
Bone age assessment is of great significance to genetic diagnosis and endocrine diseases. Traditional bone age diagnosis mainly relies on experienced radiologists to examine the regions of interest in hand radiography, but it is time-consuming and may even lead to a vast error between the diagnosis result and the reference. The existing computer-aided methods predict bone age based on general regions of interest but do not explore specific regions of interest in hand radiography. This paper aims to solve such problems by performing bone age prediction on the articular surface and epiphysis from hand radiography using deep convolutional neural networks. The articular surface and epiphysis datasets are established from the Radiological Society of North America (RSNA) pediatric bone age challenge, where the specific feature regions of the articular surface and epiphysis are manually segmented from hand radiography. Five convolutional neural networks, i.e., ResNet50, SENet, DenseNet-121, EfficientNet-b4, and CSPNet, are employed to improve the accuracy and efficiency of bone age diagnosis in clinical applications. Experiments show that the best-performing model can yield a mean absolute error (MAE) of 7.34 months on the proposed articular surface and epiphysis datasets, which is more accurate and fast than the radiologists. The project is available at https://github.com/YameiDeng/BAANet/, and the annotated dataset is also published at https://doi.org/10.5281/zenodo.7947923.
Citation: Yamei Deng, Yonglu Chen, Qian He, Xu Wang, Yong Liao, Jue Liu, Zhaoran Liu, Jianwei Huang, Ting Song. Bone age assessment from articular surface and epiphysis using deep neural networks[J]. Mathematical Biosciences and Engineering, 2023, 20(7): 13133-13148. doi: 10.3934/mbe.2023585
[1] | Guowei Wang, Yan Fu . Spatiotemporal patterns and collective dynamics of bi-layer coupled Izhikevich neural networks with multi-area channels. Mathematical Biosciences and Engineering, 2023, 20(2): 3944-3969. doi: 10.3934/mbe.2023184 |
[2] | Bingrui Zhang, Jin-E Zhang . Fixed-deviation stabilization and synchronization for delayed fractional-order complex-valued neural networks. Mathematical Biosciences and Engineering, 2023, 20(6): 10244-10263. doi: 10.3934/mbe.2023449 |
[3] | Pan Yang, Jianwen Feng, Xinchu Fu . Cluster collective behaviors via feedback pinning control induced by epidemic spread in a patchy population with dispersal. Mathematical Biosciences and Engineering, 2020, 17(5): 4718-4746. doi: 10.3934/mbe.2020259 |
[4] | A. Bernardini, J. Bragard, H. Mancini . Synchronization between two Hele-Shaw Cells. Mathematical Biosciences and Engineering, 2004, 1(2): 339-346. doi: 10.3934/mbe.2004.1.339 |
[5] | Wenjing Wang, Jingjing Dong, Dong Xu, Zhilian Yan, Jianping Zhou . Synchronization control of time-delay neural networks via event-triggered non-fragile cost-guaranteed control. Mathematical Biosciences and Engineering, 2023, 20(1): 52-75. doi: 10.3934/mbe.2023004 |
[6] | Ricardo López-Ruiz, Danièle Fournier-Prunaret . Complex Behavior in a Discrete Coupled Logistic Model for the Symbiotic Interaction of Two Species. Mathematical Biosciences and Engineering, 2004, 1(2): 307-324. doi: 10.3934/mbe.2004.1.307 |
[7] | Karim El Laithy, Martin Bogdan . Synaptic energy drives the information processing mechanisms in spiking neural networks. Mathematical Biosciences and Engineering, 2014, 11(2): 233-256. doi: 10.3934/mbe.2014.11.233 |
[8] | Biwen Li, Xuan Cheng . Synchronization analysis of coupled fractional-order neural networks with time-varying delays. Mathematical Biosciences and Engineering, 2023, 20(8): 14846-14865. doi: 10.3934/mbe.2023665 |
[9] | Harindri Chaudhary, Mohammad Sajid, Santosh Kaushik, Ali Allahem . Stability analysis of chaotic generalized Lotka-Volterra system via active compound difference anti-synchronization method. Mathematical Biosciences and Engineering, 2023, 20(5): 9410-9422. doi: 10.3934/mbe.2023413 |
[10] | Mahtab Mehrabbeik, Fatemeh Parastesh, Janarthanan Ramadoss, Karthikeyan Rajagopal, Hamidreza Namazi, Sajad Jafari . Synchronization and chimera states in the network of electrochemically coupled memristive Rulkov neuron maps. Mathematical Biosciences and Engineering, 2021, 18(6): 9394-9409. doi: 10.3934/mbe.2021462 |
Bone age assessment is of great significance to genetic diagnosis and endocrine diseases. Traditional bone age diagnosis mainly relies on experienced radiologists to examine the regions of interest in hand radiography, but it is time-consuming and may even lead to a vast error between the diagnosis result and the reference. The existing computer-aided methods predict bone age based on general regions of interest but do not explore specific regions of interest in hand radiography. This paper aims to solve such problems by performing bone age prediction on the articular surface and epiphysis from hand radiography using deep convolutional neural networks. The articular surface and epiphysis datasets are established from the Radiological Society of North America (RSNA) pediatric bone age challenge, where the specific feature regions of the articular surface and epiphysis are manually segmented from hand radiography. Five convolutional neural networks, i.e., ResNet50, SENet, DenseNet-121, EfficientNet-b4, and CSPNet, are employed to improve the accuracy and efficiency of bone age diagnosis in clinical applications. Experiments show that the best-performing model can yield a mean absolute error (MAE) of 7.34 months on the proposed articular surface and epiphysis datasets, which is more accurate and fast than the radiologists. The project is available at https://github.com/YameiDeng/BAANet/, and the annotated dataset is also published at https://doi.org/10.5281/zenodo.7947923.
This article examines the solutions to the initial boundary value problem of connected Euler-Bernoulli beam equation, expressed as
{φtt(x,t)+φxxxx(x,t)+β(φ−ψ)(x,t)=0,(x,t)∈(0,L)×(0,+∞),ψtt(x,t)+ψxxxx(x,t)+β(ψ−φ)(x,t)=0,(x,t)∈(0,L)×(0,+∞), | (1.1) |
accompanied by the following boundary conditions:
{φ(0,t)=ψ(0,t)=φx(0,t)=ψx(0,t)=0,in(0,+∞),φxx(L,t)=ψxx(L)=ψxxx(L,t)=0,in(0,+∞),φxxx(L,t)=γ∂α,ηtφ(L,t),in(0,+∞), | (1.2) |
where γ and β are positive constants, and the following variables have specific engineering meanings: φ and ψ denote the vertical displacements, φt and ψt represent the velocities, φx and ψx signify the rotations, and φxt and ψxt indicate the angular velocities.
The system's initial conditions are specified as
φ(x,0)=φ0(x),φt(x,0)=φ1(x),ψ(x,0)=ψ0(x),ψt(x,0)=ψ1(x),x∈(0,L), | (1.3) |
where the initial data (φ0,φ1) and (ψ0,ψ1) belong to an appropriate function space.
The notation ∂α,ηt denotes the tempered Caputo time-fractional derivative of order α with respect to time t, where 0<α<1, as introduced in [1,2,3,4]. It is defined as
∂α,ηtf(t)=1Γ(1−α)∫t0(t−s)−αe−η(t−s)dfds(s)ds,η≥0, |
where Γ(⋅) represents the Gamma function.
The tempered Caputo time-fractional derivative, a generalization of the classical Caputo derivative, incorporates an exponential tempering factor that allows for a finite memory horizon, making it particularly suitable for processes with decaying memory effects. These exponentially modified fractional integro-differential operators were first proposed by Choi and MacCamy in [5], presenting a slightly different approach from the classical Caputo derivative formulated by Michele Caputo in [6]. Unlike the standard Caputo derivative, the tempered version is especially suited to applications where the memory of past states diminishes over time, as observed in fields such as anomalous diffusion and viscoelasticity. This choice provides both improved model stability and greater accuracy in representing transient dynamics, offering a more realistic approach to systems with decaying temporal correlations.
A basic model that describes the transverse vibration of a system of non-homogeneous connected Euler-Bernoulli beams, as presented in [7], is represented by the following system:
φtt(x,t)+φxxxx(x,t)=0,(x,t)∈(0,L)×(0,+∞). |
While this model provides a foundational description of transverse vibrations in such systems, it primarily addresses the standard beam dynamics without incorporating advanced effects such as memory and complex dissipation. To enhance the model's applicability to real-world scenarios, we extend the classical Euler-Bernoulli framework by introducing fractional derivatives in the boundary conditions. This extension allows for a more accurate representation of systems exhibiting memory effects and intricate dissipative behavior. By integrating these fractional derivatives, we capture more realistic dissipative dynamics and gain deeper insights into the stability properties of the system.
In recent years, the scientific community has shown growing interest in unraveling the intricate dynamics and practical applications of wave equations. The behavior of waves, whether occurring naturally, such as seismic waves in the Earth's crust, or in engineered systems, such as acoustic waves in materials, captivates both researchers and practitioners alike. Considerable research effort has been devoted to investigating wave equations with diverse damping types and exploring their stability and controllability. These waves emerge when a vibrating source disrupts the surrounding medium. Researchers have shown a keen interest in addressing damping-related challenges, whether local or global, and have illustrated different forms of stability.
In [4], B. Mbodje investigated the asymptotic behavior of solutions using the system:
{∂2tu(x,t)−u2x(x,t)=0,(x,t)∈(0,1)×(0,+∞),u(0,t)=0,∂xu(1,t)=−k∂α,ηtu(1,t),α∈(0,1),η≥0,k>0,u(x,0)=u0(x),∂tu(x,0)=v0(x). |
He established that the corresponding semigroup lacks exponential stability, showing only strong asymptotic stability. Additionally, the system's energy diminishes over time, approaching a decay proportional to t−1 as time extends to infinity.
In [8], Akil and Wehbe investigated a multidimensional wave equation featuring boundary fractional damping applied to a portion of the domain's boundary:
{utt(x,t)−Δu(x,t)=0,(x,t)∈Ω×(0,+∞),u(x,t)=0,(x,t)∈Γ0×(0,+∞),∂u∂ν(x,t)+γ∂α,ηtu(x,t)=0,(x,t)∈Γ1×(0,+∞). |
They demonstrated the system's strong stability while establishing that it lacked uniform stability. In addition, they derived a polynomial energy decay for smooth solutions of the form t−11−α. This analysis assumed specific geometric conditions for the boundary control region and leveraged the exponential decay of the wave equation with standard damping.
Recently, in [9], Beniani et al. examined a system comprising coupled wave equations featuring a diffusive internal control of a general nature:
{∂ttu−Δxu+ζ∫+∞−∞ϱ(ω)ϕ(x,ω,t)dω+βv=0,∂ttv−Δxv+ζ∫+∞−∞ϱ(ω)φ(x,ω,t)dω+βu=0,u=v=0,on∂Ω,ϕt(x,ω,t)+(ω2+η)ϕ(x,ω,t)−∂tuϱ(ω)=0,φt(x,ω,t)+(ω2+η)φ(x,ω,t)−∂tvϱ(ω)=0,u(x,0)=u0(x),∂tu(x,0)=u1(x),v(x,0)=v0(x),∂tv(x,0)=v1(x),ϕ(x,ω,0)=ϕ0(x,ω),andφ(x,ω,0)=φ0(x,ω). |
They showed the absence of exponential stability and explored the asymptotic stability of the model, establishing a general decay rate that depends on the density function ϱ. See also the references in [7,10,11,12,13] for further existing results related to the stability and numerical analysis of (1.1).
In this investigation, we extend the traditional Euler-Bernoulli beam model, commonly used to describe deformations under external loads, by incorporating fractional derivatives in the boundary conditions. This extension is essential, as fractional derivatives offer more precise models for systems exhibiting memory effects and complex dissipation phenomena frequently observed in practical applications such as the analysis of material microstructures. Incorporating fractional derivatives into the boundary conditions constitutes a novel advancement within the Euler-Bernoulli framework, enabling a more accurate representation of dissipative effects in diverse materials and structures. The findings from this study provide new insights into the stability characteristics of systems with fractional boundary dissipation, enhancing both theoretical understanding and the numerical experiments used to validate these observations.
The organization of the paper is outlined below: In Section 2, we transform the system (1.1) into an augmented system by integrating the wave equations with compatible diffusion equations. In Section 3, we derive global solutions to the problem (1.1) using arguments based on semigroup theory. In Section 4, we establish the strong stability of our system, despite the lack of resolvent compactness, by applying general criteria from Arendt and Batty. In Section 5, we propose a numerical scheme capable of replicating various decay rate profiles and confirm numerically the known stability outcomes related to the energy decay through several examples.
This section focuses on transforming the model (1.1) into an augmented system. The following theorem is required:
Theorem 2.1. [4] Consider μ be the function given by
μ(ξ)=|ξ|(2α−1)/2,ξ∈R,0<α<1. |
Then the relationship between the "input" U and the "output" O of the system
∂tφ(ξ,t)+(|ξ|2+η)φ(ξ,t)−U(t)μ(ξ)=0,ξ∈R,η≥0,t>0, |
φ(ξ,0)=0, |
and
O(t)=sin(απ)π∫Rμ(ξ)φ(ξ,t)dξ, |
can be described as follows:
O=I1−α,ηU=Dα,ηU, |
where
[Iα,ηU](t)=1Γ(α)∫t0(t−τ)α−1e−η(t−τ)U(τ)dτ. |
Lemma 2.1. [4] If λ∈D={λ∈C/ℜλ+η>0}∪{ℑλ≠0}, then
F1(λ):=∫+∞−∞μ2(ξ)λ+η+ξ2dξ=πsinαπ(λ+η)α−1, |
and
F2(λ):=∫+∞−∞μ2(ξ)(λ+η+ξ2)2dξ=(1−α)πsinαπ(λ+η)α−2. |
Based on the previous theorem, the system (1.1) can be expressed as the following augmented model:
{φtt(x,t)+φxxxx+β(φ−ψ)(x,t)=0,(x,t)∈(0,L)×(0,+∞),ψtt(x,t)+ψxxxx+β(ψ−φ)(x,t)=0,(x,t)∈(0,L)×(0,+∞),∂tϕ(ξ,t)+(ξ2+η)ϕ(ξ,t)−φt(L,t)μ(ξ)=0,(ξ,t)∈R×(0,+∞),φxx(L,t)=ψxx(L,t)=ψxxx(L,t)=0,t∈(0,+∞),φ(0,t)=ψ(0,t)=φx(0,t)=ψx(0,t)=0,t∈(0,+∞),φxxx(L,t)=ζ∫+∞−∞μ(ξ)ϕ(ξ,t)dξ,t∈(0,+∞), | (2.1) |
where ζ=γπ−1sin(απ). The energy associated with the solution to the system (2.1) is expressed by
E(t)=12(‖φt‖22+‖φxx‖22+‖ψt‖22+‖ψxx‖22+β‖ψ−φ‖22+ζ‖ϕ‖22), | (2.2) |
where ‖⋅‖2 refers to the L2(0,L) norm.
Then, we have the following lemma:
Lemma 2.2. Given that (φ,ψ,ϕ) is a solution to the problem (2.1), the energy functional defined by (2.2) fulfills
ddtE(t)=−ζ∫+∞−∞(ξ2+η)ϕ2(ξ,t)dξ≤0. | (2.3) |
Proof. Multiplying Eqs (2.1)1 and (2.1)2 by φt and ψt, respectively, using integration by parts over (0,L) and Eq (2.1)6, and combining the two equations, we derive
12ddt[‖φt‖22+‖φxx‖22+‖ψt‖22+‖ψxx‖22+β‖ψ−φ‖22]+ζφt(L)∫+∞∞μ(ξ)ϕ(ξ,t)dξ=0. | (2.4) |
Scaling Eq (2.1)3 by ζϕ and integrating over R yields
ζ2ddt‖ϕ‖22+ζ∫+∞−∞(ξ2+η)ϕ2(ξ,t)dξ−ζφt(L)∫+∞−∞μ(ξ)ϕ(ξ,t)dξ=0. | (2.5) |
By combining (2.4) and (2.5), we obtain (2.3). The proof of the lemma is thereby finished.
Within this section, we establish the well-posedness of (2.1). For this purpose, we define the following Hilbert space (the energy space):
H=(H2L(0,L)×L2(0,L))2×L2(−∞,+∞), |
where
H2L(0,L)={u∈H2(0,L):u(0)=ux(0)=0}. |
For U=(φ,u,ψ,v,ϕ)T and ¯U=(¯φ,¯u,¯ψ,¯v,¯ϕ)T, we introduce the inner product in the space H as follows:
⟨U,¯U⟩H=∫L0(u¯u+φxx¯φxx)dx+∫L0(v¯v+ψxx¯ψxx)dx+ζ∫+∞−∞ϕ¯ϕdξ. |
We then reformulate (2.1) in the framework of semigroup theory: By introducing the vector function U=(φ,u,ψ,v,ϕ)T, the system (2.1) can be expressed as
{U′=AU,t>0,U(0)=U0, |
where U0:=(φ0,φ1,ψ0,ψ1,ϕ0)T. The operator A is linear and specified as follows:
A(φuψvϕ)=(u−φxxxx−β(ψ−φ)v−ψxxxx−β(φ−ψ)−(ξ2+η)ϕ(ξ,t)+u(L,t)μ(ξ)). |
The domain of A is then
D(A)={(φ,u,ψ,v,ϕ)T∈H:φ,ψ∈H4(0,L)∩H2L(0,L),u,v∈H2L(0,L),|ξ|ϕ∈L2(−∞,+∞),−(ξ2+η)ϕ+u(L)μ(ξ)∈L2(−∞,+∞),φxx(L)=ψxx(L)=ψxxx(L)=0,φxxx(L)−ζ∫+∞−∞μ(ξ)ϕ(ξ)dξ=0}. | (3.1) |
We state the following theorem on existence and uniqueness:
Theorem 3.1. (1) If U0∈D(A), then there exists a unique, strong solution to the system (2.1):
U∈C0(R+,D(A))∩C1(R+,H). |
(2) If U0∈H, then system (2.1) admits a unique weak solution U∈C0(R+,H).
Proof. Initially, we prove that the operator A exhibits dissipative properties. For any U∈D(A), where U=(φ,u,ψ,v,ϕ)T, we find
R⟨AU,U⟩H=−ζ∫+∞−∞(ξ2+η)|ϕ(ξ,t)|2dξ≤0. | (3.2) |
Thus, A is dissipative.
Next, we prove that the operator λI−A is surjective for λ>0.
Given F=(f1,f2,f3,f4,f5)∈H, we verify the existence of U∈D(A) such that
{λφ−u=f1,λu+φxxxx+β(ψ−φ)=f2,λψ−v=f3,λv+ψxxxx+β(φ−ψ)=f4,λϕ+(ξ2+η)ϕ(ξ,t)−u(L,t)μ(ξ)=f5. | (3.3) |
Then, from (3.3)1 and (3.3)3, we find that
{u=λφ−f1,v=λψ−f3. | (3.4) |
It is straightforward to see that u∈H2L(0,L). Furthermore, from (3.3)5, we can identify ϕ as
ϕ=f5+u(L,t)μ(ξ)λ+ξ2+η. | (3.5) |
By inserting (3.4)1 into (3.3)2 and (3.4)2 into (3.3)4, we obtain
{λ2φ+φxxxx+β(ψ−φ)=λf1+f2,λ2ψ+ψxxxx+β(φ−ψ)=λf3+f4. | (3.6) |
Finding the solution to system (3.6) is the same as identifying φ,ψ∈H4(0,L)∩H2L(0,L) such that
{∫L0(λ2φ¯w+φxxxx¯w+β(ψ−φ)¯w)dx=∫L0(λf1+f2)¯wdx,∫L0(λ2ψ¯χ+ψxxxx¯χ+β(φ−ψ)¯χ)dx=∫L0(λf3+f4)¯χdx, | (3.7) |
for all w,χ∈H2L(0,L). By utilizing Eqs (3.5) and (3.7) and then performing integration by parts, we arrive at
∫L0(λ2φ¯w+λ2ψχ+φxx¯wxx+ψxx¯χxx+β(φ−ψ)(¯w−¯χ))dx+¯ζu(L)¯w(L)=∫L0(λf1+f2)¯w+∫L0(λf3+f4)¯χdx−ζ¯w(L)∫+∞−∞μ(ξ)ξ2+η+λf5(ξ)dξ, | (3.8) |
where ¯ζ=ζ∫+∞−∞μ2(ξ)ξ2+η+λdξ. By applying (3.4) once more, we infer that
{u(L)=λφ(L)−f1(L),v(L)=λψ(L)−f3(L). | (3.9) |
Substituting (3.9) into (3.8), we arrive at
∫L0(λ2φ¯w+λ2ψ¯χ+φxx¯wxx+ψxx¯χxx+β(φ−ψ)(¯w−¯χ))dx+¯ζλφ(L)¯w(L)=∫L0(λf1+f2)¯wdx+∫L0(λf3+f4)¯χdx−ζ¯w(L)∫+∞−∞μ(ξ)ξ2+η+λf5(ξ)dξ+¯ζf1(L)¯w(L). | (3.10) |
Consequently, the equivalent form of problem (3.10) is given by the following:
a((φ,ψ),(w,χ))=L(w,χ), | (3.11) |
where the sesquilinear form a:[H2L(0,L)×H2L(0,L)]2→C and the antilinear form L:[H2L(0,L)]2→C are expressed as
a((φ,ψ),(w,χ))=∫L0(λ2φ¯w+λ2ψ¯χ+φxx¯wxx+ψxx¯χxx+β(φ−ψ)(¯w−¯χ))dx+¯ζλφ(L)¯w(L) |
and
L(w,χ)=∫L0(λf1+f2)¯wdx+∫L0(λf3+f4)¯χdx−ζ¯w(L)∫+∞−∞μ(ξ)ξ2+η+λf5(ξ)dξ+¯ζf1(L)¯w(L). |
One can readily verify that the bilinear form a is continuous and coercive, while L is continuous. Therefore, by applying the Lax-Milgram theorem, we deduce that for all w,χ∈H2L(0,L), the problem (3.11) admits a unique solution (φ,ψ)∈H2L(0,L)×H2L(0,L). Furthermore, by invoking classical elliptic regularity, it follows from (3.10) that (φ,ψ)∈[H4(0,L)]2. As a result, the operator λI−A is surjective for any λ>0. As a result, the proof of Theorem 3.1 is completed by applying the Hille-Yosida theorem.
In this part, we make use of a general criterion of Arendt and Batty (see [14] or [15]), which states that a C0-semi-group of contractions eAt on a Banach space is strongly stable if A has no purely imaginary eigenvalues and σ(A)∩iR comprises only a countable set of elements. We express our primary result in the form of the following theorem.
Theorem 4.1. [16] The C0-semigroup eAt is strongly stable in H; i.e., for all U0∈H, the solution of (2.1) satisfies
limt⟶∞‖eAtU0‖H=0. |
To prove Theorem 4.1, we need the lemmas listed below:
Lemma 4.1. A does not have eigenvalues on iR.
Proof. We consider two cases: iλ=0 and iλ≠0.
Case 1. Solving AU=0 under the boundary conditions given in (3.1) leads to the conclusion that U=0. Therefore, iλ=0 is not an eigenvalue of the operator A.
Case 2. We proceed by contradiction. Assume there exists λ∈R,λ≠0, and U=(φ,u,ψ,v,ϕ1,ϕ2)≠0, for which AU=iλU. This gives the following system:
{iλφ−u=0,iλu+φxxxx+β(ψ−φ)=0,iλψ−v=0,iλv+ψxxxx+β(φ−ψ)=0,iλϕ+(ξ2+η)ϕ(ξ,t)−u(L,t)μ(ξ)=0. | (4.1) |
From (3.2), we deduce that ϕ≡0. Consequently, from (4.1)5, we derive u(L)=0.
Consequently, from (4.1)1 and (3.1), we obtain
φ(L)=ψ(L)=φxxx(L)=ψxxx(L)=0. |
Inserting (4.1)1 into (4.1)3 and (4.1)2 into (4.1)4, we obtain
{−λ2φ+φxxxx+β(ψ−φ)=0,−λ2ψ+ψxxxx+β(φ−ψ)=0. |
Then, Φ=φ+ψ and Ψ=φ−ψ satisfy
{−λ2Φ+Φxxxx=0,−(λ2+2β)Ψ+Ψxxxx=0. | (4.2) |
By applying Lemma 5.2 and Corollary 5.1 from [7], we obtain the following boundary conditions:
Φ(L)=Φx(L)=Φxx(L)=Φxxx(L)=0. | (4.3) |
Similarly, the same conditions hold for Ψ, specifically:
Ψ(L)=Ψx(L)=Ψxx(L)=Ψxxx(L)=0. | (4.4) |
Now, consider X=(Φ,Φx,Φxx,Φxxx,Ψ,Ψx,Ψxx,Ψxxx). We can rewrite (4.2)–(4.4) as the initial value problem:
{ddxX=BX,X(L)=0, | (4.5) |
where
B=(010000100001λ20000⋯0⋮⋱⋮0⋯00⋯0⋮⋱⋮0⋯0010000100001λ2+2β000). |
By virtue of the Picard theorem for ordinary differential equations, the system (4.5) has a unique solution X=0. Hence, Φ=Ψ≡0. Therefore, φ=0 and ψ=0, i.e., U=0. As a result, A does not have purely imaginary eigenvalues.
Lemma 4.2. For λ≠0 or λ=0 and η≠0, the operator iλI−A is surjective.
Proof. We distinguish the following cases:
Case 1: λ≠0.
Our goal is to prove that the operator iλI−A is surjective for λ≠0. To this end, let F=(f1,f2,f3,f4,f5)∈H. We aim to identify X=(u,φ,v,ψ,ϕ)∈D(A) such that the following equation holds:
(iλI−A)X=F. |
Alternatively,
{iλφ−u=f1,iλu+φxxxx+β(ψ−φ)=f2,iλψ−v=f3,iλv+ψxxxx+β(φ−ψ)=f4,iλϕ+(ξ2+η)ϕ(ξ,t)−u(L,t)μ(ξ)=f5. | (4.6) |
Inserting (4.6)1 into (4.6)2 and (4.6)3 into (4.6)4, we obtain
{−λ2φ+φxxxx+β(ψ−φ)=f2+iλf1,−λ2ψ+ψxxxx+β(φ−ψ)=f4+iλf3. | (4.7) |
Finding a solution to system (4.7) is the same as finding φ,ψ∈H4(0,L)∩H2L(0,L) such that
{∫L0(−λ2φ¯w+φxxxx¯w+β(ψ−φ)¯w)dx=∫L0(iλf1+f2)¯wdx,∫L0(−λ2ψ¯χ+ψxxxx¯χ+β(φ−ψ)¯χ)dx=∫L0(iλf3+f4)¯χdx, | (4.8) |
for all w,χ∈H2L(0,L). Using (4.8) and (3.5), followed by integration by parts, yields
∫L0(−λ2φ¯w−λ2ψ¯χ+φxx¯wxx+ψxx¯χxx−β(φ−ψ)(¯w−¯χ))dx+¯ζλφ(L)¯w(L)=∫L0(iλf1+f2)¯w+∫L0(iλf3+f4)¯χdx−ζ¯w(L)∫+∞−∞μ(ξ)ξ2+η+iλf5(ξ)dξ+¯ζf1(L)¯w(L), | (4.9) |
where ¯ζ=ζ∫+∞−∞μ2(ξ)ξ2+η+λdξ. Consequently, problem (4.9) can be reduced to the problem
a((φ,ψ),(w,χ))=L(w,χ), | (4.10) |
where the sesquilinear form a:[H2L(0,L)×H2L(0,L)]2→C and the antilinear form L:[H2L(0,L)]2→C are given by
a((φ,ψ),(w,χ))=∫L0(−λ2φ¯w−λ2ψ¯χ+φxx¯wxx+ψxx¯χxx−β(φ−ψ)(¯w−¯χ))dx |
and
L(w,χ)=∫L0(iλf1+f2)¯w+∫L0(iλf3+f4)¯χdx−ζ¯w(L)∫+∞−∞μ(ξ)ξ2+η+iλf5(ξ)dξ+¯ζf1(L)¯w(L), |
where
H2L(0,L)={(φ,ψ)∈H10(0,L)×H2(0,L)/φxxx(L,t)=γπ−1sin(απ)∫+∞−∞μ(ξ)ϕ(ξ)dξ,andψxxx(L,t)=0}. |
It is straightforward to confirm that the bilinear form a is continuous and coercive and that the functional L is continuous. By implementing the Lax-Milgram theorem, we gather that for all w,χ∈H2L(0,L), the problem (4.10) admits a unique solution (φ,ψ)∈H2L(0,L)×H2L(0,L). Furthermore, using classical elliptic regularity, it results from (4.9) that (φ,ψ)∈[H4(0,L)]2. Accordingly, the operator λI−A is surjective for any λ>0. Hence, by invoking the Hille–Yosida theorem, we obtain the desired result.
Case 2: λ=0 and η≠0.
The problem in system (4.6) can be expressed as
{−u=f1,φxxxx+β(ψ−φ)=f2,−v=f3,ψxxxx+β(φ−ψ)=f4,(ξ2+η)ϕ(ξ,t)−u(L,t)μ(ξ)=f5. |
Then, from (4.9), we obtain
∫L0(φxx¯wxx+ψxx¯χxx−β(φ−ψ)(¯w−¯χ))dx+¯ζλφ(L)¯w(L)=∫L0f2¯wdx+∫L0f4¯χdx−ζ¯w(L)∫+∞−∞μ(ξ)ξ2+ηf5(ξ)dξ+¯ζf1(L)¯w(L), | (4.11) |
for all w,χ∈H2L(0,L).
Thus, the system (4.11) can be expressed as the problem:
aη((u,v),(χ,ζ))=Lη(χ,ζ), | (4.12) |
where the sesquilinear form aη:[H2L(0,L)×H2L(0,L)]2→C and the antilinear form Lη:[H2L(0,L)]2→C are given by
aη((φ,ψ),(w,χ))=∫L0(φxx¯wxx+ψxx¯χxx−β(φ−ψ)(¯w−¯χ))dx |
and
Lη(w,χ)=∫L0f2¯w+∫L0f4¯χdx−ζ¯w(L)∫+∞−∞μ(ξ)ξ2+ηf5(ξ)dξ+¯ζf1(L)¯w(L). |
It is straightforward to confirm that aη is continuous and coercive, and Lη is continuous. Then, by Lax-Milgram's theorem, the variational problem (4.12) admits a unique solution (φ,ψ)∈H2L(0,L)×H2L(0,L). We then deduce from (4.11) that (φ,ψ)∈[H4(0,L)]2. Hence, the operator A is surjective.
The proof of the Lemma is thus complete.
Proof of Theorem 4.1. According to Lemma 4.1, the operator A lacks purely imaginary eigenvalues. Additionally, Lemma 4.2 establishes that R(iλ−A)=H for all λ∈R∗ and R(iλ−A)=H for λ=0 when η>0. Consequently, by the closed graph theorem of Banach, we conclude that σ(A)∩iR=∅ for η>0, and σ(A)∩iR={0} when η=0. This completes the proof.
We will start making use of finite difference methods (FDM) to derive a discrete representation of the solution of (2.1). More precisely, we use the classical finite difference discretization for the time variable and the implicit compact finite difference method of fourth-order for discretization of the space variable. The comparison between different finite difference methods (implicit, explicit, and semi-implicit) will be considered in a subsequent work.
Consider the discrete domain Ωh of Ω=[0,L], with a uniform grid given by xi=iΔx,i=0,1,…,M,(M≥5) (see Figure 1). The time discretization of the interval I=[0,T] is expressed as tj=jΔt,j=0,1,…,N; Δt=tj−tj−1=TN, where M and N are positive integers. Denote by φ(xi;tj)=φji and ψ(xi;tj)=ψji the value of the functions φ and ψ evaluated at the point xi and the instant tj.
Now, we define the following approximation of the derivatives of φ and ψ, respectively:
φtt(xi,tj+1)≈φj+1i−2φji+φj−1iΔt2, | (5.1) |
ψtt(xi,tj+1)≈ψj+1i−2ψji+ψj−1iΔt2, | (5.2) |
φxxxx(xi,tj+1)≈φj+1i−2−4φj+1i−1+6φj+1i−4φj+1i+1+φj+1i+2Δx4, | (5.3) |
and
ψxxxx(xi,tj+1)≈ψj+1i−2−4ψj+1i−1+6ψj+1i−4ψj+1i+1+ψj+1i+2Δx4. | (5.4) |
Then, using (5.1)–(5.4), we obtain the discrete formulation of the initial condition of the system (1.1) as follows:
{φj+1i−2φji+φj−1iΔt2+φj+1i−2−4φj+1i−1+6φj+1i−4φj+1i+1+φj+1i+2Δx4+β(φj+1i−ψj+1i)=0,ψj+1i−2ψji+ψj−1iΔt2+ψj+1i−2−4ψj+1i−1+6ψj+1i−4ψj+1i+1+ψj+1i+2Δx4+β(ψj+1i−φj+1i)=0, | (5.5) |
for j=¯1,N−1, and i=¯2,M−2. The discrete formulation of the initial conditions (1.3) reads as follows:
φ0i=φ0(xi),ψ0i=ψ0(xi),φ1i−φ0iΔt=φ1(xi),ψ1i−ψ0iΔt=ψ1(xi), | (5.6) |
for i=¯0,M. Next, we seek the discrete formulation of the boundary conditions.
Lemma 5.1. The discrete formulation of the boundary conditions (1.2)3 reads as follows:
φj+1M−3φj+1M−1+3φj+1M−2−φj+1M−3Δx3=ηα−1γΔtΓ(1−α)j∑k=0(φk+1M−φkM)(Γ(1−α,η(tj+1−tk))−Γ(1−α,η(tj+1−tk+1))). | (5.7) |
Proof. For j=¯1,N−1, we have φxxx(L,tj+1)=γ∂α,ηtφ(L,tj+1). By using finite differences, we arrive at
φj+1M−3φj+1M−1+3φj+1M−2−φj+1M−3Δx3=γΓ(1−α)j∑k=0∫tk+1tk(tj+1−r)−αe−η(tj+1−r)φk+1M−φkMΔtdr=γΔtΓ(1−α)j∑k=0(φk+1M−φkM)∫tk+1tk(tj+1−r)−αe−η(tj+1−r)dr. | (5.8) |
On the other hand, using the change of variable u=η(tj+1−r), we obtain
∫tk+1tk(tj+1−r)−αe−η(tj+1−r)dr=ηα−1∫η(tj+1−tk)η(tj+1−tk+1)u1−(1+α)e−udu.=ηα−1(Γ((1−α,η(tj+1−tk))−Γ(1−α,η(tj+1−tk+1))). | (5.9) |
Inserting (5.9) in (5.8), we obtain (5.7).
Hence, the discrete formulation of boundary condition in the system (1.1) follows as
{φj0=0,ψj0=0,φj+11−φj+10Δx=0,ψj+11−ψj+10Δx=0,φj+1M−2φj+1M−1+φj+1M−2Δx2=0,ψj+1M−2ψj+1M−1+ψj+1M−2Δx2=0,φj+1M−3φj+1M−1+3φj+1M−2−φj+1M−3Δx3=ηα−1γΔtΓ(1−α)j∑k=0(φk+1M−φkM)(Γ(1−α,η(tj+1−tk))−Γ(1−α,η(tj+1−tk+1))),ψj+1M−3ψj+1M−1+3ψj+1M−2−ψj+1M−3Δx3=0. |
In the following, we reformulate the system (5.5)-(5.6) in abstract form as follows:
K×W=F, |
where we denoted by W=(φj+10,φj+11,⋯,φj+1M,ψj+10,ψj+11,⋯,ψj+1M)t,
F=(φj+10,φj+11,2φj2−φj−12Δt2,⋯,2φjM−2−φj−1M−2Δt2,0,λ,ψj+10,ψj+11,2ψj2−ψj−12Δt2,⋯,2ψjM−2−ψj−1M−2Δt2,0,0)t |
with
λ=ηα−1γΔtΓ(1−α)j−1∑k=0(φk+1M−φkM)(Γ(1−α,η(tj+1−tk))−Γ(1−α,η(tj+1−tk+1)))−ηα−1φjMγΓ(1−α)Δt(Γ(1−α,ηΔt)) |
and K=(A+CBBA)∈M2M(R), where A is an M×M matrix:
A=[1000000⋯000100000⋯00c1−4c1c2−4c1c100⋯000c1−4c1c2−4c1c10⋯0000c1−4c1c2−4c1c1⋯00000⋱⋱⋱⋱⋱⋱⋮⋮⋮⋮⋱⋱⋱⋱⋱⋱⋮0000⋯c1−4c1c2−4c1c10000⋯00c3−2c3c30000⋯0−c43c4−3c4c4] |
with c1=1Δx4,c2=β+1Δt2+6c1,c3=1Δx2,c4=1Δx3 and the matrices B and C are given as follows:
Bij={−β,if i=j and 3≤i≤M−2,0,otherwise, |
Cij={−ηα−1γΓ(1−α)Δt(Γ(1−α,ηΔt)),if i=j=M,0,otherwise. |
Algorithm 1 provides a summary of all the steps for the calculation of φj and ψj using matrix decomposition techniques:
Algorithm 1. Compute φj,ψj from [A+CBBA][φjψj]=[f1f2] |
Require: Matrices A, C, B, {φ0,φ1}, {ψ0,ψ1}, vectors f1, f2 |
Ensure: φj+1, ψj+1 |
for j=1 to N−1 do |
Update f1 and f2 for the current iteration j+1 |
φj+1=(A+C−BA−1B)−1(f1−BA−1f2) |
ψj+1=(A−B(A+C)−1B)−1(f2−B(A+C)−1f1) |
end for |
Recalling that the energy E(t) is expressed as
E(t)=12(‖φt‖22+‖φxx‖22+‖ψt‖22+‖ψxx‖22+β‖ψ−φ‖22). |
Now, we define the following approximations of the derivatives of φ and ψ, respectively:
φt≈φj+1i−φjiΔt,ψt≈ψj+1i−ψjiΔt,φxx≈φji+1−2φji+φji−1Δx2, and ψxx≈ψji+1−2ψji+ψji−1Δx2. |
The L2 norm of a discretized function is approximated by
‖f‖22≈ΔxN∑i=0f(xi,tj)2. | (5.10) |
Thus, the discrete energy of the system (1.1)-(1.2) at time tj is as follows:
E(tj)≈12Δx∑Mi=0((φj+1i−φjiΔt)2+(φji+1−2φji+φji−1Δx2)2+(ψj+1i−ψjiΔt)2+(ψji+1−2ψji+ψji−1Δx2)2+β(φji−ψji)2). |
Algorithm 2 summarizes all the steps for calculating of the discrete Energy E(t) using the L2 norm define by (5.10).
Algorithm 2. Calculation of the discrete energy E(tj) |
Require: {φj}, {ψj} for each j, Δt, Δx, β |
Ensure: Ej (calculated energy) |
for j=0 to N−1 do |
Compute time derivatives |
φj+1t←φj+1−φjΔt |
ψj+1t←ψj+1−ψjΔt |
Compute second spatial derivatives |
for i=1 to M−1 do |
(φxx)j+1i←φj+1i+1−2φj+1i+φj+1i−1Δx2 |
(ψxx)j+1i←ψj+1i+1−2ψj+1i+ψj+1i−1Δx2 |
end for |
Compute L2 norms at time step j |
norm_phi_t←Δx∑(φj+1t)2 |
norm_phi_xx←Δx∑(φj+1xx)2 |
norm_psi_t←Δx∑(ψj+1t)2 |
norm_psi_xx←Δx∑(ψj+1xx)2 |
norm_diff←Δx∑((ψj+1−φj+1)2) |
Compute energy |
Ej+1←0.5×(norm_phi_t+norm_phi_xx+norm_psi_t+norm_psi_xx+β×norm_diff) |
end for |
To evaluate the asymptotic properties of the solutions to the system (1.1), we consider the following data: α=0.5, β=η=1, Δt=10−2,Δx=10−1, L=1, and the initial conditions:
φ(x,0)=3x2(1−x),φt(x,0)=0,ψ(x,0)=x(1−x)2,ψt(x,0)=0,x∈(0,L). |
Figures 2–4 illustrate the decay of the discrete energy for the chosen initial data. Specifically, Figure 3 demonstrates that when T is sufficiently large, the decay becomes more rapid. Figure 5 demonstrates the exponential decay of the discrete energy for the chosen initial data, as well as for other initial conditions. The conjecture of exponential decay arises from the observation that the plot of the logarithm of the discrete energy forms a straight line with a negative slope, indicating an exponential decay pattern. This behavior has been consistently observed for various initial data, further reinforcing the conjecture. This observation also poses an open question regarding the exponential decay of the energy in the system (1.1).
Figures 6 and 7 illustrate the exponential decay of the discrete energy for the chosen initial data and for other initial conditions as well, for different values of α: α=0.75 and α=0.95. We observe a slight increase in the decay rate as α approaches 1.
Now, we provide an example to test the correctness and effectiveness of the numerical scheme. This example includes a comparison with the exact solution to demonstrate the scheme's convergence rate and validate its accuracy. These results ensure the reliability of the analysis and conclusions presented in Section 5.2.
We consider the exact solutions to the problem as follows:
φ(x,t)=(x−1)4t3andψ(x,t)=xet. |
The problem is governed by the equations:
{φtt(x,t)+φxxxx(x,t)+(φ−ψ)(x,t)=f1(x,t),(x,t)∈(0,1)×(0,+∞),ψtt(x,t)+ψxxxx(x,t)+(ψ−φ)(x,t)=f2(x,t),(x,t)∈(0,1)×(0,+∞), |
with the following boundary conditions:
{φ(0,t)=t3,ψ(0,t)=0,φx(0,t)=−4t3,ψx(0,t)=et,fort∈(0,+∞),φxx(1,t)=ψxx(1,t)=ψxxx(1,t)=0,fort∈(0,+∞),φxxx(1,t)=0,fort∈(0,+∞), |
and the initial conditions:
{φ(x,0)=0,φt(x,0)=0,x∈(0,1),ψ(x,0)=x,ψt(x,0)=x,x∈(0,1). |
The source terms are given by
f1(x,t)=t3(x−1)4+24t3+6t(x−1)4−xetandf2(x,t)=−t3(x−1)4+2xet. |
Figures 8 and 9 show the comparison between the exact solutions and the numerical solutions for φ and ψ.
Now, we define the root mean square error (RMSE) as
RMSE=√1Nm∑i=0n∑j=0(uji−ˆuji)2, | (5.11) |
where u and ˆu denote the exact and the numerical solutions, respectively.
The data presented in Table 1 below were obtained by comparing the exact solutions φ and ψ with their respective numerical approximations ˆφ and ˆψ using different discretization values of Δx and Δt.
Δx | Δt | RMSEφ | RMSEψ |
0.2 | 0.1 | 0.066421 | 0.000768 |
0.1 | 0.01 | 0.026582 | 0.000373 |
0.1 | 0.001 | 0.026449 | 0.000434 |
Finally, to present the rate of convergence, we begin by defining the pointwise rate of convergence, which corresponds to the opposite of the slope of the logarithmic energy curve. This calculation is performed for the example provided at the beginning of Section 5.2.
Define the pointwise rate of convergence as
λj=−ln(E(tj+1))−ln(E(tj))Δt,for j=0,…,N−1, |
and the rate of convergence as
λ≈1N−1N−1∑i=1λi. |
In Figure 10, we observe the behavior of the decay rate over time. We note that the curve becomes horizontal, indicating that the rate is constant, which supports the notion that the energy decay is exponential. In Table 2, we provide the values of the energy decay rate for different values of Δx and Δt.
Δx | Δt | λ |
0.1 | 0.1 | 2.302732 |
0.1 | 0.01 | 2.425248 |
0.01 | 0.01 | 2.382123 |
0.01 | 0.001 | 2.412021 |
In the present work, we studied the strong stabilization of a coupled Euler-Bernoulli beam system with one boundary dissipation of fractional derivative type. First, we analyzed the strong stability of the system (1.1)-(1.2). Next, we employed a finite difference scheme to compute the numerical solutions and demonstrated the stability of the discrete energy. Although we achieved exponential energy decay in one example, the question of exponential energy decay remains open.
In advancing the finite difference methods (FDM) for solving complex fractional and coupled PDEs, this manuscript makes a considerable impact on numerical analysis and applied mathematics, offering a significant resource for researchers and practitioners working on advanced modeling challenges.
Boumediene Boukhari, Ahmed Bchatnia and Abderrahmane Beniani: Visualization; Foued Mtiri: Investigation; Ahmed Bchatnia: Supervision, Validation; Ahmed Bchatnia and Abderrahmane Beniani: Conceptualization, Methodology; Boumediene Boukhari, Foued Mtiri and Abderrahmane Beniani: Resources; Boumediene Boukhari, Foued Mtiri, Ahmed Bchatnia and Abderrahmane Beniani: Writing–original draft. All authors have read and approved the final version of the manuscript for publication.
The authors extend their appreciation to the Deanship of Research and Graduate Studies at King Khaled University for funding this work through Large Research Project under grant number RGP2/368/45.
The authors declare that there are no conflicts of interest in this paper.
[1] | P. Hao, S. Chokuwa, X. Xie, F. Wu, J. Wu, C. Bai, Skeletal bone age assessments for young children based on regression convolutional neural networks, Math. Biosci. Eng., 16 (2019), 6454–6466. |
[2] |
W. Lin, F. Yang, Computational analysis of cutting parameters based on gradient voronoi model of cancellous bone, Math. Biosci. Eng., 19 (2022), 11657–11674. https://doi.org/10.3934/mbe.2022542 doi: 10.3934/mbe.2022542
![]() |
[3] | S. L Truesdell, M. M Saunders, Bone remodeling platforms: Understanding the need for multicellular lab-on-a-chip systems and predictive agent-based models, Math. Biosci. Eng., 17 (2020). |
[4] | A. Nordenstrom, H. Falhammar, Management of endocrine disease: Diagnosis and management of the patient with non-classic CAH due to 21-hydroxylase deficiency, Eur. J. Endocrinol., 180 (2019), R127–R145. |
[5] |
A. C. Morani, C. T. Jensen, M. A. Habra, M. M. Agrons, C. O. Menias, N. A. Wagner-Bartak, et al., Adrenocortical hyperplasia: a review of clinical presentation and imaging, Abdominal Radiol., 45 (2020), 917–927. https://doi.org/10.1007/s00261-019-02048-6 doi: 10.1007/s00261-019-02048-6
![]() |
[6] |
J.-C. Carel, J. Léger, Precocious puberty, New England J. Med., 358 (2008), 2366–2377. https://doi.org/10.1056/NEJMcp0800459 doi: 10.1056/NEJMcp0800459
![]() |
[7] |
K. Kyostila, J. E. Niskanen, M. Arumilli, J. Donner, M. K. Hytönen, H. Lohi, Intronic variant in POU1F1 associated with canine pituitary dwarfism, Human Gene., 140 (2021), 1553–1562. https://doi.org/10.1007/s00439-021-02259-2 doi: 10.1007/s00439-021-02259-2
![]() |
[8] |
K. C. Lee, K. H. Lee, C. H. Kang, K. S. Ahn, L. Y. Chung, J. J. Lee, et al., Clinical validation of a deep learning-based hybrid (Greulich-Pyle and modified Tanner-Whitehouse) method for bone age assessment, Korean J. Radiol., 22 (2021), 2017–2025. https://doi.org/10.3348/kjr.2020.1468 doi: 10.3348/kjr.2020.1468
![]() |
[9] | P. Lv, C. Zhang, Tanner–Whitehouse skeletal maturity score derived from ultrasound images to evaluate bone age, Eur. Radiol., 2022 (2022), 1–8. |
[10] | S. Zhang, L. Liu, The skeletal development standards of hand and wrist for Chinese Children¡ªChina 05 I. TW_3-C RUS, TW_3-C Carpal, and RUS-CHN methods, Chin. J. Sports Med., 2023 (2023), 6–13. |
[11] |
J. R. Kim, Y. S. Lee, J. Yu, Assessment of bone age in prepubertal healthy Korean children: comparison among the Korean standard bone age chart, Greulich-Pyle method, and Tanner-Whitehouse method, Korean J. Radiol., 16 (2015), 201–205. https://doi.org/10.3348/kjr.2015.16.1.201 doi: 10.3348/kjr.2015.16.1.201
![]() |
[12] |
C. Gao, Q. Qian, Y. Li, X. Xing, X. He, M. Lin, et al., A comparative study of three bone age assessment methods on Chinese preschool-aged children, Front. Pediatr., 10 (2022), 1–12. https://doi.org/10.3389/fped.2022.976565 doi: 10.3389/fped.2022.976565
![]() |
[13] |
H. Liu, M. Liu, D. Li, W. Zheng, L. Yin, R. Wang, Recent advances in pulse-coupled neural networks with applications in image processing, Electronics, 11 (2022), 3264. https://doi.org/10.3390/electronics11203264 doi: 10.3390/electronics11203264
![]() |
[14] |
Y. Ban, Y. Wang, S. Liu, B. Yang, M. Liu, L. Yin, et al., 2d/3d multimode medical image alignment based on spatial histograms, Appl. Sci., 12 (2022), 8261. https://doi.org/10.3390/app12168261 doi: 10.3390/app12168261
![]() |
[15] | S. Xiong, B. Li, S. Zhu, DCGNN: A single-stage 3D object detection network based on density clustering and graph neural network, Complex Intell. Syst., 2022 (2022), 1–10. |
[16] |
C. Spampinato, S. Palazzo, D. Giordano, M. Aldinucci, R. Leonardi, Deep learning for automated skeletal bone age assessment in X-ray images, Med. Image Anal., 36 (2017), 41–51. https://doi.org/10.1016/j.media.2016.10.010 doi: 10.1016/j.media.2016.10.010
![]() |
[17] |
D. B. Larson, M. C. Chen, M. P. Lungren, S. S. Halabi, N. V. Stence, C. P. Langlotz, Performance of a deep-learning neural network model in assessing skeletal maturity on pediatric hand radiographs, Radiology, 287 (2018), 313–322. https://doi.org/10.1148/radiol.2017170236 doi: 10.1148/radiol.2017170236
![]() |
[18] |
Y. Liu, C. Zhang, J. Cheng, X. Chen, Z. J. Wang, A multi-scale data fusion framework for bone age assessment with convolutional neural networks, Comput. Biol. Med., 108 (2019), 161–173. https://doi.org/10.1016/j.compbiomed.2019.03.015 doi: 10.1016/j.compbiomed.2019.03.015
![]() |
[19] |
Q. H. Nguyen, B. P. Nguyen, M. T. Nguyen, M. C. Chua, T. T. Do, N. Nghiem, Bone age assessment and sex determination using transfer learning, Expert Syst. Appl., 200 (2022), 116926. https://doi.org/10.1016/j.eswa.2022.116926 doi: 10.1016/j.eswa.2022.116926
![]() |
[20] |
X. Ren, T. Li, X. Yang, S. Wang, S. Ahmad, L. Xiang, et al., Regression convolutional neural network for automated pediatric bone age assessment from hand radiograph, IEEE J. Biomed. Health Inf., 23 (2018), 2030–2038. https://doi.org/10.1109/JBHI.2018.2876916 doi: 10.1109/JBHI.2018.2876916
![]() |
[21] |
C. Liu, H. Xie, Y. Zhang, Self-supervised attention mechanism for pediatric bone age assessment with efficient weak annotation, IEEE Trans. Med. Imaging, 40 (2020), 2685–2697. https://doi.org/10.1109/TMI.2020.3046672 doi: 10.1109/TMI.2020.3046672
![]() |
[22] |
N. E. Reddy, J. C. Rayan, A. V. Annapragada, N. F. Mahmood, A. E. Scheslinger, W. Zhang, et al., Bone age determination using only the index finger: a novel approach using a convolutional neural network compared with human radiologists, Pediatr. Radiol., 50 (2020), 516–523. https://doi.org/10.1007/s00247-019-04587-y doi: 10.1007/s00247-019-04587-y
![]() |
[23] |
S. Mutasa, P. D. Chang, C. Ruzal-Shapiro, R. Ayyala, MABAL: A novel deep-learning architecture for machine-assisted bone age labeling, J. Dig. Imaging, 31 (2018), 513–519. https://doi.org/10.1007/s10278-018-0053-3 doi: 10.1007/s10278-018-0053-3
![]() |
[24] | X. Li, Y. Jiang, Y. Liu, J. Zhang, S. Yin, H. Luo, RAGCN: Region aggregation graph convolutional network for bone age assessment from X-ray images, IEEE Trans. Instru. Measure., 71 (2022), 1–12. |
[25] |
S. S. Halabi, L. M. Prevedello, J. Kalpathy-Cramer, A. B. Mamonov, A. Bilbily, M. Cicero, et al., The RSNA pediatric bone age machine learning challenge, Radiology, 290 (2019), 498–503. https://doi.org/10.1148/radiol.2018180736 doi: 10.1148/radiol.2018180736
![]() |
[26] |
B. C. Russell, A. Torralba, K. P. Murphy, W. T. Freeman, LabelMe: a database and web-based tool for image annotation, Int. J. Comput. Vision, 77 (2008), 157–173. https://doi.org/10.1007/s11263-007-0090-8 doi: 10.1007/s11263-007-0090-8
![]() |
[27] | K. He, X. Zhang, S. Ren, J. Sun, Deep residual learning for image recognition, in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016,770–778. |
[28] | J. Hu, L. Shen, G. Sun, Squeeze-and-excitation networks, in Proceedings of the IEEE conference on computer vision and pattern recognition (CVPR), 2018, 7132–7141. |
[29] | G. Huang, Z. Liu, L. Van Der Maaten, K. Q. Weinberger, Densely connected convolutional networks, in Proceedings of the IEEE conference on computer vision and pattern recognition (CVPR), 2017, 4700–4708. |
[30] | M. Tan, Q. Le, Efficientnet: Rethinking model scaling for convolutional neural networks, in Proceedings of the International conference on machine learning (ICLR), 2019, 6105–6114. |
[31] | C. Y. Wang, H. Y. M. Liao, Y. H. Wu, P. Y. Chen, J. W. Hsieh, I. H. Yeh, CSPNet: A new backbone that can enhance learning capability of CNN, in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops (CVPR), 2020,390–391. |
[32] |
D. Giordano, C. Spampinato, G. Scarciofalo, R. Leonardi, An automatic system for skeletal bone age measurement by robust processing of carpal and epiphysial/metaphysial bones, IEEE Trans. Instru. Measure., 59 (2010), 2539–2553. https://doi.org/10.1109/TIM.2010.2058210 doi: 10.1109/TIM.2010.2058210
![]() |
[33] | D. Knapik, J. Sanders, A. Gilmore, D. Weber, D. Cooperman, R. Liu, A quantitative method for the radiological assessment of skeletal maturity using the distal femur, Bone Joint J., 100 (2018), 1106–1111. |
[34] | A. K. Bhat, A. M. Acharya, M. Pai, Radiology of the wrist and hand, in Clinical Examination of the Hand, CRC Press, (2022), 275–299. |
[35] |
K. S. Ahn, B. Bae, W. Y. Jang, J. H. Lee, J. H. Lee, S. Oh, B. H. Kim, et al., Assessment of rapidly advancing bone age during puberty on elbow radiographs using a deep neural network model, Eur. Radiol., 31 (2021), 8947–8955. https://doi.org/10.1007/s00330-021-08096-1 doi: 10.1007/s00330-021-08096-1
![]() |
[36] |
U. Nemec, S. F. Nemec, M. Weber, et al., Human long bone development in vivo: analysis of the distal femoral epimetaphysis on MR images of fetuses, Radiology, 267 (2013), 570–580. https://doi.org/10.1148/radiol.13112441 doi: 10.1148/radiol.13112441
![]() |
[37] | A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, et al., Attention is all you need, in Proceedings of the 31st International Conference on Neural Information Processing Systems (NIPS), 2017, 5998–6008. |
[38] |
Z. Q. Zhao, P. Zheng, S. Xu, X. Wu, Object detection with deep learning: A review, IEEE Trans. Neural Networks Learn. syst., 30 (2019), 3212–3232. https://doi.org/10.1109/TNNLS.2018.2876865 doi: 10.1109/TNNLS.2018.2876865
![]() |
[39] |
S. Ren, K. He, R. Girshick, J. Sun, Faster R-CNN: Towards real-time object detection with region proposal networks, IEEE Trans. Pattern Anal. Mach. Intell., 39 (2017), 1137–1149. https://doi.org/10.1109/TPAMI.2016.2577031 doi: 10.1109/TPAMI.2016.2577031
![]() |
[40] | A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, et al., An image is worth 16x16 words: Transformers for image recognition at scale, in Proceedings of the International Conference on Learning Representations (ICLR), 2020, 1–21. |
[41] | M. Roberts, D. Driggs, M. Thorpe, J. Gilbey, M. Yeung, S. Ursprung, et al., Common pitfalls and recommendations for using machine learning to detect and prognosticate for COVID-19 using chest radiographs and CT scans, Nat. Mach. Intell., 3 (2021), 199–217. |
[42] | Z. Liu, Y. Lin, Y. Cao, H. Hu, Y. Wei, Z. Zhang, et al., Swin transformer: Hierarchical vision transformer using shifted windows, in Proceedings of the IEEE/CVF international conference on computer vision (CVPR), 2021 (2021), 10012–10022. |
[43] | C. Y. Wang, A. Bochkovskiy, H. Y. M. Liao, Scaled-yolov4: Scaling cross stage partial network, in Proceedings of the IEEE/cvf conference on computer vision and pattern recognition (CVPR), (2021), 13029–13038. |
[44] | A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, et al., Pytorch: An imperative style, high-performance deep learning library, in Proceedings of the Advances in Neural Information Processing Systems (NIPS), (2019), 8026–8037. |
[45] |
L. Bottou, F. E. Curtis, J. Nocedal, Optimization methods for large-scale machine learning, Siam Rev., 60 (2018), 223–311. https://doi.org/10.1137/16M1080173 doi: 10.1137/16M1080173
![]() |
[46] | C. Gonzalez, M. Escobar, L. Daza, F. Torres, G. Triana, P. Arbelaez, SIMBA: Specific identity markers for bone age assessment, in Proceeding of Medical Image Computing and Computer Assisted Intervention (MICCAI), (2020), 753–763. |
[47] | C. Chen, Z. Chen, X. Jin, L. Li, W. Speier, C. W. Arnold, Attention-guided discriminative region localization and label distribution learning for bone age assessment, IEEE J. Biomed. Health Inf., 26 (2021), 1208–1218. |
[48] |
B. D. Lee, M. S. Lee, Automated bone age assessment using artificial intelligence: the future of bone age assessment, Korean J. Radiol., 22 (2021), 792–800. https://doi.org/10.3348/kjr.2020.0941 doi: 10.3348/kjr.2020.0941
![]() |
Δx | Δt | RMSEφ | RMSEψ |
0.2 | 0.1 | 0.066421 | 0.000768 |
0.1 | 0.01 | 0.026582 | 0.000373 |
0.1 | 0.001 | 0.026449 | 0.000434 |
Δx | Δt | λ |
0.1 | 0.1 | 2.302732 |
0.1 | 0.01 | 2.425248 |
0.01 | 0.01 | 2.382123 |
0.01 | 0.001 | 2.412021 |