Improved breast ultrasound tumor classification using dual-input CNN with GAP-guided attention loss

Xiao Zou; Jintao Zhai; Shengyou Qian; Ang Li; Feng Tian; Xiaofei Cao; Runmin Wang; Xiao Zou; Jintao Zhai; Shengyou Qian; Ang Li; Feng Tian; Xiaofei Cao; Runmin Wang

doi:10.3934/mbe.2023682

Mathematical Biosciences and Engineering

2023, Volume 20, Issue 8: 15244-15264. doi: 10.3934/mbe.2023682

Previous Article Next Article

Research article

Improved breast ultrasound tumor classification using dual-input CNN with GAP-guided attention loss

1.
School of Physics and Electronics, Hunan Normal University, Changsha 410081, China
2.
College of Information Science and Engineering, Hunan Normal University, Changsha 410081, China

Academic Editor: Vladimir Mityushev

Received: 29 May 2023 Revised: 29 June 2023 Accepted: 04 July 2023 Published: 20 July 2023

Ultrasonography is a widely used medical imaging technique for detecting breast cancer. While manual diagnostic methods are subject to variability and time-consuming, computer-aided diagnostic (CAD) methods have proven to be more efficient. However, current CAD approaches neglect the impact of noise and artifacts on the accuracy of image analysis. To enhance the precision of breast ultrasound image analysis for identifying tissues, organs and lesions, we propose a novel approach for improved tumor classification through a dual-input model and global average pooling (GAP)-guided attention loss function. Our approach leverages a convolutional neural network with transformer architecture and modifies the single-input model for dual-input. This technique employs a fusion module and GAP operation-guided attention loss function simultaneously to supervise the extraction of effective features from the target region and mitigate the effect of information loss or redundancy on misclassification. Our proposed method has three key features: (i) ResNet and MobileViT are combined to enhance local and global information extraction. In addition, a dual-input channel is designed to include both attention images and original breast ultrasound images, mitigating the impact of noise and artifacts in ultrasound images. (ii) A fusion module and GAP operation-guided attention loss function are proposed to improve the fusion of dual-channel feature information, as well as supervise and constrain the weight of the attention mechanism on the fused focus region. (iii) Using the collected uterine fibroid ultrasound dataset to train ResNet18 and load the pre-trained weights, our experiments on the BUSI and BUSC public datasets demonstrate that the proposed method outperforms some state-of-the-art methods. The code will be publicly released at https://github.com/425877/Improved-Breast-Ultrasound-Tumor-Classification.

Keywords:

Citation: Xiao Zou, Jintao Zhai, Shengyou Qian, Ang Li, Feng Tian, Xiaofei Cao, Runmin Wang. Improved breast ultrasound tumor classification using dual-input CNN with GAP-guided attention loss[J]. Mathematical Biosciences and Engineering, 2023, 20(8): 15244-15264. doi: 10.3934/mbe.2023682

Related Papers:

[1]	Mahmoud Saleh, Endre Kovács, Nagaraja Kallur . Adaptive step size controllers based on Runge-Kutta and linear-neighbor methods for solving the non-stationary heat conduction equation. Networks and Heterogeneous Media, 2023, 18(3): 1059-1082. doi: 10.3934/nhm.2023046
[2]	Danielle Hilhorst, Hideki Murakawa . Singular limit analysis of a reaction-diffusion system with precipitation and dissolution in a porous medium. Networks and Heterogeneous Media, 2014, 9(4): 669-682. doi: 10.3934/nhm.2014.9.669
[3]	Changli Yuan, Mojdeh Delshad, Mary F. Wheeler . Modeling multiphase non-Newtonian polymer flow in IPARS parallel framework. Networks and Heterogeneous Media, 2010, 5(3): 583-602. doi: 10.3934/nhm.2010.5.583
[4]	Robert Carlson . Myopic models of population dynamics on infinite networks. Networks and Heterogeneous Media, 2014, 9(3): 477-499. doi: 10.3934/nhm.2014.9.477
[5]	Verónica Anaya, Mostafa Bendahmane, David Mora, Ricardo Ruiz Baier . On a vorticity-based formulation for reaction-diffusion-Brinkman systems. Networks and Heterogeneous Media, 2018, 13(1): 69-94. doi: 10.3934/nhm.2018004
[6]	Thomas Geert de Jong, Georg Prokert, Alef Edou Sterk . Reaction–diffusion transport into core-shell geometry: Well-posedness and stability of stationary solutions. Networks and Heterogeneous Media, 2025, 20(1): 1-14. doi: 10.3934/nhm.2025001
[7]	Steinar Evje, Aksel Hiorth, Merete V. Madland, Reidar I. Korsnes . A mathematical model relevant for weakening of chalk reservoirs due to chemical reactions. Networks and Heterogeneous Media, 2009, 4(4): 755-788. doi: 10.3934/nhm.2009.4.755
[8]	Narcisa Apreutesei, Vitaly Volpert . Reaction-diffusion waves with nonlinear boundary conditions. Networks and Heterogeneous Media, 2013, 8(1): 23-35. doi: 10.3934/nhm.2013.8.23
[9]	Matthieu Alfaro, Thomas Giletti . Varying the direction of propagation in reaction-diffusion equations in periodic media. Networks and Heterogeneous Media, 2016, 11(3): 369-393. doi: 10.3934/nhm.2016001
[10]	Panpan Xu, Yongbin Ge, Lin Zhang . High-order finite difference approximation of the Keller-Segel model with additional self- and cross-diffusion terms and a logistic source. Networks and Heterogeneous Media, 2023, 18(4): 1471-1492. doi: 10.3934/nhm.2023065

Abstract

1. Introduction

Reaction-diffusion phenomena is ubiquitous occurring mostly in different fields of science and engineering in which the system components interact ^[1,2,3,4,5]. They are used to describe population growth, propagation of travelling waves, pattern formation, other intriguing phenomena arising in combustion theory, tumor invasion, and neural networks spanning across various temporal and spatial scales ^[6,7,8,9]. Mathematically, the reaction-diffusion model is a parabolic partial differential equation (PDE) involving terms which denote diffusion and local reaction kinetics as given below:

$\begin{equation} \partial_{t}\mathrm{Y}-D\nabla^{2}\mathrm{Y} = Q(\mathrm{Y}), \end{equation}$

(1.1)

where the first term on the left side represents the temporal part, the second term represents the diffusion component, and the term on the right side of equation denotes the local reaction kinetics. For the choice of $Q(\mathrm{Y}) = \mathrm{Y} (1-\mathrm{Y})$ , the resultant model is Fisher's equation that can be used to describe population evolution and how the wave propagates in certain medium ^[10]. If $Q(\mathrm{Y}) = \mathrm{Y} (1-\mathrm{Y}^{2})$ , the reaction-diffusion is called Newell-Whitehead-Segel (NWS) equation which is primarily used to describe convection phenomena in fluid thermodynamics ^[11]. Zeldovich-Frank-Kamenetskii equation (ZFK) is obtained for the choice of $Q(\mathrm{Y}) = \mathrm{Y} (1-\mathrm{Y})e^{-\beta (1-\mathrm{Y})}$ . In combustion theory, the ZFK equation is used to describe how flames propagate ^[12].

In theoretical neuroscience, the most well-studied reaction diffusion equation is the FitzHugh-Nagumo (FHN) model ^[13,14]. The FHN model is the simplification of the Hodgkin-Huxley model for the action potentials in squid giant axon ^[15,16,17]. Although FHN is not phenomenological (the involved parameters are not biophysical), it is a good candidate for studying how an action potential is generated and propagated. The main advantage of the FHN is that the solution space is two-dimensional and hence geometrical phase plane tools can be utilized to show how the trajectory evolves, which gives rise to the excitability and spike mechanism of spike generation. In the phase plane diagram of the FHN neuronal model, one of the nullclines has cubic nonlinearity. Over the last few decades, due to its model simplicity, FHN gained a lot of attention in the scientific community. Besides the generation and propagation of a train of spikes, FHN is also used to explained the dynamics of various physical systems in several fields of science, typically in the propagation of flame, Brownian motion process, autocatalytic chemical reaction, growth of logistic population, and neurophysiology ^[18,19].

Consider the following scaler reaction diffusion model with constant coefficients:

$\begin{equation} \partial_{t}\mathrm{Y}-\partial_{\xi\xi}\mathrm{Y} = \mathrm{Y}(1-\mathrm{Y})(\mathrm{Y}-\eta), \end{equation}$

(1.2)

where $\eta$ is a known real number and $\mathrm{Y}(\xi, t)$ is an unknown function to be determined. In this equation, the local reaction kinetics is denoted by the cubic function $\mathrm{Y}(1-\mathrm{Y})(\mathrm{Y}-\eta)$ . For $\eta = -1$ , Eq (1.2) reduces to the NWS equation.

In literature, various analytical and numerical strategies have been used to solve FHN-type equation. Shih et al. ^[20] used approximate conditional Lie point symmetry method for the numerical solution of the perturbed FHN-type equations. Mehta et al. ^[21] solved such nonlinear time dependent problems by using a novel block method coupled with compact finite difference schemes. In the perturbed model, the term $\epsilon \mathrm{Y}$ has been added to the cubic term and for $\epsilon = 0,$ the problem reduces to an unperturbed FHN-type model. Li and Guo implemented an integral approach for the exact solution of FHN-type equations ^[22]. Abbasbandy proposed a homotopy analysis method (HAM) to calculate a solitary-type solution of the nonlinear FHN-type model ^[23]. Kawahara and Tanaka derived the exact solution for the interaction of traveling fronts ^[24]. Gorder and Vajravelua obtained exact solutions for FHN-type equations and Nagumo telegraph equation using a variational method for fixed initial conditions in order to control the error ^[25]. Ali et al. proposed a Galerkin finite element method for the numerical solutions of FHN-type equation^[26]. Mehran and Sadegh used a non-conventional finite difference scheme for the numerical solutions of FHN-type equations ^[27]. Gorder implemented HAM for the approximate solutions of FHN-type reaction diffusion equations ^[28].

Reaction-diffusion of FHN-type equations discusses previously involved constant coefficients. However, in general, time-dependent coefficients and dispersion-reaction terms involved in such models are more realistic physically ^[29,30,31]. Due to variation in problem geometry and taking into account the factor of heterogeneity in the propagating medium, FHN-type models with time-dependent coefficients are quite sophisticated to study.

The main goal of this work is to demonstrate the numerical solutions of the generalized FitzHugh Nagumo (GFHN) equation with a linear dispersion term and time-dependent coefficients, which are given below:

$\begin{equation} \partial_{t}\mathrm{Y}+\alpha(t)\partial_{\xi}\mathrm{Y}-\vartheta(t)\partial_{\xi\xi}\mathrm{Y} -\delta(t)\mathrm{Y}(1-\mathrm{Y})(\mathrm{Y}-\eta) = 0,\; (\xi,t)\in[a,b]\times[0,t], \end{equation}$

(1.3)

where $\alpha(t)$ , $\vartheta(t)$ , and $\delta(t)$ are time-dependent coefficients. The associated initial condition is:

$\begin{equation} \mathrm{Y}(\xi,0) = \psi_{0}(\xi), \; \xi\in[a,b], \end{equation}$

(1.4)

and the boundary conditions are:

$\begin{equation} \mathrm{Y}(a,t) = \psi_{a}(t),\; \mathrm{Y}(b,t) = \psi_{b}(t), \,\, t\geq 0, \end{equation}$

(1.5)

where $\psi_{0}(\xi), \psi_{a}(t),$ and $\psi_{b}(t)$ are the known functions.

In recent literature, various authors computed the numerical solutions of GFHN equation with time-dependent coefficients. For instance, tanh and the Jacobi Gauss Lobatto collocation technique ^[18,29], the polynomial differential quadrature method ^[32], and the mixed-types discontinuous Galerkin method ^[33,34] were used.

Recently, wavelet numerical procedures attracted different researchers in various scientific communities because of their good properties and easy implementation. Some interesting aspects of wavelet methods include, compact support, localization in time and space, and multi-resistant analysis. Among different families of wavelet, Haar wavelet (HW) is simple and popular. The importance of the HW has been explored in various articles: Lepik proposed HW-based methods for different problems ^[35], Jiwari ^[36] utilized the HW method for the solution of Burger's equation, Kumar and Pandit ^[37] solved coupled Burger's equations using the HW method, Shiralashetti et al. ^[38] used the HW scheme for the numerical solutions of the initial valued problems. For further analysis of the HW, interested readers are referred to ^[39,40,41]. The remaining parts of the paper are presented in the following pattern:

● The underlying motivation and preliminaries are given in Sections 2 and 3, respectively.

● The proposed methodology is presented in Section 4.

● Numerical results and stability are reported in Sections 5 and 6.

● Finally, conclusions are drawn in Section 7.

2. Main motive

The analytical solution of the time-dependent coefficients partial differential equations is quite complicated to compute. Therefore, numerical treatment is an alternative and an efficient way to cope with this issue. According to our analysis, the method of lines using the HW is not proposed for FHN-type models. In this work, we propose an HW-based method of lines for the solutions of FHN-type equations. The theoretical stability and its numerical verification will also be a part of this work.

3. Haar wavelet and its integrals

Here, we address some basic results. Consider an arbitrary interval $[a, b)$ and divide it into $2L$ equal subintervals of length $\partial \xi = \frac{b-a}{2L}$ , where $L = 2^\mathcal{J}$ represents the maximum level of resolution. Define the dilation and translational parameters as $j = 0, 1, ..., \mathcal{J}$ , $\mathcal{K} = 0, 1, ..., l-1$ , where $l = 2^{j}$ , respectively. Using the dilation and translational parameters, define the wavelet number $\iota = l+\mathcal{K}+1$ . Now the first and $\iota$ -th HW are given below ^[35]:

$\begin{eqnarray} \mathrm{h}_{1}(\xi) = \left\{ \begin{array}{ll} 1, & { \xi\in[a,b) },\\ 0, & \mbox{elsewhere} .\end{array} \right. \end{eqnarray}$

(3.1)

$\begin{eqnarray} \mathrm{h}_{\iota}(\xi) = \left\{ \begin{array}{lll} 1, & { \xi\in[\zeta_{1}(\iota), \zeta_{2}(\iota)) },\\ -1, & { \xi\in[\zeta_{2}(\iota), \zeta_{3}(\iota)) },\\ 0, & \mbox{elsewhere,}\end{array} \right. \end{eqnarray}$

(3.2)

where $\zeta_{1}(\iota) = a+2\mathcal{K}\omega\partial \xi$ , $\zeta_{1}(\iota) = a+(2\mathcal{K}+1)\omega\delta \xi$ , $\zeta_{3}(\iota) = a+2(\mathcal{K}+1)\omega\partial \xi,$ and $\omega = \frac{\mathrm{L}}{l}$ .

In our analysis, we will use the following integrals:

$\begin{equation} \beta_{\iota,\alpha}(\xi) = \displaystyle {\int}_{a}^{\xi}\displaystyle {\int}_{a}^{\xi},...,\displaystyle {\int}_{a}^{\xi}\mathrm{h}_{\iota}(z) dz^{\alpha} = \frac{1}{(\alpha-1)!}\displaystyle {\int}_{a}^{\xi}(\xi-z)^{\alpha-1}\mathrm{h}_{\iota}(z) dz, \end{equation}$

(3.3)

where $\alpha = 1, 2, 3, ..., n$ and $\iota = 1, 2, 3, ..., 2L$ . Analytically, the evaluation of these integrals for Eqs (3.1) and (3.2) are given below:

$\begin{equation} \beta_{1,\alpha}(\xi) = \frac{(\xi-a)^\alpha}{\alpha!}, \end{equation}$

(3.4)

$\begin{eqnarray} \begin{gathered} \beta_{\iota,\alpha}(\xi) = \left\{ \begin{array}{llll} 0, & { \xi < \zeta_{1}(\iota) },\\ \frac{1}{\alpha!}[\xi-\zeta_{1}(\iota)]^{\alpha}, & { \xi\in[\zeta_{1}(\iota),\zeta_{2}(\iota)] },\\ \frac{1}{\alpha!}[(\xi-\zeta_{1}(\iota))^{\alpha}-2(\xi-\zeta_{2}(\iota))^{\alpha}], & { \xi\in[\zeta_{2}(\iota),\zeta_{3}(\iota)] },\\ \frac{1}{\alpha!}[(\xi-\zeta_{1}(\iota))^{\alpha}-2(\xi-\zeta_{2}(\iota))^{\alpha}+(\xi-\zeta_{3}(\iota))^{\alpha}],& { \xi > \zeta_{3}(\iota) }.\end{array} \right. \ \end{gathered} \end{eqnarray}$

(3.5)

4. Proposed scheme

In this section, the proposed method is described in detail. Here, an integral approach is utilized, therefore, the greatest order spatial derivative in Eq (1.3) can be approximated via the HW series as:

$\begin{equation} \partial_{\xi\xi}\mathrm{Y}(\xi,t) = \sum\limits_{\iota = 1}^{2L}\lambda_\iota(t) \mathrm{h}_\iota(\xi), \end{equation}$

(4.1)

where $\lambda_\iota (t)$ denotes the unknown HW coefficients and $\mathrm{h}_\iota(\xi)$ are HW basis. From Eq (4.1), we deduce the following equations via twice integration:

$\begin{equation} \partial_{\xi}\mathrm{Y}(\xi,t) = \sum\limits_{\iota = 1}^{2L}\lambda_\iota(t) \beta_{1,\iota}(\xi)+\partial_\xi\mathrm{Y}(\xi,t)|_{\xi = a}, \end{equation}$

(4.2)

$\begin{equation} \mathrm{Y}(\xi,t) = \sum\limits_{\iota = 1}^{2L}\lambda_\iota \beta_{2,\iota}(\xi)+(\xi-a)\partial_\xi\mathrm{Y}(\xi,t)|_{\xi = a}+\mathrm{Y}(\xi,t)|_{\xi = a}. \end{equation}$

(4.3)

The evaluation of Eq (4.3) at $\xi = b$ gives:

$\begin{equation} \partial _{\xi}\mathrm{Y}(a,t) = \frac{\psi_{b}(t)-\psi_{a}(t)}{b-a}-\frac{1}{b-a}\sum\limits_{\iota = 1}^{2L}\lambda_\iota(t) \beta_{2,\iota}(b). \end{equation}$

(4.4)

Plugging $\partial _{\xi}\mathrm{Y}(a, t)$ in Eqs (4.2) and (4.3), we get:

$\begin{equation} \partial_{\xi}\mathrm{Y}(\xi,t) = \sum\limits_{\iota = 1}^{2L}\lambda_\iota(t)\big[\beta_{1,\iota}(\xi)-\frac{1}{(b-a)}\beta_{2,\iota}(b)\big]+\frac{1}{(b-a)}\big(\psi_{b}(t) -\psi_{a}(t) \big), \end{equation}$

(4.5)

$\begin{equation} \mathrm{Y}(\xi,t) = \sum\limits_{\iota = 1}^{2L}\lambda_\iota(t)\big[\beta_{2,\iota}(\xi)-\frac{(\xi-a)}{(b-a)}\beta_{2,\iota}(b)\big]+\frac{(\xi-a)}{(b-a)}\psi_{b}(t)+\frac{(b-\xi)}{(b-a)}\psi_{a}(t). \end{equation}$

(4.6)

Now, the corresponding matrix form of Eqs (4.1)–(4.5), and (4.6) by using $\xi \rightarrow \xi_{m}$ are:

$\begin{equation} \partial_{\xi\xi}\mathrm{Y}(\xi,t) = H(\xi_{m})\lambda(t), \end{equation}$

(4.7)

$\begin{equation} \partial_{\xi}\mathrm{Y}(\xi,t) = \mathrm{M}_{1}(\xi_{m})\lambda(t)+\mathrm{M}_{2}(t), \end{equation}$

(4.8)

$\begin{equation} \mathrm{Y}(\xi,t) = \mathrm{N}_{1}(\xi_{m})\lambda(t)+\mathrm{N}_{2}(t), \end{equation}$

(4.9)

where

$\mathrm{M}_{1}(\xi_{m}) = \sum\limits_{\iota = 1}^{2L}\big[\beta_{1,\iota}(\xi_{m})-\frac{1}{(b-a)}\beta_{2,\iota}(b)\big],\; \mathrm{M}_{2}(t) = \frac{1}{(b-a)}\big(\psi_{b}(t) - \psi_{a}(t) \big),$

$\mathrm{N}_{1}(\xi_{m}) = \big[\beta_{2,\iota}(\xi_{m})-\frac{(\xi_{m}-a)}{(b-a)}\beta_{2,\iota}(b)\big],\; \mathrm{N}_{2}(t) = \frac{(\xi_{m}-a)}{(b-a)}\psi_{b}(t)+\frac{(b-\xi_{m})}{(b-a)}\psi_{a}(t).$

From Eqs (4.7)–(4.9) and (1.3), the following equation can be obtained:

$\begin{equation} \begin{split} &\frac{d\mathrm{Y}_{m}(t)}{dt} = -\alpha(t)\Big(\mathrm{M}_{1}(\xi_{m})\lambda(t)-\mathrm{M}_{2}(t)\Big) +\vartheta(t)H(\xi_{m}) \big)+\delta(t)\Big[(\mathrm{N}_{1}(\xi_{m})\lambda(t) +\mathrm{N}_{2}(t))*\\ &\Big(1-(\mathrm{N}_{1}(\xi_{m})\lambda(t)+\mathrm{N}_{2}(t))\Big)*\Big((\mathrm{N}_{1}(\xi_{m})\lambda(t)+\mathrm{N}_{2}(t))-\eta \Big)\Big],\; \xi\in[a,b], \; t > 0, \end{split} \end{equation}$

(4.10)

where "*" represents an element wise product. In a more compact form, Eq (4.10) can be written as:

$\begin{equation} \frac{d \mathrm{Y}}{dt} = \mathbf{F}(t,\mathrm{Y}),\; \mathrm{Y}(0) = \psi_{0}, \end{equation}$

(4.11)

where $\mathbf{F}(t, \mathrm{Y}) = \vartheta(t)\big(\mathrm{h}\lambda(t) \big)- \alpha(t)\Big(\mathrm{M}_{1}\lambda(t)+\mathrm{M}_{2}(t)\Big)$ - $\delta(t) \Big[(\mathrm{N}_{1}\lambda(t)+\mathrm{N}_{2}(t))*\Big(1-(\mathrm{N}_{1}\lambda(t) +\mathrm{N}_{2}(t))\Big)*\Big((\mathrm{N}_{1}\lambda(t)+\mathrm{N}_{2}(t))-\eta \Big)\Big].$

Here, Eq (4.11) represents the system of first-order ordinary differential equations, which can be solved via the RK-4 scheme discussed in a later section.

4.1. Temporal discretization

To obtain the solutions of Eq (4.11), we use the following $\mbox{RK-4}$ scheme:

$\begin{gathered} \mathcal{K}_1 = \mathbf{F}(t^m ,\mathrm{Y}^m), \\ \mathcal{K}_2 = \mathbf{F}(t^m + \frac{\partial t}{2} , \mathrm{Y}^m+\frac{\mathcal{K}_1}{2}), \\ \mathcal{K}_3 = \mathbf{F}(t^m + \frac{\partial t}{2},\mathrm{Y}^m+\frac{\mathcal{K}_2}{2}), \\ \mathcal{K}_4 = \mathbf{F}(t^m + {\partial t},\mathrm{Y}+{\mathcal{K}_3}).\\ \mathrm{Y}^{m+1} = \mathrm{Y}^{m}+\frac{\partial t}{6}\big(\mathcal{K}_1 +2(\mathcal{K}_2+\mathcal{K}_3)+\mathcal{K}_4\big), \,\,\, m\geq 0, \end{gathered}$

(4.12)

where $\partial t$ is the time step size.

5. Simulations and comparison

In this section, numerical solutions of GFHN with constant and time-dependent coefficient are listed. The efficiency of the proposed scheme is checked by applying different error measures, namely: $L_2$ , $L_\infty$ , and $L_{rms}$ as described below:

$\begin{equation} L_2 = \|\mathrm{Y}^E-\mathrm{Y}^N\|_{2}\cong\sqrt{\delta \xi \sum\limits_{\iota = 1}^{2L} {\mid{\mathrm{Y}_{\iota}^E-\mathrm{Y}_{\iota}^N}\mid}^2}, \end{equation}$

(5.1)

$\begin{equation} L_\infty = \|\mathrm{Y}^E-\mathrm{Y}^N\|_{\infty}\cong max_{\infty} {\mid{\mathrm{Y}_{\iota}^E-\mathrm{Y}_{\iota}^N}\mid}, \end{equation}$

(5.2)

$\begin{equation} L_{rms} = \sqrt{\sum\limits_{\iota = 1}^{2L}\big(\frac{\mathrm{Y}^E-\mathrm{Y}^N}{\sqrt{2L}} \big)^2}. \end{equation}$

(5.3)

5.1. Problem 1

Here, we consider Eq (1.2) with constant coefficients and various choices of the exact solutions:

$\begin{equation*} \label{e1.1} \begin{split} &\mbox{case}\; (i) :\; \mathrm{Y}(\xi,t) = \frac{1}{2}+\frac{1}{2}\tanh\Big[\frac{1}{2\sqrt{2}}\big( \xi-\frac{2\eta-1}{\sqrt{2}}t\big) \Big],\\ &\mbox{case}\; (ii) :\; \mathrm{Y}(\xi,t) = \frac{1}{2}(1+\alpha)+\frac{1}{2}(1-\alpha)\tanh\Big[\frac{(1-\alpha)}{2\sqrt{2}} \xi+\frac{(1-\alpha^2)}{2\sqrt{2}}t \Big],\\ &\mbox{case}\; (iii) :\; \mathrm{Y}(\xi,t) = \frac{1}{1+\exp\big[\frac{-\xi+(\sqrt{2}(\frac{1}{2}-\eta))t}{\sqrt{2}} \big]},\\ &\mbox{case}\; (iv) :\; \mathrm{Y}(\xi,t) = \Big[\frac{1}{2}\Big]\Big[1-\coth\big(\frac{\xi}{2\sqrt{2}}+\frac{2\eta-1}{4}t+\frac{\pi}{4}\big)\Big].\\ \end{split} \end{equation*}$

The associated initial and boundary conditions for all cases are used from the exact solutions. Simulations are done in different spatial domains to test the technique, and a comparison is made with the previously published work. The spatial domains used for case $(i)$ are $[0, 1]$ and $[-10, 10]$ , for case $(ii)$ is $[-22, 22]$ , for case $(iii)$ are $[0, 1]$ and $[-10, 10]$ , while for case $(iv)$ is the interval $[-10, 10]$ . The obtained results of case $(i)$ in $[-10, 10]$ with different values of $\eta$ are given in –, while the outcomes in the interval $[0, 1]$ with various values of $\eta$ are given in Table 4.

Table 1. Outcomes of problem 1 for case

$(i)$ with

$\eta = 0.75$ .

	Present ( $\partial t=0.0001$ )			Present ( $\partial t=0.001$ )			Present ( $\partial t=0.1$ )
$t$	$L_\infty$	$L_{rms}$	CPU time (Seconds)	$L_\infty$	$L_{rms}$	CPU time (Seconds)	$L_\infty$	$L_{rms}$	CPU time (Seconds)
0.2	$6.387\times10^{-6}$	$1.844\times10^{-7}$	6.240240	$1.203\times10^{-5}$	$1.782\times10^{-6}$	2.195968	$8.339\times10^{-4}$	$1.835\times10^{-4}$	1.624090
0.5	$1.414\times10^{-5}$	$4.959\times10^{-7}$	8.821217	$2.809\times10^{-5}$	$4.515\times10^{-6}$	2.954488	$2.027\times10^{-3}$	$4.626\times10^{-4}$	1.805044
1.0	$2.419\times10^{-5}$	$1.124\times10^{-6}$	13.053880	$5.211\times10^{-5}$	$9.234\times10^{-6}$	3.111479	$3.917\times10^{-3}$	$9.362\times10^{-4}$	1.839039
1.5	$3.230\times10^{-5}$	$1.899\times10^{-6}$	18.199838	$7.429\times10^{-5}$	$1.415\times10^{-5}$	5.149131	$5.822\times10^{-3}$	$1.418\times10^{-3}$	1.942381
2.0	$3.926\times10^{-5}$	$2.816\times10^{-6}$	23.565382	$9.548\times10^{-5}$	$1.924\times10^{-5}$	4.562245	$7.718\times10^{-3}$	$1.906\times10^{-3}$	1.919129
5.0	$7.276\times10^{-5}$	$1.0461\times10^{-5}$	54.294368	$2.174\times10^{-4}$	$5.217\times10^{-5}$	8.126805	$1.840\times10^{-2}$	$4.896\times10^{-3}$	2.256312
	^[42] ( $\partial t=0.0001$ )			^[32] ( $\partial t=0.001$ )			^[43] ( $\partial t=0.1$ )
$t$	$L_\infty$	$L_{rms}$		$L_\infty$	$L_{rms}$		$L_\infty$	$L_{rms}$
0.2	$1.889\times10^{-5}$	$2.196\times10^{-7}$		$4.741\times10^{-5}$	$1.588\times10^{-5}$		$1.887\times10^{-5}$	$7.455\times10^{-6}$
0.5	$4.155\times10^{-5}$	$1.569\times10^{-6}$		$1.231\times10^{-4}$	$3.843\times10^{-5}$		$4.151\times10^{-5}$	$1.641\times10^{-5}$
1.0	$6.989\times10^{-5}$	$7.144\times10^{-6}$		$2.626\times10^{-4}$	$8.187\times10^{-5}$		$6.973\times10^{-5}$	$2.743\times10^{-5}$
1.5	$9.168\times10^{-5}$	$1.726\times10^{-5}$		$4.209\times10^{-4}$	$1.338\times10^{-4}$		$9.118\times10^{-5}$	$3.534\times10^{-5}$
2.0	$1.096\times10^{-4}$	$3.185\times10^{-5}$		$5.999\times10^{-3}$	$1.943\times10^{-4}$		$1.085\times10^{-4}$	$4.128\times10^{-5}$
5.0	$1.896\times10^{-4}$	$1.880\times10^{-4}$		$2.305\times10^{-3}$	$7.863\times10^{-4}$		$1.800\times10^{-4}$	$6.134\times10^{-5}$

| Show Table

DownLoad: CSV

Table 2. Outcomes of problem 1 for case

$(i)$ with

$\eta = -1$ .

		$t=0.1$			$t=0.5$			$t=1.0$
	$\xi$	Exact	Numerical	Error	Exact	Numerical	Error	Exact	Numerical	Error
Present
	-10	$1.013383\times10^{-3}$	$1.013382\times10^{-3}$	$8.713\times10^{-10}$	$1.844968\times10^{-3}$	1.844963 $\times10^{-3}$	$4.687\times10^{-9}$	$3.897765\times10^{-3}$	$3.897749\times10^{-3}$	$1.573\times10^{-8}$
	-7	$8.209472\times10^{-3}$	$8.209430\times10^{-3}$	$4.275\times10^{-8}$	$1.485835\times10^{-2}$	1.485800 $\times10^{-2}$	$3.444\times10^{-7}$	$3.094160\times10^{-2}$	$3.094052\times10^{-2}$	$1.078\times10^{-6}$
	-3	$1.216556\times10^{-1}$	$1.216557\times10^{-1}$	$1.920\times10^{-7}$	$2.015162\times10^{-1}$	2.015188 $\times10^{-1}$	$2.553\times10^{-6}$	$3.482263\times10^{-1}$	$3.482340\times10^{-1}$	$7.646\times10^{-6}$
	0	$5.305565\times10^{-1}$	$5.305566\times10^{-1}$	$1.619\times10^{-7}$	$6.731305\times10^{-1}$	6.731300 $\times10^{-1}$	$5.918\times10^{-7}$	$8.134186\times10^{-1}$	$8.134174\times10^{-1}$	$1.247\times10^{-6}$
	3	$9.069410\times10^{-1}$	$9.069410\times10^{-1}$	$2.120\times10^{-7}$	$9.466898\times10^{-1}$	9.466899 $\times10^{-1}$	$1.552\times10^{-7}$	$9.740892\times10^{-1}$	$9.740895\times10^{-1}$	$3.233\times10^{-7}$
	7	$9.939053\times10^{-1}$	$9.939053\times10^{-1}$	$3.713\times10^{-8}$	$9.966459\times10^{-1}$	9.966460 $\times10^{-1}$	$1.017\times10^{-7}$	$9.984128\times10^{-1}$	$9.984129\times10^{-1}$	$9.444\times10^{-8}$
	10	$9.992490\times10^{-1}$	$9.992490\times10^{-1}$	$7.370\times10^{-10}$	$9.995877\times10^{-1}$	9.995877 $\times10^{-1}$	$1.214\times10^{-9}$	$9.998052\times10^{-1}$	$9.998052\times10^{-1}$	$9.589\times10^{-10}$
^[33]
	-10	$0.123172E\times10^{-2}$	$0.123147\times10^{-2}$	$2.974\times10^{-6}$	$0.273123\times10^{-2}$	0.273299 $\times10^{-2}$	$1.086\times10^{-6}$	$0.639709\times10^{-2}$	$0.639771\times10^{-2}$	$4.544\times10^{-6}$
	-7	$0.799954\times10^{-2}$	$0.800058\times10^{-2}$	$1.005\times10^{-6}$	$0.149620\times10^{-1}$	0.149662 $\times10^{-1}$	$1.086\times10^{-6}$	$0.639709\times10^{-2}$	$0.639771\times10^{-2}$	$4.544\times10^{-6}$
	-3	$0.122814\times10^{+0}$	$0.122808\times10^{+0}$	$1.352\times10^{-6}$	$0.198331\times10^{+0}$	0.198340 $\times10^{+0}$	$2.984\times10^{-6}$	$0.351539\times10^{+0}$	$0.351533\times10^{+0}$	$9.543\times10^{-3}$
	0	$0.537579\times10^{+0}$	$0.537585\times10^{+0}$	$1.550\times10^{-6}$	$0.680017\times10^{+0}$	0.680007 $\times10^{+0}$	$8.287\times10^{-6}$	$0.819036\times10^{+0}$	$0.819048\times10^{+0}$	$4.715\times10^{-6}$
	3	$0.909059\times10^{+0}$	$0.909004\times10^{+0}$	$2.442\times10^{-6}$	$0.9487161\times10^{+0}$	0.948717 $\times10^{+0}$	$9.415\times10^{-6}$	$0.975932\times10^{+0}$	$0.975967\times10^{+0}$	$3.343\times10^{-6}$
	7	$0.994343\times10^{+0}$	$0.994331\times10^{+0}$	$2.683\times10^{-6}$	$0.997526\times10^{+0}$	0.997569 $\times10^{+0}$	$9.821\times10^{-7}$	$0.999603\times10^{+0}$	$0.999694\times10^{+0}$	$1.900\times10^{-6}$
	10	$0.999376\times10^{+0}$	$0.999372\times10^{+0}$	$9.138\times10^{-7}$	$0.9999983\times10^{+0}$	0.100002 $\times10^{+1}$	$7.363\times10^{-7}$	$0.999998\times10^{+0}$	$0.100001\times10^{+1}$	$1.930\times10^{-7}$

| Show Table

DownLoad: CSV

Table 3. Solutions of problem 1 for case

$(i)$ with

$\eta = -2$ .

	$t=0.1$			$t=0.5$			$t=1.0$
$\xi$	Exact	Numerical	Error	Exact	Numerical	Error	Exact	Numerical	Error
-10	$1.119842\times10^{-3}$	$1.119841\times10^{-3}$	$9.172\times10^{-10}$	$3.038202\times10^{-3}$	3.038195 $\times10^{-3}$	$7.248\times10^{-9}$	$1.052473\times10^{-2}$	$1.052469\times10^{-2}$	$3.677\times10^{-8}$
-7	$9.065044\times10^{-3}$	$9.064999\times10^{-3}$	$4.448\times10^{-8}$	$2.426341\times10^{-2}$	2.426292 $\times10^{-2}$	$4.832\times10^{-7}$	$7.986202\times10^{-2}$	$7.986054\times10^{-2}$	$1.479\times10^{-6}$
-3	$1.327517\times10^{-1}$	$1.327520\times10^{-1}$	$2.791\times10^{-7}$	$2.938320\times10^{-1}$	2.938363 $\times10^{-1}$	$4.239\times10^{-6}$	$5.922212\times10^{-1}$	$9.922280\times10^{-1}$	$6.850\times10^{-6}$
0	$5.553666\times10^{-1}$	$5.553666\times10^{-1}$	$2.417\times10^{-7}$	$7.724818\times10^{-1}$	7.724813 $\times10^{-1}$	$4.934\times10^{-7}$	$9.221826\times10^{-1}$	$9.221826\times10^{-1}$	$3.423\times10^{-8}$
3	$9.150444\times10^{-1}$	$9.150445\times10^{-1}$	$4.741\times10^{-8}$	$9.669729\times10^{-1}$	9.669732 $\times10^{-1}$	$2.824\times10^{-7}$	$9.903092\times10^{-1}$	$9.903095\times10^{-1}$	$2.634\times10^{-7}$
7	$9.944820\times10^{-1}$	$9.944821\times10^{-1}$	$3.521\times10^{-8}$	$9.979629\times10^{-1}$	9.979630 $\times10^{-1}$	$6.559\times10^{-8}$	$9.994155\times10^{-1}$	$9.994155\times10^{-1}$	$3.733\times10^{-8}$
10	$9.993204\times10^{-1}$	$9.993204\times10^{-1}$	$6.943\times10^{-10}$	$9.997499\times10^{-1}$	9.997477 $\times10^{-1}$	$7.674\times10^{-10}$	$9.999283\times10^{-1}$	$9.999283\times10^{-1}$	$3.684\times10^{-10}$

| Show Table

DownLoad: CSV

Table 4.

$L_{\infty}$ norm of problem 1 for case

$(i)$ with various values of

$\eta$ .

	$\eta=0.25$		${\eta=0.5}$		$\eta=0.75$
$t$	Present	^[44]	Present	^[44]	Present	^[44]
0.01	$4\times10^{-9}$	$5\times10^{-8}$	$7\times10^{-11}$	$5\times10^{-8}$	$4\times10^{-10}$	$5\times10^{-8}$
0.1	$3\times10^{-8}$	$3\times10^{-7}$	$4\times10^{-9}$	$3\times10^{-7}$	$3\times10^{-8}$	$3\times10^{-7}$
1.0	$1\times10^{-7}$	$1\times10^{-6}$	$6\times10^{-9}$	$5\times10^{-7}$	$1\times10^{-7}$	$1\times10^{-6}$
$L_{2}$
	$\eta=0.25$		${\eta=0.5}$		$\eta=0.75$
$t$	Present	^[44]	Present	^[44]	Present	^[44]
0.01	$5\times10^{-8}$	$5\times10^{-8}$	$8\times10^{-10}$	$5\times10^{-8}$	$5\times10^{-9}$	$5\times10^{-8}$
0.1	$4\times10^{-7}$	$3\times10^{-7}$	$4\times10^{-8}$	$3\times10^{-7}$	$4\times10^{-7}$	$3\times10^{-7}$
1.0	$2\times10^{-6}$	$1\times10^{-6}$	$7\times10^{-8}$	$5\times10^{-7}$	$2\times10^{-7}$	$9\times10^{-7}$

| Show Table

DownLoad: CSV

In these tables, the obtained solutions are compared with the existing results in the literature ^{[32,42,43,44]}. Through comparison, it is obvious that the present technique shows better performance than the cited work ^[32,42,44]. The present method is based on RK-4 which gives good results using a small step size.

In , the computed values of the constants $\mathcal{K}_{1}, \mathcal{K}_{2}, \mathcal{K}_{3}, \mbox{and}\; \mathcal{K}_{4},$ for $\eta$ = 0.75, $\partial$ t = 0.001, and $L$ = 2 in the spatial domain $[-10, 10]$ at distinct time levels are reported.

Table 5. The coefficients values

$\mathcal{K}_{1}, \mathcal{K}_{2}, \mathcal{K}_{3}, \mbox{and}\; \mathcal{K}_{4}$ of problem 1 for case

$(i)$ with

$\eta = 0.75$ ,

$\partial t = 0.001$ , and

$L = 2$ .

$t$	$\mathcal{K}_1$	$\mathcal{K}_2$	$\mathcal{K}_3$	$\mathcal{K}_4$	Exact solution	Approximate Solution
0.1	$-0.000550127435708$	$-0.000550134504350$	-0.000550134504441	$-0.000550141573170$	$0.002000842010538$	0.001998228498937
	$-0.003944143032133$	$-0.003944200185466$	-0.003944200186294	$-0.003944257340322$	$0.011607626920875$	0.011488174877494
	$-0.013201078042744$	$-0.013201383972046$	-0.013201383979136	$-0.013201689914171$	$0.064365301508760$	0.064570374111189
	$-0.051365677465119$	$-0.051368265656579$	-0.051368265786992	$-0.051370854097156$	$0.287228366819168$	0.287197745055223
	$-0.051813540724938$	$-0.051816722723024$	-0.051816722918439	$-0.051819905116762$	$0.702427327811607$	0.702470418192179
	$-0.017916749446516$	$-0.017917560559869$	-0.017917560596589	$-0.017918371748347$	$0.932557185568827$	0.932334691176706
	$-0.001827678052741$	$-0.001827749308296$	-0.001827749311074	$-0.001827820569427$	$0.987804495301507$	0.987929599908556
	$-0.000469739827336$	$-0.000469757565601$	-0.000469757566270	$-0.000469775305206$	$0.997896788384635$	0.997899530906305
0.5	$-0.000577059786102$	$-0.000577067177572$	-0.000577067177666	$-0.000577074569228$	$0.001810781501093$	0.001767463721627
	$-0.003051640090534$	$-0.003051683586188$	-0.003051683586808	$-0.003051727083001$	$0.010514629756167$	0.010105627339338
	$-0.013080724367551$	$-0.013081016714371$	-0.013081016720905	$-0.013081309072915$	$0.058599062346241$	0.059295267046785
	$-0.048329726367191$	$-0.048332073109897$	-0.048332073223848	$-0.048334420069440$	$0.267198413052228$	0.267262778290452
	$-0.053763565954023$	$-0.053766904699846$	-0.053766904907184	$-0.053770243864621$	$0.681111275379947$	0.681347627307882
	$-0.019155289006543$	$-0.019156170889244$	-0.019156170929845	$-0.019157052855031$	$0.925989123684090$	0.924926841938356
	$-0.002566491484946$	$-0.002566591819423$	-0.002566591823345	$-0.002566692161784$	$0.986539147932092$	0.987046273168589
	$-0.000386965519072$	$-0.000386980139663$	-0.000386980140215	$-0.000386994761360$	$0.997676105725876$	0.997732428060447
1.0	$-0.000492580802813$	$-0.000492587089040$	-0.000492587089121	-0.000492593375426	0.001598349150613	0.001497994290761
	$-0.002423286399790$	$-0.002423320374624$	-0.002423320375100	-0.002423354350360	0.009290606750817	0.008752036768966
	$-0.012362148938809$	$-0.012362412752656$	-0.012362412758286	$-0.012362676576548$	0.052072036738247	0.052916126005743
	$-0.044718979445185$	$-0.044721048624543$	-0.044721048720285	$-0.044723117985214$	0.243445131323980	0.244007643848355
	$-0.055982628284796$	$-0.055986143692980$	-0.055986143913729	$-0.055989659546008$	0.653369506436006	0.653900188317205
	$-0.020946593575349$	$-0.020947579113019$	-0.020947579159389	$-0.020948564745616$	0.916953070453016	0.914910265106938
	$-0.003363560269563$	$-0.003363692368992$	-0.003363692374180	$-0.003363824478866$	0.984774145473912	0.985560005959086
	$-0.000411345006904$	$-0.000411360558612$	-0.000411360559200	$-0.000411376111496$	0.997367497354861	0.997536946415740

| Show Table

DownLoad: CSV

In , numerical and exact solutions are plotted for various values of $\eta$ for case $(i)$ while in , the absolute error at distinct points are plotted with various time levels. Similar simulations are obtained for case ( $ii$ ) and its absolute error at distinct points are compared with the existing results ^[33,45,46] in Table 6, which show the superiority of the method.

Figure 1. Exact and numerical measures of problem 1 for case

$(i)$ . The arrow shows the direction of the travelling wave.

	$t=0.001$				$t=0.002$				$t=0.003$
$\xi$	^[45]	^[46]	^[33]	Present	^[45]	^[46]	^[33]	Present	^[45]	^[46]	^[33]	Present
-22	$1.680\times10^{-7}$	$1.367\times10^{-7}$	$1.435\times10^{-7}$	$2.139\times10^{-10}$	$3.362\times10^{-7}$	2.888 $\times10^{-7}$	$2.955\times10^{-7}$	$3.158\times10^{-10}$	$5.045\times10^{-7}$	$4.561\times10^{-7}$	$4.097\times10^{-7}$	$3.955\times10^{-10}$
-6	$5.184\times10^{-6}$	$4.611\times10^{-6}$	$4.844\times10^{-7}$	$4.987\times10^{-6}$	$1.037\times10^{-5}$	9.740 $\times10^{-6}$	$9.990\times10^{-6}$	$9.980\times10^{-6}$	$1.556\times10^{-5}$	$1.538\times10^{-5}$	$1.567\times10^{-5}$	$1.497\times10^{-5}$
2	$2.956\times10^{-5}$	$2.669\times10^{-5}$	$2.751\times10^{-5}$	$2.924\times10^{-5}$	$5.915\times10^{-5}$	5.633 $\times10^{-5}$	$5.811\times10^{-5}$	$5.847\times10^{-5}$	$8.876\times10^{-5}$	$8.875\times10^{-5}$	$8.620\times10^{-5}$	$8.768\times10^{-5}$
10	$1.073\times10^{-6}$	$9.152\times10^{-7}$	$1.006\times10^{-7}$	$5.511\times10^{-7}$	$2.146\times10^{-6}$	1.931 $\times10^{-6}$	$2.011\times10^{-6}$	$1.101\times10^{-6}$	$3.218\times10^{-6}$	$3.047\times10^{-6}$	$4.992\times10^{-6}$	$1.651\times10^{-6}$
18	$5.458\times10^{-7}$	$4.272\times10^{-7}$	$5.130\times10^{-7}$	$6.036\times10^{-9}$	$1.091\times10^{-6}$	9.016 $\times10^{-7}$	$9.215\times10^{-7}$	$1.206\times10^{-8}$	$1.636\times10^{-6}$	$1.423\times10^{-6}$	$1.610\times10^{-6}$	$1.808\times10^{-8}$

	$\partial t=10^{-4}$		$\partial t=10^{-5}$
$t$	Present	^[44]	Present	^[44]
0.01	$4\times10^{-8}$	$2\times10^{-7}$	$6\times10^{-9}$	$2\times10^{-7}$
0.1	$3\times10^{-7}$	$9\times10^{-7}$	$4\times10^{-8}$	$9\times10^{-7}$
10.0	$9\times10^{-8}$	$3\times10^{-7}$	$7\times10^{-9}$	$3\times10^{-7}$
$L_{2}$
	$\partial t=10^{-4}$		$\partial t=10^{-5}$
$t$	Present	^[44]		Present	^[44]
0.01	$2\times10^{-7}$	$1\times10^{-7}$	$3\times10^{-8}$	$1\times10^{-7}$
0.1	$1\times10^{-6}$	$6\times10^{-7}$	$2\times10^{-7}$	$6\times10^{-7}$
10.0	$3\times10^{-7}$	$2\times10^{-7}$	$4\times10^{-8}$	$2\times10^{-7}$

	$\eta=-1$			$\eta=0.75$			$\eta=4$
$t$	$L_{\infty}$	$L_{2}$	$L_{rms}$	$L_{\infty}$	$L_{2}$	$L_{rms}$	$L_{\infty}$	$L_{2}$	$L_{rms}$
0.1	$9.677\times10^{-7}$	$4.969\times10^{-6}$	$5.559\times10^{-8}$	$7.961\times10^{-7}$	4.699 $\times10^{-6}$	$9.270\times10^{-9}$	$3.662\times10^{-6}$	$1.361\times10^{-5}$	$1.464\times10^{-7}$
0.2	$1.890\times10^{-6}$	$9.619\times10^{-6}$	$1.166\times10^{-7}$	$1.514\times10^{-76}$	8.965 $\times10^{-6}$	$1.950\times10^{-8}$	$7.200\times10^{-6}$	$2.596\times10^{-5}$	$3.349\times10^{-7}$
0.3	$2.773\times10^{-6}$	$1.399\times10^{-5}$	$1.834\times10^{-7}$	$2.168\times10^{-6}$	1.286 $\times10^{-5}$	$3.081\times10^{-8}$	$1.042\times10^{-6}$	$3.687\times10^{-5}$	$5.497\times10^{-7}$
0.4	$3.617\times10^{-6}$	$1.814\times10^{-5}$	$2.555\times10^{-7}$	$2.771\times10^{-6}$	1.643 $\times10^{-5}$	$4.330\times10^{-8}$	$1.320\times10^{-5}$	$4.624\times10^{-5}$	$7.665\times10^{-7}$
0.5	$4.424\times10^{-6}$	$2.208\times10^{-5}$	$3.325\times10^{-7}$	$3.326\times10^{-6}$	1.973 $\times10^{-5}$	$5.702\times10^{-8}$	$1.544\times10^{-5}$	$5.399\times10^{-5}$	$9.573\times10^{-7}$
1.0	$7.881\times10^{-6}$	$3.949\times10^{-5}$	$7.631\times10^{-7}$	$5.641\times10^{-6}$	3.310 $\times10^{-5}$	$1.460\times10^{-7}$	$1.650\times10^{-5}$	$6.607\times10^{-5}$	$9.066\times10^{-7}$
1.5	$1.031\times10^{-5}$	$5.316\times10^{-5}$	$1.186\times10^{-7}$	$7.452\times10^{-6}$	4.300 $\times10^{-5}$	$2.707\times10^{-7}$	$8.097\times10^{-6}$	$2.564\times10^{-5}$	$1.411\times10^{-7}$
2.0	$1.168\times10^{-5}$	$6.187\times10^{-5}$	$1.525\times10^{-6}$	$8.979\times10^{-6}$	5.088 $\times10^{-5}$	$4.305\times10^{-7}$	$5.428\times10^{-6}$	$1.746\times10^{-5}$	$3.026\times10^{-7}$

		$\eta=-1$			$\eta=1$			$\eta=0.75$
$t$	L	$L_{\infty}$	$L_{2}$	$L_{rms}$	$L_{\infty}$	$L_{2}$	$L_{rms}$	$L_{\infty}$	$L_{2}$	$L_{rms}$
0.01	1	$2.145\times10^{-4}$	$2.173\times10^{-4}$	$6.515\times10^{-5}$	$2.022\times10^{-4}$	2.049 $\times10^{-4}$	$6.193\times10^{-5}$	$2.307\times10^{-4}$	$2.064\times10^{-4}$	$6.232\times10^{-5}$
	3	$3.599\times10^{-5}$	$5.573\times10^{-5}$	$9.275\times10^{-6}$	$3.319\times10^{-5}$	5.171 $\times10^{-5}$	$8.652\times10^{-6}$	$3.352\times10^{-5}$	$5.219\times10^{-5}$	$8.727\times10^{-6}$
	5	$2.625\times10^{-6}$	$8.187\times10^{-6}$	$6.643\times10^{-7}$	$2.234\times10^{-6}$	7.016 $\times10^{-6}$	$5.751\times10^{-7}$	$2.281\times10^{-6}$	$7.155\times10^{-6}$	$5.857\times10^{-7}$
	7	$3.131\times10^{-7}$	$1.946\times10^{-6}$	$7.687\times10^{-8}$	$9.539\times10^{-8}$	5.991 $\times10^{-7}$	$2.520\times10^{-8}$	$1.209\times10^{-7}$	$7.573\times10^{-7}$	$3.128\times10^{-8}$
0.1	1	$2.132\times10^{-3}$	$2.288\times10^{-3}$	$8.423\times10^{-4}$	$1.142\times10^{-3}$	1.270 $\times10^{-3}$	$4.984\times10^{-4}$	$1.228\times10^{-3}$	$1.359\times10^{-3}$	$5.304\times10^{-4}$
	3	$2.519\times10^{-4}$	$5.146\times10^{-4}$	$9.770\times10^{-5}$	$1.177\times10^{-4}$	2.562 $\times10^{-4}$	$5.065\times10^{-5}$	$1.284\times10^{-4}$	$2.774\times10^{-4}$	$5.461\times10^{-5}$
	5	$1.753\times10^{-5}$	$7.158\times10^{-5}$	$6.734\times10^{-6}$	$7.536\times10^{-6}$	3.263 $\times10^{-5}$	$3.211\times10^{-6}$	$8.298\times10^{-6}$	$3.571\times10^{-5}$	$3.497\times10^{-6}$
	7	$2.065\times10^{-6}$	$1.679\times10^{-5}$	$7.777\times10^{-7}$	$3.194\times10^{-7}$	2.818 $\times10^{-6}$	$1.422\times10^{-7}$	$4.636\times10^{-7}$	$3.783\times10^{-6}$	$1.870\times10^{-7}$
0.2	1	$5.512\times10^{-3}$	$6.009\times10^{-3}$	$2.238\times10^{-3}$	$1.147\times10^{-3}$	1.763 $\times10^{-3}$	$7.469\times10^{-4}$	$1.693\times10^{-3}$	$2.009\times10^{-3}$	$8.449\times10^{-4}$
	3	$6.614\times10^{-4}$	$1.396\times10^{-3}$	$2.723\times10^{-4}$	$1.335\times10^{-4}$	3.243 $\times10^{-4}$	$6.846\times10^{-5}$	$1.565\times10^{-4}$	$3.765\times10^{-4}$	$7.901\times10^{-5}$
	5	$4.613\times10^{-5}$	$1.948\times10^{-4}$	$1.887\times10^{-5}$	$8.413\times10^{-6}$	4.091 $\times10^{-5}$	$4.311\times10^{-6}$	$9.990\times10^{-6}$	$4.802\times10^{-5}$	$5.027\times10^{-6}$
	7	$5.276\times10^{-6}$	$4.482\times10^{-5}$	$2.163\times10^{-6}$	$3.610\times10^{-7}$	3.597 $\times10^{-6}$	$1.931\times10^{-7}$	$5.271\times10^{-7}$	$5.111\times10^{-6}$	$2.695\times10^{-7}$
0.3	1	$1.295\times10^{-2}$	$1.394\times10^{-2}$	$4.660\times10^{-3}$	$1.537\times10^{-3}$	1.926 $\times10^{-3}$	$8.404\times10^{-4}$	$1.886\times10^{-3}$	$2.331\times10^{-3}$	$1.008\times10^{-3}$
	3	$1.796\times10^{-3}$	$3.661\times10^{-3}$	$6.980\times10^{-4}$	$1.295\times10^{-4}$	3.375 $\times10^{-4}$	$7.358\times10^{-5}$	$1.636\times10^{-4}$	$4.182\times10^{-4}$	$9.059\times10^{-5}$
	5	$1.306\times10^{-4}$	$5.218\times10^{-4}$	$4.930\times10^{-5}$	$8.177\times10^{-6}$	4.242 $\times10^{-5}$	$4.621\times10^{-6}$	$1.040\times10^{-5}$	$5.313\times10^{-5}$	$5.748\times10^{-6}$
	7	$1.400\times10^{-5}$	$1.144\times10^{-4}$	$5.458\times10^{-6}$	$3.564\times10^{-7}$	3.791 $\times10^{-6}$	$2.094\times10^{-7}$	$5.505\times10^{-7}$	$5.677\times10^{-6}$	$3.088\times10^{-7}$
0.4	1	$3.446\times10^{-2}$	$3.905\times10^{-2}$	$7.273\times10^{-3}$	$1.487\times10^{-3}$	1.917 $\times10^{-3}$	$8.488\times10^{-4}$	$1.941\times10^{-3}$	$2.461\times10^{-3}$	$1.080\times10^{-3}$
	3	$7.093\times10^{-3}$	$1.220\times10^{-2}$	$2.140\times10^{-3}$	$1.211\times10^{-4}$	3.264 $\times10^{-4}$	$7.239\times10^{-5}$	$1.616\times10^{-4}$	$4.305\times10^{-4}$	$9.485\times10^{-5}$
	5	$5.414\times10^{-4}$	$1.867\times10^{-3}$	$1.605\times10^{-4}$	$7.588\times10^{-6}$	4.095 $\times10^{-5}$	$4.540\times10^{-6}$	$1.029\times10^{-5}$	$5.458\times10^{-5}$	$6.009\times10^{-6}$
	7	$5.169\times10^{-5}$	$3.698\times10^{-4}$	$1.635\times10^{-5}$	$3.368\times10^{-7}$	3.721 $\times10^{-6}$	$2.085\times10^{-7}$	$5.465\times10^{-7}$	$5.851\times10^{-6}$	$3.236\times10^{-7}$
0.5	1	$1.921\times10^{-1}$	$2.391\times10^{-1}$	$5.616\times10^{-1}$	$1.386\times10^{-3}$	1.821 $\times10^{-3}$	$8.130\times10^{-4}$	$1.921\times10^{-3}$	$2.477\times10^{-3}$	$1.095\times10^{-3}$
	3	$6.222\times10^{-2}$	$7.995\times10^{-2}$	$1.088\times10^{-2}$	$1.103\times10^{-4}$	3.042 $\times10^{-4}$	$6.809\times10^{-5}$	$1.570\times10^{-4}$	$4.261\times10^{-4}$	$9.479\times10^{-5}$
	5	$7.028\times10^{-3}$	$1.727\times10^{-2}$	$1.144\times10^{-3}$	$6.897\times10^{-6}$	3.813 $\times10^{-5}$	$4.268\times10^{-6}$	$9.935\times10^{-6}$	$5.395\times10^{-5}$	$5.999\times10^{-6}$
	7	$5.741\times10^{-4}$	$2.903\times10^{-3}$	$9.948\times10^{-5}$	$3.121\times10^{-7}$	3.525 $\times10^{-6}$	$1.990\times10^{-7}$	$5.295\times10^{-7}$	$5.804\times10^{-6}$	$3.139\times10^{-7}$

[1]	N. Wu, J. Phang, J. Park, Y. Shen, Z. Huang, M. Zorin, Deep neural networks improve radiologists' performance in breast cancer screening, IEEE Trans. Med. Imaging, 39 (2019), 1184–1194. https://doi.org/10.1109/TMI.2019.2945514 doi: 10.1109/TMI.2019.2945514
[2]	D. M. van der Kolk, G. H. de Bock, B. K. Leegte, M. Schaapveld, M. J. Mourits, J. de Vries, et al., Penetrance of breast cancer, ovarian cancer and contralateral breast cancer in BRCA1 and BRCA2 families: high cancer incidence at older age, Breast Cancer Res. Treat., 124 (2010), 643–651. https://doi.org/10.1007/s10549-010-0805-3 doi: 10.1007/s10549-010-0805-3
[3]	Q. Xia, Y. Cheng, J. Hu, J. Huang, Y. Yu, H. Xie, et al., Differential diagnosis of breast cancer assisted by s-detect artificial intelligence system, Math. Biosci. Eng., 18 (2021), 3680–3689. https://doi.org/10.3934/mbe.2021184 doi: 10.3934/mbe.2021184
[4]	S. Williamson, K. Vijayakumar, V. J. Kadam, Predicting breast cancer biopsy outcomes from bi-rads findings using random forests with chi-square and mi features, Multimedia Tools Appl., 81 (2022), 36869–36889. https://doi.org/10.1007/s11042-021-11114-5 doi: 10.1007/s11042-021-11114-5
[5]	D. J. Gavaghan, J. P. Whiteley, S. J. Chapman, J. M. Brady, P. Pathmanathan, Predicting tumor location by modeling the deformation of the breast, IEEE Trans. Biomed. Eng., 55 (2008), 2471–2480. https://doi.org/10.1109/TBME.2008.925714 doi: 10.1109/TBME.2008.925714
[6]	M. M. Ghiasi, S. Zendehboudi, Application of decision tree-based ensemble learning in the classification of breast cancer, Comput. Biol. Med., 128 (2021), 104089. https://doi.org/10.1016/j.compbiomed.2020.104089 doi: 10.1016/j.compbiomed.2020.104089
[7]	S. Liu, J. Zeng, H. Gong, H. Yang, J. Zhai, Y. Cao, et al., Quantitative analysis of breast cancer diagnosis using a probabilistic modelling approach, Comput. Biol. Med., 92 (2018), 168–175. https://doi.org/10.1016/j.compbiomed.2017.11.014 doi: 10.1016/j.compbiomed.2017.11.014
[8]	Y. Dong, J. Wan, L. Si, Y. Meng, Y. Dong, S. Liu, et al., Deriving polarimetry feature parameters to characterize microstructural features in histological sections of breast tissues, IEEE Trans. Biomed. Eng., 68 (2020), 881–892. https://doi.org/10.1109/TBME.2020.3019755 doi: 10.1109/TBME.2020.3019755
[9]	I. Elyasi, M. A. Pourmina, M. S. Moin, Speckle reduction in breast cancer ultrasound images by using homogeneity modified bayes shrink, Measurement, 91 (2016), 55–65. https://doi.org/10.1016/j.measurement.2016.05.025 doi: 10.1016/j.measurement.2016.05.025
[10]	H. H. Xu, Y. C. Gong, X. Y. Xia, D. Li, Z. Z. Yan, J. Shi, et al., Gabor-based anisotropic diffusion with lattice boltzmann method for medical ultrasound despeckling., Math. Biosci. Eng., 16 (2019), 7546–7561. https://doi.org/10.3934/mbe.2019379 doi: 10.3934/mbe.2019379
[11]	J. Levman, T. Leung, P. Causer, D. Plewes, A. L. Martel, Classification of dynamic contrast-enhanced magnetic resonance breast lesions by support vector machines, IEEE Trans. Biomed. Eng., 27 (2008), 688–696. https://doi.org/10.1109/TMI.2008.916959 doi: 10.1109/TMI.2008.916959
[12]	A. Ed-daoudy, K. Maalmi, Breast cancer classification with reduced feature set using association rules and support vector machine, Network Modeling Analysis in Health Informatics and Bioinformatics, 9 (2020), 1–10. https://doi.org/10.1007/s13721-020-00237-8 doi: 10.1007/s13721-020-00237-8
[13]	R. Ranjbarzadeh, S. Dorosti, S. J. Ghoushchi, A. Caputo, E. B. Tirkolaee, S. S. Ali, et al., Breast tumor localization and segmentation using machine learning techniques: Overview of datasets, findings, and methods, Comput. Biol. Med., (2022), 106443. https://doi.org/10.1016/j.compbiomed.2022.106443
[14]	P. Sathiyanarayanan, S. Pavithra, M. S. Saranya, M. Makeswari, Identification of breast cancer using the decision tree algorithm, in 2019 IEEE International Conference on System, Computation, Automation and Networking (ICSCAN), IEEE, (2019), 1–6. https://doi.org/10.1109/ICSCAN.2019.8878757
[15]	J. X. Tian, J. Zhang, Breast cancer diagnosis using feature extraction and boosted c5. 0 decision tree algorithm with penalty factor, Math. Biosci. Eng., 19 (2022), 2193–205. https://doi.org/10.3934/mbe.2022102 doi: 10.3934/mbe.2022102
[16]	S. Wang, Y. Wang, D. Wang, Y. Yin, Y. Wang, Y. Jin, An improved random forest-based rule extraction method for breast cancer diagnosis, Appl. Soft Comput., 86 (2020), 105941. https://doi.org/10.1016/j.asoc.2019.105941 doi: 10.1016/j.asoc.2019.105941
[17]	T. Octaviani, d. Z. Rustam, Random forest for breast cancer prediction, in AIP Conference Proceedings, AIP Publishing LLC, 2168 (2019), 020050. https://doi.org/10.1063/1.5132477
[18]	S. Das, O. R. R. Aranya, N. N. Labiba, Brain tumor classification using convolutional neural network, in 2019 1st International Conference on Advances in Science, Engineering and Robotics Technology (ICASERT), IEEE, (2019), 1–5. https://doi.org/10.1007/978-981-10-9035-6_33
[19]	R. Hao, K. Namdar, L. Liu, F. Khalvati, A transfer learning–based active learning framework for brain tumor classification, Front. Artif. Intell., 4 (2021), 635766. https://doi.org/10.3389/frai.2021.635766 doi: 10.3389/frai.2021.635766
[20]	Q. Zhang, C. Bai, Z. Liu, L. T. Yang, H. Yu, J. Zhao, et al., A gpu-based residual network for medical image classification in smart medicine, Inf. Sci., 536 (2020), 91–100. https://doi.org/10.1016/j.ins.2020.05.013 doi: 10.1016/j.ins.2020.05.013
[21]	Y. Dai, Y. Gao, F. Liu, Transmed: Transformers advance multi-modal medical image classification, Diagnostics, 11 (2021), 1384. https://doi.org/10.3390/diagnostics11081384 doi: 10.3390/diagnostics11081384
[22]	S. Aladhadh, M. Alsanea, M. Aloraini, T. Khan, S. Habib, M. Islam, An effective skin cancer classification mechanism via medical vision transformer, Sensors, 22 (2022), 4008. https://doi.org/10.3390/s22114008 doi: 10.3390/s22114008
[23]	S. Yu, K. Ma, Q. Bi, C. Bian, M. Ning, N. He, et al., Mil-vt: Multiple instance learning enhanced vision transformer for fundus image classification, in Medical Image Computing and Computer Assisted Intervention–MICCAI 2021: 24th International Conference, Strasbourg, France, September 27–October 1, 2021, Proceedings, Part VIII 24, Springer, (2021), 45–54. https://doi.org/10.1007/978-3-030-87237-3_5
[24]	F. Almalik, M. Yaqub, K. Nandakumar, Self-ensembling vision transformer (sevit) for robust medical image classification, in Medical Image Computing and Computer Assisted Intervention–MICCAI 2022: 25th International Conference, Singapore, September 18–22, 2022, Proceedings, Part III, Springer, (2022), 376–386. https://doi.org/10.1007/978-3-031-16437-8_36
[25]	Y. Wu, S. Qi, Y. Sun, S. Xia, Y. Yao, W. Qian, A vision transformer for emphysema classification using ct images, Phys. Med. Biol., 66 (2021), 245016. https://doi.org/10.1088/1361-6560/ac3dc8 doi: 10.1088/1361-6560/ac3dc8
[26]	B. Hou, G. Kaissis, R. M. Summers, B. Kainz, Ratchet: Medical transformer for chest x-ray diagnosis and reporting, in Medical Image Computing and Computer Assisted Intervention–MICCAI 2021: 24th International Conference, Strasbourg, France, September 27–October 1, 2021, Proceedings, Part VII 24, Springer, (2021), 293–303. https://doi.org/10.1007/978-3-030-87234-2_28
[27]	F. A. Spanhol, L. S. Oliveira, C. Petitjean, L. Heutte, Breast cancer histopathological image classification using convolutional neural networks, in 2016 International Joint Conference on Neural Networks (IJCNN), IEEE, (2016), 2560–2567. https://doi.org/10.1109/IJCNN.2016.7727519
[28]	W. Lotter, G. Sorensen, D. Cox, A multi-scale cnn and curriculum learning strategy for mammogram classification, in Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support, Springer, (2017), 169–177. https://doi.org/10.1007/978-3-319-67558-9_20
[29]	A. A. Nahid, M. A. Mehrabi, Y. Kong, Histopathological breast cancer image classification by deep neural network techniques guided by local clustering, Biomed Res. Int., 2018 (2018). https://doi.org/10.1155/2018/2362108
[30]	H. K. Mewada, A. V. Patel, M. Hassaballah, M. H. Alkinani, K. Mahant, Spectral–spatial features integrated convolution neural network for breast cancer classification, Sensors, 20 (2020), 4747. https://doi.org/10.3390/s20174747 doi: 10.3390/s20174747
[31]	W. Al-Dhabyani, M. Gomaa, H. Khaled, A. Fahmy, Dataset of breast ultrasound images, Data Brief, 28 (2020), 104863. https://doi.org/10.1016/j.dib.2019.104863 doi: 10.1016/j.dib.2019.104863
[32]	P. S. Rodrigues, Breast ultrasound image, Mendeley Data, 1 (2017). https://doi.org/10.17632/wmy84gzngw.1
[33]	J. Virmani, R. Agarwal, Deep feature extraction and classification of breast ultrasound images, Multimedia Tools Appl., 79 (2020), 27257–27292. https://doi.org/10.1007/s11042-020-09337-z doi: 10.1007/s11042-020-09337-z
[34]	W. Al-Dhabyani, M. Gomaa, H. Khaled, F. Aly, Deep learning approaches for data augmentation and classification of breast masses using ultrasound images, Int. J. Adv. Comput. Sci. Appl., 10 (2019), 1–11. https://doi.org/10.14569/IJACSA.2019.0100579 doi: 10.14569/IJACSA.2019.0100579
[35]	N. Vigil, M. Barry, A. Amini, M. Akhloufi, X. P. Maldague, L. Ma, et al., Dual-intended deep learning model for breast cancer diagnosis in ultrasound imaging, Cancers, 14 (2022), 2663. https://doi.org/10.3390/cancers14112663 doi: 10.3390/cancers14112663
[36]	T. Xiao, L. Liu, K. Li, W. Qin, S. Yu, Z. Li, Comparison of transferred deep neural networks in ultrasonic breast masses discrimination, Biomed Res. Int., 2018 (2018). https://doi.org/10.1155/2018/4605191
[37]	W. X. Liao, P. He, J. Hao, X. Y. Wang, R. L. Yang, D. An, et al., Automatic identification of breast ultrasound image based on supervised block-based region segmentation algorithm and features combination migration deep learning model, IEEE J. Biomed. Health. Inf., 24 (2019), 984–993. https://doi.org/10.1109/JBHI.2019.2960821 doi: 10.1109/JBHI.2019.2960821
[38]	W. K. Moon, Y. W. Lee, H. H. Ke, S. H. Lee, C. S. Huang, R. F. Chang, Computer-aided diagnosis of breast ultrasound images using ensemble learning from convolutional neural networks, Comput. Methods Programs Biomed., 190 (2020), 105361. https://doi.org/10.1016/j.cmpb.2020.105361 doi: 10.1016/j.cmpb.2020.105361
[39]	S. Acharya, A. Alsadoon, P. Prasad, S. Abdullah, A. Deva, Deep convolutional network for breast cancer classification: enhanced loss function (elf), J. Supercomput., 76 (2020), 8548–8565. https://doi.org/10.1007/s11227-020-03157-6 doi: 10.1007/s11227-020-03157-6
[40]	E. Y. Kalafi, A. Jodeiri, S. K. Setarehdan, N. W. Lin, K. Rahmat, N. A. Taib, et al., Classification of breast cancer lesions in ultrasound images by using attention layer and loss ensemble in deep convolutional neural networks, Diagnostics, 11 (2021), 1859. https://doi.org/10.3390/diagnostics11101859 doi: 10.3390/diagnostics11101859
[41]	G. S. Tran, T. P. Nghiem, V. T. Nguyen, C. M. Luong, J. C. Burie, Improving accuracy of lung nodule classification using deep learning with focal loss, J. Healthcare Eng., 2019 (2019). https://doi.org/10.1155/2019/5156416
[42]	L. Ma, R. Shuai, X. Ran, W. Liu, C. Ye, Combining dc-gan with resnet for blood cell image classification, Med. Biol. Eng. Comput., 58 (2020), 1251–1264. https://doi.org/10.1007/s11517-020-02163-3 doi: 10.1007/s11517-020-02163-3
[43]	C. Zhao, R. Shuai, L. Ma, W. Liu, D. Hu, M. Wu, Dermoscopy image classification based on stylegan and densenet201, IEEE Access, 9 (2021), 8659–8679. https://doi.org/10.1109/ACCESS.2021.3049600 doi: 10.1109/ACCESS.2021.3049600
[44]	D. Sarwinda, R. H. Paradisa, A. Bustamam, P. Anggia, Deep learning in image classification using residual network (resnet) variants for detection of colorectal cancer, Procedia Comput. Sci., 179 (2021), 423–431. https://doi.org/10.1016/j.procs.2021.01.025 doi: 10.1016/j.procs.2021.01.025
[45]	Y. Chen, Q. Zhang, Y. Wu, B. Liu, M. Wang, Y. Lin, Fine-tuning resnet for breast cancer classification from mammography, in Proceedings of the 2nd International Conference on Healthcare Science and Engineering 2nd, Springer, (2019), 83–96. https://doi.org/10.1007/978-981-13-6837-0_7
[46]	F. Almalik, M. Yaqub, K. Nandakumar, Self-ensembling vision transformer (sevit) for robust medical image classification, in Medical Image Computing and Computer Assisted Intervention-MICCAI 2022, Springer, (2022), 376–386. https://doi.org/10.1007/978-3-031-16437-8_36
[47]	B. Gheflati, H. Rivaz, Vision transformers for classification of breast ultrasound images, in 2022 44th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), IEEE, (2022), 480–483. https://doi.org/10.1109/EMBC48229.2022.9871809
[48]	L. Yuan, X. Wei, H. Shen, L. L. Zeng, D. Hu, Multi-center brain imaging classification using a novel 3d cnn approach, IEEE Access, 6 (2018), 49925–49934. https://doi.org/10.1109/ACCESS.2018.2868813 doi: 10.1109/ACCESS.2018.2868813
[49]	J. Zhang, Y. Xie, Y. Xia, C. Shen, Attention residual learning for skin lesion classification, IEEE Trans. Med. Imaging, 38 (2019), 2092–2103. https://doi.org/10.1109/TMI.2019.2893944 doi: 10.1109/TMI.2019.2893944
[50]	B. Xu, J. Liu, X. Hou, B. Liu, J. Garibaldi, I. O. Ellis, et al., Attention by selection: A deep selective attention approach to breast cancer classification, IEEE Trans. Med. Imaging, 39 (2019), 1930–1941. https://doi.org/10.1109/TMI.2019.2962013 doi: 10.1109/TMI.2019.2962013
[51]	Z. Zhang, M. Sabuncu, Generalized cross entropy loss for training deep neural networks with noisy labels, Adv. Neural Inf. Process. Syst., 31 (2018).
[52]	R. R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, D. Batra, Grad-cam: Visual explanations from deep networks via gradient-based localization, in Proceedings of the IEEE International Conference on Computer Vision, (2017), 618–626. https://doi.org/10.1109/ICCV.2017.74
[53]	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, (2016), 770–778. https://doi.org/10.1109/CVPR.2016.90
[54]	A. Howard, M. Sandler, G. Chu, L. C. Chen, B. Chen, M. Tan, et al., Searching for mobilenetv3, in Proceedings of the IEEE/CVF International Conference on Computer Vision, (2019), 1314–1324. https://doi.org/10.1109/ICCV.2019.00140
[55]	X. Zhang, X. Zhou, M. Lin, J. Sun, Shufflenet: An extremely efficient convolutional neural network for mobile devices, in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, (2018), 6848–6856. https://doi.org/10.1109/CVPR.2018.00716
[56]	S. H. Gao, M. M. Cheng, K. Zhao, X. Y. Zhang, M. H. Yang, P. Torr, Res2net: A new multi-scale backbone architecture, IEEE Trans. Pattern Anal. Mach. Intell., 43 (2019), 652–662. https://doi.org/10.1109/TPAMI.2019.2938758 doi: 10.1109/TPAMI.2019.2938758
[57]	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, (2021), 10012–10022.
[58]	A. Trockman, J. Z. Kolter, Patches are all you need?, preprint, arXiv: 2201.09792. https://doi.org/10.48550/arXiv.2201.09792
[59]	Z. Peng, W. Huang, S. Gu, L. Xie, Y. Wang, J. Jiao, et al., Conformer: Local features coupling global representations for visual recognition, in Proceedings of the IEEE/CVF International Conference on Computer Vision, (2021), 367–376. https://doi.org/10.1109/ICCV48922.2021.00042

	$L_{\infty}$			$L_{2}$			$L_{rms}$
$t$	Present	^[32]	^[33]	Present	^[32]	^[33]	Present	^[32]	^[33]
0.2	$5.325\times10^{-7}$	$1.235\times10^{-5}$	$1.122\times10^{-5}$	$3.626\times10^{-6}$	1.312 $\times10^{-6}$	$7.154\times10^{-6}$	$1.124\times10^{-8}$	$4.567\times10^{-5}$	....
0.5	$1.203\times10^{-6}$	$5.198\times10^{-4}$	$7.584\times10^{-5}$	$8.234\times10^{-6}$	6.099 $\times10^{-6}$	$7.489\times10^{-6}$	$6.344\times10^{-8}$	$5.642\times10^{-5}$	....
1.0	$2.158\times10^{-6}$	$6.328\times10^{-4}$	$9.127\times10^{-5}$	$1.521\times10^{-5}$	2.121 $\times10^{-5}$	$8.689\times10^{-5}$	$2.164\times10^{-7}$	$8.167\times10^{-5}$	....
1.5	$2.996\times10^{-6}$	$8.538\times10^{-4}$	$9.242\times10^{-5}$	$2.213\times10^{-5}$	$3.234\times10^{-5}$	$7.027\times10^{-5}$	$4.160\times10^{-7}$	$2.368\times10^{-4}$	....

Mathematical Biosciences and Engineering

Improved breast ultrasound tumor classification using dual-input CNN with GAP-guided attention loss

Related Papers:

Abstract

1. Introduction

2. Main motive

3. Haar wavelet and its integrals

4. Proposed scheme

4.1. Temporal discretization

5. Simulations and comparison

5.1. Problem 1

5.2. Problem 2

6. Stability analysis

7. Concluding remarks

Author contributions

Use of AI tools declaration

Acknowledgments

Conflict of interest

References

Reader Comments

通讯作者: 陈斌, bchen63@163.com

Metrics

Figures and Tables

Other Articles By Authors

Related pages

Tools

Export File

Citation

Format

Content

Catalog