrasts in impedance tomographyrdquo InverseProblems vol 27 Article ID 095005 2011
[20] B Gebauer and N Hyvonen ldquoFactorization method and irreg-ular inclusions in electrical impedance tomographyrdquo InverseProblems vol 23 no 5 pp 2159ndash2170 2007
[21] B Gebauer ldquoLocalized potentials in electrical impedancetomographyrdquo Inverse Problems and Imaging vol 2 pp 251ndash2692008
[22] H Kang J K Seo and D Sheen ldquoThe inverse conductivityproblem with one measurement Stability and estimation ofsizerdquo SIAM Journal on Mathematical Analysis vol 28 no 6 pp1389ndash1405 1997
[23] M Ikehata ldquoSize estimation of inclusionrdquo Journal of Inverse andIll-Posed Problems vol 6 no 2 pp 127ndash140 1998
[24] T Ide H Isozaki S Nakata S Siltanen and G UhlmannldquoProbing for electrical inclusions with complex sphericalwavesrdquo Communications on Pure and Applied Mathematics vol60 no 10 pp 1415ndash1442 2007
[25] B Harrach and J K Seo ldquoExact shape-reconstruction by one-step linearization in electrical impedance tomographyrdquo TheSIAM Journal onMathematical Analysis vol 42 no 4 pp 1505ndash1518 2010
[26] N Bourbaki Elements of Mathematics Topological VectorSpaces chapters 1ndash5 Springer Berlin Germany 2003
[27] F Fruhauf B Gebauer andO Scherzer ldquoDetecting interfaces ina parabolic-elliptic problem from surfacemeasurementsrdquo SIAMJournal on Numerical Analysis vol 45 no 2 pp 810ndash836 2007
[28] B Harrach and M Ullrich ldquoMonotony based shape recon-struction in electrical impedance tomographyrdquo submitted forpublication
[29] S Kusiak and J Sylvester ldquoThe scattering supportrdquoCommunica-tions on Pure and Applied Mathematics vol 56 no 11 pp 1525ndash1548 2003
[30] B Gebauer and N Hyvonen ldquoFactorization method andinclusions of mixed type in an inverse elliptic boundary valueproblemrdquo Inverse Problems and Imaging vol 2 pp 355ndash3722008
[31] N I Grinberg and A Kirsch ldquoThe factorization method forobstacles with a-priori separated sound-soft and sound-hardpartsrdquo Mathematics and Computers in Simulation vol 66 no4-5 pp 267ndash279 2004
[32] A Lechleiter ldquoA regularization technique for the factorizationmethodrdquo Inverse Problems vol 22 no 5 pp 1605ndash1625 2006
Submit your manuscripts athttpwwwhindawicom
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MEDIATORSINFLAMMATION
of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Behavioural Neurology
EndocrinologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Disease Markers
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
OncologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Oxidative Medicine and Cellular Longevity
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
PPAR Research
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Immunology ResearchHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
ObesityJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational and Mathematical Methods in Medicine
OphthalmologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Diabetes ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentAIDS
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Gastroenterology Research and Practice
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Parkinsonrsquos Disease
Evidence-Based Complementary and Alternative Medicine
Volume 2014Hindawi Publishing Corporationhttpwwwhindawicom
6 Computational and Mathematical Methods in Medicine
Theorem 10 Let 1205900= 1 and 120590 = 1 + 120594
119863+ minus (12)120594
119863minus where
119863+
119863minus
sube Ω are open Let 119864 sube Ω be an open set
(a) If 119863+ sube 119864 and 119863minus cup 119864 sube Ω has a connectedcomplement then for all 119911 isin Ω 119911 notin 120597(119863
minus
cup119864) and alldipole directions 119889 isin R119899 119889 = 1
119911 isin 119863minus
cup 119864 iff Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1) + 119871119864119871lowast
119864
1003816100381610038161003816
12
)
(51)
(b) If 119863minus sube 119864 and 119863+ cup 119864 sube Ω has a connectedcomplement then for all 119911 isin Ω 119911 notin 120597(119863
+
cup119864) and alldipole directions 119889 isin R119899 119889 = 1
119911 isin 119863+
cup 119864 iff Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1) minus 119871119864119871lowast
119864
1003816100381610038161003816
12
)
(52)
Proof Themonotony Lemma 1 yields that for all119892 isin 1198712
(120597Ω)
1
2int119863minus
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 minus int119863+
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
le intΣ
119892 (Λ (120590) minus Λ (1)) 119892d119904
le int119863minus
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 minus1
2int119863+
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(53)
Since (cf Lemma 2)
intΣ
119892 (119871119864119871lowast
119864) 119892d119904 = 1003817100381710038171003817119871
lowast
1198641198921003817100381710038171003817
2
= int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 (54)
it follows for case (a) that
1
2int119863minuscup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 le intΣ
119892 (Λ (120590) minus Λ (1) + 2119871119864119871lowast
119864) 119892d119904
le int119863minuscup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909(55)
Using the functional analytic Lemma 3 this implies that
R (119871119863minuscup119864) = R (
1003816100381610038161003816Λ (120590) minus Λ (1) + 2119871119864119871lowast
119864
1003816100381610038161003816
12
) (56)
so that the assertion (a) follows from Lemma 4In case (b) we obtain that
3
2int119863+cup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 ge intΣ
119892 (Λ (1) minus Λ (120590) + 2119871119864119871lowast
119864) 119892d119904
ge1
2int119863+cup119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909(57)
and the same arguments as above yield the assertion
We can also extend these ideas to the general setting ofSection 32
Theorem 11 Let 120590 isin 119871infin
+(Ω) and let 120590
0isin 119871
infin
+(Ω) be a
piecewise analytic function Let 119864 sube Ω be a measurable setChoose 120572 120573 isin R such that
120572 gt1003817100381710038171003817120590 minus 120590
0
1003817100381710038171003817119871infin(Ω) 120573 gt
1003817100381710038171003817100381710038171003817
1205900
120590(1205900minus 120590)
1003817100381710038171003817100381710038171003817119871infin(Ω) (58)
(a) If 120590 le 1205900on Ω 119864 then for all 119911 isin Ω that have a
neighborhood in which 1205900is analytic as well as all unit
vectors 119889 isin R119899 119889 = 1
119911 isin inn supp (120590 minus 1205900) cup 119864 implies
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590)minusΛ (1205900)+120572119871
119864119871lowast
119864
1003816100381610038161003816
12
)
(59)
and
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864
1003816100381610038161003816
12
) implies
119911 isin outΣ(supp (120590 minus 120590
0) cup 119864)
(60)
(b) If 120590 ge 1205900on Ω 119864 then for all 119911 isin Ω that have a
neighborhood in which 1205900is analytic as well as all unit
vectors 119889 isin R119899 119889 = 1
119911 isin inn supp (120590 minus 1205900) cup 119864 implies
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900) minus 120573119871
119864119871lowast
119864
1003816100381610038161003816
12
)
(61)
and
Φ119911119889|Σisin R (
1003816100381610038161003816Λ (120590) minus Λ (1205900) minus 120573119871
119864119871lowast
119864
1003816100381610038161003816
12
) implies
119911 isin outΣ(supp (120590 minus 120590
0) cup 119864)
(62)
Proof For every 119911 isin inn supp(120590minus1205900)with 119911 notin 119864 there exists
a small ball 119861120598(119911) and 120575 gt 0 so that 120590
0minus120590 ge 120575 on 119861
120598(119911) Using
the monotony Lemma 1 it follows that for all 119892 isin 1198712
(120597Ω)
intΣ
119892 (Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864) 119892d119904
ge intΩ
(1205900minus 120590
1)1003816100381610038161003816nabla1199060
1003816100381610038161003816
2d119909 + 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
ge 120575int119861120598(119911)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909+(120572 minus1003817100381710038171003817120590 minus 120590
0
1003817100381710038171003817119871infin(Ω))int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(63)
Computational and Mathematical Methods in Medicine 7
As in the previous proofs we obtain from Lemmas 2 and 3that
R (1003816100381610038161003816Λ (120590) minus Λ (120590
0) + 120572119871
119864119871lowast
119864
1003816100381610038161003816
12
) supe R (119871119861120598(119911)cup119864
) (64)
so that the first implication of (a) follows from Lemma 7The monotony Lemma 1 also implies that
intΣ
119892 (Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864) 119892d119904
le intΩ
1205900
120590(1205900minus 120590)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 + 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
le
1003817100381710038171003817100381710038171003817
1205900
120590(1205900minus 120590)
1003817100381710038171003817100381710038171003817119871infin(Ω)intsupp(120590minus1205900)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
+ 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(65)
so that the second implication of (a) follows from Lemmas 23 and 7 Assertion (b) can be proven analogously
Remark 9 also applies to this case
4 Conclusions and Remarks
The Factorization Method can be used to detect regionsin which a conductivity differs from a known referenceconductivity In this work we summarized the progresson the methodrsquos theoretical foundation We formulated themethod for general piecewise analytic conductivities andgave comparatively simple and self-contained proofsWe alsoshowed how the idea of excluding a part of the imaging regioncan be incorporated into this formulation
The regularity assumptions can beweakened even furtherOur proofs only require unique continuation arguments forthe reference conductivity 120590
0and the existence of the dipole
functionsTwo major open theoretical questions still exist in the
context of the FactorizationMethodThe theoretical justifica-tion of the method requires a definiteness condition (on thewhole domain or after excluding an a priori known part of thedomain) It is unknownwhether themethodrsquos range criterionholds without such a definiteness conditionThe second openquestion concerns the numerical stability of the methodrsquosrange criterion So far there are no rigorous convergenceresults for numerical implementations of this range criterion(see however Lechleiter [32] for a first step in this direction)As a promising approach to overcome both problems wewould like to point out the recent work on monotony-basedmethods [28]
Acknowledgments
This paper was funded by the German Research Foundation(DFG) and the University of Wurzburg in the fundingprogramme Open Access Publishing
References
[1] A Kirsch ldquoCharacterization of the shape of a scattering obstacleusing the spectral data of the far field operatorrdquo Inverse Prob-lems vol 14 no 6 pp 1489ndash1512 1998
[2] M Bruhl Gebietserkennung in der elektrischen Impedanztomo-graphie [PhD thesis] Universitat Karlsruhe 1999
[3] M Bruhl and M Hanke ldquoNumerical implementation of twononiterative methods for locating inclusions by impedancetomographyrdquo Inverse Problems vol 16 no 4 pp 1029ndash10422000
[4] M Bruhl ldquoExplicit characterization of inclusions in electricalimpedance tomographyrdquo SIAM Journal on Mathematical Anal-ysis vol 32 no 6 pp 1327ndash1341 2001
[5] M Hanke and M Bruhl ldquoRecent progress in electricalimpedance tomographyrdquo Inverse Problems vol 19 no 6 ppS65ndashS90 2003
[6] A Kirsch andN GrinbergTheFactorizationMethod for InverseProblems vol 36 ofOxford Lecture Series inMathematics and ItsApplications Oxford University Press Oxford UK 2007
[7] M Hanke and A Kirsch ldquoSampling methodsrdquo in Handbook ofMathematical Models in Imaging O Scherzer Ed pp 501ndash550Springer 2011
[8] A Kirsch ldquoThe factorization method for a class of inverseelliptic problemsrdquo Mathematische Nachrichten vol 278 pp258ndash277 2005
[9] B Gebauer ldquoThe Factorization Method for real elliptic prob-lemsrdquo Zeitschrift fur Analysis und ihre Anwendung vol 25 no1 pp 81ndash102 2006
[10] A I Nachman L Paivarinta and A Teirila ldquoOn imagingobstacles inside inhomogeneous mediardquo Journal of FunctionalAnalysis vol 252 no 2 pp 490ndash516 2007
[11] M Hanke and B Schappel ldquoThe factorization method forelectrical impedance tomography in the half-spacerdquo SIAMJournal on Applied Mathematics vol 68 no 4 pp 907ndash9242008
[12] N Hyvonen ldquoComplete electrode model of electricalimpedance tomography approximation properties andcharacterization of inclusionsrdquo SIAM Journal on AppliedMathematics vol 64 pp 902ndash931 2004
[13] A Lechleiter N Hyvonen and H Hakula ldquoThe factorizationmethod applied to the complete electrode model of impedancetomographyrdquo SIAM Journal onAppliedMathematics vol 68 no4 pp 1097ndash1121 2008
[14] N Hyvonen H Hakula and S Pursiainen ldquoNumerical imple-mentation of the factorization method within the completeelectrode model of electrical impedance tomographyrdquo InverseProblems and Imaging vol 1 pp 299ndash317 2007
[15] HHakula andNHyvonen ldquoOn computation of test dipoles forfactorization methodrdquo BIT Numerical Mathematics vol 49 no1 pp 75ndash91 2009
[16] B Harrach and J K Seo ldquoDetecting inclusions in electri-cal impedance tomography without reference measurementsrdquoSIAM Journal on Applied Mathematics vol 69 no 6 pp 1662ndash1681 2009
[17] B Harrach J K Seo and E J Woo ldquoFactorization methodand its physical justification in frequency-difference electricalimpedance tomographyrdquo IEEE Transactions on Medical Imag-ing vol 29 no 11 pp 1918ndash1926 2010
[18] S Schmitt ldquoThe factorization method for EIT in the case ofmixed inclusionsrdquo Inverse Problems vol 25 no 6 Article ID065012 2009
8 Computational and Mathematical Methods in Medicine
[19] S Schmitt and A Kirsch ldquoA factorization scheme for determin-ing conductivity contrasts in impedance tomographyrdquo InverseProblems vol 27 Article ID 095005 2011
[20] B Gebauer and N Hyvonen ldquoFactorization method and irreg-ular inclusions in electrical impedance tomographyrdquo InverseProblems vol 23 no 5 pp 2159ndash2170 2007
[21] B Gebauer ldquoLocalized potentials in electrical impedancetomographyrdquo Inverse Problems and Imaging vol 2 pp 251ndash2692008
[22] H Kang J K Seo and D Sheen ldquoThe inverse conductivityproblem with one measurement Stability and estimation ofsizerdquo SIAM Journal on Mathematical Analysis vol 28 no 6 pp1389ndash1405 1997
[23] M Ikehata ldquoSize estimation of inclusionrdquo Journal of Inverse andIll-Posed Problems vol 6 no 2 pp 127ndash140 1998
[24] T Ide H Isozaki S Nakata S Siltanen and G UhlmannldquoProbing for electrical inclusions with complex sphericalwavesrdquo Communications on Pure and Applied Mathematics vol60 no 10 pp 1415ndash1442 2007
[25] B Harrach and J K Seo ldquoExact shape-reconstruction by one-step linearization in electrical impedance tomographyrdquo TheSIAM Journal onMathematical Analysis vol 42 no 4 pp 1505ndash1518 2010
[26] N Bourbaki Elements of Mathematics Topological VectorSpaces chapters 1ndash5 Springer Berlin Germany 2003
[27] F Fruhauf B Gebauer andO Scherzer ldquoDetecting interfaces ina parabolic-elliptic problem from surfacemeasurementsrdquo SIAMJournal on Numerical Analysis vol 45 no 2 pp 810ndash836 2007
[28] B Harrach and M Ullrich ldquoMonotony based shape recon-struction in electrical impedance tomographyrdquo submitted forpublication
[29] S Kusiak and J Sylvester ldquoThe scattering supportrdquoCommunica-tions on Pure and Applied Mathematics vol 56 no 11 pp 1525ndash1548 2003
[30] B Gebauer and N Hyvonen ldquoFactorization method andinclusions of mixed type in an inverse elliptic boundary valueproblemrdquo Inverse Problems and Imaging vol 2 pp 355ndash3722008
[31] N I Grinberg and A Kirsch ldquoThe factorization method forobstacles with a-priori separated sound-soft and sound-hardpartsrdquo Mathematics and Computers in Simulation vol 66 no4-5 pp 267ndash279 2004
[32] A Lechleiter ldquoA regularization technique for the factorizationmethodrdquo Inverse Problems vol 22 no 5 pp 1605ndash1625 2006
Submit your manuscripts athttpwwwhindawicom
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MEDIATORSINFLAMMATION
of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Behavioural Neurology
EndocrinologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Disease Markers
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
OncologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Oxidative Medicine and Cellular Longevity
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
PPAR Research
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Immunology ResearchHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
ObesityJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational and Mathematical Methods in Medicine
OphthalmologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Diabetes ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentAIDS
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Gastroenterology Research and Practice
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Parkinsonrsquos Disease
Evidence-Based Complementary and Alternative Medicine
Volume 2014Hindawi Publishing Corporationhttpwwwhindawicom
Computational and Mathematical Methods in Medicine 7
As in the previous proofs we obtain from Lemmas 2 and 3that
R (1003816100381610038161003816Λ (120590) minus Λ (120590
0) + 120572119871
119864119871lowast
119864
1003816100381610038161003816
12
) supe R (119871119861120598(119911)cup119864
) (64)
so that the first implication of (a) follows from Lemma 7The monotony Lemma 1 also implies that
intΣ
119892 (Λ (120590) minus Λ (1205900) + 120572119871
119864119871lowast
119864) 119892d119904
le intΩ
1205900
120590(1205900minus 120590)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909 + 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
le
1003817100381710038171003817100381710038171003817
1205900
120590(1205900minus 120590)
1003817100381710038171003817100381710038171003817119871infin(Ω)intsupp(120590minus1205900)
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
+ 120572int119864
1003816100381610038161003816nabla11990601003816100381610038161003816
2d119909
(65)
so that the second implication of (a) follows from Lemmas 23 and 7 Assertion (b) can be proven analogously
Remark 9 also applies to this case
4 Conclusions and Remarks
The Factorization Method can be used to detect regionsin which a conductivity differs from a known referenceconductivity In this work we summarized the progresson the methodrsquos theoretical foundation We formulated themethod for general piecewise analytic conductivities andgave comparatively simple and self-contained proofsWe alsoshowed how the idea of excluding a part of the imaging regioncan be incorporated into this formulation
The regularity assumptions can beweakened even furtherOur proofs only require unique continuation arguments forthe reference conductivity 120590
0and the existence of the dipole
functionsTwo major open theoretical questions still exist in the
context of the FactorizationMethodThe theoretical justifica-tion of the method requires a definiteness condition (on thewhole domain or after excluding an a priori known part of thedomain) It is unknownwhether themethodrsquos range criterionholds without such a definiteness conditionThe second openquestion concerns the numerical stability of the methodrsquosrange criterion So far there are no rigorous convergenceresults for numerical implementations of this range criterion(see however Lechleiter [32] for a first step in this direction)As a promising approach to overcome both problems wewould like to point out the recent work on monotony-basedmethods [28]
Acknowledgments
This paper was funded by the German Research Foundation(DFG) and the University of Wurzburg in the fundingprogramme Open Access Publishing
References
[1] A Kirsch ldquoCharacterization of the shape of a scattering obstacleusing the spectral data of the far field operatorrdquo Inverse Prob-lems vol 14 no 6 pp 1489ndash1512 1998
[2] M Bruhl Gebietserkennung in der elektrischen Impedanztomo-graphie [PhD thesis] Universitat Karlsruhe 1999
[3] M Bruhl and M Hanke ldquoNumerical implementation of twononiterative methods for locating inclusions by impedancetomographyrdquo Inverse Problems vol 16 no 4 pp 1029ndash10422000
[4] M Bruhl ldquoExplicit characterization of inclusions in electricalimpedance tomographyrdquo SIAM Journal on Mathematical Anal-ysis vol 32 no 6 pp 1327ndash1341 2001
[5] M Hanke and M Bruhl ldquoRecent progress in electricalimpedance tomographyrdquo Inverse Problems vol 19 no 6 ppS65ndashS90 2003
[6] A Kirsch andN GrinbergTheFactorizationMethod for InverseProblems vol 36 ofOxford Lecture Series inMathematics and ItsApplications Oxford University Press Oxford UK 2007
[7] M Hanke and A Kirsch ldquoSampling methodsrdquo in Handbook ofMathematical Models in Imaging O Scherzer Ed pp 501ndash550Springer 2011
[8] A Kirsch ldquoThe factorization method for a class of inverseelliptic problemsrdquo Mathematische Nachrichten vol 278 pp258ndash277 2005
[9] B Gebauer ldquoThe Factorization Method for real elliptic prob-lemsrdquo Zeitschrift fur Analysis und ihre Anwendung vol 25 no1 pp 81ndash102 2006
[10] A I Nachman L Paivarinta and A Teirila ldquoOn imagingobstacles inside inhomogeneous mediardquo Journal of FunctionalAnalysis vol 252 no 2 pp 490ndash516 2007
[11] M Hanke and B Schappel ldquoThe factorization method forelectrical impedance tomography in the half-spacerdquo SIAMJournal on Applied Mathematics vol 68 no 4 pp 907ndash9242008
[12] N Hyvonen ldquoComplete electrode model of electricalimpedance tomography approximation properties andcharacterization of inclusionsrdquo SIAM Journal on AppliedMathematics vol 64 pp 902ndash931 2004
[13] A Lechleiter N Hyvonen and H Hakula ldquoThe factorizationmethod applied to the complete electrode model of impedancetomographyrdquo SIAM Journal onAppliedMathematics vol 68 no4 pp 1097ndash1121 2008
[14] N Hyvonen H Hakula and S Pursiainen ldquoNumerical imple-mentation of the factorization method within the completeelectrode model of electrical impedance tomographyrdquo InverseProblems and Imaging vol 1 pp 299ndash317 2007
[15] HHakula andNHyvonen ldquoOn computation of test dipoles forfactorization methodrdquo BIT Numerical Mathematics vol 49 no1 pp 75ndash91 2009
[16] B Harrach and J K Seo ldquoDetecting inclusions in electri-cal impedance tomography without reference measurementsrdquoSIAM Journal on Applied Mathematics vol 69 no 6 pp 1662ndash1681 2009
[17] B Harrach J K Seo and E J Woo ldquoFactorization methodand its physical justification in frequency-difference electricalimpedance tomographyrdquo IEEE Transactions on Medical Imag-ing vol 29 no 11 pp 1918ndash1926 2010
[18] S Schmitt ldquoThe factorization method for EIT in the case ofmixed inclusionsrdquo Inverse Problems vol 25 no 6 Article ID065012 2009
8 Computational and Mathematical Methods in Medicine
[19] S Schmitt and A Kirsch ldquoA factorization scheme for determin-ing conductivity contrasts in impedance tomographyrdquo InverseProblems vol 27 Article ID 095005 2011
[20] B Gebauer and N Hyvonen ldquoFactorization method and irreg-ular inclusions in electrical impedance tomographyrdquo InverseProblems vol 23 no 5 pp 2159ndash2170 2007
[21] B Gebauer ldquoLocalized potentials in electrical impedancetomographyrdquo Inverse Problems and Imaging vol 2 pp 251ndash2692008
[22] H Kang J K Seo and D Sheen ldquoThe inverse conductivityproblem with one measurement Stability and estimation ofsizerdquo SIAM Journal on Mathematical Analysis vol 28 no 6 pp1389ndash1405 1997
[23] M Ikehata ldquoSize estimation of inclusionrdquo Journal of Inverse andIll-Posed Problems vol 6 no 2 pp 127ndash140 1998
[24] T Ide H Isozaki S Nakata S Siltanen and G UhlmannldquoProbing for electrical inclusions with complex sphericalwavesrdquo Communications on Pure and Applied Mathematics vol60 no 10 pp 1415ndash1442 2007
[25] B Harrach and J K Seo ldquoExact shape-reconstruction by one-step linearization in electrical impedance tomographyrdquo TheSIAM Journal onMathematical Analysis vol 42 no 4 pp 1505ndash1518 2010
[26] N Bourbaki Elements of Mathematics Topological VectorSpaces chapters 1ndash5 Springer Berlin Germany 2003
[27] F Fruhauf B Gebauer andO Scherzer ldquoDetecting interfaces ina parabolic-elliptic problem from surfacemeasurementsrdquo SIAMJournal on Numerical Analysis vol 45 no 2 pp 810ndash836 2007
[28] B Harrach and M Ullrich ldquoMonotony based shape recon-struction in electrical impedance tomographyrdquo submitted forpublication
[29] S Kusiak and J Sylvester ldquoThe scattering supportrdquoCommunica-tions on Pure and Applied Mathematics vol 56 no 11 pp 1525ndash1548 2003
[30] B Gebauer and N Hyvonen ldquoFactorization method andinclusions of mixed type in an inverse elliptic boundary valueproblemrdquo Inverse Problems and Imaging vol 2 pp 355ndash3722008
[31] N I Grinberg and A Kirsch ldquoThe factorization method forobstacles with a-priori separated sound-soft and sound-hardpartsrdquo Mathematics and Computers in Simulation vol 66 no4-5 pp 267ndash279 2004
[32] A Lechleiter ldquoA regularization technique for the factorizationmethodrdquo Inverse Problems vol 22 no 5 pp 1605ndash1625 2006
Submit your manuscripts athttpwwwhindawicom
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MEDIATORSINFLAMMATION
of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Behavioural Neurology
EndocrinologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Disease Markers
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
OncologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Oxidative Medicine and Cellular Longevity
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
PPAR Research
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Immunology ResearchHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
ObesityJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational and Mathematical Methods in Medicine
OphthalmologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Diabetes ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentAIDS
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Gastroenterology Research and Practice
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Parkinsonrsquos Disease
Evidence-Based Complementary and Alternative Medicine
Volume 2014Hindawi Publishing Corporationhttpwwwhindawicom
8 Computational and Mathematical Methods in Medicine
[19] S Schmitt and A Kirsch ldquoA factorization scheme for determin-ing conductivity contrasts in impedance tomographyrdquo InverseProblems vol 27 Article ID 095005 2011
[20] B Gebauer and N Hyvonen ldquoFactorization method and irreg-ular inclusions in electrical impedance tomographyrdquo InverseProblems vol 23 no 5 pp 2159ndash2170 2007
[21] B Gebauer ldquoLocalized potentials in electrical impedancetomographyrdquo Inverse Problems and Imaging vol 2 pp 251ndash2692008
[22] H Kang J K Seo and D Sheen ldquoThe inverse conductivityproblem with one measurement Stability and estimation ofsizerdquo SIAM Journal on Mathematical Analysis vol 28 no 6 pp1389ndash1405 1997
[23] M Ikehata ldquoSize estimation of inclusionrdquo Journal of Inverse andIll-Posed Problems vol 6 no 2 pp 127ndash140 1998
[24] T Ide H Isozaki S Nakata S Siltanen and G UhlmannldquoProbing for electrical inclusions with complex sphericalwavesrdquo Communications on Pure and Applied Mathematics vol60 no 10 pp 1415ndash1442 2007
[25] B Harrach and J K Seo ldquoExact shape-reconstruction by one-step linearization in electrical impedance tomographyrdquo TheSIAM Journal onMathematical Analysis vol 42 no 4 pp 1505ndash1518 2010
[26] N Bourbaki Elements of Mathematics Topological VectorSpaces chapters 1ndash5 Springer Berlin Germany 2003
[27] F Fruhauf B Gebauer andO Scherzer ldquoDetecting interfaces ina parabolic-elliptic problem from surfacemeasurementsrdquo SIAMJournal on Numerical Analysis vol 45 no 2 pp 810ndash836 2007
[28] B Harrach and M Ullrich ldquoMonotony based shape recon-struction in electrical impedance tomographyrdquo submitted forpublication
[29] S Kusiak and J Sylvester ldquoThe scattering supportrdquoCommunica-tions on Pure and Applied Mathematics vol 56 no 11 pp 1525ndash1548 2003
[30] B Gebauer and N Hyvonen ldquoFactorization method andinclusions of mixed type in an inverse elliptic boundary valueproblemrdquo Inverse Problems and Imaging vol 2 pp 355ndash3722008
[31] N I Grinberg and A Kirsch ldquoThe factorization method forobstacles with a-priori separated sound-soft and sound-hardpartsrdquo Mathematics and Computers in Simulation vol 66 no4-5 pp 267ndash279 2004
[32] A Lechleiter ldquoA regularization technique for the factorizationmethodrdquo Inverse Problems vol 22 no 5 pp 1605ndash1625 2006
Submit your manuscripts athttpwwwhindawicom
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MEDIATORSINFLAMMATION
of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Behavioural Neurology
EndocrinologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Disease Markers
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
OncologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Oxidative Medicine and Cellular Longevity
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
PPAR Research
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Immunology ResearchHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
ObesityJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational and Mathematical Methods in Medicine
OphthalmologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Diabetes ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentAIDS
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Gastroenterology Research and Practice
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Parkinsonrsquos Disease
Evidence-Based Complementary and Alternative Medicine
Volume 2014Hindawi Publishing Corporationhttpwwwhindawicom
Submit your manuscripts athttpwwwhindawicom
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MEDIATORSINFLAMMATION
of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Behavioural Neurology
EndocrinologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Disease Markers
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
OncologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Oxidative Medicine and Cellular Longevity
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
PPAR Research
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Immunology ResearchHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
ObesityJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational and Mathematical Methods in Medicine
OphthalmologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Diabetes ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentAIDS
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Gastroenterology Research and Practice
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Parkinsonrsquos Disease
Evidence-Based Complementary and Alternative Medicine
Volume 2014Hindawi Publishing Corporationhttpwwwhindawicom