The hippocampus is a small, yet intricate seahorse-shaped tiny structure located deep within the brain's medial temporal lobe. It is a crucial component of the limbic system, which is responsible for regulating emotions, memory, and spatial navigation. This research focuses on automatic hippocampus segmentation from Magnetic Resonance (MR) images of a human head with high accuracy and fewer false positive and false negative rates. This segmentation technique is significantly faster than the manual segmentation methods used in clinics. Unlike the existing approaches such as UNet and Convolutional Neural Networks (CNN), the proposed algorithm generates an image that is similar to a real image by learning the distribution much more quickly by the semi-supervised iterative learning algorithm of the Deep Neuro-Fuzzy (DNF) technique. To assess its effectiveness, the proposed segmentation technique was evaluated on a large dataset of 18,900 images from Kaggle, and the results were compared with those of existing methods. Based on the analysis of results reported in the experimental section, the proposed scheme in the Semi-Supervised Deep Neuro-Fuzzy Iterative Learning System (SS-DNFIL) achieved a 0.97 Dice coefficient, a 0.93 Jaccard coefficient, a 0.95 sensitivity (true positive rate), a 0.97 specificity (true negative rate), a false positive value of 0.09 and a 0.08 false negative value when compared to existing approaches. Thus, the proposed segmentation techniques outperform the existing techniques and produce the desired result so that an accurate diagnosis is made at the earliest stage to save human lives and to increase their life span.
Citation: M Nisha, T Kannan, K Sivasankari. A semi-supervised deep neuro-fuzzy iterative learning system for automatic segmentation of hippocampus brain MRI[J]. Mathematical Biosciences and Engineering, 2024, 21(12): 7830-7853. doi: 10.3934/mbe.2024344
[1] | Shishi Wang, Yuting Ding, Hongfan Lu, Silin Gong . Stability and bifurcation analysis of SIQR for the COVID-19 epidemic model with time delay. Mathematical Biosciences and Engineering, 2021, 18(5): 5505-5524. doi: 10.3934/mbe.2021278 |
[2] | Hongfan Lu, Yuting Ding, Silin Gong, Shishi Wang . Mathematical modeling and dynamic analysis of SIQR model with delay for pandemic COVID-19. Mathematical Biosciences and Engineering, 2021, 18(4): 3197-3214. doi: 10.3934/mbe.2021159 |
[3] | Sarita Bugalia, Jai Prakash Tripathi, Hao Wang . Mathematical modeling of intervention and low medical resource availability with delays: Applications to COVID-19 outbreaks in Spain and Italy. Mathematical Biosciences and Engineering, 2021, 18(5): 5865-5920. doi: 10.3934/mbe.2021295 |
[4] | Honghua Bin, Daifeng Duan, Junjie Wei . Bifurcation analysis of a reaction-diffusion-advection predator-prey system with delay. Mathematical Biosciences and Engineering, 2023, 20(7): 12194-12210. doi: 10.3934/mbe.2023543 |
[5] | Yuting Ding, Gaoyang Liu, Yong An . Stability and bifurcation analysis of a tumor-immune system with two delays and diffusion. Mathematical Biosciences and Engineering, 2022, 19(2): 1154-1173. doi: 10.3934/mbe.2022053 |
[6] | Fang Liu, Yanfei Du . Spatiotemporal dynamics of a diffusive predator-prey model with delay and Allee effect in predator. Mathematical Biosciences and Engineering, 2023, 20(11): 19372-19400. doi: 10.3934/mbe.2023857 |
[7] | Xin-You Meng, Yu-Qian Wu . Bifurcation analysis in a singular Beddington-DeAngelis predator-prey model with two delays and nonlinear predator harvesting. Mathematical Biosciences and Engineering, 2019, 16(4): 2668-2696. doi: 10.3934/mbe.2019133 |
[8] | Jinhu Xu, Yicang Zhou . Bifurcation analysis of HIV-1 infection model with cell-to-cell transmission and immune response delay. Mathematical Biosciences and Engineering, 2016, 13(2): 343-367. doi: 10.3934/mbe.2015006 |
[9] | Xinyu Liu, Zimeng Lv, Yuting Ding . Mathematical modeling and stability analysis of the time-delayed SAIM model for COVID-19 vaccination and media coverage. Mathematical Biosciences and Engineering, 2022, 19(6): 6296-6316. doi: 10.3934/mbe.2022294 |
[10] | Zuolin Shen, Junjie Wei . Hopf bifurcation analysis in a diffusive predator-prey system with delay and surplus killing effect. Mathematical Biosciences and Engineering, 2018, 15(3): 693-715. doi: 10.3934/mbe.2018031 |
The hippocampus is a small, yet intricate seahorse-shaped tiny structure located deep within the brain's medial temporal lobe. It is a crucial component of the limbic system, which is responsible for regulating emotions, memory, and spatial navigation. This research focuses on automatic hippocampus segmentation from Magnetic Resonance (MR) images of a human head with high accuracy and fewer false positive and false negative rates. This segmentation technique is significantly faster than the manual segmentation methods used in clinics. Unlike the existing approaches such as UNet and Convolutional Neural Networks (CNN), the proposed algorithm generates an image that is similar to a real image by learning the distribution much more quickly by the semi-supervised iterative learning algorithm of the Deep Neuro-Fuzzy (DNF) technique. To assess its effectiveness, the proposed segmentation technique was evaluated on a large dataset of 18,900 images from Kaggle, and the results were compared with those of existing methods. Based on the analysis of results reported in the experimental section, the proposed scheme in the Semi-Supervised Deep Neuro-Fuzzy Iterative Learning System (SS-DNFIL) achieved a 0.97 Dice coefficient, a 0.93 Jaccard coefficient, a 0.95 sensitivity (true positive rate), a 0.97 specificity (true negative rate), a false positive value of 0.09 and a 0.08 false negative value when compared to existing approaches. Thus, the proposed segmentation techniques outperform the existing techniques and produce the desired result so that an accurate diagnosis is made at the earliest stage to save human lives and to increase their life span.
Consider with the following fourth-order elliptic Navier boundary problem
{Δ2u+cΔu=λa(x)|u|s−2u+f(x,u)inΩ,u=Δu=0on∂Ω, | (1.1) |
where Δ2:=Δ(Δ) denotes the biharmonic operator, Ω⊂RN(N≥4) is a smooth bounded domain, c<λ1 (λ1 is the first eigenvalue of −Δ in H10(Ω)) is a constant, 1<s<2,λ≥0 is a parameter, a∈L∞(Ω),a(x)≥0,a(x)≢0, and f∈C(ˉΩ×R,R). It is well known that some of these fourth order elliptic problems appear in different areas of applied mathematics and physics. In the pioneer paper Lazer and Mckenna [13], they modeled nonlinear oscillations for suspensions bridges. It is worth mentioning that problem (1.1) can describe static deflection of an elastic plate in a fluid, see [21,22]. The static form change of beam or the motion of rigid body can be described by the same problem. Equations of this type have received more and more attentions in recent years. For the case λ=0, we refer the reader to [3,7,11,14,16,17,20,23,27,29,34,35,36,37] and the reference therein. In these papers, existence and multiplicity of solutions have been concerned under some assumptions on the nonlinearity f. Most of them considered the case f(x,u)=b[(u+1)+−1] or f having asymptotically linear growth at infinity or f satisfying the famous Ambrosetti-Rabinowitz condition at infinity. Particularly, in the case λ≠0, that is, the combined nonlinearities for the fourth-order elliptic equations, Wei [33] obtained some existence and multiplicity by using the variational method. However, the author only considered the case that the nonlinearity f is asymptotically linear. When λ=1, Pu et al. [26] did some similar work. There are some latest works for problem (1.1), for example [10,18] and the reference therein. In this paper, we study problem (1.1) from two aspects. One is that we will obtain two multiplicity results when the nonlinearity f is superlinear at infinity and has the standard subcritical polynomial growth but not satisfy the Ambrosetti-Rabinowitz condition, the other is we can establish some existence results of multiple solutions when the nonlinearity f has the exponential growth but still not satisfy the Ambrosetti-Rabinowitz condition. In the first case, the standard methods for the verification of the compactness condition will fail, we will overcome it by using the functional analysis methods, i.e., Hahn-Banach Theorem combined the Resonance Theorem. In the last case, although the original version of the mountain pass theorem of Ambrosetti-Rabinowitz [1] is not directly applied for our purpose. Therefore, we will use a suitable version of mountain pass theorem and some new techniques to finish our goal.
When N>4, there have been substantial lots of works (such as [3,7,11,16,17,26,34,35,36,37]) to study the existence of nontrivial solutions or the existence of sign-changing for problem (1.1). Furthermore, almost all of the works involve the nonlinear term f(x,u) of a standard subcritical polynomial growth, say:
(SCP): There exist positive constants c1 and q∈(1,p∗−1) such that
|f(x,t)|≤c1(1+|t|q)for allt∈Randx∈Ω, |
where p∗=2NN−4 expresses the critical Sobolev exponent. In this case, people can deal with problem (1.1) variationally in the Sobolev space H2(Ω)∩H10(Ω) owing to the some critical point theory, such as, the method of invariant sets of descent flow, mountain pass theorem and symmetric mountain pass theorem. It is worth while to note that since Ambrosetti and Rabinowitz presented the mountain pass theorem in their pioneer paper [1], critical point theory has become one of the main tools on looking for solutions to elliptic equation with variational structure. One of the important condition used in many works is the so-called Ambrosetti-Rabinowitz condition:
(AR) There exist θ>2 and R>0 such that
0<θF(x,t)≤f(x,t)t,forx∈Ωand|t|≥R, |
where F(x,t)=∫t0f(x,s)ds. A simple computation explains that there exist c2,c3>0 such that F(x,t)≥c2|t|θ−c3 for all (x,t)∈ˉΩ×R and f is superlinear at infinity, i.e., limt→∞f(x,t)t=+∞ uniformly in x∈Ω. Thus problem (1.1) is called strict superquadratic if the nonlinearity f satisfies the (AR) condition. Notice that (AR) condition plays an important role in ensuring the boundedness of Palais-Smale sequences. However, there are many nonlinearities which are superlinear at infinity but do not satisfy above (AR) condition such as f(x,t)=tln(1+|t|2)+|sint|t.
In the recent years many authors tried to study problem (1.1) with λ=0 and the standard Laplacian problem where (AR) is not assumed. Instead, they regard the general superquadratic condition:
(WSQC) The following limit holds
lim|t|→+∞F(x,t)t2=+∞,uniformly forx∈Ω |
with additional assumptions (see [3,5,7,11,12,15,17,19,24,26,31,37] and the references therein). In the most of them, there are some kind of monotonicity restrictions on the functions F(x,t) or f(x,t)t, or some convex property for the function tf(x,t)−2F(x,t).
In the case N=4 and c=0, motivated by the Adams inequality, there are a few works devoted to study the existence of nontrivial solutions for problem (1.1) when the nonlinearity f has the exponential growth, for example [15] and the references therein.
Now, we begin to state our main results: Let μ1 be the first eigenvalue of (Δ2−cΔ,H2(Ω)∩H10(Ω)) and suppose that f(x,t) satisfies:
(H1) f(x,t)t≥0,∀(x,t)∈Ω×R;
(H2) limt→0f(x,t)t=f0 uniformly for a.e. x∈Ω, where f0∈[0,+∞);
(H3) limt→∞F(x,t)t2=+∞ uniformly for a.e. x∈Ω, where F(x,t)=∫t0f(x,s)ds.
In the case of N>4, our results are stated as follows:
Theorem 1.1. Assume that f has the standard subcritical polynomial growth on Ω (condition (SCP)) and satisfies (H1)–(H3). If f0<μ1 and a(x)≥a0 (a0 is a positive constant ), then there exists Λ∗>0 such that for λ∈(0,Λ∗), problem (1.1) has five solutions, two positive solutions, two negative solutions and one nontrivial solution.
Theorem 1.2. Assume that f has the standard subcritical polynomial growth on Ω (condition (SCP)) and satisfies (H2) and (H3). If f(x,t) is odd in t.
a) For every λ∈R, problem (1.1) has a sequence of solutions {uk} such that Iλ(uk)→∞,k→∞, definition of the functional Iλ will be seen in Section 2.
b) If f0<μ1, for every λ>0, problem (1.1) has a sequence of solutions {uk} such that Iλ(uk)<0 and Iλ(uk)→0,k→∞.
Remark. Since our the nonlinear term f(x,u) satisfies more weak condition (H3) compared with the classical condition (AR), our Theorem 1.2 completely contains Theorem 3.20 in [32].
In case of N=4, we have p∗=+∞. So it's necessary to introduce the definition of the subcritical exponential growth and critical exponential growth in this case. By the improved Adams inequality (see [28] and Lemma 2.2 in Section 2) for the fourth-order derivative, namely,
supu∈H2(Ω)∩H10(Ω),‖ |
So, we now define the subcritical exponential growth and critical exponential growth in this case as follows:
(SCE): f satisfies subcritical exponential growth on \Omega , i.e., \lim\limits_{t\rightarrow \infty}\frac{|f(x, t)|}{\exp(\alpha t^2)} = 0 uniformly on x\in \Omega for all \alpha > 0 .
(CG): f satisfies critical exponential growth on \Omega , i.e., there exists \alpha_0 > 0 such that
\lim\limits_{t\rightarrow \infty}\frac{|f(x, t)|}{\exp\left(\alpha t^2\right)} = 0, \; \text{uniformly on} \; x\in \Omega, \; \forall \alpha > \alpha_0, |
and
\lim\limits_{t\rightarrow \infty}\frac{|f(x, t)|}{\exp\left(\alpha t^2\right)} = +\infty, \; \text{uniformly on} \; x\in \Omega, \; \forall \alpha < \alpha_0. |
When N = 4 and f satisfies the subcritical exponential growth (SCE), our work is still to consider problem (1.1) where the nonlinearity f satisfies the (WSQC)-condition at infinity. As far as we know, this case is rarely studied by other people for problem (1.1) except for [24]. Hence, our results are new and our methods are technique since we successfully proved the compactness condition by using the Resonance Theorem combined Adams inequality and the truncated technique. In fact, the new idea derives from our work [25]. Our results are as follows:
Theorem 1.3. Assume that f satisfies the subcritical exponential growth on \Omega (condition \mathrm{(SCE)} ) and satisfies (H_1) – (H_3) . If f_0 < \mu_1 and a(x)\geq a_0 (a_0 is a positive constant ) , then there exists \Lambda^* > 0 such that for \lambda\in (0, \Lambda^*), problem (1.1) has five solutions, two positive solutions, two negative solutions and one nontrivial solution.
Remark. Let F(x, t) = t^2e^{\sqrt{|t|}}, \forall (x, t) \in \Omega\times \mathbb{R}. Then it satisfies that our conditions (H_1) – (H_3) but not satisfy the condition \mathrm{(AR)} . It's worth noting that we do not impose any monotonicity condition on \frac{f(x, t)}{t} or some convex property on tf(x, t)-2F(x, t) . Hence, our Theorem 1.3 completely extends some results contained in [15,24] when \lambda = 0 in problem (1.1).
Theorem 1.4. Assume that f satisfies the subcritical exponential growth on \Omega (condition \mathrm{(SCE)} ) and satisfies (H_2) and (H_3) . If f_0 < \mu_1 and f(x, t) is odd in t .
a) For \lambda > 0 small enough, problem (1.1) has a sequence of solutions \{u_k\} such that I_\lambda(u_k)\rightarrow \infty, k\rightarrow \infty.
b) For every \lambda > 0 , problem (1.1) has a sequence of solutions \{u_k\} such that I_\lambda(u_k) < 0 and I_\lambda(u_k)\rightarrow 0, k\rightarrow \infty.
When N = 4 and f satisfies the critical exponential growth \mathrm{(CG)} , the study of problem (1.1) becomes more complicated than in the case of subcritical exponential growth. Similar to the case of the critical polynomial growth in \mathbb{R}^N\; (N\geq 3) for the standard Laplacian studied by Brezis and Nirenberg in their pioneering work [4], our Euler-Lagrange functional does not satisfy the Palais-Smale condition at all level anymore. For the class standard Laplacian problem, the authors [8] used the extremal function sequences related to Moser-Trudinger inequality to complete the verification of compactness of Euler-Lagrange functional at some suitable level. Here, we still adapt the method of choosing the testing functions to study problem (1.1) without (AR) condition. Our result is as follows:
Theorem 1.5. Assume that f has the critical exponential growth on \Omega (condition \mathrm{(CG)} ) and satisfies (H_1) – (H_3) . Furthermore, assume that
(H_4) \lim\limits_{t\rightarrow \infty}f(x, t) \exp(-\alpha_0t^2)t\geq \beta > \frac{64}{\alpha_0r_0^4} , uniformly in (x, t), where r_0 is the inner radius of \Omega , i.e., r_0: = radius of the largest open ball \subset \Omega. and
(H_5) f is in the class (H_0) , i.e., for any \{u_n\} in H^2(\Omega)\cap H_0^1(\Omega) , if u_n\rightharpoonup 0 in H^2(\Omega)\cap H_0^1(\Omega) and f(x, u_n)\rightarrow 0 in L^1(\Omega), then F(x, u_n)\rightarrow 0 in L^1(\Omega) (up to a subsequence).
If f_0 < \mu_1, then there exists \Lambda^* > 0 such that for \lambda\in (0, \Lambda^*), problem (1.1) has at least four nontrivial solutions.
Remark. For standard biharmonic problems with Dirichlet boundary condition, Lam and Lu [15] have recently established the existence of nontrivial nonnegative solutions when the nonlinearity f has the critical exponential growth of order \exp(\alpha u^2) but without satisfying the Ambrosetti- Rabinowitz condition. However, for problem (1.1) with Navier boundary condition involving critical exponential growth and the concave term, there are few works to study it. Hence our result is new and interesting.
The paper is organized as follows. In Section 2, we present some necessary preliminary knowledge and some important lemmas. In Section 3, we give the proofs for our main results. In Section 4, we give a conclusion.
We let \lambda_k (k = 1, 2, \cdot\cdot\cdot) denote the eigenvalue of -\Delta in H_0^1(\Omega) , then 0 < \mu_1 < \mu_2 < \cdot\cdot\cdot < \mu_k < \cdot\cdot\cdot be the eigenvalues of (\Delta^2-c\Delta, H^2(\Omega)\cap H_0^1(\Omega)) and \varphi_k(x) be the eigenfunction corresponding to \mu_k . Let X_k denote the eigenspace associated to \mu_k. In fact, \mu_k = \lambda_k(\lambda_k-c). Throughout this paper, we denote by \|\cdot\|_p the L^p(\Omega) norm, c < \lambda_1 in \Delta^2-c\Delta and the norm of u in H^2(\Omega)\cap H_0^1(\Omega) will be defined by the
\|u\|: = \left(\int_\Omega (|\Delta u|^2-c|\nabla u|^2)dx\right)^{\frac{1}{2}}. |
We also denote E = H^2(\Omega)\cap H_0^1(\Omega) .
Definition 2.1. Let ( \mathbb{E}, ||\cdot||_\mathbb{E}) be a real Banach space with its dual space (\mathbb{E}^*, ||\cdot||_{\mathbb{E}^*}) and I\in C^1(\mathbb{E}, \mathbb{R}) . For c^*\in \mathbb{ R}, we say that I satisfies the \mathrm{(PS)_{c^*}} condition if for any sequence \{x_n\}\subset \mathbb{E} with
I(x_n)\rightarrow c^*, I'(x_n)\rightarrow 0 \ \mbox{in}\ \mathbb{E}^*, |
there is a subsequence \{x_{n_k}\} such that \{x_{n_k}\} converges strongly in \mathbb{E} . Also, we say that I satisfy the \mathrm{(C)_{c^*}} condition if for any sequence \{x_n\}\subset \mathbb{E} with
I(x_n)\rightarrow c^*, \ ||I'(x_n)||_{\mathbb{E}^*}(1+||x_n||_\mathbb{E})\rightarrow 0, |
there exists subsequence \{ x_{n_k}\} such that \{ x_{n_k}\} converges strongly in \mathbb{E}.
Definition 2.2. We say that u\in E is the solution of problem (1.1) if the identity
\int_\Omega (\Delta u \Delta \varphi-c\nabla u \nabla \varphi)dx = \lambda\int_\Omega a(x)|u|^{s-2}u\varphi dx+\int_\Omega f(x, u)\varphi dx |
holds for any \varphi\in E.
It is obvious that the solutions of problem (1.1) are corresponding with the critical points of the following C^1 functional:
I_\lambda(u) = \frac{1}{2}\|u\|^2-\frac{\lambda}{s}\int_\Omega a(x)|u|^sdx-\int_\Omega F(x, u)dx, \quad u\in E. |
Let u^+ = \max\{u, 0\}, u^- = \min\{u, 0\}.
Now, we concern the following problem
\begin{equation} \left\{\begin{array}{ll} \Delta^2u+c\Delta u = \lambda a(x)|u^+|^{s-2}u^++f^+(x, u) \; &\text{in}\; \Omega, \\ u = \Delta u = 0 \; &\text{on}\; \partial \Omega, \end{array}\right. \end{equation} | (2.1) |
where
f^+(x, t) = \left\{\begin{array}{ll} f(x, t) \; &t\geq 0, \\ 0, \; &t < 0. \end{array}\right. |
Define the corresponding functional I_\lambda^+:E\rightarrow \mathbb{R} as follows:
I_\lambda^+(u) = \frac{1}{2}\|u\|^2-\frac{\lambda}{s}\int_\Omega a(x)|u^+|^sdx-\int_\Omega F^+(x, u)dx, |
where F^+(x, u) = \int_0^u f^+(x, s)ds. Obviously, the condition \mathrm{(SCP)} or \mathrm{(SCE)} ( \mathrm{(CG)} ) ensures that I_\lambda^+\in C^1(E, \mathbb{R}). Let u be a critical point of I_\lambda^+, which means that u is a weak solution of problem (2.1). Furthermore, since the weak maximum principle (see [34]), it implies that u\geq 0 in \Omega . Thus u is also a solution of problem (1.1) and I_\lambda^+(u) = I_\lambda(u) .
Similarly, we define
f^-(x, t) = \left\{\begin{array}{ll} f(x, t) \; &t\leq 0, \\ 0, \; &t > 0, \end{array}\right. |
and
I_\lambda^-(u) = \frac{1}{2}\|u\|^2-\frac{\lambda}{s}\int_\Omega a(x)|u^-|^sdx-\int_\Omega F^-(x, u)dx, |
where F^-(x, u) = \int_0^u f^-(x, s)ds. Similarly, we also have I_\lambda^-\in C^1(E, \mathbb{R}) and if v is a critical point of I_\lambda^- then it is a solution of problem (1.1) and I_\lambda^-(v) = I_\lambda(v) .
Prosition 2.1. ([6,30]). Let \mathbb{E} be a real Banach space and suppose that I \in C^1(\mathbb{E}, \mathbb{R}) satisfies the condition
\max\{I(0), I(u_1)\}\leq\alpha < \beta\leq\inf\limits_{||u|| = \rho}I(u), |
for some \alpha < \beta, \rho > 0 and u_1\in \mathbb{E} with ||u_1|| > \rho. Let c^*\geq\beta be characterized by
c^* = \inf\limits_{\gamma\in \Gamma}\max\limits_{0\leq t\leq1}I(\gamma(t)), |
where \Gamma = \{\gamma\in C([0, 1], \mathbb{E}), \gamma(0) = 0, \gamma(1) = u_1\} is the set of continuous paths joining 0 and u_1. Then, there exists a sequence \{u_n\}\subset \mathbb{E} such that
I(u_n)\rightarrow c^*\geq\beta\ \mathit{\mbox{and}}\ (1+||u_n|| )||I'(u_n)||_{\mathbb{E}^*}\rightarrow0\ \mathit{\mbox{as}}\ n\rightarrow \infty. |
Lemma 2.1. ([28]). Let \Omega\subset \mathbb{R}^4 be a bounded domain. Then there exists a constant C > 0 such that
\sup\limits_{u\in E, \|\Delta u\|_2\leq 1}\int_\Omega e^{32\pi^2u^2}dx < C|\Omega|, |
and this inequality is sharp.
Next, we introduce the following a revision of Adams inequality:
Lemma 2.2. Let \Omega\subset \mathbb{R}^4 be a bounded domain. Then there exists a constant C^* > 0 such that
\sup\limits_{u\in E, \| u\|\leq 1}\int_\Omega e^{32\pi^2 u^2}dx < C^*|\Omega|, |
and this inequality is also sharp.
Proof. We will give a summarize proof in two different cases. In the case of c\leq0 in the definition of \|.\| , if \|u\|\leq 1, we can deduce that \|\Delta u\|_2\leq 1 and by using Lemma 2.1 combined with the Proposition 6.1 in [28], the conclusion holds.
In the case of 0 < c < \lambda_1 in the definition of \|.\| , from Lemma 2.1, the proof and remark of Theorem 1 in [2] and the proof of Proposition 6.1 in [28], we still can establish this revised Adams inequality.
Lemma 2.3. Assume (H_{1}) and (H_{3}) hold. If f has the standard subcritical polynomial growth on \Omega (condition \mathrm{(SCP)} ), then I_\lambda^+ ( I_\lambda^- ) satisfies \mathrm{(C)}_{c^*} .
Proof. We only prove the case of I_\lambda^+ . The arguments for the case of I_\lambda^- are similar. Let \{u_n\}\subset E be a \mathrm{(C)}_{c^*} sequence such that
\begin{equation} I_\lambda^+(u_n) = \frac{1}{2}||u_n||^2-\frac{\lambda}{s}\int_\Omega a(x)|u_n^+|^sdx-\int_\Omega F^+(x, u_n)dx = c^*+\circ(1), \end{equation} | (2.2) |
\begin{equation} (1+||u_n||)||I_\lambda^{+'}(u_n)||_{E^*}\rightarrow 0\ \mbox{as}\ n\rightarrow \infty. \end{equation} | (2.3) |
Obviously, (2.3) implies that
\begin{equation} \langle I_\lambda^{+'}(u_n), \varphi\rangle = \langle u_n, \varphi \rangle-\lambda\int_\Omega a(x)|u_n^+|^{s-2}u_n^+\varphi dx-\int_\Omega f^+(x, u_n(x))\varphi dx = \circ(1). \end{equation} | (2.4) |
Step 1. We claim that \{u_n\} is bounded in E . In fact, assume that
\|u_n\|\rightarrow \infty, \ \ \text{as}\ n\rightarrow \infty. |
Define
v_n = \frac{u_n}{\|u_n\|}. |
Then, \|v_n\| = 1 , \forall n\in {\bf N} and then, it is possible to extract a subsequence (denoted also by \{v_n\} ) converges weakly to v in E , converges strongly in L^p(\Omega) (1\leq p < p^*) and converges v a.e. x\in\Omega .
Dividing both sides of (2.2) by \|u_n\|^2 , we get
\begin{equation} \int_{\Omega} \frac{F^+(x, u_n)}{\|u_n\|^2}dx\rightarrow \frac{1}{2}. \end{equation} | (2.5) |
Set
\Omega_+ = \{x\in \Omega: v(x) > 0\}. |
By (H_3) , we imply that
\begin{equation} \frac{F^+(x, u_n)}{u_n^2}v_n^2\rightarrow \infty, \; \; x\in \Omega_+. \end{equation} | (2.6) |
If |\Omega_+| is positive, since Fatou's lemma, we get
\lim\limits_{n\rightarrow \infty}\int_{\Omega} \frac{F^+(x, u_n)}{\|u_n\|^2}dx\geq\lim\limits_{n\rightarrow \infty}\int_{\Omega_+} \frac{F^+(x, u_n)}{u_n^2}v_n^2dx = +\infty, |
which contradicts with (2.5). Thus, we have v\leq 0 . In fact, we have v = 0 . Indeed, again using (2.3), we get
(1+\|u_n\|) |\langle I_\lambda^{+'}(u_n), v\rangle|\leq \circ(1)\|v\|. |
Thus, we have
\begin{eqnarray*} \int_\Omega (\Delta u_n \Delta v-c\nabla u_n \nabla v)dx &\leq& \int_\Omega (\Delta u_n \Delta v-c\nabla u \nabla v)dx-\lambda\int_\Omega a(x)|u_n^+|^{s-2}u_n^+vdx\\ &-&\int_\Omega f^+(x, u_n)vdx\leq \frac{\circ(1)\|v\|}{1+\|u_n\|}, \end{eqnarray*} |
by noticing that since v\leq 0, f^+(x, u_n)v\leq 0 \ \text{a.e.} \ x\in\Omega, thus -\int_\Omega f^+(x, u_n)vdx\geq 0. So we get
\int_\Omega (\Delta v_n \Delta v-c\nabla v_n \nabla v)dx\rightarrow 0. |
On the other hand, from v_n\rightharpoonup v in E , we have
\int_\Omega (\Delta v_n \Delta v-c\nabla v_n \nabla v)dx\rightarrow \|v\|^2 |
which implies v = 0 .
Dividing both sides of (2.4) by \|u_n\| , for any \varphi \in E , then there exists a positive constant M(\varphi) such that
\begin{equation} \left|\int_\Omega \frac{f^+(x, u_n)}{\|u_n\|}\varphi dx\right|\leq M(\varphi), \; \forall n\in {\bf N}. \end{equation} | (2.7) |
Set
{\bf f}_n(\varphi) = \int_\Omega \frac{f^+(x, u_n)}{\|u_n\|}\varphi dx, \; \varphi \in E. |
Thus, by (SCP), we know that \{{\bf f}_n\} is a family bounded linear functionals defined on E . Combing (2.7) with the famous Resonance Theorem, we get that \{|{\bf f}_n|\} is bounded, where |{\bf f}_n| denotes the norm of {\bf f}_n . It means that
\begin{equation} |{\bf f}_n|\leq C_*. \end{equation} | (2.8) |
Since E\subset L^{\frac{p^*}{p^*-q}}(\Omega) , using the Hahn-Banach Theorem, there exists a continuous functional \hat{{\bf f}}_n defined on L^{\frac{p^*}{p^*-q}}(\Omega) such that \hat{{\bf f}}_n is an extension of {\bf f}_n , and
\begin{equation} \hat{{\bf f}}_n(\varphi) = {\bf f}_n(\varphi), \; \varphi\in E, \end{equation} | (2.9) |
\begin{equation} \|\hat{{\bf f}}_n\|_{\frac{p^*}{q}} = |{\bf f}_n|, \end{equation} | (2.10) |
where \|\hat{{\bf f}}_n\|_{\frac{p^*}{q}} denotes the norm of \hat{{\bf f}}_n(\varphi) in L^{\frac{p^*}{q}}(\Omega) which is defined on L^{\frac{p^*}{p^*-q}}(\Omega) .
On the other hand, from the definition of the linear functional on L^{\frac{p^*}{p^*-q}}(\Omega), we know that there exists a function S_n(x)\in L^{\frac{p^*}{q}}(\Omega) such that
\begin{equation} \hat{{\bf f}}_n(\varphi) = \int_\Omega S_n(x)\varphi (x)dx, \; \varphi \in L^{\frac{p^*}{p^*-q}}(\Omega). \end{equation} | (2.11) |
So, from (2.9) and (2.11), we obtain
\int_\Omega S_n(x)\varphi (x)dx = \int_\Omega \frac{f^+(x, u_n)}{\|u_n\|}\varphi dx, \; \varphi \in E, |
which implies that
\int_\Omega \left(S_n(x)-\frac{f^+(x, u_n)}{\|u_n\|}\right)\varphi dx = 0, \; \varphi \in E. |
According to the basic lemma of variational, we can deduce that
S_n(x) = \frac{f^+(x, u_n)}{\|u_n\|}\; \; \text{a.e.}\; x\in \Omega. |
Thus, by (2.8) and (2.10), we have
\begin{equation} \|\hat{{\bf f}}_n\|_{\frac{p^*}{q}} = \|S_n\|_{\frac{p^*}{q}} = |{\bf f}_n| < C_*. \end{equation} | (2.12) |
Now, taking \varphi = v_n-v in (2.4), we get
\begin{equation} \langle A(v_n), v_n-v \rangle-\lambda\int_\Omega a(x)|u_n^+|^{s-2}u_n^+v_ndx-\int_\Omega \frac{f^+(x, u_n)}{\|u_n\|}v_ndx\rightarrow 0, \end{equation} | (2.13) |
where A: E\rightarrow E^* defined by
\langle A(u), \varphi\rangle = \int_\Omega \Delta u \Delta \varphi dx-c\int_\Omega \nabla u \nabla \varphi dx, \ u, \varphi \in E. |
By the H \mathrm{\ddot{o}} lder inequality and (2.12), we obtain
\int_\Omega \frac{f^+(x, u_n)}{\|u_n\|}v_ndx\rightarrow 0. |
Then from (2.13), we can conclude that
v_n\rightarrow v\; \; \text{in} \; E. |
This leads to a contradiction since \|v_n\| = 1 and v = 0. Thus, \{u_n\} is bounded in E .
Step 2. We show that \{u_n\} has a convergence subsequence. Without loss of generality, we can suppose that
\begin{align*} &u_n\rightharpoonup u\ \mbox{ in }\ E, \\ &u_n \rightarrow u \ \mbox{in}\ L^\gamma(\Omega), \ \forall1\leq \gamma < p^*, \\ &u_n(x)\rightarrow u(x)\ \mbox{a.e.}\ x\in \Omega. \end{align*} |
Now, it follows from f satisfies the condition (SCP) that there exist two positive constants c_4, c_5 > 0 such that
f^+(x, t)\leq c_4 +c_5|t|^{q}, \ \forall (x, t)\in \Omega\times \mathbb{R}, |
then
\begin{eqnarray*} &&\left|\int_\Omega f^+(x, u_n)(u_n-u)dx\right|\\ && \leq c_4 \int_\Omega |u_n-u|dx+ c_5\int_\Omega |u_n-u||u_n|^{q}dx \\&&\leq c_4\int_\Omega |u_n-u|dx+c_5\left( \int_\Omega \left(|u_n|^{q}\right)^{\frac{p^*}{q}}dx\right)^{\frac{q}{p^*}}\left(\int_\Omega |u_n-u|^{\frac{p^*}{p^*-q}}dx\right)^{\frac{p^*-q}{p^*}}. \end{eqnarray*} |
Similarly, since u_n\rightharpoonup u in E, \int_\Omega |u_n-u|dx\rightarrow 0 and \int_\Omega |u_n-u|^{\frac{p^*}{p^*-q}}dx\rightarrow 0.
Thus, from (2.4) and the formula above, we obtain
\langle A(u_n), u_n-u\rangle\rightarrow 0, \; \text{as}\; n\rightarrow \infty. |
So, we get \|u_n\|\rightarrow \|u\| . Thus we have u_n\rightarrow u in E which implies that I_\lambda^+ satisfies \mathrm{(C)}_{c^*} .
Lemma 2.4. Let \varphi _{1} > 0 be a \mu_{1} -eigenfunction with \| \varphi_{1} \| = 1 and assume that (H_{1}) – (H_{3}) and \mathrm{(SCP)} hold. If f_{0} < \mu _{1}, then:
\mathrm{(i)}\ For \lambda > 0 small enough, there exist \rho, \alpha > 0 such that I_\lambda^{\pm}(u)\geq \alpha for all u\in E with \| u \| = \rho,
\mathrm{(ii)}\ I_\lambda^{\pm}(t\varphi_1)\rightarrow -\infty as t\rightarrow +\infty .
Proof. Since \mathrm{(SCP)} and (H_1) – (H_3), for any \varepsilon > 0, there exist A = A(\varepsilon), M large enough and B = B(\varepsilon) such that for all (x, s)\in \Omega\times \mathbb{R},
\begin{equation} F^{\pm}(x, s)\leq \frac{1}{2}(f_0+\epsilon)s^2+A|s|^q, \end{equation} | (2.14) |
\begin{equation} F^{\pm}(x, s)\geq \frac{M}{2}s^2-B. \end{equation} | (2.15) |
Choose \varepsilon > 0 such that (f_0+\varepsilon) < \mu_1. By (2.14), the Poincaré inequality and the Sobolev embedding, we obtain
\begin{eqnarray*} I_\lambda^{\pm}(u)&\geq&\frac{1}{2}\|u\|^2 -\frac{\lambda \|a\|_\infty}{s}\int_\Omega |u|^s dx-\int_\Omega{F^{\pm}(x, u)}dx\\&\geq& \frac{1}{2}\|u\|^2-\frac{\lambda \|a\|_\infty}{s}\int_\Omega |u|^s dx-\frac{f_0+\varepsilon}{2}\|u\|_2^2-A\int_\Omega |u|^qdx\\ &\geq&\frac{1}{2}\left(1-\frac{f_0+\varepsilon}{\mu_1}\right)\|u\|^2-\lambda K\|u\|^s-C^{**}\|u\|^q \\ &\geq&\|u\|^2\left(\frac{1}{2}\left(1-\frac{f_0+\varepsilon}{\mu_1}\right)-\lambda K\|u\|^{s-2}-C^{**}\|u\|^{q-2}\right), \end{eqnarray*} |
where K, C^{**} are constant.
Write
h(t) = \lambda Kt^{s-2}+ C^{**}t^{q-2}. |
We can prove that there exists t^* such that
h(t^*) < \frac{1}{2}\left(1-\frac{f_0+\varepsilon}{\mu_1}\right). |
In fact, letting h'(t) = 0 , we get
t^* = \left(\frac{\lambda K(2-s)}{C^{**}(q-2)} \right)^{\frac{1}{q-s}}. |
According to the knowledge of mathematical analysis, h(t) has a minimum at t = t^* . Denote
\vartheta = \frac{K(2-s)}{C^{**}(q-2)}, \ \hat{\vartheta} = \frac{s-2}{q-s}, \ \bar{\vartheta} = \frac{q-2}{q-s}, \ \nu = \frac{1}{2}\left(1-\frac{f_0+\varepsilon}{\mu_1}\right). |
Taking t^* in h(t) , we get
h(t^*) < \nu, \; 0 < \lambda < \Lambda^*, |
where \Lambda^* = (\frac{\nu}{K\vartheta^{\hat{\vartheta}}+C^{**}\vartheta^{\bar{\vartheta}}})^{\frac{1}{\bar{\vartheta}}}. So, part (i) holds if we take \rho = t^* .
On the other hand, from (2.15), we get
I_\lambda^+(t\varphi_1)\leq\frac{1}{2}\left(1-\frac{M}{\mu_1}\right)t^2-t^s\frac{\lambda}{s}\int_\Omega a(x)|\varphi_1|^sdx +B|\Omega|\rightarrow -\infty\ \mbox{as}\ t\rightarrow +\infty. |
Similarly, we have
I_\lambda^-(t(-\varphi_1))\rightarrow -\infty, \ \text{as}\ t\rightarrow +\infty. |
Thus part \mathrm{(ii)} holds.
Lemma 2.5. Let \varphi _{1} > 0 be a \mu_{1} -eigenfunction with \| \varphi_{1} \| = 1 and assume that (H_{1}) – (H_{3}) and \mathrm{(SCE)} (or \mathrm{(CG)} ) hold. If f_{0} < \mu _{1}, then:
\mathrm{(i)}\ For \lambda > 0 small enough, there exist \rho, \alpha > 0 such that I_\lambda^{\pm}(u)\geq \alpha for all u\in E with \| u \| = \rho,
\mathrm{(ii)}\ I_\lambda^{\pm}(t\varphi_1)\rightarrow -\infty as t\rightarrow +\infty .
Proof. From \mathrm{(SCE)} (or (CG)) and (H_1) - (H_3), for any \varepsilon > 0, there exist A_1 = A_1(\varepsilon), M_1 large enough, B_1 = B_1(\varepsilon), \kappa_1 > 0 and q_1 > 2 such that for all (x, s)\in \Omega\times \mathbb{R},
\begin{equation} F^{\pm}(x, s)\leq \frac{1}{2}(f_0+\epsilon)s^2+A_1\exp(\kappa_1 s^{2})|s|^{q_1}, \end{equation} | (2.16) |
\begin{equation} F^{\pm}(x, s)\geq \frac{M_1}{2}s^2-B_1. \end{equation} | (2.17) |
Choose \varepsilon > 0 such that (f_0+\varepsilon) < \mu_1. By (2.16), the Hölder inequality and the Adams inequality (see Lemma 2.2), we obtain
\begin{eqnarray*} I_\lambda^{\pm}(u)&\geq&\frac{1}{2}\|u\|^2-\frac{\lambda \|a\|_\infty}{s}\int_\Omega |u|^s dx -\int_\Omega{F^{\pm}(x, u)}dx\\&\geq& \frac{1}{2}\|u\|^2-\frac{\lambda \|a\|_\infty}{s}\int_\Omega |u|^s dx-\frac{f_0+\varepsilon}{2}\|u\|_2^2-A_1\int_\Omega \exp(\kappa_1 u^{2})|u|^{q_1}dx\\ &\geq&\frac{1}{2}\left(1-\frac{f_0+\varepsilon}{\mu_1}\right)\|u\|^2-\lambda K\|u\|^s -A_1\left(\int_\Omega \exp(\kappa_1 r_1\|u\|^{2}(\frac{|u|}{\|u\|})^{2}\right)dx)^{\frac{1}{r_1}}\left(\int_\Omega |u|^{r_1'q}dx\right)^{\frac{1}{r_1'}}\\ &\geq&\frac{1}{2}\left(1-\frac{f_0+\varepsilon}{\mu_1}\right)\|u\|^2-\lambda K\|u\|^s-\hat{C}^{**}\|u\|^{q_1}, \end{eqnarray*} |
where r_1 > 1 sufficiently close to 1 , \|u\|\leq \sigma and \kappa_1 r_1\sigma^{2} < 32\pi^2. Remained proof is completely similar to the proof of part (ⅰ) of Lemma 2.4, we omit it here. So, part (ⅰ) holds if we take \|u\| = \rho > 0 small enough.
On the other hand, from (2.17), we get
I_\lambda^+(t\varphi_1)\leq\frac{1}{2}\left(1-\frac{M_1}{\mu_1}\right)t^2-t^s\frac{\lambda}{s}\int_\Omega a(x)|\varphi_1|^sdx +B_1|\Omega|\rightarrow -\infty\ \mbox{as}\ t\rightarrow +\infty. |
Similarly, we have
I_\lambda^-(t(-\varphi_1))\rightarrow -\infty, \ \text{as}\ t\rightarrow +\infty. |
Thus part \mathrm{(ii)} holds.
Lemma 2.6. Assume (H_{1}) and (H_{3}) hold. If f has the subcritical exponential growth on \Omega (condition \mathrm{(SCE)} ), then I_\lambda^+ ( I_\lambda^- ) satisfies \mathrm{(C)}_{c^*} .
Proof. We only prove the case of I_\lambda^+ . The arguments for the case of I_\lambda^- are similar. Let \{u_n\}\subset E be a \mathrm{(C)}_{c^*} sequence such that the formulas (2.2)–(2.4) in Lemma 2.3 hold.
Now, according to the previous section of Step 1 of the proof of Lemma 2.3, we also obtain that the formula (2.7) holds. Set
{\bf f}_n(\varphi) = \int_{\Omega} \frac{f^+(x, u_n)}{\|u_n\|}\varphi dx, \; \varphi \in E. |
Then from for any u\in E , e^{\alpha u^2}\in L^1(\Omega) for all \alpha > 0 , we can draw a conclusion that \{{\bf f}_n\} is a family bounded linear functionals defined on E . Using (2.7) and the famous Resonance Theorem, we know that \{|{\bf f}_n|\} is bounded, where |{\bf f}_n| denotes the norm of {\bf f}_n . It means that the formula (2.8) (see the proof of Lemma 2.3) holds.
Since E\subset L^{q_0}(\Omega) for some q_0 > 1 , using the Hahn-Banach Theorem, there exists a continuous functional \hat{{\bf f}}_n defined on L^{q_0}(\Omega) such that \hat{{\bf f}}_n is an extension of {\bf f}_n , and
\begin{equation} \hat{{\bf f}}_n(\varphi) = {\bf f}_n(\varphi), \; \varphi\in E, \end{equation} | (2.18) |
\begin{equation} \|\hat{{\bf f}}_n\|_{q_0^*} = |{\bf f}_n|, \end{equation} | (2.19) |
where \|\hat{{\bf f}}_n\|_{q_0^*} is the norm of \hat{{\bf f}}_n(\varphi) in L^{q_0^*}(\Omega) which is defined on L^{q_0}(\Omega) and q_0^* is the dual number of q_0 .
By the definition of the linear functional on L^{q_0}(\Omega), we know that there is a function S_n(x)\in L^{q_0^*}(\Omega) such that
\begin{equation} \hat{{\bf f}}_n(\varphi) = \int_{\Omega} S_n(x)\varphi (x)dx, \; \varphi \in L^{q_0}(\Omega). \end{equation} | (2.20) |
Similarly to the last section of the Step 1 of the proof of Lemma 2.3, we can prove that \mathrm{(C)}_{c^*} sequence \{u_n\} is bounded in E . Next, we show that \{u_n\} has a convergence subsequence. Without loss of generality, assume that
\begin{align*} & \|u_n\|\leq \beta^*, \\ &u_n\rightharpoonup u\ \mbox{ in }\; E, \\ &u_n \rightarrow u \ \mbox{in}\ L^\gamma(\Omega), \ \forall \gamma\geq 1, \\ &u_n(x)\rightarrow u(x)\ \mbox{a.e.}\ x\in \Omega. \end{align*} |
Since f has the subcritical exponential growth \mathrm{(SCE)} on \Omega , we can find a constant C_{\beta^*} > 0 such that
|f^+(x, t)|\leq C_{\beta^*} \exp\left(\frac{32\pi^2}{k(\beta^*)^{2}}t^{2}\right), \ \forall (x, t)\in \Omega \times \mathbb{R}. |
Thus, from the revised Adams inequality (see Lemma 2.2),
\begin{eqnarray*} &&\left|\int_{\Omega} f^+(x, u_n)(u_n-u)dx\right|\\ && \leq C_{\beta^*}\left(\int_{\Omega} \exp\left(\frac{32\pi^2}{(\beta^*)^{2}}u_n^{2}\right)dx\right)^{\frac{1}{k}}|u_n-u|_{k'} \\ &&\leq C_{**}|u_n-u|_{k'}\rightarrow0, \end{eqnarray*} |
where k > 1 and k' is the dual number of k . Similar to the last proof of Lemma 2.3, we have u_n\rightarrow u in E which means that I_\lambda^+ satisfies \mathrm{(C)}_{c^*} .
Lemma 2.7. Assume (H_{3}) holds. If f has the standard subcritical polynomial growth on \Omega (condition \mathrm{(SCP)} ), then I_\lambda satisfies \mathrm{(PS)}_{c^*} .
Proof. Let \{u_n\}\subset E be a \mathrm{(PS)}_{c^*} sequence such that
\begin{equation} \frac{\|u_n\|^2}{2}-\frac{\lambda}{s}\int_\Omega a(x)|u_n|^sdx-\int_\Omega F(x, u_n)dx\rightarrow c^*, \end{equation} | (2.21) |
\begin{equation} \int_\Omega \Delta u_n\Delta\varphi dx-c\int_\Omega \nabla u_n \nabla \varphi dx-\lambda \int_\Omega a(x)|u_n|^{s-2}u_n\varphi dx-\int_\Omega f(x, u_n)\varphi dx = \circ(1)\|\varphi\|, \ \varphi\in E. \end{equation} | (2.22) |
Step 1. To prove that \{u_n\} has a convergence subsequence, we first need to prove that it is a bounded sequence. To do this, argue by contradiction assuming that for a subsequence, which is still denoted by \{u_n\} , we have
\|u_n\|\rightarrow \infty. |
Without loss of generality, assume that \|u_n\|\geq 1 for all n\in {\bf N} and let
v_n = \frac{u_n}{\|u_n\|}. |
Clearly, \|v_n\| = 1 , \forall n\in {\bf N} and then, it is possible to extract a subsequence (denoted also by \{v_n\} ) converges weakly to v in E , converges strongly in L^p(\Omega) (1\leq p < p^*) and converges v a.e. x\in\Omega .
Dividing both sides of (2.21) by \|u_n\|^2 , we obtain
\begin{equation} \int_{\Omega} \frac{F(x, u_n)}{\|u_n\|^2}dx\rightarrow \frac{1}{2}. \end{equation} | (2.23) |
Set
\Omega_0 = \{x\in \Omega: v(x)\neq 0\}. |
By (H_3) , we get that
\begin{equation} \frac{F(x, u_n)}{u_n^2}v_n^2\rightarrow \infty, \; \; x\in \Omega_0. \end{equation} | (2.24) |
If |\Omega_0| is positive, from Fatou's lemma, we obtain
\lim\limits_{n\rightarrow \infty}\int_{\Omega} \frac{F(x, u_n)}{\|u_n\|^2}dx\geq\lim\limits_{n\rightarrow \infty}\int_{\Omega_0} \frac{F(x, u_n)}{u_n^2}v_n^2dx = +\infty, |
which contradicts with (2.23).
Dividing both sides of (2.22) by \|u_n\| , for any \varphi \in E , then there exists a positive constant M(\varphi) such that
\begin{equation} \left|\int_\Omega \frac{f(x, u_n)}{\|u_n\|}\varphi dx\right|\leq M(\varphi), \; \forall n\in {\bf N}. \end{equation} | (2.25) |
Set
{\bf f}_n(\varphi) = \int_\Omega \frac{f(x, u_n)}{\|u_n\|}\varphi dx, \; \varphi \in E. |
Thus, by (SCP), we know that \{{\bf f}_n\} is a family bounded linear functionals defined on E . By (2.25) and the famous Resonance Theorem, we get that \{|{\bf f}_n|\} is bounded, where |{\bf f}_n| denotes the norm of {\bf f}_n . It means that
\begin{equation} |{\bf f}_n|\leq \tilde{C}_*. \end{equation} | (2.26) |
Since E\subset L^{\frac{p^*}{p^*-q}}(\Omega) , using the Hahn-Banach Theorem, there exists a continuous functional \hat{{\bf f}}_n defined on L^{\frac{p^*}{p^*-q}}(\Omega) such that \hat{{\bf f}}_n is an extension of {\bf f}_n , and
\begin{equation} \hat{{\bf f}}_n(\varphi) = {\bf f}_n(\varphi), \; \varphi\in E, \end{equation} | (2.27) |
\begin{equation} \|\hat{{\bf f}}_n\|_{\frac{p^*}{q}} = |{\bf f}_n|, \end{equation} | (2.28) |
where \|\hat{{\bf f}}_n\|_{\frac{p^*}{q}} denotes the norm of \hat{{\bf f}}_n(\varphi) in L^{\frac{p^*}{q}}(\Omega) which is defined on L^{\frac{p^*}{p^*-q}}(\Omega) .
Remained proof is completely similar to the last proof of Lemma 2.3, we omit it here.
Lemma 2.8. Assume (H_{3}) holds. If f has the subcritical exponential growth on \Omega (condition \mathrm{(SCE)} ), then I_\lambda satisfies \mathrm{(PS)}_{c^*} .
Proof. Combining the previous section of the proof of Lemma 2.7 with slightly modifying the last section of the proof of Lemma 2.6, we can prove it. So we omit it here.
To prove the next Lemma, we firstly introduce a sequence of nonnegative functions as follows. Let \Phi(t)\in C^\infty[0, 1] such that
\Phi(0) = \Phi'(0) = 0, |
\Phi(1) = \Phi'(1) = 0. |
We let
H(t) = \begin{cases} \frac{1}{n}\Phi(nt), &\quad \text{if}\; t\leq \frac{1}{n}, \\ t, &\quad \text{if}\; \frac{1}{n} < t < 1-\frac{1}{n}, \\ 1-\frac{1}{n}\Phi(n(1-t)), &\quad \text{if}\; 1-\frac{1}{n}\leq t\leq1, \\ 1, &\quad \text{if}\; 1\leq t, \end{cases} |
and \psi_n(r) = H((ln n)^{-1}ln \frac{1}{r}). Notice that \psi_n(x)\in E , B the unit ball in \mathbb{R}^N , \psi_n(x) = 1 for |x|\leq \frac{1}{n} and, as it was proved in [2],
\|\Delta \psi_n\|_2 = 2\sqrt{2}\pi(ln n)^{-\frac{1}{2}}A_n = \|\psi_n\|+\circ(1), \; \text{as}\; n\rightarrow \infty. |
where 0\leq \lim\limits_{n\rightarrow \infty} A_n\leq 1. Thus, we take x_0\in \Omega and r_0 > 0 such that B(x_0, r)\subset \Omega, denote
\Psi_n(x) = \begin{cases} \frac{\psi_n(|x-x_0|)}{\|\psi_n\|}, &\quad \text{if}\; x\in B(x_0, r_0), \\ 0, &\quad \text{if}\; x\in \Omega\backslash B(x_0, r_0).\\ \end{cases} |
Lemma 2.9. Assume (H_{1}) and (H_{4}) hold. If f has the critical exponential growth on \Omega (condition \mathrm{(CG)} ), then there exists n such that
\max\{I_\lambda^{\pm}(\pm t\Psi_n):t\geq 0\} < \frac{16\pi^2}{\alpha_0}. |
Proof. We only prove the case of I_\lambda^+ . The arguments for the case of I_\lambda^- are similar. Assume by contradiction that this is not the case. So, for all n , this maximum is larger or equal to \frac{16\pi^2}{\alpha_0}. Let t_n > 0 be such that
\begin{equation} \mathcal{I_\lambda^+}(t_n\Psi_n)\geq\frac{16\pi^2}{\alpha_0}. \end{equation} | (2.29) |
From (H_1) and (2.29), we conclude that
\begin{equation} t_n^{2}\geq \frac{32\pi^2}{\alpha_0}. \end{equation} | (2.30) |
Also at t = t_n , we have
t_n-t_n^{s-1}\lambda\int_\Omega a(x)|\Psi_n|^{s}dx-\int_\Omega f(x, t_n\Psi_n)\Psi_ndx = 0, |
which implies that
\begin{equation} t_n^{2}\geq t_n^{s}\lambda\int_\Omega a(x)|\Psi_n|^{s}dx+\int_{B(x_0, r_0)} f(x, t_n\Psi_n)t_n\Psi_ndx. \end{equation} | (2.31) |
Since (H_4) , for given \epsilon > 0 there exists R_\epsilon > 0 such that
tf(x, t)\geq (\beta-\epsilon)\exp\left(\alpha_0t^{2}\right), \ t\geq R_\epsilon. |
So by (2.31), we deduce that, for large n
\begin{equation} t_n^{2}\geq t_n^{s}\lambda\int_\Omega a(x)|\Psi_n|^{s}dx+ (\beta-\epsilon)\frac{\pi^2}{2}r_0^4\exp\left[ \left((\frac{t_n}{ A_n})^2\frac{\alpha_0}{32\pi^2}-1\right)4ln n\right]. \end{equation} | (2.32) |
By (2.30), the inequality above is true if, and only if
\begin{equation} \lim\limits_{n\rightarrow \infty}A_n = 1 \ \ \text{and}\ \ t_n\rightarrow \left( \frac{32\pi^2}{\alpha_0}\right)^{\frac{1}{2}}. \end{equation} | (2.33) |
Set
A_n^* = \{x\in B(x_0, r_0): t_n \Psi_n(x)\geq R_{\epsilon}\}, \; \; B_n = B(x_0, r_0)\setminus A_n^*, |
and break the integral in (2.31) into a sum of integrals over A_n^* and B_n . By simple computation, we have
\begin{equation} \left[\frac{32\pi^2}{\alpha_0}\right]\geq (\beta-\epsilon) \lim\limits_{n\rightarrow \infty} \int_{B(x_0, r_0)}\exp\left[\alpha_0t_n^2|\Psi_n(x)|^2\right]dx-(\beta-\epsilon)r_0^4\frac{\pi^2}{2}. \end{equation} | (2.34) |
The last integral in (2.34), denote I_n is evaluated as follows:
I_n\geq (\beta-\epsilon)r_0^4\pi^2. |
Thus, finally from (2.34) we get
\left[\frac{32\pi^2}{\alpha_0}\right]\geq(\beta-\epsilon)r_0^4\frac{\pi^2}{2}, |
which means \beta\leq \frac{64}{\alpha_0r_0^4}. This results in a contradiction with (H_4) .
To conclude this section we state the Fountain Theorem of Bartsch [32].
Define
\begin{equation} Y_k = \oplus_{j = 1}^kX_j, \ \ Z_k = \overline{\oplus_{j\geq k}X_j}. \end{equation} | (2.35) |
Lemma 2.10. (Dual Fountain Theorem). Assume that I_\lambda\in C^1(\mathbb{E}, \mathbb{R}) satisfies the \mathrm{(PS)_c^*} condition (see [32]), I_\lambda(-u) = I_\lambda(u) . If for almost every k\in {\bf N}, there exist \rho_k > r_k > 0 such that
\mathrm{(i)} a_k: = \inf\limits_{u\in Z_k, \|u\| = \rho_k}I_\lambda(u)\geq 0,
\mathrm{(ii)} b_k: = \max\limits_{u\in Y_k, \|u\| = r_k}I_\lambda(u) < 0,
\mathrm{(iii)} b_k = \inf\limits_{u\in Z_k, \|u\| = \rho_k}I_\lambda(u)\rightarrow 0, \ \mathit{\text{as}}\ k\rightarrow \infty,
then I_\lambda has a sequence of negative critical values converging 0 .
Proof of Theorem 1.1. For I_\lambda^{\pm}, we first demonstrate that the existence of local minimum v_{\pm} with I_\lambda^{\pm}(v_{\pm}) < 0 . We only prove the case of I_\lambda^+. The arguments for the case of I_\lambda^- are similar.
For \rho determined in Lemma 2.4, we write
\bar{B}(\rho) = \{u\in E, \ \|u\|\leq \rho\}, \ \ \partial B(\rho) = \{u\in E, \ \|u\| = \rho\}. |
Then \bar{B}(\rho) is a complete metric space with the distance
\text{dist}(u, v) = \|u-v\|, \quad \forall u, v\in \bar{B}(\rho). |
From Lemma 2.4, we have for 0 < \lambda < \Lambda^*,
I_\lambda^+(u)|_{\partial B(\rho)}\geq \alpha > 0. |
Furthermore, we know that I_\lambda^+\in C^1(\bar{B}(\rho), \mathbb{R}), hence I_\lambda^+ is lower semi-continuous and bounded from below on \bar{B}(\rho) . Set
c_1^* = \inf\{ I_\lambda^+(u), u\in \bar{B}(\rho)\}. |
Taking \tilde{\phi}\in C_0^\infty (\Omega) with \tilde{\phi} > 0, and for t > 0 , we get
\begin{eqnarray*} I_\lambda^+(t\tilde{\phi})& = &\frac{t^2}{2}\|\tilde{\phi}\|^2-\frac{\lambda t^s}{s}\int_\Omega a(x) |\tilde{\phi}|^s dx -\int_\Omega{F^+(x, t\tilde{\phi})}dx\\ &\leq& \frac{t^2}{2}\|\tilde{\phi}\|^2-\frac{\lambda t^s}{s}\int_\Omega a(x) |\tilde{\phi}|^s dx\\ & < & 0, \end{eqnarray*} |
for all t > 0 small enough. Hence, c_1^* < 0 .
Since Ekeland's variational principle and Lemma 2.4, for any m > 1 , there exists u_m with \|u_m\| < \rho such that
I_\lambda^+(u_m)\rightarrow c_1^*, \quad I_\lambda^{+'}(u_m)\rightarrow 0. |
Hence, there exists a subsequence still denoted by \{u_m\} such that
u_m\rightarrow v_+, \quad I_\lambda^{+'}(v_+) = 0. |
Thus v_+ is a weak solution of problem (1.1) and I_\lambda^+(v_+) < 0 . In addition, from the maximum principle, we know v_+ > 0 . By a similar way, we obtain a negative solution v_- with I_\lambda^-(v_-) < 0 .
On the other hand, from Lemmas 2.3 and 2.4, the functional I_\lambda^+ has a mountain pass-type critical point u_+ with I_\lambda^+(u_+) > 0. Again using the maximum principle, we have u_+ > 0 . Hence, u_+ is a positive weak solution of problem (1.1). Similarly, we also obtain a negative mountain pass-type critical point u_- for the functional I_\lambda^- . Thus, we have proved that problem (1.1) has four different nontrivial solutions. Next, our method to obtain the fifth solution follows the idea developed in [33] for problem (1.1). We can assume that v_+ and v_- are isolated local minima of I_\lambda . Let us denote by b_\lambda the mountain pass critical level of I_\lambda with base points v_+, v_-:
b_\lambda = \inf\limits_{\gamma\in \Gamma}\max\limits_{0\leq t\leq1}I_\lambda(\gamma(t)), |
where \Gamma = \{\gamma\in C([0, 1], E), \gamma(0) = v_+, \gamma(1) = v_-\} . We will show that b_\lambda < 0 if \lambda is small enough. To this end, we regard
I_\lambda(tv_{\pm}) = \frac{t^2}{2}\|v_{\pm}\|^2-\frac{\lambda t^s}{s}\int_\Omega a(x)|v_{\pm}|^sdx-\int_\Omega F(x, tv_{\pm})dx. |
We claim that there exists \delta > 0 such that
\begin{equation} I_\lambda(tv_{\pm}) < 0, \ \forall t\in (0, 1), \ \forall \lambda\in (0, \delta). \end{equation} | (3.1) |
If not, we have t_0\in (0, 1) such that I_\lambda(t_0v_{\pm})\geq 0 for \lambda small enough. Similarly, we also have I_\lambda(tv_{\pm}) < 0 for t > 0 small enough. Let \rho_0 = t_0\|v_{\pm}\| and \check{c}_*^{\pm} = \inf \{I_\lambda^{\pm}(u), u\in \bar{B}(\rho_0)\}. Since previous arguments, we obtain a solution v_{\pm}^* such that I_\lambda(v_{\pm}^*) < 0, a contradiction. Hence, (3.1) holds.
Now, let us consider the 2 -dimensional plane \Pi_2 containing the straightlines tv_- and tv_+ , and take v\in \Pi_2 with \|v\| = \epsilon. Note that for such v one has \|v\|_s = c_s\epsilon. Then we get
I_\lambda(v)\leq \frac{\epsilon^2}{2}-\frac{\lambda}{s}c_s^sh_0\epsilon^s. |
Thus, for small \epsilon ,
\begin{equation} I_\lambda(v) < 0. \end{equation} | (3.2) |
Consider the path \bar{\gamma} obtained gluing together the segments \{tv_-:\epsilon \|v_-\|^{-1}\leq t\leq 1\}, \{tv_+:\epsilon \|v_+\|^{-1}\leq t\leq 1\} and the arc \{v\in \Pi_2: \|v\| = \epsilon\} . by (3.1)and (3.2), we get
b_\lambda\leq \max\limits_{v\in \bar{\gamma}}I_\lambda(v) < 0, |
which verifies the claim. Since the (PS) condition holds because of Lemma 2.3, the level \{I_\lambda(v) = b_\lambda\} carries a critical point v_3 of I_\lambda , and v_3 is different from v_{\pm} .
Proof of Theorem 1.2. We first use the symmetric mountain pass theorem to prove the case of a) . It follows from our assumptions that the functional I_\lambda is even. Since the condition (SCP), we know that (I_1') of Theorem 9.12 in [30] holds. Furthermore, by condition (H_3) , we easily verify that (I_2') of Theorem 9.12 also holds. Hence, by Lemma 2.7, our theorem is proved.
Next we use the dual fountain theorem (Lemma 2.10) to prove the case of b) . Since Lemma 2.7, we know that the functional I_\lambda satisfies \mathrm{(PS)_c^*} condition. Next, we just need to prove the conditions (ⅰ)-(ⅲ) of Lemma 2.10.
First, we verify (ⅰ) of Lemma 2.10. Define
\beta_k: = \sup\limits_{u\in Z_k, \|u\| = 1}\|u\|_s. |
From the conditions (SCP) and (H_2) , we get, for u\in Z_k, \|u\|\leq R,
\begin{align} I_\lambda(u)&\geq \frac{\|u\|^2}{2}-\lambda \beta_k^s\frac{\|u\|^s}{s}-\frac{f_0+\epsilon}{2}\|u\|_2^2-c_6\|u\|^q\\ &\geq \frac{1}{4}(1-\frac{f_0+\epsilon}{\mu_1})\|u\|^2-\lambda \beta_k^s\frac{\|u\|^s}{s}. \end{align} | (3.3) |
Here, R is a positive constant and \epsilon > 0 small enough. We take \rho_k = (4\mu_1\lambda \beta_k^s/[(\mu_1-f_0-\epsilon)s])^{\frac{1}{2-s}}. Since \beta_k\rightarrow 0, k\rightarrow \infty, it follows that \rho_k\rightarrow 0, k\rightarrow \infty. There exists k_0 such that \rho_k\leq R when k\geq k_0 . Thus, for k\geq k_0, u\in Z_k and \|u\| = \rho_k, we have I_\lambda(u)\geq 0 and (ⅰ) holds. The verification of (ⅱ) and (ⅲ) is standard, we omit it here.
Proof of Theorem 1.3. According to our assumptions, similar to previous section of the proof of Theorem 1.1, we obtain that the existence of local minimum v_{\pm} with I_\lambda^{\pm}(v_{\pm}) < 0 . In addition, by Lemmas 2.5 and 2.6, for I_\lambda^{\pm} , we obtain two mountain pass type critical points u_+ and u_- with positive energy. Similar to the last section of the proof of Theorem 1.1, we can also get another solution u_3 , which is different from v_{\pm} and u_{\pm} . Thus, this proof is completed.
Proof of Theorem 1.4. We first use the symmetric mountain pass theorem to prove the case of a) . It follows from our assumptions that the functional I_\lambda is even. Since the condition (SCE), we know that (I_1') of Theorem 9.12 in [30] holds. In fact, similar to the proof of (ⅰ) of Lemma 2.5, we can conclude it. Furthermore, by condition (H_3) , we easily verify that (I_2') of Theorem 9.12 also holds. Hence, by Lemma 2.8, our theorem is proved.
Next we use the dual fountain theorem (Lemma 2.10) to prove the case of b) . Since Lemma 2.8, we know that the functional I_\lambda satisfies \mathrm{(PS)_c^*} condition. Next, we just need to prove the conditions (ⅰ)-(ⅲ) of Lemma 2.10.
First, we verify (ⅰ) of Lemma 2.10. Define
\beta_k: = \sup\limits_{u\in Z_k, \|u\| = 1}\|u\|_s. |
From the conditions (SCE), (H_2) and Lemma 2.2, we get, for u\in Z_k, \|u\|\leq R,
\begin{align} I_\lambda(u)&\geq \frac{\|u\|^2}{2}-\lambda \beta_k^s\frac{\|u\|^s}{s}-\frac{f_0+\epsilon}{2}\|u\|_2^2-c_7\|u\|^q\\ &\geq \frac{1}{4}(1-\frac{f_0+\epsilon}{\mu_1})\|u\|^2-\lambda \beta_k^s\frac{\|u\|^s}{s}. \end{align} | (3.4) |
Here, R is a positive constant small enough and \epsilon > 0 small enough. We take \rho_k = (4\mu_1\lambda \beta_k^s/[(\mu_1-f_0-\epsilon)s])^{\frac{1}{2-s}}. Since \beta_k\rightarrow 0, k\rightarrow \infty, it follows that \rho_k\rightarrow 0, k\rightarrow \infty. There exists k_0 such that \rho_k\leq R when k\geq k_0 . Thus, for k\geq k_0, u\in Z_k and \|u\| = \rho_k, we have I_\lambda(u)\geq 0 and (ⅰ) holds. The verification of (ⅱ) and (ⅲ) is standard, we omit it here.
Proof of Theorem 1.5. According to our assumptions, similar to previous section of the proof of Theorem 1.1, we obtain that the existence of local minimum v_{\pm} with I_\lambda^{\pm}(v_{\pm}) < 0 . Now, we show that I_\lambda^+ has a positive mountain pass type critical point. Since Lemmas 2.5 and 2.9, then there exists a \mathrm{(C)_{c_M}} sequence \{u_n\} at the level 0 < c_M\leq \frac{16\pi^2}{\alpha_0} . Similar to previous section of the proof of Lemma 2.6, we can prove that \mathrm{(C)_{c_M}} sequence \{u_n\} is bounded in E . Without loss of generality, we can suppose that
u_n\rightharpoonup u_+\; \; \text{in}\; E. |
Following the proof of Lemma 4 in [9], we can imply that u_+ is weak of problem (1.1). So the theorem is proved if u_+ is not trivial. However, we can get this due to our technical assumption (H_5) . Indeed, assume u_+ = 0 , similarly as in [9], we obtain f^+(x, u_n)\rightarrow 0 in L^1(\Omega) . Since (H_5), F^+(x, u_n)\rightarrow 0 in L^1(\Omega) and we get
\lim\limits_{n\rightarrow \infty}\|u_n\|^{2} = 2c_M < \frac{32\pi^2}{\alpha_0} , |
and again following the proof in [9], we get a contradiction.
We claim that v_+ and u_+ are distinct. Since the previous proof, we know that there exist sequence \{u_n\} and \{v_n\} in E such that
\begin{equation} u_n\rightarrow v_+, \ I_\lambda^+(u_n)\rightarrow c_*^+ < 0, \ \langle I_\lambda^{+'}(u_n), u_n\rangle\rightarrow 0, \end{equation} | (3.5) |
and
\begin{equation} v_n\rightharpoonup u_+, \ I_\lambda^+(v_n)\rightarrow c_M > 0, \ \langle I_\lambda^{+'}(v_n), v_n\rangle\rightarrow 0. \end{equation} | (3.6) |
Now, argue by contradiction that v_+ = u_+. Since we also have v_n\rightharpoonup v_+ in E , up to subsequence, \lim\limits_{n\rightarrow \infty}\|v_n\|\geq \|v_+\| > 0. Setting
w_n = \frac{v_n}{\|v_n\|}, \ \ w_0 = \frac{v_+}{\lim\limits_{n\rightarrow \infty}\|v_n\|}, |
we know that \|w_n\| = 1 and w_n\rightharpoonup w_0 in E .
Now, we consider two possibilities:
\mathrm{ (i)} \ \|w_0\| = 1, \quad \mathrm{(ii)}\ \|w_0\| < 1. |
If (ⅰ) happens, we have v_n\rightarrow v_+ in E , so that I_\lambda^+(v_n)\rightarrow I_\lambda^+(v_+) = c_*^+. This is a contradiction with (3.5) and (3.6).
Now, suppose that (ⅱ) happens. We claim that there exists \delta > 0 such that
\begin{equation} h\alpha_0\|v_n\|^2\leq \frac{32\pi^2}{1-\|w_0\|^2}-\delta \end{equation} | (3.7) |
for n large enough. In fact, by the proof of v_+ and Lemma 2.9, we get
\begin{equation} 0 < c_M < c_*^++ \frac{16\pi^2}{\alpha_0}. \end{equation} | (3.8) |
Thus, we can choose h > 1 sufficiently close to 1 and \delta > 0 such that
h\alpha_0\|v_n\|^2\leq \frac{16\pi^2}{c_M-I_\lambda^+(v_+)}\|v_n\|^2-\delta. |
Since v_n\rightharpoonup v_+, by condition (H_5) , up to a subsequence, we conclude that
\begin{equation} \frac{1}{2}\|v_n\|^2 = c_M+\frac{\lambda}{s}\int_\Omega a(x) v_+^s dx +\int_\Omega F^+(x, v_+)dx+\circ (1). \end{equation} | (3.9) |
Thus, for n sufficiently large we get
\begin{equation} h\alpha_0\|v_n\|^2\leq 32\pi^2\frac{ c_M+\frac{\lambda}{s}\int_\Omega a(x) v_+^s dx +\int_\Omega F^+(x, v_+)dx+\circ (1)}{c_M-I_\lambda^+(v_+)}-\delta. \end{equation} | (3.10) |
Thus, from (3.9) and the definition of w_0 , (3.10) implies (3.7) for n large enough.
Now, taking \tilde{h} = (h+\epsilon)\alpha_0\|v_n\|^2, it follows from (3.7) and a revised Adams inequality (see [28]), we have
\begin{equation} \int_\Omega \exp((h+\epsilon)\alpha_0\|v_n\|^2|w_n|^2dx\leq C \end{equation} | (3.11) |
for \epsilon > 0 small enough. Thus, from our assumptions and the Hölder inequality we get v_n\rightarrow v_+ and this is absurd.
Similarly, we can find a negative mountain pass type critical point u_- which is different that v_- . Thus, the proof is completed.
In this research, we mainly studied the existence and multiplicity of nontrivial solutions for the fourth-order elliptic Navier boundary problems with exponential growth. Our method is based on the variational methods, Resonance Theorem together with a revised Adams inequality.
The authors would like to thank the referees for valuable comments and suggestions in improving this article. This research is supported by the NSFC (Nos. 11661070, 11764035 and 12161077), the NSF of Gansu Province (No. 22JR11RE193) and the Nonlinear mathematical physics Equation Innovation Team (No. TDJ2022-03).
There is no conflict of interest.
[1] |
A. Obenaus, C. J. Yong-Hing, K. A. Tong, G. E. Sarty, A reliable method for measurement and normalization of pediatric hippocampal volumes, J. Pediatr. Res., 50 (2001), 124–132. https://doi.org/10.1203/00006450-200107000-00022 doi: 10.1203/00006450-200107000-00022
![]() |
[2] |
D. Shen, S. Moffat, S. M. Resnick, C. Davatzikos, Measuring size and shape of the hippocampus in MR images using a deformable shape model, Neuroimage, 15 (2002), 422–434. https://doi.org/10.1006/nimg.2001.0987 doi: 10.1006/nimg.2001.0987
![]() |
[3] |
S. Li, F. Shi, F. Pu, X. Li, T. Jiang, S. Xie, Hippocampal shape analysis of Alzheimer disease based on machine learning methods, J. Neuroradiol., 28 (2007), 1339–1345. https://doi.org/10.3174/ajnr.A0620 doi: 10.3174/ajnr.A0620
![]() |
[4] |
J. H. Morra, Z. Tu, L. G. Apostolova, A. E. Green, A. W. Toga, P. M. Thompson, Comparison of AdaBoost and support vector machines for detecting Alzheimer's disease through automated hippocampal segmentation, IEEE Trans. Med. Imaging, 29 (2010), 30–43. https://doi.org/10.1109/TMI.2009.2021941 doi: 10.1109/TMI.2009.2021941
![]() |
[5] |
H. Wang, J. W. Suh, S. R. Das, J. B. Pluta, C. Craige; P. A. Yushkevich, Multi-atlas segmentation with joint label fusion, IEEE Trans. Pattern Anal. Mach. Intell., 35 (2012), 611–623. https://doi.org/10.1109/TPAMI.2012.143 doi: 10.1109/TPAMI.2012.143
![]() |
[6] |
M. Kim, G. Wu, D. Shen, Unsupervised deep learning for hippocampus segmentation in 7.0 Tesla MR images, Int. Workshop Mach. Learn. Med. Imaging, (2013), 1–8. https://doi.org/10.1007/978-3-319-02267-3_1 doi: 10.1007/978-3-319-02267-3_1
![]() |
[7] |
J. Kim, M. C. Valdes-Hernandez, N. A. Royle, J. Park, Hippocampal shape modeling based on a progressive template surface deformation and its verification, IEEE Trans. Med. Imaging, 34 (2015), 1242–1261. https://doi.org/10.1109/TMI.2014.2382581 doi: 10.1109/TMI.2014.2382581
![]() |
[8] |
D. Zarpalas, P. Gkontra, P. Daras, N. Maglaveras, Accurate and fully automatic hippocampus segmentation using subject-specific 3D optimal local maps into a hybrid active contour model, IEEE J. Transl. Eng. Health Med., 2 (2016), 1–16. https://doi.org/10.1109/JTEHM.2014.2297953 doi: 10.1109/JTEHM.2014.2297953
![]() |
[9] |
S. Sri Devi, A. Mano, R. Asha, MRI brain tumor segmentation and feature extraction using GLCM, Int. J. Res. Appl. Sci. Eng. Technol., 6 (2018), 1911–1916. https://doi.org/10.22214/ijraset.2018.1297 doi: 10.22214/ijraset.2018.1297
![]() |
[10] |
V. Dill, P. C. Klein, A. R. Franco, M. S. Pinho, Atlas selection for hippocampus segmentation: Relevance evaluation of three meta-information parameters, J. Comput. Biol. Med., 95 (2018), 90–98. https://doi.org/10.1016/j.compbiomed.2018.02.005 doi: 10.1016/j.compbiomed.2018.02.005
![]() |
[11] |
N. Varuna Shree, T. N. R. Kumar, Identification and classification of brain tumor MRI images with feature extraction using DWT and probabilistic neural network, Brain Inform., 5 (2018), 23–30. https://doi.org/10.1007/s40708-017-0075-5 doi: 10.1007/s40708-017-0075-5
![]() |
[12] |
E. Gibson, W. Li, C. Sudre, L. Fidon, D. I. Shakir, G. Wang, et al., NiftyNet: a deep-learning platform for medical imaging, Comput. Methods Programs Biomed., 158 (2018), 113–122. https://doi.org/10.1016/j.cmpb.2018.01.025 doi: 10.1016/j.cmpb.2018.01.025
![]() |
[13] |
Y. Shao, J. Kim, Y. Gao, Q. Wang, W. Lin, D. Shen, Hippocampal segmentation from longitudinal infant brain MR images via classification-guided boundary regression, IEEE Access, 7 (2019), 33728–33740. https://doi.org/10.1109/ACCESS.2019.2904143 doi: 10.1109/ACCESS.2019.2904143
![]() |
[14] |
A. Basher, K. Y. Choi, J. J. Lee, B. Lee, B. C. Kim, K. H. Lee, et al., Hippocampus localization using a two-stage ensemble Hough convolutional neural network, IEEE Access, 7 (2019), 73436–73447. https://doi.org/10.1109/ACCESS.2019.2920005 doi: 10.1109/ACCESS.2019.2920005
![]() |
[15] |
S. Liu, Y. Wang, X. Yang, B. Lei, L. Liu, S. X. Li, Deep learning in medical ultrasound analysis: a review, Engineering, 5 (2019), 261–275. https://doi.org/10.1016/j.eng.2018.11.020 doi: 10.1016/j.eng.2018.11.020
![]() |
[16] |
A. Gumaei, M. M. Hassan, M. R. Hassan, A. Alelaiwi, G. Fortino, A hybrid feature extraction method with regularized extreme learning machine for brain tumor classification, IEEE Access, 7 (2019), 36266–36273. https://doi.org/10.1109/ACCESS.2019.2904145 doi: 10.1109/ACCESS.2019.2904145
![]() |
[17] |
Y. Shi, K. Cheng, Z. Liu, Hippocampal subfields segmentation in brain MR images using generative adversarial networks, Biomed. Eng. Online, 18 (2019), 1–12. https://doi.org/10.1186/s12938-019-0623-8 doi: 10.1186/s12938-019-0623-8
![]() |
[18] |
A. S. Lundervold, A. Lundervold, An overview of deep learning in medical imaging focusing on MRI, J. Med. Phys., 29 (2019), 102–127. https://doi.org/10.1016/j.zemedi.2018.11.002 doi: 10.1016/j.zemedi.2018.11.002
![]() |
[19] | S. M. Nisha, A novel computer-aided diagnosis scheme for breast tumor classification, Int. Res. J. Eng. Technol., 7 (2020), 718–724. |
[20] |
N. Safavian, S. A. H. Batouli, M. A. Oghabian, An automatic level set method for hippocampus segmentation in MR images, Comput. Methods Biomech. Biomed. Eng. Imaging Vis., 8 (2020), 400–410. https://doi.org/10.1080/21681163.2019.1706054 doi: 10.1080/21681163.2019.1706054
![]() |
[21] |
M. Liu, F. Li, H. Yan, K. Wang, Y. Ma, L. Shen, et al., A multi-model deep convolutional neural network for automatic hippocampus segmentation and classification in Alzheimer's disease, Neuroimage, 208 (2020), 116459. https://doi.org/10.1016/j.neuroimage.2019.116459 doi: 10.1016/j.neuroimage.2019.116459
![]() |
[22] |
M. K. Singh, K. K. Singh, A review of publicly available automatic brain segmentation methodologies, machine learning models, recent advancements, and their comparison, Ann. Neurosci., 28 (2021), 82–93. https://doi.org/10.1177/0972753121990 doi: 10.1177/0972753121990
![]() |
[23] |
L. Liu, L. Kuang, Y. Ji, Multimodal MRI brain tumor image segmentation using sparse subspace clustering algorithm, Comput. Math. Methods Med., (2020), 8620403. https://doi.org/10.1155/2020/8620403 doi: 10.1155/2020/8620403
![]() |
[24] |
D. Carmo, B. Silva, C. Yasuda, L. Rittner, R. Lotufo, Hippocampus segmentation on epilepsy and Alzheimer's disease studies with multiple convolutional neural networks, Heliyon, 7 (2021), e06226. https://doi.org/10.1016/j.heliyon.2021.e06226 doi: 10.1016/j.heliyon.2021.e06226
![]() |
[25] |
R. De Feo, E. Hämäläinen, E. Manninen, R. Immonen, J. M. Valverde, X. E. Ndode-Ekane, et al., Convolutional neural networks enable robust automatic segmentation of the rat hippocampus in mri after traumatic brain injury, Front. Neurol., 13 (2022), 820267. https://doi.org/10.3389/fneur.2022.820267 doi: 10.3389/fneur.2022.820267
![]() |
[26] |
M. Nisha, T. Kannan, K. Sivasankari, M. Sabrigiriraj, Automatic hippocampus segmentation model for MRI of human head through semi-supervised generative adversarial networks, Neuroquantology, 20 (2022), 5222–5232. https://doi.org/10.14704/nq.2022.20.6.NQ22528 doi: 10.14704/nq.2022.20.6.NQ22528
![]() |
[27] |
K. S. Chuang, H. L. Tzeng, S. Chen, J. Wu, T. J. Chen, Fuzzy c-means clustering with spatial information for image segmentation, Comput. Med. Imaging Graph., 30 (2006), 9–15. https://doi.org/10.1016/j.compmedimag.2005.10.001 doi: 10.1016/j.compmedimag.2005.10.001
![]() |
[28] |
B. N. Li, C. K. Chui, S. Chang, S. H. Ong, Integrating spatial fuzzy clustering with level set methods for automated medical image segmentation, Comput. Biol. Med., 41 (2011), 1–10. https://doi.org/10.1016/j.compbiomed.2010.10.007 doi: 10.1016/j.compbiomed.2010.10.007
![]() |
[29] |
C. Militello, L. Rundo, M. Dimarco, A. Orlando, V. Conti, R. Woitek, et al., Semi-automated and interactive segmentation of contrast-enhancing masses on breast DCE-MRI using spatial fuzzy clustering, Biomed. Signal Process. Control, 71 (2022), 103113. https://doi.org/10.1016/j.bspc.2021.103113 doi: 10.1016/j.bspc.2021.103113
![]() |
[30] | Z. Zhou, M. M. R. Siddiquee, N. Tajbakhsh, J. Liang, U-Net++: A nested U-Net architecture for medical image segmentation, in Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support, Springer, (2018), 3–11. https://doi.org/10.1007/978-3-030-00889-5_1 |
[31] | K. He, X. Zhang, S. Ren, J. Sun, Deep residual learning for image recognition, in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., (2016), 770–778. https://doi.org/10.1109/CVPR.2016.90 |
[32] | G. Huang, Z. Liu, L. Van Der Maaten., K. Q. Weinberger, Densely connected convolutional networks, in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., (2017), 4700–4708. https://doi.org/10.1109/CVPR.2017.243 |
[33] | Z. Liu, Y. Lin, Y. Cao, H. Hu, Y. Wei, Z. Zhang, et al., Swin transformer: Hierarchical vision transformer using shifted windows, in Proc. IEEE/CVF Int. Conf. Comput. Vis., (2021), 10012–10022. https://doi.org/10.1109/ICCV48922.2021.00986 |
[34] | J. Hu, L. Shen, G. Sun, Squeeze-and-Excitation networks, in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., (2018), 7132–7141. https://doi.org/10.48550/arXiv.1709.01507 |
[35] |
J. Wang, K. Sun, T. Cheng, B. Jiang, C. Deng, Y. Zhao, et al., Deep high-resolution representation learning for visual recognition, IEEE Trans. Pattern Anal. Mach. Intell., 43 (2020), 3349–3364. https://doi.org/10.1109/TPAMI.2020.2983686 doi: 10.1109/TPAMI.2020.2983686
![]() |
[36] |
Z. Szentimrey, A. Al‐Hayali, S. de Ribaupierre, A. Fenster, E. Ukwatta, Semi‐supervised learning framework with shape encoding for neonatal ventricular segmentation from 3D ultrasound, Med. Phys., 2024. https://doi.org/10.1002/mp.17242 doi: 10.1002/mp.17242
![]() |
[37] | Z. Wang, C. Ma, Dual-contrastive dual-consistency dual-transformer: A semi-supervised approach to medical image segmentation, in Proc. 2023 IEEE/CVF Int. Conf. Comput. Vis. Workshops, (2023), 870–879. https://doi.org/10.1109/ICCVW60793.2023.00094 |
[38] |
L. Huang, S. Ruan, T. Denœux, Semi-supervised multiple evidence fusion for brain tumor segmentation, Neurocomputing, 535 (2023), 40–52. https://doi.org/10.1016/j.neucom.2023.02.047 doi: 10.1016/j.neucom.2023.02.047
![]() |
[39] | Z. Wang, I. Voiculescu, Exigent examiner and mean teacher: An advanced 3d cnn-based semi-supervised brain tumor segmentation framework, in Med. Image Learn. Limited Noisy Data: 2nd Int. Workshop MILLanD 2023, (2023), 181–190. https://doi.org/10.1007/978-3-031-44917-8_17 |
[40] |
G. Qu, B. Lu, J. Shi, Z. Wang, Y. Yuan, Y. Xia, et al., Motion-artifact-augmented pseudo-label network for semi-supervised brain tumor segmentation, Phys. Med. Biol., 69 (2024), 5. https://doi.org/10.1088/1361-6560/ad2634 doi: 10.1088/1361-6560/ad2634
![]() |
[41] |
R. A. Hazarika, A. K. Maji, R. Syiem, S. N. Sur, D. Kandar, Hippocampus segmentation using U-net convolutional network from brain magnetic resonance imaging (MRI), J. Digit. Imaging, 35 (2022), 893–909. https://doi.org/10.1007/s10278-022-00613-y doi: 10.1007/s10278-022-00613-y
![]() |
[42] |
D. Ataloglou, A. Dimou, D. Zarpalas, P. Daras, Fast and precise hippocampus segmentation through deep convolutional neural network ensembles and transfer learning, Neuroinformatics, 17 (2019), 563–582. https://doi.org/10.1007/s12021-019-09417-y doi: 10.1007/s12021-019-09417-y
![]() |
[43] |
M. Nisha, T. Kannan, K. Sivasankari, Deep integration model: A robust autonomous segmentation technique for hippocampus in MRI images of human head, Int. J. Health Sci., 6 (2022), 13745–13758. https://doi.org/10.53730/ijhs.v6nS2.8756 doi: 10.53730/ijhs.v6nS2.8756
![]() |
[44] | N. Allinson, H. Yin, L. Allinson, J. Slack, Advances in Self-Organising Maps, Springer, 2001. https://doi.org/10.1007/978-1-4471-0715-6. |
[45] | S. N. Sivanandam, S. Sumathi, S. N. Deepa, Applications of Fuzzy Logic: Introduction to Fuzzy Logic Using MATLAB, Springer, 2007. https://doi.org/10.1007/978-3-540-35781-0_8 |
[46] |
V. Conti, C. Militello, L. Rundo, S. Vitabile, A novel bio-inspired approach for high-performance management in service-oriented networks, IEEE Trans. Emerg. Top. Comput., 9 (2021), 1709–1722. https://doi.org/10.1109/TETC.2020.3018312 doi: 10.1109/TETC.2020.3018312
![]() |
1. | Yongxiang Li, Yanyan Wang, The Existence and Uniqueness of Radial Solutions for Biharmonic Elliptic Equations in an Annulus, 2024, 13, 2075-1680, 383, 10.3390/axioms13060383 | |
2. | Danni Zhang, Ziheng Zhang, Sign-changing Solutions for Fourth Order Elliptic Equation with Concave-convex Nonlinearities, 2024, 11, 2409-5761, 1, 10.15377/2409-5761.2024.11.1 |