Research article Special Issues

Brain tissue segmentation via non-local fuzzy c-means clustering combined with Markov random field

  • Received: 24 October 2021 Accepted: 14 December 2021 Published: 20 December 2021
  • The segmentation and extraction of brain tissue in magnetic resonance imaging (MRI) is a meaningful task because it provides a diagnosis and treatment basis for observing brain tissue development, delineating lesions, and planning surgery. However, MRI images are often damaged by factors such as noise, low contrast and intensity brightness, which seriously affect the accuracy of segmentation. A non-local fuzzy c-means clustering framework incorporating the Markov random field for brain tissue segmentation is proposed in this paper. Firstly, according to the statistical characteristics that MRF can effectively describe the local spatial correlation of an image, a new distance metric with neighborhood constraints is constructed by combining probabilistic statistical information. Secondly, a non-local regularization term is integrated into the objective function to utilize the global structure feature of the image, so that both the local and global information of the image can be taken into account. In addition, a linear model of inhomogeneous intensity is also built to estimate the bias field in brain MRI, which has achieved the goal of overcoming the intensity inhomogeneity. The proposed model fully considers the randomness and fuzziness in the image segmentation problem, and obtains the prior knowledge of the image reasonably, which reduces the influence of low contrast in the MRI images. Then the experimental results demonstrate that the proposed method can eliminate the noise and intensity inhomogeneity of the MRI image and effectively improve the image segmentation accuracy.

    Citation: Jianhua Song, Lei Yuan. Brain tissue segmentation via non-local fuzzy c-means clustering combined with Markov random field[J]. Mathematical Biosciences and Engineering, 2022, 19(2): 1891-1908. doi: 10.3934/mbe.2022089

    Related Papers:

    [1] Lukáš Pichl, Taisei Kaizoji . Volatility Analysis of Bitcoin Price Time Series. Quantitative Finance and Economics, 2017, 1(4): 474-485. doi: 10.3934/QFE.2017.4.474
    [2] Andres Fernandez, Norman R. Swanson . Further Evidence on the Usefulness of Real-Time Datasets for Economic Forecasting. Quantitative Finance and Economics, 2017, 1(1): 2-25. doi: 10.3934/QFE.2017.1.2
    [3] Samuel Asante Gyamerah . Modelling the volatility of Bitcoin returns using GARCH models. Quantitative Finance and Economics, 2019, 3(4): 739-753. doi: 10.3934/QFE.2019.4.739
    [4] Guillermo Peña . Interest rates affect public expenditure growth. Quantitative Finance and Economics, 2023, 7(4): 622-645. doi: 10.3934/QFE.2023030
    [5] Abdul Haque, Huma Fatima, Ammar Abid, Muhammad Ali Jibran Qamar . Impact of firm-level uncertainty on earnings management and role of accounting conservatism. Quantitative Finance and Economics, 2019, 3(4): 772-794. doi: 10.3934/QFE.2019.4.772
    [6] Arifenur Güngör, Hüseyin Taştan . On macroeconomic determinants of co-movements among international stock markets: evidence from DCC-MIDAS approach. Quantitative Finance and Economics, 2021, 5(1): 19-39. doi: 10.3934/QFE.2021002
    [7] Cemile Özgür, Vedat Sarıkovanlık . An application of Regular Vine copula in portfolio risk forecasting: evidence from Istanbul stock exchange. Quantitative Finance and Economics, 2021, 5(3): 452-470. doi: 10.3934/QFE.2021020
    [8] Md Qamruzzaman, Jianguo Wei . Do financial inclusion, stock market development attract foreign capital flows in developing economy: a panel data investigation. Quantitative Finance and Economics, 2019, 3(1): 88-108. doi: 10.3934/QFE.2019.1.88
    [9] David Melkuev, Danqiao Guo, Tony S. Wirjanto . Applications of random-matrix theory and nonparametric change-point analysis to three notable systemic crises. Quantitative Finance and Economics, 2018, 2(2): 413-467. doi: 10.3934/QFE.2018.2.413
    [10] Fredrik Hobbelhagen, Ioannis Diamantis . A comparative study of symbolic aggregate approximation and topological data analysis. Quantitative Finance and Economics, 2024, 8(4): 705-732. doi: 10.3934/QFE.2024027
  • The segmentation and extraction of brain tissue in magnetic resonance imaging (MRI) is a meaningful task because it provides a diagnosis and treatment basis for observing brain tissue development, delineating lesions, and planning surgery. However, MRI images are often damaged by factors such as noise, low contrast and intensity brightness, which seriously affect the accuracy of segmentation. A non-local fuzzy c-means clustering framework incorporating the Markov random field for brain tissue segmentation is proposed in this paper. Firstly, according to the statistical characteristics that MRF can effectively describe the local spatial correlation of an image, a new distance metric with neighborhood constraints is constructed by combining probabilistic statistical information. Secondly, a non-local regularization term is integrated into the objective function to utilize the global structure feature of the image, so that both the local and global information of the image can be taken into account. In addition, a linear model of inhomogeneous intensity is also built to estimate the bias field in brain MRI, which has achieved the goal of overcoming the intensity inhomogeneity. The proposed model fully considers the randomness and fuzziness in the image segmentation problem, and obtains the prior knowledge of the image reasonably, which reduces the influence of low contrast in the MRI images. Then the experimental results demonstrate that the proposed method can eliminate the noise and intensity inhomogeneity of the MRI image and effectively improve the image segmentation accuracy.



    1. Introduction

    The Caginalp phase-field system

    utΔu+f(u)=θ, (1.1)
    θtΔθ=ut, (1.2)

    has been introduced in [1] in order to describe the phase transition phenomena in certain class of material. In this context, θ denotes the relative temperature (relative to the equilibrium melting temperature), and u is the phase-field or order parameter, f is a given function (precisely, the derivaritve of a double-well potential F). This system has received much attention (see for example, [2], [3], [4], [5], [6], [7], [8], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [22], [29], [33] and [41]). These equations can be derived by introducing the (total Ginzburg-Landau) free energy:

    ψ=Ω(12|u|2+F(u)uθ12θ2)dx, (1.3)

    where Ω is the domain occupied by the system (here, we assume that it is a bounded and smooth domain of Rn, n=1,2 or 3, with boundary Ω), and the enthalpy

    H=u+θ. (1.4)

    Then, the evolution equation for the order parameter u is given by:

    ut=δuψ, (1.5)

    where δu stands for the variational derivative with respect to u, which yields (1.1). Then, we have the energy equation

    Ht=divq, (1.6)

    where q is the heat flux. Assuming finally the classical Fourier law for heat conduction, which prescribes the heat flux as

    q=θ, (1.7)

    we obtain (1.2). Now, a well-known side effect of the Fourier heat law is the infinite speed of propagation of thermal disturbances, deemed physically unreasonable and thus called paradox of heat conduction (see, for example, [9]). In order to account for more realistic features, several variations of (1.7), based, for example, on the Maxwell-Cattaneo law or recent laws from thermomechanics, have been proposed in the context of the Caginalp phase-field system (see, for example, [19], [20], [21], [23], [24], [25], [26], [27], [28], [30], [31], [35], [36], [37], [38], [44], [45] and [46]).

    A different approach to heat conduction was proposed in the Sixties (see, [47], [48] and [49]), where it was observed that two temperatures are involved in the definition of the entropy: the conductive temperature θ, influencing the heat conduction contribution, and the thermodynamic temperature, appearing in the heat supply part. For time-independent models, it appears that these two temperatures coincide in absence of heat supply. Actually, they are different generally in the time depedent case see, for example, [19] and references therein for more discussion on the subject. In particular, this happens for non-simple materials. In that case, the two temperatures are related as follows (see [42], [43]):

    θ=φΔφ. (1.8)

    Our aim in this paper is to study a generalization of the Caginalp phase-field system based on this two temperatures theory and the usual Fourier law with a nonlinear coupling.

    The purpose of our study is the following initial and boundary value problem

    utΔu+f(u)=g(u)(φΔφ), (1.9)
    φtΔφtΔφ=g(u)ut, (1.10)
    u=φ=0onΩ, (1.11)
    u|t=0=u0, φ|t=0=φ0. (1.12)

    The paper is organized as follows. In Section 2, we give the derivation of the model. The Section 3 states existence, regularity and uniqueness results. In Section 4, we address the question of dissipativity properties of the system. The last section, analyzes the spatial behavior of solutions in a semi-infinite cylinder, assuming their existence.

    Thoughout this paper, the same letters c,c,c, and sometimes c denote constants which may change from line to line and also .p will denote the usual Lp norm and (.,.) the usual L2 scalar product. More generally, we will denote by .X the norm in the Banach space X. When there is no possible confusion, . will be noted instead of .2.


    2. Derivation of the model

    In our case, to obtain equations (1.9) and (1.10), the total free energy reads in terms of the conductive temperature θ,

    ψ(u,θ)=Ω(12|u|2+F(u)G(u)θ12θ2)dx, (2.1)

    where f=F and g=G, and (1.5) yields, in view of (1.8), the evolution equation for the order parameter (1.9). Furthermore, the enthalpy now reads

    H=G(u)+θ=G(u)+φΔφ, (2.2)

    which yields thanks to (1.6), the energy equation,

    φtΔφt+divq=g(u)ut. (2.3)

    Considering the usual Fourier law (q=φ), one has (1.10).

    Remark 2.1. We can note that we still have an infinite speed of propagation here.


    3. Existence and uniqueness of solutions

    Before stating the existence result, we make some assumptions on nonlinearities f and g:

    |G(s)|2c1F(s)+c2,c0,c1,c20, (3.1)
    |g(s)s|c3(|G(s)|2+1),c30, (3.2)
    c4sk+2c5F(s)f(s)s+c0c6sk+2c7,c4,c6>0,c5,c70, (3.3)
    |g(s)|c8(|s|+1),|g(s)|c9c8,c90, (3.4)
    |f(s)|c10(|s|k+1),c100, (3.5)

    where k is an integer.

    Theorem 3.1. We assume that (3.1)-(3.4) hold true. For all initial data (u0,φ0)H10(Ω)Lk+2(Ω)×H10(Ω)H2(Ω), the problem (1.9)-(1.12) possesses at least one solution (u,φ) with the following regularity uL(0,T;H10(Ω))Lk+2(Ω),utL2(0,T;L2(Ω)),φL(0,T;H10(Ω)H2(Ω)) and φtL2(0,T;H10(Ω)).

    Proof. The proof is based on the Galerkin scheme. Here, we just make formally computations to get a priori estimates, having in mind that these estimates can be rigourously justified using the Galerkin scheme see, for example, [10], [11] and [40] for details.

    Multiplying (1.9) by ut and integrating over Ω, we get

    12ddt(u2+2ΩF(u)dx)+ut2=Ωg(u)ut(φΔφ)dx. (3.6)

    Multiplying (1.10) by φΔφ and integrating over Ω, we have

    12ddt(φ2+2φ2+Δφ2)+φ2+Δφ2=Ωg(u)ut(φΔφ)dx. (3.7)

    Now, summing (3.6) and (3.7), we are led to,

    ddt(u2+2ΩF(u)dx+φ2+2φ2+Δφ2)+2(ut2+φ2+Δφ2)=0. (3.8)

    Multiplying (1.9) by u and integrating over Ω, we obtain

    12ddtu2+u2+Ωf(u)udx=Ωg(u)u(φΔφ)dx. (3.9)

    Using (3.2)-(3.3), (3.9) becomes

    12ddtu2+u2+cΩF(u)dxcΩ|G(u)|2dx+12(φ2+Δφ2)+c. (3.10)

    Adding (3.8) and (3.10), one has

    dE1dt+2(u2+cΩF(u)dx+ut2+φ2)+Δφ2cΩ|G(u)|2dx+φ2+c, (3.11)

    where

    E1=u2+u2+2ΩF(u)dx+φ2+2φ2+Δφ2 (3.12)

    enjoys

    E1c(u2H1(Ω)+uk+2k+2+φ2H2(Ω))c (3.13)

    and

    E1c(u2H1(Ω)+uk+2k+2+φ2H2(Ω))c. (3.14)

    Multiplying now (1.10) by φt and integrating over Ω, we have

    12ddtφ2+φt2+φt2=Ωg(u)utφtdx. (3.15)

    Taking into account (3.4) and using Hölder's inequality, we get

    12ddtφ2+12φt2+φt2c(u2+1)ut2 (3.16)

    and then, summing (3.11) and (3.16), we have

    dE2dt+2(u2+cΩF(u)dx+ut2+φ2+12Δφ2+12φt2+φt2)cΩ|G(u)|2dx+φ2+c(u2+1)ut2+c, (3.17)

    where

    E2=E1+φ2 (3.18)

    satisfies similar estimates as E1.

    We deduce from (3.1) and (3.17)

    dE2dt+c(φt2+φt2)cE2+c, (3.19)

    which achieve the proof.

    For more regularity on solutions, we make following additional assumptions:

    f(0)=0andf(s)c,c0. (3.20)

    We have:

    Theorem 3.2. Under assumptions of Theorem 3.1 and assuming that (3.20) is satisfied. For every initial data (u0,φ0)H10(Ω)Lk+2(Ω)×H10(Ω)H2(Ω), the problem (1.9)-(1.12) admits at least one solution (u,φ) such that uL(0,T;H10(Ω))Lk+2(Ω),utL(0,T;L2(Ω))L2(0,T;H10(Ω)), φL(0,T;H10(Ω)H2(Ω)) and φtL2(0,T;H10(Ω)H2(Ω)).

    Proof. As above proof, we focus on a priori estimates.

    We multiply (1.10) by Δφt and have, integrating over Ω,

    12ddtφ2+φt2+Δφt2=Ωg(u)utΔφtdx. (3.21)

    Thanks to (3.4) and Hölder's inequality:

    Ωg(u)utΔφtdxcΩ(|u|+1)|ut||Δφt|dxc(u2+1)ut2+12Δφt2 (3.22)

    and then,

    12ddtφ2+φt2+12Δφt2c(u2+1)ut2. (3.23)

    Differentiating (1.9) with respect to time, we get

    2ut2Δut+f(u)ut=g(u)ut(φΔφ)+g(u)(φtΔφt). (3.24)

    Multiplying (3.24) by ut and integrating over Ω, we obtain

    12ddtut2+ut2+Ωf(u)|ut|2dx=Ωg(u)|ut|2(φΔφ)dx+Ωg(u)ut(φtΔφt)dx. (3.25)

    Using (1.10), we write,

    Ωg(u)ut(φtΔφt)dx=Ωg(u)ut(g(u)ut+Δφ)dx=Ω|g(u)ut|2dx+Ωg(u)utΔφdx. (3.26)

    Owing to (3.26), (3.25) reads

    12ddtut2+ut2+Ωf(u)|ut|2dx=Ωg(u)|ut|2(φΔφ)dx+Ωg(u)utΔφdxΩ|g(u)ut|2dx, (3.27)

    since

    Ωg(u)|ut|2(φΔφ)dxcΩ|ut|2(|φ|+|Δφ|)dx12ut2+c(φ2+Δφ2), (3.28)
    Ωg(u)utΔφdx=Ωg(u)uutφdxΩg(u)utφdx (3.29)

    and then,

    |Ωg(u)uutφdx|cΩ|u||ut||φ|dx16ut2+cu2Δφ2 (3.30)

    and

    |Ωg(u)utφdx|cΩ(|u|+1)|ut||φ|dx16ut2+c(u2+1)φ2. (3.31)

    Furthemore,

    Ω|g(u)ut|2dxcΩ(|u|+1)2|ut|2dxc(u2+u2+1)ut2. (3.32)

    Now, collecting (3.27)–(3.32) and owing to (3.20), we are led to

    ddtut2+cut2c(u2H1(Ω)+1)(ut2+φ2H2(Ω)). (3.33)

    Adding (3.19), ε1(3.23) and ε2(3.33), with εi>0,i=1,2, small enough, we obtain

    dE3dt+c(ut2H1(Ω)+φt2H2(Ω))cE3+c, (3.34)

    where

    E3=E2+ε1φ2+ε2ut2 (3.35)

    enjoys

    E3c(u2H(Ω)+uk+2k+2+φ2H2(Ω))c (3.36)

    and

    E3c(u2H(Ω)+uk+2k+2+φ2H2(Ω))c. (3.37)

    We complete the proof applying Gronwall's lemma.

    We now give a uniqueness result

    Theorem 3.3. Under assumptions of Theorem 3.2 and assuming that (3.5) holds true. The problem (1.9)-(1.12) has a unique solution (u,φ), with the above regularity.

    Proof. We suppose the existence of two solutions (u1,φ1) and (u2,φ2) to problem (1.9)-(1.11) associated to initial conditions (u01,φ01) and (u02,φ02), respectively. We then have

    utΔu+f(u1)f(u2)=g(u1)(φΔφ)+(g(u1)g(u2))(φ2Δφ2), (3.38)
    φtΔφtΔφ=g(u1)ut(g(u1)g(u2))u2t, (3.39)
    u|Ω=φ|Ω=0, (3.40)
    u|t=0=u01u02,φ|t=0=φ01φ02, (3.41)

    with u=u1u2, φ=φ1φ2, u0=u01u02 and φ0=φ01φ02.

    Multiplying (3.38) by ut and integrating over Ω, we have

    12ddtu2+ut2+Ω(f(u1f(u2)))utdx=Ωg(u1)(φΔφ)utdx+Ω(g(u1)g(u2))(φ2Δφ2)utdx. (3.42)

    Multiplying (3.39) by φ and integrating over Ω, one has

    12ddt(φ2+φ2)+φ2=Ωg(u1)utφdxΩ(g(u1)g(u2))u2tφdx. (3.43)

    Multiplying (3.39) by Δφ and integrating over Ω, we obtain

    12ddt(φ2+Δφ2)+Δφ2=Ωg(u1)utΔφdx+Ω(g(u1)g(u2))u2tΔφdx. (3.44)

    Finally, adding (3.42), (3.43) and (3.44), we get

    dE4dt+ut2+φ2+Δφ2+Ω(f(u1)f(u2))utdx=Ω(g(u1)g(u2))(φ2Δφ2)utdxΩ(g(u1)g(u2))(φΔφ)u2tdx, (3.45)

    where

    E4=u2+φ2+2φ2+Δφ2. (3.46)

    Now, owing to (3.5), and applying Hölder's inequality for k=2, when n=3, we can write

    Ω(f(u1)f(u2))utdxcΩ(|u2|k+1)|u||ut|dxc(u22k+1)u2+ut2, (3.47)

    we also get, thanks to (3.4), and applying Hölder's inequality,

    Ω(g(u1)g(u2))(φ2Δφ2)utdxcΩ|u||φ2Δφ2||ut|dxcu2(φ22+Δφ22)+ut2 (3.48)

    and

    Ω(g(u1)g(u2))(φΔφ)u2tdxcΩ|u||ut||φΔφ|dxcu2t2(φ2+Δφ2)+u2. (3.49)

    From (3.45)-(3.49), we deduce a differential inequality of the type

    dE4dt+cut2c(u22k+u2t2+φ22+Δφ22+1)E4. (3.50)

    In particular,

    dE4dtcE4 (3.51)

    and then applying the Gronwall's lemma to (3.51), we end the proof.


    4. Dissipativity properties of the system

    This section is devoted to the existence of bounded absorbing sets for the semigroup S(t),t0. To this end, we consider a more restrictive assumption on G, namely,

    ϵ>0,|G(u)|2ϵF(s)+cϵ,sR. (4.1)

    We then have

    Theorem 4.1. Under the assumptions of the Theorem 3.3 and assuming that (4.1) holds true. Then, uL(R+;H10(Ω))Lk+2(Ω), φL(R+;H10(Ω)H2(Ω)).

    Proof. Going from (3.8) and (3.10), we get, summing (3.8) and δ(3.10), with δ>0, as small as we need,

    dE5dt+2(cu2+δΩF(u)dx+ut2+φ2+Δφ2)2cδΩ|G(u)|2dx+δ(φ2+Δφ2)+c2cδΩ|G(u)|2dx+δ(cφ2+Δφ2)+c, (4.2)

    where

    E5=δu2+u2+2ΩF(u)dx+φ2+2φ2+Δφ2 (4.3)

    satisfies

    E5c(u2H1(Ω)+uk+2k+2+φ2H2(Ω))c (4.4)

    and

    E5c(u2H1(Ω)+uk+2k+2+φ2H2(Ω))c. (4.5)

    From (4.2) and owing to (4.1), we obtain

    dE5dt+2(cu2+δΩF(u)dx+ut2+φ2+Δφ2)CϵΩF(u)dx+δ(cφ2+Δφ2)+Cϵ, (4.6)

    where Cϵ and Cϵ are positive constants which depend on ϵ. Now, choosing ϵ and δ such that:

    2δCϵand2>cδ, (4.7)

    we then deduce from (4.6),

    dE5dt+c(E5+ut2)c, (4.8)

    we complete the proof applying the Gronwall's lemma.

    Remark 4.2. It follows from theorems 3.1, 3.2 and 4.1 that we can define the family solving operators:

    S(t):ΦΦ,(u0,φ0)(u(t),φ(t)),t0, (4.9)

    where Φ=H10(Ω)×H10(Ω)H2(Ω), and (u,φ) is the unique solution to the problem (1.9)-(1.12). Moreover, this family of solving operators forms a continuous semigroup i.e., S(0)=Id and S(t+τ)=S(t)S(τ),t,τ0. And then, it follows from (4.8) that S(t) is dissipative in Φ, it means that it possesses a bounded absorbing set B0Φ i.e., BΦ(bounded),t0=t0(B) suchthat tt0 implies S(t)BB0. (see, e.g., [32], [34] for details).


    5. Spatial behavior of solutions

    The aim of this section is to study the spatial behavior of solutions in a semi-infinite cylinder, assuming that such solutions exist. This study is motivated by the possibility of extending results obtained above to the case of unbounded domains like semi-infinite cylinders. To do so, we will study the behavior of solutions in a semi-infinite cylinder denoted R=(0,+)×D, where D is a smooth bounded domain of Rn1, n being the space dimension. We then consider the problem defined by the system (1.9)-(1.10) in the semi-infinite R, with n=3. Furthermore, we endow to this system following boundary conditions:

    u=φ=0on(0,+)×D×(0,T) (5.1)

    and

    u(0,x2,x3;t)=h(x2,x3;t),φ(0,x2,x3;t)=l(x2,x3;t)on{0}×D×(0,T), (5.2)

    where T>0 is a given final time.

    We also consider following initial data

    u|t=0=φ|t=0=0onR. (5.3)

    Let us suppose that such solutions exist. We consider the function

    Fw(z,t)=t0D(z)ews(usu,1+φ(φ,1+φ,1s)+φsφ,1)dads, (5.4)

    where D(z)={xR:x1=z}, u,1=ux1, us=us and w is a positive constant. Using the divergence theorem and owing to (5.1), we have

    Fw(z+h,t)Fw(z,t)=ewt2R(z,z+h)(|u|2+2F(u)+|φ|2+2|φ|2+|Δφ|2)dx+t0R(z,z+h)ews(|us|2+|φ|2+|Δφ|2)dxds+w2t0R(z,z+h)ews(|u|2+2F(u)+|φ|2+2|φ|2+|Δφ|2)dxds, (5.5)

    where R(z,z+h)={xR:z<x1<z+h}.

    Hence,

    Fwt(z,t)=ewt2D(z)(|u|2+2F(u)+|φ|2+2|φ|2+|Δφ|2)da+t0D(z)ews(|us|2+|φ|2+|Δφ|2)dads+w2t0D(z)ews(|u|2+2F(u)+|φ|2+2|φ|2+|Δφ|2)dads. (5.6)

    We consider a second function, namely,

    Gw(z,t)=t0D(z)ews(usu,1+φ(θ,1+φ,1s))dads, (5.7)

    where θ=t0φ(s)ds.

    Similarly, we have

    Gw(z+h,t)Gw(z,t)=ewt2R(z,z+h)(|u|2+|θ|2)dx+t0R(z,z+h)ews(|u|2+f(u)u+uΔφ+|φ|2+|φ|2)dxds+w2t0R(z,z+h)ews(|u|2+|θ|2)dxds+t0R(z,z+h)ews(G(u)g(u)u)φdxds (5.8)

    and then

    Gwt(z,t)=ewt2D(z)(|u|2+|θ|2)da+t0D(z)ews(|u|2+f(u)u+uΔφ+|φ|2+|φ|2)dads+w2t0D(z)ews(|u|2+|θ|2)dads+t0D(z)ews(G(u)g(u)u)φdads. (5.9)

    We choose τ large enough such as

    2F(u)+τu2C1u2,C1>0. (5.10)

    Now, we focus on the nonliear part i.e.,

    w(F(u)+τ2|u|2)+τf(u)u+τ(G(u)g(u)u)φ+w2|φ|2. (5.11)

    We assume that G(s)g(s)sc(|s|k+2+s2).

    For τ large enough, we have F(u)+τ2|u|2C2(|u|k+2+|u|2),C2>0. Thus, for wτ, we deduce that

    w(F(u)+τ2|u|2)+τf(u)u+τ(G(u)g(u)u)φ+w2|φ|2C3(|u|2+|φ|2+|Δφ|2). (5.12)

    Taking into account previous choices, it clearly appears that the following function

    Hw=Fw+τGw (5.13)

    satisfies

    Hwt(z,t)C4t0D(z)ews(|u|2+|u|2+|us|2+|φ|2+|φ|2+|Δφ|2+|θ|2)dads. (5.14)

    We give now an estimate of |Hw| in terms of Hwt. Applying Cauchy-Schwarz's inequality, one has

    |Fw|(t0D(z)ewsu2sdads)1/2(ewsu2,1)1/2+(t0D(z)ewsφ2dads)1/2(ewsφ2,1)1/2+(t0D(z)ewsφ2dads)1/2(ewsφ2,1s)1/2+(t0D(z)ewsφ2sdads)1/2(ewsφ2,1)1/2C5t0D(z)ews(|u|2+|us|2+|φ|2+|φ|2+|φs|2+|φs|2)dads,C5>0. (5.15)

    Similarly,

    |Gw|(t0D(z)ewsu2dads)1/2(t0D(z)ewsu2,1dads)1/2+(t0D(z)ewsφ2dads)1/2(t0D(z)ewsθ2,1dads)1/2+(t0D(z)ewsφ2sdads)1/2(t0D(z)ewsφ2,1dads)1/2C6t0D(z)ews(|u|2+|u|2+|φ|2+|φ|2+|θ|2)dads,C6>0. (5.16)

    We then deduce the existence of a positive constant C7=C5+τC6C4 such that

    |Hw|C7Hwz. (5.17)

    Remark 5.1. The inequality (5.17) is well known in the study of spatial estimates and leads to the Phragmén-Lindelöf alternative (see, e.g., [9], [39]).

    In particular, if there exist z00 such that Fw(z0,t)>0, then the solution satisfies

    Hw(z,t)Hw(z0,t)eC17(zz0),zz0. (5.18)

    The estimate (5.18) gives information in terms of measure defined in the cylinder. Actually, from (5.18), we deduce that

    ewt2R(0,z)(|u|2+2F(u)+|φ|2+2|φ|2+|Δφ|2)dx+τewt2R(0,z)(|u|2+|θ|2)dx+t0R(0,z)ews(|us|2+|φ|2+|Δφ|2)dxds+τt0R(0,z)ews(|u|2+f(u)u+g(u)uΔφ+|φ|2+2|φ|2)dxds+w2t0R(0,z)ews(|u|2+2F(u)+|φ|2+2|φ|2+|Δφ|2)dxds+τw2t0R(0,z)ews(|u|2+|θ|2)dx+τt0R(0,z)ews(G(u)g(u)u)φdxds (5.19)

    tends to infinity exponentially fast. On the other hand, if Hw(z,t)0, for every z0, we deduce that the solution decreases and we get an inequality of the type

    Hw(z,t)Hw(0,t)eC17z,z0, (5.20)

    where

    Ew(z,t)=ewt2R(z)(|u|2+2F(u)+|φ|2+2|φ|2+|Δφ|2)dx+τewt2R(z)(|u|2+|θ|2)dx+t0R(z)ews(|us|2+|φ|2+|Δφ|2)dxds+τt0R(z)ews(|u|2+f(u)u+g(u)uΔφ+|φ|2+2|φ|2)dxds+w2t0R(z)ews(|u|2+2F(u)+|φ|2+2|φ|2+|Δφ|2)dxds+τw2t0R(z)ews(|u|2+|θ|2)dx+τt0R(z)ews(G(u)g(u)u)φdxds (5.21)

    and R(z)={xR:x1>z}.

    Finally, setting

    Ew(z,t)=12R(z)(|u|2+2F(u)+|φ|2+2|φ|2+|Δφ|2)dx+τ12R(z)(|u|2+|θ|2)dx+t0R(z)(|us|2+|φ|2+|Δφ|2)dxds+τt0R(z)(|u|2+f(u)u+g(u)uΔφ+|φ|2+2|φ|2)dxds+w2t0R(z)(|u|2+2F(u)+|φ|2+2|φ|2+|Δφ|2)dxds+τw2t0R(z)(|u|2+|θ|2)dx+τt0R(z)(G(u)g(u)u)φdxds. (5.22)

    We have the following result

    Theorem 5.2. Let (u,φ) be a solution to the problem given by (1.9)-(1.10), boundary conditions (5.1)-(5.2) and initial data (5.3). Then, either this solution satisfies (5.18), or it satisfies

    Ew(z,t)Ew(0,t)ewtC17z,z0, (5.23)

    where the energy Ew is given by (5.22).


    Acknowledgments

    The author would like to thank Alain Miranville for his advices and for his careful reading of this paper.


    Conflict of interest

    The author declares no conflicts of interest in this paper.




    [1] P. Moeskops, M. A.Viergever, A. M. Mendrik, L. S. De Vries, M. J. Benders, I. Isgum, Automatic segmentation of MR brain images with a Convolutional Neural Network, IEEE Trans. Med. Imaging, 35 (2016), 1252–1261. doi: 10.1109/TMI.2016.2548501. doi: 10.1109/TMI.2016.2548501
    [2] A. Makropoulos, S. J. Counsell, D. Rueckert, A review on automatic fetal and neonatal brain MRI segmentation, NeuroImage, 170 (2017), 231–248. doi: 10.1016/j.neuroimage.2017.06.074 doi: 10.1016/j.neuroimage.2017.06.074
    [3] Y. Yang, W. Jia, Y. Yang, Multi-atlas segmentation and correction model with level set formulation for 3D brain MR images, Pattern Recogn., 90 (2019), 450–463. doi: 10.1016/j.patcog.2019.01.031 doi: 10.1016/j.patcog.2019.01.031
    [4] J. Song, Z. Zhang, Magnetic resonance imaging segmentation via weighted level set model based on local kernel metric and spatial constraint, Entropy, 23 (2021), 1196. doi: 10.3390/e23091196 doi: 10.3390/e23091196
    [5] L. Sun, L. Zhang, D. Zhang, Multi-Atlas based methods in brain MR image segmentation, Chin. Med. Sci. J., 34 (2019), 110–119. doi: 10.24920/003576 doi: 10.24920/003576
    [6] J. Song, Z. Zhang, A modified robust FCM model with spatial constraints for brain MR image segmentation, Information, 10 (2019), 74. doi: 10.3390/info10020074 doi: 10.3390/info10020074
    [7] Z. Zhang, J. Song, An adaptive fuzzy level set Model with local spatial information for medical image segmentation and bias correction, IEEE Access, 7 (2019), 27322–27338. doi: 10.1109/ACCESS.2019.2900089 doi: 10.1109/ACCESS.2019.2900089
    [8] S. Roy, P. Maji, Medical image segmentation by partitioning spatially constrained fuzzy approximation spaces, IEEE Trans. Fuzzy Syst., 28 (2020), 965–977. doi: 10.1109/TFUZZ.2020.2965896 doi: 10.1109/TFUZZ.2020.2965896
    [9] D. L. Pham, Spatial models for fuzzy clustering, Comput. Vis. Image Und., 84 (2001), 285–297, doi: 10.1006/cviu.2001.0951. doi: 10.1006/cviu.2001.0951
    [10] W. Cui, Y. Wang, Y. Fan, Y. Feng, T. Lei, Localized FCM clustering with spatial information for medical image segmentation and bias field estimation, J. BiomMed. Imaging, 2013 (2013), 930301. doi: 10.1155/2013/930301 doi: 10.1155/2013/930301
    [11] A. Jaiswal, M. A. Williams, A. Bhalerao, M. K. Tiwari, J. M. Warnett, Markov random field segmentation for industrial computed tomography with metal artefacts, J. X-Ray Sci. Technol., 26 (2018), 573–591. doi: 10.3233/XST-17322 doi: 10.3233/XST-17322
    [12] H. Liu, G. Dai, F. Pu, Hip-Joint CT image segmentation based on hidden Markov model with gauss regression constraints, J. Med. Syst., 43 (2019), 309. doi: 10.1007/s10916-019-1439-6 doi: 10.1007/s10916-019-1439-6
    [13] X. Xu, Y. Guan, H. Gong, Z. Feng, Q. Luo, Automated brain region segmentation for single cell resolution histological images based on markov random Field, Neuroinformatics, 18 (2020), 181–197. doi: 10.1007/s12021-019-09432-z doi: 10.1007/s12021-019-09432-z
    [14] A. Chen, Y. Zhang, Image segmentation based on a robust fuzzy c means algorithm, J. Med. Imag. Health In., 9 (2019), 1464–1468. doi: 10.1166/jmihi.2019.2745 doi: 10.1166/jmihi.2019.2745
    [15] J. Besag, Spatial interaction and the statistical analysis of lattice systems, J. Royal Stat. Soc., 36 (1974), 192–236. https://www.jstor.org/stable/2984812
    [16] C. Sutton, A. Mccallum, An introduction to conditional random fields, Found. Trends, 4 (2012), 267–373. doi: 10.1561/2200000013 doi: 10.1561/2200000013
    [17] H. Rue, H. Tjelmeland, Fitting Gaussian Markov random fields to Gaussian fields, Scand. J. Stat., 29 (2002), 31–49. doi: 10.1111/1467-9469.00058 doi: 10.1111/1467-9469.00058
    [18] F. Wu, The Potts model, Rev. Mod. Phys., 54 (1982), 235–268. doi: 10.1103/RevModPhys.54.235 doi: 10.1103/RevModPhys.54.235
    [19] S. Ribes, D. Didierlaurent, N. Decoster, E. Gonneau, L. Risser, V. Feillel, Automatic segmentation of breast MR images through a Markov random field statistical model, IEEE Trans. Med. Imaging, 33 (2014), 1986–1996. doi: 10.1109/TMI.2014.2329019 doi: 10.1109/TMI.2014.2329019
    [20] R. C. Dubes, A. K. Jain, Clustering techniques: The user's dilemma, Pattern Recogn., 8 (1976), 247–260. doi: 10.1016/0031-3203(76)90045-5 doi: 10.1016/0031-3203(76)90045-5
    [21] J. C. Bezdek, Pattern recognition with fuzzy objective function Algorithms, Adv. Appl. Pattern Recogn., 22 (1981), 203–239. doi: 10.1007/978-1-4757-0450-1 doi: 10.1007/978-1-4757-0450-1
    [22] B. N. Subudhi, F. Bovolo, A. Ghosh, L. Bruzzone, Spatio-contextual fuzzy clustering with Markov random field model for change detection in remotely sensed images, Opt. Laser Technol., 57 (2014), 284–292. doi: 10.1016/j.optlastec.2013.10.003 doi: 10.1016/j.optlastec.2013.10.003
    [23] T. K. Palani, B. Parvathavarthini, K. Chitra, Segmentation of brain regions by integrating meta heuristic multilevel threshold with Markov random field, Current Med. Imaging Rev., 12 (2016), 4–12. doi: 10.2174/1573394711666150827203434 doi: 10.2174/1573394711666150827203434
    [24] O. Salih, S. Viriri, Skin lesion segmentation using Sto-chastic region-merging and pixel-based Markov random field, Symmetry, 12 (2020), 1224. doi: 10.3390/sym12081224 doi: 10.3390/sym12081224
    [25] M. Hao, M. Zhou, J. Jin, W. Shi, An advanced superpixel-based Markov random field model for unsupervised change detection, IEEE Geosci. Remote S., 17 (2020), 1401–1405. doi: 10.1109/LGRS.2019.2948660 doi: 10.1109/LGRS.2019.2948660
    [26] X. Li, J. Chen, L. Zhao, S. Guo, X. Zhao, Adaptive distance-weighted Voronoi tessellation for remote sensing image segmentation, Remote Sens., 12 (2020), 4115. doi: 10.3390/rs12244115 doi: 10.3390/rs12244115
    [27] J. Song, Z. Zhang, Brain tissue segmentation and bias field correction of MR image based on spatially coherent FCM with nonlocal constraints, Comput. Math. Method Med., 2019 (2019), 4762490. doi: 10.1155/2019/4762490 doi: 10.1155/2019/4762490
    [28] Y. Chen, J. Li, H. Zhang, Y. Zheng, B. Jeon, Q. J. Wu, Non-local-based spatially constrained hierarchical fuzzy C-means method for brain magnetic resonance imaging segmentation, Iet Image Process., 10 (2016), 865–876. doi: 10.1049/iet-ipr.2016.0271 doi: 10.1049/iet-ipr.2016.0271
    [29] BrainWeb: Simulated Brain Database, Available online: http://brainweb.bic.mni.mcgill.ca/brainweb/. (accessed on 28 August 2021).
    [30] IBSR: The Internet Brain Segmentation Repository, Available online: http://www.nitrc.org/projects/ibsr. (accessed on 29 August 2021)
    [31] C. Li, J. Gore, C. Davatzikos, Multiplicative intrinsic component optimization (MICO) for MRI bias field estimation and tissue segmentation, Magn. Reson. Imaging, 32 (2014), 913–923. doi: 10.1016/j.mri.2014.03.010 doi: 10.1016/j.mri.2014.03.010
    [32] A. Elazab, C. Wang, F. Jia, Q. Hu, Segmentation of brain tissues from magnetic resonance images using adaptively regularized kernel-based fuzzy c-means clustering, Comput. Math. Method. M., 2015 (2015), 485495. doi: 10.1155/2015/485495 doi: 10.1155/2015/485495
    [33] S. Zhan, X. Yang, MR image bias field harmonic approximation with histogram statistical analysis, Pattern Recogn. Lett., 83 (2016), 91–98. doi: 10.1016/j.patrec.2016.02.009 doi: 10.1016/j.patrec.2016.02.009
  • Reader Comments
  • © 2022 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(3021) PDF downloads(99) Cited by(10)

Article outline

Figures and Tables

Figures(9)  /  Tables(2)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog