COVID-19 disease identification network based on weakly supervised feature selection

Jingyao Liu; Qinghe Feng; Yu Miao; Wei He; Weili Shi; Zhengang Jiang; Jingyao Liu; Qinghe Feng; Yu Miao; Wei He; Weili Shi; Zhengang Jiang

doi:10.3934/mbe.2023409

Mathematical Biosciences and Engineering

2023, Volume 20, Issue 5: 9327-9348. doi: 10.3934/mbe.2023409

Previous Article Next Article

Research article Special Issues

COVID-19 disease identification network based on weakly supervised feature selection

1.
School of Computer Science and Technology, Changchun University of Science and Technology, Changchun, Jilin 130022, China
2.
School of Computer and Information Engineering, Chuzhou University, Chuzhou 239000, China
3.
Zhongshan Institute of Changchun University of Science and Technology, Zhongshan 528437, China
4.
School of Intelligent Engineering, Henan Institute of Technology, Xinxiang 453003, China

Academic Editor: Yang Kuang

Received: 13 January 2023 Revised: 23 February 2023 Accepted: 06 March 2023 Published: 16 March 2023

The coronavirus disease 2019 (COVID-19) outbreak has resulted in countless infections and deaths worldwide, posing increasing challenges for the health care system. The use of artificial intelligence to assist in diagnosis not only had a high accuracy rate but also saved time and effort in the sudden outbreak phase with the lack of doctors and medical equipment. This study aimed to propose a weakly supervised COVID-19 classification network (W-COVNet). This network was divided into three main modules: weakly supervised feature selection module (W-FS), deep learning bilinear feature fusion module (DBFF) and Grad-CAM++ based network visualization module (Grad-Ⅴ). The first module, W-FS, mainly removed redundant background features from computed tomography (CT) images, performed feature selection and retained core feature regions. The second module, DBFF, mainly used two symmetric networks to extract different features and thus obtain rich complementary features. The third module, Grad-Ⅴ, allowed the visualization of lesions in unlabeled images. A fivefold cross-validation experiment showed an average classification accuracy of 85.3%, and a comparison with seven advanced classification models showed that our proposed network had a better performance.

Keywords:

Citation: Jingyao Liu, Qinghe Feng, Yu Miao, Wei He, Weili Shi, Zhengang Jiang. COVID-19 disease identification network based on weakly supervised feature selection[J]. Mathematical Biosciences and Engineering, 2023, 20(5): 9327-9348. doi: 10.3934/mbe.2023409

Related Papers:

[1]	Jagdev Singh, Jitendra Kumar, Devendra kumar, Dumitru Baleanu . A reliable numerical algorithm based on an operational matrix method for treatment of a fractional order computer virus model. AIMS Mathematics, 2024, 9(2): 3195-3210. doi: 10.3934/math.2024155
[2]	Khalaf M. Alanazi . The asymptotic spreading speeds of COVID-19 with the effect of delay and quarantine. AIMS Mathematics, 2024, 9(7): 19397-19413. doi: 10.3934/math.2024945
[3]	Hui Miao . Global stability of a diffusive humoral immunity viral infection model with time delays and two modes of transmission. AIMS Mathematics, 2025, 10(6): 14122-14139. doi: 10.3934/math.2025636
[4]	Wajaree Weera, Thongchai Botmart, Teerapong La-inchua, Zulqurnain Sabir, Rafaél Artidoro Sandoval Núñez, Marwan Abukhaled, Juan Luis García Guirao . A stochastic computational scheme for the computer epidemic virus with delay effects. AIMS Mathematics, 2023, 8(1): 148-163. doi: 10.3934/math.2023007
[5]	Xin Jiang . Threshold dynamics of a general delayed HIV model with double transmission modes and latent viral infection. AIMS Mathematics, 2022, 7(2): 2456-2478. doi: 10.3934/math.2022138
[6]	Kalpana Umapathy, Balaganesan Palanivelu, Renuka Jayaraj, Dumitru Baleanu, Prasantha Bharathi Dhandapani . On the decomposition and analysis of novel simultaneous SEIQR epidemic model. AIMS Mathematics, 2023, 8(3): 5918-5933. doi: 10.3934/math.2023298
[7]	Omar Kahouli, Imane Zouak, Ma'mon Abu Hammad, Adel Ouannas . Chaos, control and synchronization in discrete time computer virus system with fractional orders. AIMS Mathematics, 2025, 10(6): 13594-13621. doi: 10.3934/math.2025612
[8]	Liang Hong, Jie Li, Libin Rong, Xia Wang . Global dynamics of a delayed model with cytokine-enhanced viral infection and cell-to-cell transmission. AIMS Mathematics, 2024, 9(6): 16280-16296. doi: 10.3934/math.2024788
[9]	Ahmed M. Elaiw, Ghadeer S. Alsaadi, Aatef D. Hobiny . Global co-dynamics of viral infections with saturated incidence. AIMS Mathematics, 2024, 9(6): 13770-13818. doi: 10.3934/math.2024671
[10]	Nada A. Almuallem, Hasanur Mollah, Sahabuddin Sarwardi . A study of a prey-predator model with disease in predator including gestation delay, treatment and linear harvesting of predator species. AIMS Mathematics, 2025, 10(6): 14657-14698. doi: 10.3934/math.2025660

Abstract

1. Introduction

Since the first personal computer came out in 1980, computers gradually appeared in our daily life. In 1995, the emergence of the Internet further promoted the computers into all fields of production and living. By June 2020, China's Internet users had reached 940 million, an increase of 36.25 million compared with March 2020, and the Internet penetration rate reached 67.0%, an increase of 2.5 percentage points compared with March 2020 ^[1]. Computer network is a sharp double-edged sword, bringing conveniences as well as disasters. In October 2019, a total of 44.23 million new viruses was found in the National Computer Virus Emergency Response Center and 218.04 million computers were infected, which was 1.78% higher than that in September, and the main transmission channels were "phishing", "webpage pegging" and loopholes ^[2]. The propagation of computer virus has become more rapid and harmful, posing serious changes. In the early October 2019, Demant, the world's largest hearing aid manufacturer, was invaded by blackmail virus, resulting in a direct economic loss of more than 95 million dollars. Fractional derivative equations are often used to study the dynamic behavior of systems, which can help us understand the evolution law of the system ^[3,4,5,6]. Consequently, it is of great practical significance to analyze the propagation of computer virus to protect computers against viruses by use fractional derivative equations.

Some mathematical models, which characterized the spread of computer viruses over the internet, were proposed to help us study the problem quantitatively. There are many similarities between computer virus and biological virus, such as infectivity, destructiveness, variability and so on. Based on these similarities, J. O. Kephart and S. R. White applied the mathematical models of epidemics to the computer virus propagation model creatively ^[7]. On this foundation, many computer virus models have been established ^[8,9,10]. Singh et al. ^[11] considered a fractional epidemiological SIR model with an arbitrary order derivative having nonsingular kernel, and discussed the existence of the solution. Considering that the recovered nodes may become susceptible again once some new viruses appear or the known computer viruses mutate, Chen et al. ^[12] presented a new SIRS model. But they all assumed the infection rate in models is bilinear. But in fact, this situation is not the case. In most realistic situations, the bilinear infection rate is always impossible to achieve due to the increase of the susceptible computers and infectious computers. In view of the nonlinear infection rate, both of the inhibition effect owing to the uncertain behavior of susceptible computers and the crowding effect of infectious computers are considered at the same time.

Considering that the network topology in the proliferation of virus may lead to nonlinear infection rate, MadhuSudanan et al. ^[13] formulated a computer viruses model with nonlinear infection rate and incubation period delay:

$\begin{equation} \left\{ \begin{array}{lll} \frac{dS(t)}{dt}& = &(1-p)b-\frac{\beta S(t-\tau)I(t-\tau)}{1+\sigma S(t-\tau)}-dS(t)+\delta R(t), \\ \frac{dI(t)}{dt}& = &\frac{\beta S(t-\tau)I(t-\tau)}{1+\sigma S(t-\tau)}-(d+\alpha+\gamma)I(t), \\ \frac{dR(t)}{dt}& = &pb+\gamma I(t)-(d+\delta)R(t), \end{array} \right. \end{equation}$

(1.1)

where $S(t)$ , $I(t)$ , $R(t)$ represent the number of susceptible computers, infected computers and recovered computers at time $t$ , respectively. The meanings of all the parameters in system (1.1) can be referred to ^[13].

Quarantine strategy generally refers to the control of individuals with abnormal performance, so as to prevent others from being infected by viruses. Quarantine strategy is an important measure for the treatment of infectious diseases. It can not only conduct centralized management and treatment for infected individuals, but also effectively control the source of infection and greatly reduce the number of contacts. Later, inspired by the biological infectious disease model, many scholars applied the quarantine strategy to the research of computer virus model, and put forward a series of models accordingly ^[14,15]. Hence, quarantine strategy should be introduced into the computer virus model. The effect of Anti-virus can protect recovered computers from the known viruses, however, as time goes on, Anti-virus may lose function as a result of the emergence of new viruses and the variation of known viruses, and the update speed of anti-virus software is always slower than that of new virus. So it needs a short time before entering susceptible state, called the temporary immune time delay. Considering the effect of quarantine strategy and the existence of temporary immune time delay, we investigate a new SIQRS computer virus model with two delays:

$\begin{equation} \left\{ \begin{array}{llll} \frac{dS(t)}{dt}& = &(1-p)b-\frac{\beta S(t-\tau_1)I(t-\tau_1)}{1+\sigma S(t-\tau_1)}-dS(t)+\delta R(t-\tau_2), \\ \frac{dI(t)}{dt}& = &\frac{\beta S(t-\tau_1)I(t-\tau_1)}{1+\sigma S(t-\tau_1)}-(d+\alpha_1+\gamma+\varepsilon)I(t), \\ \frac{dQ(t)}{dt}& = &\varepsilon I(t)-(\eta+d+\alpha_2)Q(t), \\ \frac{dR(t)}{dt}& = &pb+\gamma I(t)+\eta Q(t)-dR(t)-\delta R(t-\tau_2), \end{array} \right. \end{equation}$

(1.2)

where $Q(t)$ is the number of quarantine computers at time $t$ ; $\alpha_1$ is the death rate of infected computers due to virus; $\alpha_2$ is the death rate of quarantine computers due to virus; $\varepsilon$ is the quarantine rate of infected computers; $\eta$ is the recovered rate of the quarantine computers; $\tau_1$ is the incubation period delay; $\tau_2$ is the temporary immune time delay before the recovered computers come into the susceptible status.

2. Local asymptotic stability of the virus-free equilibrium

When the system (1.2) reaches the virus-free equilibrium, there is no virus, namely $I_0^* = 0$ . Let us equate system (1.2) to be zero, we can obtain:

$\begin{equation} \left\{ \begin{array}{llll} \delta R_0^*+(1-p)b-\frac{\beta S_0^*I_0^*}{1+\sigma S_0^*}-dS_0^*& = &0, \\ \frac{\beta S_0^*I_0^*}{1+\sigma S_0^*}-(d+\alpha_1+\gamma+\varepsilon)I_0^*& = &0, \\ \varepsilon I_0^*-(\eta+d+\alpha_2)Q_0^*& = &0, \\ \gamma I_0^*+\eta Q_0^*+pb-(d+\delta)R_0^*& = &0, \end{array} \right. \end{equation}$

(2.1)

Then, then system (1.2) has a virus-free equilibrium $E_0^*(S_0^*, I_0^*, Q_0^*, R_0^*)$ . Here,

$\begin{eqnarray} S_0^*& = &\frac{\delta b+bd(1-p)}{d(d+\delta)}, \\ I_0^*& = &0, \\ Q_0^*& = &0, \\ R_0^*& = &\frac{p b}{d+\delta}. \end{eqnarray}$

The basic regeneration number is the critical threshold to determine whether there is a virus in system (1.2). According to the way in ^[16], it is easy to obtain the basic regeneration number of system (1.2). Let $X = (I, S, Q, R)^T$ , then system (1.2) can be equivalent to $\frac{dX(t)}{dt} = F-V$ , where

$F = \begin{pmatrix} \frac{\beta SI}{1+\sigma S}\\ {0}\\ {0}\\ {0} \end{pmatrix}, V = \begin{pmatrix} (d+\alpha_1+\eta+\gamma)I\\ -(1-p)b+\frac{\beta SI}{1+\sigma S}+dS+\delta R\\ (\eta+d+\alpha_2)Q-\varepsilon I\\ dR+\delta R-\eta Q-\gamma I-pb \end{pmatrix}.$

The infected compartment is $I$ , giving $m = 1$ , then the Jacobian matrixes of $F$ and $V$ at $E_0^*(S_0^*, I_0^*, Q_0^*, R_0^*)$ are

$F^\prime = \begin{pmatrix} \frac{\beta S_0^*}{1+\sigma S_0^*} \end{pmatrix}, V^\prime = \begin{pmatrix} d+\alpha_1+\eta+\gamma \end{pmatrix}.$

Then

$\begin{equation} R_0 = \frac{\beta S_0^*}{(1+\sigma S_0^*)(d+\alpha_1+\gamma+\varepsilon)}. \end{equation}$

(2.2)

If $R_0 < 1$ , then system (1.2) has a virus-free equilibrium $E_0^*(S_0^*, I_0^*, Q_0^*, R_0^*)$ . The Jacobian matrix of system (1.2) at $E_0^*(S_0^*, I_0^*, Q_0^*, R_0^*)$ is

$J(E_0^*) = \begin{pmatrix} -d \quad & -\frac{\beta S_0^*}{1+\sigma S_0^*} \quad & {0} \quad & \delta\\ {0} \quad & \frac{\beta S_0^*}{1+\sigma S_0^*}-(d+\alpha_1+\gamma+\varepsilon) \quad & {0} \quad & {0}\\ {0} \quad & \varepsilon \quad & -(\eta+d+\alpha_2) \quad & {0}\\ {0} \quad & \gamma \quad & \eta \quad & -(d+\delta) \quad \end{pmatrix},$

The corresponding characteristic equation becomes

$\begin{equation} (\lambda+d)(\lambda-\frac{\beta S_0^*}{1+\sigma S_0^*}+d+\alpha_1+\gamma+\varepsilon)(\lambda+d+\eta+\alpha_2)(\lambda+d+\delta) = 0. \end{equation}$

(2.3)

Then the eigenvalues of Eq.(2.3) are

$\begin{eqnarray} \lambda_1& = &-d < 0, \\ \lambda_2& = &\frac{\beta S_0^*}{1+\sigma S_0^*}-(d+\alpha_1+\gamma+\varepsilon) < 0, \\ \lambda_3& = &-(d+\eta+\alpha_2) < 0, \\ \lambda_4& = &-(d+\delta) < 0, \end{eqnarray}$

So, when all the eigenvalues are less than zero, the virus-free equilibrium of system (1.2) is locally stable according to Routh-Hurwitz criteria.

3. Local asymptotic stability of the virus-existence equilibrium and the occurrence of Hopf bifurcation

If $R_0 = \frac{\beta S_0^*}{(1+\sigma S_0^*)(d+\alpha_1+\gamma+\varepsilon)} > 1$ , then system (1.2) has a unique virus-existence equilibrium $E^*(S^*, I^*, Q^*, R^*)$ . Here,

$\begin{eqnarray} S^*&=&\frac{d+\alpha_1+\gamma+\varepsilon}{\beta-\sigma(d+\alpha_1+\gamma+\varepsilon)},\nonumber\\ I^*&=&\frac{(d+\delta)dS^*-(d+\delta)(1-p)b-\delta pb}{k_1+\delta\gamma-k_2},\nonumber\\ Q^*&=&\frac{\varepsilon}{\eta+d+\alpha_2}I^*,\nonumber\\ R^*&=&\frac{pb+\gamma I^*+\eta Q^*}{d+\delta},\nonumber \end{eqnarray}$

where

$\begin{eqnarray} k_1& = &\frac{\delta\eta\varepsilon}{\eta+d+\alpha_2}, \\ k_2& = &\frac{(\delta+d)\beta S^*}{1+\sigma S^*}. \end{eqnarray}$

The linearized part of system (1.2) is

$\begin{equation} \left\{ \begin{array}{llll} \frac{dS(t)}{dt}& = &l_{11}S(t)+m_{11}S(t-\tau_1)+m_{12}I(t-\tau_1)+n_{14}R(t-\tau_2), \\ \frac{dI(t)}{dt}& = &m_{21}S(t-\tau_1)+l_{22}I(t)+m_{22}I(t-\tau_1), \\ \frac{dQ(t)}{dt}& = &l_{32}I(t)+l_{33}Q(t), \\ \frac{dR(t)}{dt}& = &l_{42}I(t)+l_{43}Q(t)+l_{44}R(t)+n_{44}R(t-\tau_2), \end{array} \right. \end{equation}$

(3.1)

where

$\begin{eqnarray} l_{11}& = &-d, m_{11} = -\frac{\beta I^*}{(1+\sigma S^*)^2}, m_{12} = -\frac{\beta S^*}{1+\sigma S^*}, n_{14} = \delta, \\ m_{21}& = &\frac{\beta I^*}{(1+\sigma S^*)^2}, m_{22} = \frac{\beta S^*}{1+\sigma S^*}, l_{22} = -(d+\alpha_1+\gamma+\varepsilon), \\ l_{32}& = &\varepsilon, l_{33} = -(\eta+d+\alpha_2), l_{42} = \gamma, l_{43} = \eta, l_{44} = -d, n_{44} = -\delta, \end{eqnarray}$

From the system (3.1), we can obtain that

$\begin{eqnarray} X_0(\lambda)+X_1(\lambda)e^{-\lambda\tau_1}&+&X_2(\lambda)e^{-\lambda\tau_2}+X_3(\lambda)e^{-\lambda(\tau_1+\tau_2)}\\ &+&X_4(\lambda)e^{-2\lambda\tau_2}+X_5(\lambda)e^{-\lambda(2\tau_1+\tau_2)} = 0, \end{eqnarray}$

(3.2)

where

$\begin{eqnarray} X_0(\lambda)& = &\lambda^4+\lambda^3(-l_{11}-l_{22}-l_{33}-l_{44})+\lambda^2(l_{11}l_{22}+l_{11}l_{33}+l_{11}l_{44}+l_{22}l_{33}+l_{22}l_{44}+l_{33}l_{44})\\ &&+\lambda(-l_{11}l_{22}l_{33}-l_{11}l_{22}l_{44}-l_{11}l_{33}l_{44}-l_{22}l_{33}l_{44})+l_{11}l_{22}l_{33}l_{44}, \\ X_1(\lambda)& = &\lambda^3(-m_{11}-m_{22})+\lambda^2(l_{11}m_{22}+l_{22}m_{11}+l_{33}m_{11}+l_{33}m_{22}+l_{44}m_{11}+l_{44}m_{22})\\ &&+\lambda(-l_{33}l_{44}m_{22}-l_{11}l_{44}m_{22}-l_{11}l_{33}m_{22}-l_{33}l_{44}m_{11}-l_{22}l_{44}m_{11}-l_{22}l_{33}m_{11})\\ &&+l_{22}l_{33}l_{44}m_{22}+l_{22}l_{33}l_{44}m_{11}, \\ X_2(\lambda)& = &-n_{44}\lambda^3+\lambda^2(l_{11}n_{44}+l_{22}n_{44}+l_{33}n_{44})+\lambda(-l_{11}l_{22}n_{44}-l_{11}l_{33}n_{44}-l_{22}l_{33}n_{44})\\ &&+l_{11}l_{22}l_{33}n_{44}, \\ X_3(\lambda)& = &\lambda^2(m_{11}n_{44}+m_{22}n_{44})+\lambda(-l_{11}m_{22}n_{44}-l_{22}m_{22}n_{44}-l_{33}m_{11}n_{44}-l_{33}m_{22}n_{44}\\ &&+l_{42}m_{21}n_{44})+(l_{11}l_{33}m_{22}n_{44}+l_{22}l_{33}m_{11}n_{44}+l_{32}l_{43}m_{21}n_{14}-l_{33}l_{42}m_{21}n_{14}), \\ X_4(\lambda)& = &\lambda^2(m_{11}m_{22}+m_{12}m_{21})+\lambda(-l_{33}m_{11}m_{22}-l_{44}m_{11}m_{22}-l_{33}m_{12}m_{21}-l_{44}m_{12}m_{21})\\ &&+l_{33}l_{44}m_{11}m_{22}+l_{33}l_{44}m_{12}m_{21}, \\ X_5(\lambda)& = &\lambda(-m_{11}m_{22}n_{44}-m_{12}m_{21}n_{44})+l_{33}m_{11}m_{22}n_{44}+l_{33}m_{12}m_{21}n_{44}. \end{eqnarray}$

Case 1. $\tau_1 = \tau_2 = 0$ , Eq (3.2) becomes

$\begin{eqnarray} \lambda^4+X_0^3\lambda^3+X_0^2\lambda^2+X_0^1\lambda+X_0^0 = 0, \end{eqnarray}$

(3.3)

where $X_i^j(i = 0, 1, 2, 3, 4, 5; j = 0, 1, 2, 3, 4)$ represents the coefficient of $\lambda^j$ in $X_i(\lambda)$ .

Lemma 1 ^[13]. According to Routh-Hurwitz criteria, when $R_0 > 1$ , the virus-existence equilibrium $E^*(S^*, I^*, Q^*, R^*)$ is locally asymptotically stable.

Case 2. $\tau_1 > 0$ , $\tau_2 = 0$ . Then, Eq (3.2) becomes

$\begin{equation} [X_0(\lambda)+X_2(\lambda)]+[X_1(\lambda)+X_3(\lambda)]e^{-\lambda\tau_1}+[X_4(\lambda)+X_5(\lambda)]e^{-2\lambda\tau_1} = 0. \end{equation}$

(3.4)

Taking $\lambda = i\omega_1$ into Eq (3.4) and separating the real and imaginary parts, we obtain

$\begin{equation} \left\{ \begin{array}{ll} A_{11}\cos\tau_1\omega_1+A_{21}\sin\tau_1\omega_1+B_{11} = -A_{31}\sin2\tau_1\omega_1+A_{41}\cos2\tau_1\omega_1, \\ A_{21}\cos\tau_1\omega_1-A_{11}\sin\tau_1\omega_1+B_{21} = -A_{31}\cos2\tau_1\omega_1-A_{41}\sin2\tau_1\omega_1, \end{array} \right. \end{equation}$

(3.5)

with

$\begin{eqnarray} A_{11}& = &X_1^0-X_1^2\omega_1^2+X_3^0-X_3^2\omega_1^2, \\ A_{21}& = &X_1^1\omega_1-X_1^3\omega_1^3+X_3^1\omega_1, \\ A_{31}& = &X_4^1\omega_1+X_5^1\omega_1, \\ A_{41}& = &X_4^2\omega_1^2-X_4^0-X_5^0, \\ B_{11}& = &\omega_1^4-X_0^2\omega_1^2+X_0^0-X_2^2\omega_1^2+X_2^0, \\ B_{21}& = &X_0^1\omega_1^1-X_0^3\omega_1^3-X_2^3\omega_1^3-X_2^1\omega_1, \end{eqnarray}$

Because $\cos^2\tau_1\omega_1+\sin^2\tau_1\omega_1 = 1$ , $\sin\tau_1\omega_1 = \pm \sqrt{1-\cos^2\tau_1\omega_1}$ .

(1) If $\sin\tau_1\omega_1 = \sqrt{1-\cos^2\tau_1\omega_1}$ , after calculation, we have

$\begin{eqnarray} A_{11}^2+A_{21}^2+B_{11}^2+B_{21}^2-A_{31}^2-A_{41}^2+2(A_{11}B_{11}+A_{21}B_{21})\cos\tau_1\omega_1\\ +2(B_{11}A_{21}-A_{11}B_{21})\sqrt{1-\cos^2\tau_1\omega_1} = 0. \end{eqnarray}$

(3.6)

Let $f_1(\omega_1) = \cos\tau_1\omega_1$ , and we suppose that $(G_1)$ : $f_1(\omega_1) = \cos\tau_1\omega_1$ has at least a positive root $\omega_{11}$ , which makes Eq (3.6) true. Thus,

$\begin{equation} \tau_{11}^{(i)} = \frac{1}{\omega_{11}}\times[\arccos(f_1(\omega_{11}))+2i\pi], i = 0, 1, 2, \cdots. \end{equation}$

(3.7)

(2) If $\sin\tau_1\omega_1 = -\sqrt{1-\cos^2\tau_1\omega_1}$ , after calculation, we have

$\begin{eqnarray} A_{11}^2+A_{21}^2+B_{11}^2+B_{21}^2-A_{31}^2-A_{41}^2+2(A_{11}B_{11}+A_{21}B_{21})\cos\tau_1\omega_1\\ +2(A_{11}B_{21}-A_{21}B_{11})\sqrt{1-\cos^2\tau_1\omega_1} = 0. \end{eqnarray}$

(3.8)

Let $g_1(\omega_1) = \cos\tau_1\omega_1$ , and we suppose that $(G_2)$ : $g_1(\omega_1) = \cos\tau_1\omega_1$ has at least a positive root $\omega_{12}$ , which makes Eq (3.8) true. Thus,

$\begin{equation} \tau_{12}^{(i)} = \frac{1}{\omega_{12}}\times[\arccos(g_1(\omega_{12}))+2i\pi], i = 0, 1, 2, \cdots. \end{equation}$

(3.9)

Define

$\begin{equation} \tau_{10} = \text{min}\{\tau_{11}^{(i)}, \tau_{12}^{(i)}\}, i = 0, 1, 2, \cdots, \end{equation}$

(3.10)

where $\tau_{11}^{(i)}$ and $\tau_{12}^{(i)}$ are defined by Eq (3.7) and Eq (3.9), respectively.

Multiplying $e^{\lambda\tau_1}$ on both sides of Eq (3.4), and then after deriving from $\tau$ to $\lambda$ , we can get

$\begin{equation} \bigg[\frac{d\lambda}{d\tau_1}\bigg]^{-1} = -\frac{[X_0^\prime(\lambda)+X_2^\prime(\lambda)]e^{\lambda\tau_1}+[X_1^\prime(\lambda)+X_3^\prime(\lambda)]+[X_4^\prime(\lambda)+X_5^\prime(\lambda)]e^{-\lambda\tau_1}}{-\lambda[X_0(\lambda)+X_2(\lambda)]e^{-\lambda\tau_1}+\lambda[X_4(\lambda)+X_5(\lambda)]e^{\lambda\tau_1}}-\frac{\tau_1}{\lambda}. \end{equation}$

(3.11)

According to the Hopf bifurcation theorem ^[17], if the surmise $(G_3)$ : $\text{Re}[d\lambda/d\tau_1]^{-1}_{\tau_1 = \tau_{10}}\neq0$ is true, the virus-existence equilibrium $E^*(S^*, I^*, Q^*, R^*)$ is locally asymptotically stable. So, we have Theorem 1.

Theorem 1. For system (1.2), when $R_0 > 1$ and the conditions $(G_1)$ - $(G_3)$ hold, then $E^*(S^*, I^*,$ $Q^*, R^*)$ is locally asymptotically stable when $\tau_1\in[0, \tau_{10})$ ; there is a Hopf bifurcation at $E^*(S^*, I^*,$ $Q^*, R^*)$ when $\tau_1 = \tau_{10}$ .

Case 3. $\tau_1 = 0$ , $\tau_2 > 0$ . Then Eq (3.2) becomes

$\begin{equation} [X_0(\lambda)+X_1(\lambda)+X_4(\lambda)]+[X_2(\lambda)+X_3(\lambda)+X_5(\lambda)]e^{-\lambda\tau_2} = 0, \end{equation}$

(3.12)

Substituting $\lambda = i\omega_2$ into Eq (3.12), we obtain

$\begin{equation} \left\{ \begin{array}{ll} C_{11}\cos\tau_2\omega_2+C_{21}\sin\tau_2\omega_2 = D_{11}, \\ C_{21}\cos\tau_2\omega_2-C_{11}\sin\tau_2\omega_2 = D_{21}, \end{array} \right. \end{equation}$

(3.13)

with

$\begin{eqnarray} C_{11}& = &-X_2^2\omega_2^2+X_2^0+X_3^0+X_5^0-X_3^2\omega_2^2, \\ C_{21}& = &-X_2^3\omega_2^3+X_2^1\omega_2+X_3^1\omega_2+X_5^1\omega_2, \\ D_{11}& = &X_0^2\omega_2^2+X_1^2\omega_2^2+X_4^2\omega_2^2-\omega_2^4-X_0^0-X_1^0-X_4^0, \\ D_{21}& = &X_0^3\omega_2^3+X_1^3\omega_2^3-X_0^1\omega_2-X_1^1\omega_2-X_4^1\omega_2, \end{eqnarray}$

Squaring both sides of two equations in Eq (3.13), and adding them up, we obtain

$\begin{eqnarray} C_{11}^2+C_{21}^2 = D_{11}^2+D_{21}^2. \end{eqnarray}$

(3.14)

We suppose that $(G_4)$ : Eq (3.14) has at least one positive real root $\omega_{20}$ . Then, from Eq (3.13), we derive

$\begin{equation} \tau_2^{(i)} = \frac{1}{\omega_{20}}\times\bigg[\arccos\frac{C_{11}D_{11}+C_{21}D_{21}}{C_{11}^2+C_{21}^2}+2i\pi\bigg], \end{equation}$

(3.15)

where $i = 0, 1, 2, \cdots$ .

Define

$\begin{equation} \tau_{20} = \text{min}\{\tau_2^{(i)}, i = 0, 1, 2, \cdots\}, \end{equation}$

(3.16)

and $\tau_2^{(i)}$ is defined by Eq (3.15).

Taking the derivative of $\lambda$ with respect to $\tau$ , we obtain

$\begin{equation} \bigg[\frac{d\lambda}{d\tau_2}\bigg]^{-1} = -\frac{X_0^\prime+X_1^\prime+X_4^\prime}{\lambda[X_0+X_1+X_4]}+\frac{X_2^\prime+X_3^\prime+X_5^\prime}{\lambda[X_2+X_3+X_5]}-\frac{\tau_2}{\lambda}, \end{equation}$

(3.17)

Thus, it is easy to obtain the expression of $\text{Re}[d\lambda/d\tau_2]^{-1}_{\tau_2 = \tau_{20}}$ . According to the Hopf bifurcation theorem ^[17], if the hypothesis $(G_5)$ : $\text{Re}[d\lambda/d\tau_2]^{-1}_{\tau_2 = \tau_{20}}\neq0$ is true, the virus-existence equilibrium $E^*(S^*, I^*, Q^*, R^*)$ is locally asymptotically stable. In conclusion, Theorem 2 can be obtained.

Theorem 2. For system (1.2), when $R_0 > 1$ and the conditions $(G_4)$ - $(G_5)$ hold, then $E^*(S^*, I^*,$ $Q^*, R^*)$ is locally asymptotically stable when $\tau_2\in[0, \tau_{20})$ ; there is a Hopf bifurcation at $E^*(S^*, I^*,$ $Q^*, R^*)$ when $\tau_2 = \tau_{20}$ .

Case 4. $\tau_1 = \tau_2 = \tau_*$ . Then Eq (3.2) becomes

$\begin{equation} X_0(\lambda)+[X_1(\lambda)+X_2(\lambda)]e^{-\lambda\tau_*}+[X_3(\lambda)+X_4(\lambda)]e^{-2\lambda\tau_*}+X_5(\lambda)e^{-3\lambda\tau_*} = 0, \end{equation}$

(3.18)

Multiplying $e^{\lambda\tau_*}$ on both sides of Eq (3.18), then we obtain

$\begin{equation} X_0(\lambda)e^{\lambda\tau_*}+[X_1(\lambda)+X_2(\lambda)]+[X_3(\lambda)+X_4(\lambda)]e^{-\lambda\tau_*}+X_5(\lambda)e^{-2\lambda\tau_*} = 0, \end{equation}$

(3.19)

Substituting $\lambda = i\omega_3$ into Eq (3.19), we obtain

$\begin{equation} \left\{ \begin{array}{ll} A_{12}\cos\tau_*\omega_3+A_{22}\sin\tau_*\omega_3 = -A_{32}\sin2\tau_*\omega_3+A_{42}\cos2\tau*\omega_3, \\ A_{22}^\prime\cos\tau_*\omega_3-A_{12}\sin\tau*\omega_3 = -A_{32}\cos2\tau*\omega_3-A_{42}\sin2\tau_*\omega_3, \end{array} \right. \end{equation}$

(3.20)

with

$\begin{eqnarray} A_{12}& = &X_0^3\omega_3^3-X_0^1\omega_3-X_3^1\omega_3-X_4^1\omega_3, \\ A_{22}& = &\omega_3^4-X_0^2\omega_3^2-X_3^2\omega_3^2+X_4^2\omega_3^2+X_0^0-X_3^0-X_4^0, \\ A_{22}^\prime& = &\omega_3^4-X_0^2\omega_3^2-X_3^2\omega_3^2-X_4^2\omega_3^2+X_0^0+X_3^0+X_4^0, \\ A_{32}& = &X_5^1\omega_3, \\ A_{42}& = &-X_5^0, \\ B_{12}& = &-X_1^2\omega_3^2+X_1^0-X_2^2\omega_3^2+X_2^0, \\ B_{22}& = &-X_1^3\omega_3^3-X_2^3\omega_3^3+X_1^1\omega_3+X_2^1\omega_3, \end{eqnarray}$

Squaring both sides of two equations in Eq (3.20), and adding them up, we obtain

$\begin{eqnarray} (A_{12}\cos\tau_*\omega_3+A_{22}\sin\tau_*\omega_3+B_{12})^2+(A_{22}^\prime\cos\tau_*\omega_3-A_{12}\sin\tau_*\omega_3+B_{22})^2\\ & = &A_{32}^2+A_{42}^2. \end{eqnarray}$

(3.21)

Because $\cos^2\tau_*\omega_3+\sin^2\tau_*\omega_3 = 1$ , $\sin\tau_*\omega_3 = \pm \sqrt{1-\cos^2\tau_*\omega_3}$ .

(1) If $\sin\tau_*\omega_3 = \sqrt{1-\cos^2\tau_*\omega_3}$ , after calculation, we have

$\begin{eqnarray} A_{12}^2+A_{22}^2+B_{12}^2+B_{22}^2-A_{32}^2-A_{42}^2+2(A_{12}A_{22}+A_{12}B_{12}+A_{22}B_{12})\cos\tau_*\omega_3\\ +2(A_{22}^\prime B_{22}-A_{22}^\prime A_{12}-A_{12}B_{22})\sqrt{1-\cos^2\tau_*\omega_3} = 0. \end{eqnarray}$

(3.22)

Let $f_2(\omega_3) = \cos\tau_*\omega_3$ , and we suppose that $(G_6)$ : $f_2(\omega_3) = \cos\tau_*\omega_3$ has at least a positive root $\omega_{31}$ , which makes Eq (3.22) true. Thus,

$\begin{equation} \tau_{*1}^{(i)} = \frac{1}{\omega_{31}}\times[\arccos(f_2(\omega_{31}))+2i\pi], i = 0, 1, 2, \cdots. \end{equation}$

(3.23)

(2) If $\sin\tau_*\omega_3 = -\sqrt{1-\cos^2\tau_*\omega_3}$ , after calculation, we have

$\begin{eqnarray} A_{12}^2+A_{22}^2+B_{12}^2+B_{22}^2-A_{32}^2-A_{42}^2+2(A_{12}A_{22}+A_{12}B_{12}+A_{22}B_{12})\cos\tau_*\omega_3\\ -2(A_{22}^\prime B_{22}-A_{22}^\prime A_{12}-A_{12}B_{22})\sqrt{1-\cos^2\tau_*\omega_3} = 0. \end{eqnarray}$

(3.24)

Let $g_2(\omega_3) = \cos\tau_*\omega_3$ , and we suppose that $(G_7)$ : $g_2(\omega_3) = \cos\tau_*\omega_3$ has at least a positive root $\omega_{32}$ , which makes Eq (3.24) true. Thus,

$\begin{equation} \tau_{*2}^{(i)} = \frac{1}{\omega_{32}}\times[\arccos(g_2(\omega_{32}))+2i\pi], i = 0, 1, 2, \cdots. \end{equation}$

(3.25)

Define

$\begin{equation} \tau_{*0} = \text{min}\{\tau_{*1}^{(i)}, \tau_{*2}^{(i)}\}, i = 0, 1, 2, \cdots, \end{equation}$

(3.26)

where $\tau_{*1}^{(i)}$ and $\tau_{*2}^{(i)}$ are defined by Eq (3.23) and Eq (3.25), respectively.

Then after deriving from $\tau$ to $\lambda$ , we can get

$\begin{equation} \bigg[\frac{d\lambda}{d\tau_*}\bigg]^{-1} = -\frac{X_0^\prime(\lambda)e^{\lambda\tau_*}+[X_1^\prime(\lambda)+X_2^\prime(\lambda)]+[X_3^\prime(\lambda)+X_4^\prime(\lambda)]e^{-\lambda\tau_*}+X_5^\prime(\lambda)e^{-2\lambda\tau_*}}{-\lambda X_0(\lambda)e^{\lambda\tau_*}+\lambda [X_3(\lambda)+X_4(\lambda)]e^{-\lambda\tau_*}+2\lambda X_5(\lambda)e^{-2\lambda\tau_*}}-\frac{\tau_*}{\lambda}. \end{equation}$

(3.27)

Based on the Hopf bifurcation theorem ^[17], if the surmise $(G_8)$ : $\text{Re}[d\lambda/d\tau_*]^{-1}_{\tau_* = \tau_{*0}}\neq0$ is true, the virus-existence equilibrium $E^*(S^*, I^*, Q^*, R^*)$ is locally asymptotically stable. Therefore, Theorem 3 can be obtained.

Theorem 3. For system (1.2), when $R_0 > 1$ and the conditions $(G_6)$ - $(G_8)$ hold, then $E^*(S^*, I^*,$ $Q^*, R^*)$ is locally asymptotically stable when $\tau_*\in[0, \tau_{*0})$ ; there is a Hopf bifurcation at $E^*(S^*, I^*,$ $Q^*, R^*)$ when $\tau_* = \tau_{*0}$ .

Case 5. $\tau_1 > 0$ , $\tau_2\in(0, \tau_{20})$ . For convenience, let $\omega_4$ be equal to $\omega_1$ . Then, this case is similar as in Case 2.

$\begin{equation} [X_0(\lambda)+X_1(\lambda)+X_4(\lambda)]+[X_1(\lambda)+X_3(\lambda)]e^{-\lambda\tau_1}+[X_4(\lambda)+X_5(\lambda)]e^{-2\lambda\tau_1} = 0. \end{equation}$

(3.28)

$\begin{equation} \left\{ \begin{array}{ll} A_{13}\cos\tau_{1*}\omega_4+A_{23}\sin\tau_{1*}\omega_4+B_{13} = -A_{33}\sin2\tau_{1*}\omega_4+A_{43}\cos2\tau_{1*}\omega_4, \\ A_{23}\cos\tau_{1*}\omega_4-A_{13}\sin\tau_{1*}\omega_4+B_{23} = -A_{33}\cos2\tau_{1*}\omega_4-A_{43}\sin2\tau_{1*}\omega_4, \end{array} \right. \end{equation}$

(3.29)

with

$\begin{eqnarray} A_{13}& = &X_1^0-X_1^2\omega_4^2+X_3^0-X_3^2\omega_4^2, \\ A_{23}& = &X_1^1\omega_4-X_1^3\omega_4^3+X_3^1\omega_4, \\ A_{33}& = &X_4^1\omega_4+X_5^1\omega_4, \\ A_{43}& = &X_4^2\omega_4^2-X_4^0-X_5^0, \\ B_{13}& = &\omega_4^4-X_0^2\omega_4^2+X_0^0-X_2^2\omega_4^2+X_2^0, \\ B_{23}& = &X_0^1\omega_4^1-X_0^3\omega_4^3-X_2^3\omega_4^3-X_2^1\omega_4, \end{eqnarray}$

Because $\cos^2\tau_{1*}\omega_4+\sin^2\tau_{1*}\omega_4 = 1$ , $\sin\tau_{1*}\omega_4 = \pm \sqrt{1-\cos^2\tau_{1*}\omega_4}$ .

(1) If $\sin\tau_{1*}\omega_4 = \sqrt{1-\cos^2\tau_{1*}\omega_4}$ , after calculation, we have

$\begin{eqnarray} A_{13}^2+A_{23}^2+B_{13}^2+B_{23}^2-A_{33}^2-A_{43}^2+2(A_{13}B_{13}+A_{23}B_{23})\cos\tau_{1*}\omega_4\\ +2(B_{13}A_{23}-A_{13}B_{23})\sqrt{1-\cos^2\tau_{1*}\omega_4} = 0. \end{eqnarray}$

(3.30)

Let $f_3(\omega_4) = \cos\tau_{1*}\omega_4$ , and we suppose that $(G_9)$ : $f_3(\omega_4) = \cos\tau_{1*}\omega_4$ has at least a positive root $\omega_{41}$ , which makes Eq (3.30) true. Thus,

$\begin{equation} \tau_{1*1}^{(i)} = \frac{1}{\omega_{41}}\times[\arccos(f_3(\omega_{41}))+2i\pi], i = 0, 1, 2, \cdots. \end{equation}$

(3.31)

(2) If $\sin\tau_{1*}\omega_4 = -\sqrt{1-\cos^2\tau_{1*}\omega_4}$ , after calculation, we have

$\begin{eqnarray} A_{13}^2+A_{23}^2+B_{13}^2+B_{23}^2-A_{33}^2-A_{43}^2+2(A_{13}B_{13}+A_{23}B_{23})\cos\tau_{1*}\omega_4\\ +2(A_{13}B_{23}-A_{23}B_{13})\sqrt{1-\cos^2\tau_{1*}\omega_4} = 0. \end{eqnarray}$

(3.32)

Let $g_3(\omega_4) = \cos\tau_{1*}\omega_4$ , and we suppose that $(G_{10})$ : $g_3(\omega_4) = \cos\tau_{1*}\omega_4$ has at least a positive root $\omega_{42}$ , which makes Eq (3.32) true. Thus,

$\begin{equation} \tau_{1*2}^{(i)} = \frac{1}{\omega_{42}}\times[\arccos(g_3(\omega_{42}))+2i\pi], i = 0, 1, 2, \cdots. \end{equation}$

(3.33)

Define

$\begin{equation} \tau_{1*0} = \text{min}\{\tau_{1*1}^{(i)}, \tau_{1*2}^{(i)}\}, i = 0, 1, 2, \cdots, \end{equation}$

(3.34)

where $\tau_{1*1}^{(i)}$ and $\tau_{1*2}^{(i)}$ are defined by Eq (3.31) and Eq (3.33), respectively. Multiplying $e^{\lambda\tau_{1*}}$ on both sides of Eq (3.28), and then after deriving from $\tau$ to $\lambda$ , we can get

$\begin{equation} \bigg[\frac{d\lambda}{d\tau_{1*}}\bigg]^{-1} = -\frac{[X_0^\prime(\lambda)+X_2^\prime(\lambda)]e^{\lambda\tau_{1*}}+[X_1^\prime(\lambda)+X_3^\prime(\lambda)]+[X_4^\prime(\lambda)+X_5^\prime(\lambda)]e^{-\lambda\tau_{1*}}}{-\lambda[X_0(\lambda)+X_2(\lambda)]e^{-\lambda\tau_{1*}}+\lambda[X_4(\lambda)+X_5(\lambda)]e^{\lambda\tau_{1*}}}-\frac{\tau_{1*}}{\lambda}. \end{equation}$

(3.35)

According to the Hopf bifurcation theorem ^[17], if the surmise $(G_{11})$ : $\text{Re}[d\lambda/d\tau_{1*}]^{-1}_{\tau_{1*} = \tau_{1*0}}\neq0$ is true, the virus-existence equilibrium $E^*(S^*, I^*, Q^*, R^*)$ is locally asymptotically stable. Thus, we have Theorem 4.

Theorem 4. For system (1.2), when $R_0 > 1$ and the conditions $(G_9)$ - $(G_{11})$ hold, then $E^*(S^*, I^*,$ $Q^*, R^*)$ is locally asymptotically stable when $\tau_{1*}\in[0, \tau_{1*0})$ ; there is a Hopf bifurcation at $E^*(S^*, I^*, Q^*, R^*)$ when $\tau_{1*} = \tau_{1*0}$ .

Case 6. $\tau_1\in[0, \tau_{10})$ , $\tau_2 > 0$ . Assume that $\lambda = i\omega_5$ is the root of Eq (3.2). For convenience, let $\omega_5$ be equal to $\omega_2$ . Then, this case is similar as in Case 3. Then we can get

$\begin{equation} [X_0(\lambda)+X_1(\lambda)+X_4(\lambda)]+[X_2(\lambda)+X_3(\lambda)+X_5(\lambda)]e^{-\lambda\tau_{2*}} = 0, \end{equation}$

(3.36)

Substituting $\lambda = i\omega_5$ into Eq.(3.36), we obtain

$\begin{equation} \left\{ \begin{array}{ll} C_{12}\cos\tau_{2*}\omega_5+C_{22}\sin\tau_{2*}\omega_5 = D_{12}, \\ C_{22}\cos\tau_{2*}\omega_5-C_{12}\sin\tau_{2*}\omega_5 = D_{22}, \end{array} \right. \end{equation}$

(3.37)

with

$\begin{eqnarray} C_{12}& = &-X_2^2\omega_5^2+X_2^0+X_3^0+X_5^0-X_3^2\omega_5^2, \\ C_{22}& = &-X_2^3\omega_5^3+X_2^1\omega_5+X_3^1\omega_5+X_5^1\omega_5, \\ D_{12}& = &X_0^2\omega_5^2+X_1^2\omega_5^2+X_4^2\omega_5^2-\omega_5^4-X_0^0-X_1^0-X_4^0, \\ D_{22}& = &X_0^3\omega_5^3+X_1^3\omega_5^3-X_0^1\omega_5-X_1^1\omega_5-X_4^1\omega_5, \end{eqnarray}$

Squaring both sides of two equations in Eq.(3.37), and adding them up, we obtain

$\begin{eqnarray} C_{12}^2+C_{22}^2 = D_{12}^2+D_{22}^2. \end{eqnarray}$

(3.38)

We suppose that $(G_{12})$ : Eq (3.38) has at least one positive real root $\omega_{50}$ . Then, from Eq (3.36), we derive

$\begin{equation} \tau_{2*}^{(i)} = \frac{1}{\omega_{50}}\times\bigg[\arccos\frac{C_{13}D_{13}+C_{13}D_{13}}{C_{13}^2+C_{23}^2}+2i\pi\bigg], \end{equation}$

(3.39)

where $i = 0, 1, 2, \cdots$ .

Define

$\begin{equation} \tau_{2*0} = \text{min}\{\tau_{2*}^{(i)}, 1 = 0, 1, 2, \cdots\}, \end{equation}$

(3.40)

and $\tau_{2*}^{(i)}$ is defined by Eq (3.39).

Taking the derivative of $\lambda$ with respect to $\tau$ , we obtain

$\begin{equation} \bigg[\frac{d\lambda}{d\tau_{2*}}\bigg]^{-1} = -\frac{X_0^\prime+X_1^\prime+X_4^\prime}{\lambda[X_0+X_1+X_4]}+\frac{X_2^\prime+X_3^\prime+X_5^\prime}{\lambda[X_2+X_3+X_5]}-\frac{\tau_{2*}}{\lambda}, \end{equation}$

(3.41)

Thus, it is easy to obtain the expression of $\text{Re}[d\lambda/d\tau_{2*}]^{-1}_{\tau_{2*} = \tau_{2*0}}$ . Based on the Hopf bifurcation theorem ^[17], when the hypothesis $(G_{13})$ : $\text{Re}[d\lambda/d\tau_{2*}]^{-1}_{\tau_{2*} = \tau_{2*0}}\neq0$ is true. In conclusion, Theorem 5 can be gotten.

Theorem 5. For system (1.2), when $R_0 > 1$ and the conditions $(G_{12})$ - $(G_{13})$ hold, then $E^*(S^*, I^*,$ $Q^*, R^*)$ is locally asymptotically stable when $\tau_{2*}\in[0, \tau_{2*0})$ ; there is a Hopf bifurcation at $E^*(S^*, I^*, Q^*, R^*)$ when $\tau_{2*} = \tau_{2*0}$ .

4. Direction and stability of Hopf bifurcation

Center manifold theory is one of the important theories for studying Hopf bifurcation. Considering this idea, in this section, we use the method in ^[18,19] to study direction and stability of Hopf bifurcation of system (1.2). We assume that $\tau_2^* < \tau_1^*$ , where $\tau_2^*\in(0, \tau_{20})$ . Let $\tau_1 = \tau_1^*+\varpi(\varpi\in \text{R})$ , $\chi_1 = S(\tau_1 t)$ , $\chi_2 = I(\tau_1 t)$ , $\chi_3 = Q(\tau_1 t)$ , $\chi_4 = R(\tau_1 t)$ . System (1.2) becomes

$\begin{equation} \dot{\chi}(t) = L_\varpi(\chi_t)+F(\varpi, \chi_t), \end{equation}$

(4.1)

where $\chi(t) = (\chi_1, \chi_2, \chi_3, \chi_4)^T \in C = C([-1, 0], \text{R}^4)$ and $L_\varpi$ : $C\rightarrow \text{R}^4$ and $F$ : $\text{R}\times C\rightarrow \text{R}^4$ are defined as

$\begin{equation} L_{\varpi}\varphi = (\tau_1^*+\varpi)\bigg(L^{\prime}\varphi(0)+M^{\prime}\varphi\bigg(-\frac{\tau_2^*}{\tau_1^*}\bigg)+N^{\prime}\varphi(-1)\bigg), \end{equation}$

(4.2)

and

$\begin{equation} F(\varpi, \varphi) = (\tau_1^*+\varpi)[F_1, F_2, 0, 0]^T, \end{equation}$

(4.3)

with

$L^{\prime} = \begin{pmatrix} l_{11} \quad & {0} \quad & {0} \quad & {0}\\ {0} \quad & l_{22} \quad & {0} \quad & {0}\\ {0} \quad & l_{32} \quad & l_{33} \quad & {0}\\ {0} \quad & l_{42} \quad & l_{43} \quad & l_{44} \end{pmatrix}, M^{\prime} = \begin{pmatrix} {0} \quad & m_{12} \quad & {0} \quad & {0}\\ m_{21} \quad & m_{22} \quad & {0} \quad & {0}\\ {0} \quad & {0} \quad & {0} \quad & {0}\\ {0} \quad & {0} \quad & {0} \quad & {0} \end{pmatrix}, N^{\prime} = \begin{pmatrix} {0} \quad & {0} \quad & {0} \quad & n_{14}\\ {0} \quad & {0} \quad & {0} \quad & {0}\\ {0} \quad & {0} \quad & {0} \quad & {0}\\ {0} \quad & {0} \quad & {0} \quad & n_{44} \end{pmatrix},$

and

$\begin{eqnarray} F_1& = &h_{11}\varphi_1(0)\varphi_2(0)+h_{12}\varphi_1^2(0)+h_{13}\varphi_1^2(0)\varphi_2(0)+h_{14}\varphi_1^3(0)+\cdots, \\ F_2& = &h_{21}\varphi_1(0)\varphi_2(0)+h_{22}\varphi_1^2(0)+h_{23}\varphi_1^2(0)\varphi_2(0)+h_{24}\varphi_1^3(0)+\cdots, \end{eqnarray}$

$\begin{eqnarray} h_{11}& = &-\frac{\beta}{(1+\sigma S^*)^2}, h_{12} = -\frac{2\sigma\beta I^*}{(1+\sigma S^*)^3}, h_{13} = \frac{2\sigma\beta}{(1+\sigma S^*)^3}, h_{14} = \frac{6\sigma^2\beta I^*}{(1+\sigma S^*)^4}, \\ h_{21}& = &\frac{\beta}{(1+\sigma S^*)^2}, h_{22} = \frac{2\sigma\beta I^*}{(1+\sigma S^*)^3}, h_{23} = -\frac{2\sigma\beta I^*}{(1+\sigma S^*)^3}, h_{24} = -\frac{6\sigma^2\beta I^*}{(1+\sigma S^*)^4}. \end{eqnarray}$

According to Riesz representation theorem, there exists $\eta(\vartheta, \varpi)$ and $\vartheta\in[-1, 0)$ such that

$\begin{equation} L_{\varpi}\varphi = \int_{-1}^0d\eta(\vartheta, \varpi)\varphi(\vartheta). \end{equation}$

(4.4)

In fact, we choose

$\begin{equation} \eta(\vartheta, \varpi) = \left\{ \begin{array}{llll} (\tau_1^*+\varpi)(L^{\prime}+M^{\prime}+N^{\prime}), \quad \vartheta = 0, \\ (\tau_1^*+\varpi)(M^{\prime}+N^{\prime}), \quad \vartheta\in[-\frac{\tau_2^*}{\tau_1^*}, 0), \\ (\tau_1^*+\varpi)(N^{\prime}), \quad \vartheta\in(-1, -\frac{\tau_2^*}{\tau_1^*}), \\ 0, \quad \vartheta = -1. \end{array} \right. \end{equation}$

(4.5)

with $\phi(\vartheta)$ is the Dirac delta function.

For $\varphi\in C([-1, 0], R^4)$ , define

$A(\varpi)\varphi = \left\{ \begin{array}{ll} \frac{d\varphi(\vartheta)}{d\vartheta}, \quad -1\leq\vartheta < 0, \\ \int_{-1}^0d\eta(\vartheta, \varpi)\varphi(\vartheta), \quad \vartheta = 0, \end{array} \right.$

and

$R(\varpi)\varphi = \left\{ \begin{array}{ll} 0, \quad -1\leq\vartheta < 0, \\ F(\varpi, \varphi), \quad \vartheta = 0. \end{array} \right.$

Then system (1.2) is equivalent to

$\begin{equation} \dot{\chi}(t) = A(\varpi)\chi_t+R(\varpi)\chi_t. \end{equation}$

(4.6)

For $\psi\in C^1([0, 1], (R^4)^*)$ , define

$A^*(\psi) = \left\{ \begin{array}{ll} -\frac{d\psi(s)}{ds}, \quad 0 < s\leq 1, \\ \int_{-1}^0d{\eta}^T(s, 0)\psi(-s), \quad s = 0, \end{array} \right.$

and the bilinear inner form for $A(0)$ and $A^*$

$\begin{equation} \langle\psi(s), \varphi(\vartheta)\rangle = \bar{\psi}(0)\varphi(0)-\int_{\vartheta = -1}^0\int_{\zeta = 0}^{\vartheta}\bar{\psi}(\zeta-\vartheta)d\eta(\vartheta)\varphi(\zeta)d\zeta, \end{equation}$

(4.7)

where $\eta(\vartheta) = \eta(\vartheta, 0)$ .

Let $u(\vartheta) = (1, u_2, u_3, u_4)^Te^{i\tau_1^*\omega_1^*\vartheta}$ and $u^*(s) = D(1, u_2^*, u_3^*, u_4^*)^Te^{i\tau_1^*\omega_1^*s}$ . Based on definitions of $A(0)$ and $A^*(0)$ , one can obtain

$\begin{eqnarray} u_2& = &\frac{m_{21}e^{-i\tau_1^*\omega_1^*}}{i\omega_1^*-l_{22}-m_{22}e^{-i\tau_1^*\omega_1^*}}, \\ u_3& = &\frac{l_{32}u_2}{i\omega_1^*-l_{33}}, \\ u_4& = &\frac{l_{42}u_2+l_{43}u_3}{i\omega_1^*-l_{44}-n_{44}e^{-i\tau_2^*\omega_1^*}}, \\ u_2^*& = &-\frac{i\omega_1^*+l_{11}}{m_{21}e^{i\tau_1^*\omega_1^*}}, \\ u_3^*& = &-\frac{l_{43}u_2^*}{i\tau_1^*\omega_1^*+l_{33}}, \\ u_4^*& = &-\frac{n_{14}e^{i\tau_2^*\omega_1^*}}{l_{44}+n_{14}e^{i\tau_2^*\omega_1^*}+i\omega_1^*}. \end{eqnarray}$

Then, we have

$\begin{eqnarray} \bar{D}& = &[1+u_2\bar{u}_2^*+u_3\bar{u}_3^*+u_4\bar{u}_4^*+\tau_1^*e^{-i\tau_1^*\omega_1^*}u_2(m_{12}+m_{22}\bar{u}_2^*)\\ &&+\tau_1^*e^{-i\tau_1^*\omega_1^*}m_{21}\bar{u}_2^*+\tau_2^*e^{-i\tau_2^*\omega_1^*}u_4(n_{14}+n_{44}\bar{u}_4^*)]^{-1}. \end{eqnarray}$

Next, $g_{20}$ , $g_{11}$ , $g_{02}$ and $g_{21}$ can be obtained with aid of the method in ^[20]:

$\begin{eqnarray} g_{20}& = &2\tau_1^*\bar{D}[h_{11}u_2+h_{12}+\bar{u}_2^*(h_{21}u_2+h_{22})], \\ g_{11}& = &\tau_1^*\bar{D}[h_{11}(u_2+\bar{u}_2)+2h_{12}+\bar{u}_2^*h_{21}(u_2+\bar{u}_2)+2h_{22}\bar{u}_2^*)], \\ g_{20}& = &2\tau_1^*\bar{D}[h_{11}\bar{u}_2+h_{12}+\bar{u}_2^*h_{21}\bar{u}_2+h_{22}\bar{u}_2^*)], \\ g_{21}& = &2\tau_1^*\bar{D}\bigg[h_{11}\bigg(u_2W_{11}^{(1)}(0)+\frac{1}{2}W_{20}^{(1)}(0)\bar{u}_2+W_{11}^{(2)}(0)+\frac{1}{2}W_{20}^{(2)}(0)\bigg)\\ &&+h_{12}(W_{11}^{(2)}(0)+W_{20}^{(1)}(0))+h_{13}(2u_2+\bar{u}_2)+3h_{14}\\ &&+\bar{u}_2^*h_{21}\bigg(W_{11}^{(1)}(0)u_2+\frac{1}{2}W_{20}^{(1)}(0)\bar{u}_2+W_{11}^{(2)}(0)+\frac{1}{2}W_{20}^{(2)}(0)\bigg)\\ &&+h_{22}\bar{u}_2^*(W_{11}^{(2)}(0)+W_{20}^{(1)}(0))+h_{22}(2u_2+\bar{u}_2)+3h_{24}\bigg], \end{eqnarray}$

with

$\begin{eqnarray} W_{20}(\vartheta)& = &\frac{ig_{20}u(0)}{\tau_1^*\omega_1^*}e^{i\tau_1^*\omega_1^*\vartheta}+\frac{i\bar{g}_{02}\bar{u}(0)}{3\tau_1^*\omega_1^*}e^{-i\tau_1^*\omega_1^*\vartheta}+P_1e^{2i\tau_1^*\omega_1^*\vartheta}, \\ W_{11}(\theta)& = &-\frac{ig_{11}u(0)}{\tau_1^*\omega_1^*}e^{i\tau_1^*\omega_1^*\vartheta}+\frac{i\bar{g}_{11}\bar{u}(0)}{\tau_1^*\omega_1^*}e^{-i\tau_1^*\omega_1^*\vartheta}+P_2. \end{eqnarray}$

$P_1$ and $P_2$ can be computed by

$P_1 = 2\left( \begin{array}{cccc} l_{11}^* \quad &-m_{12}e^{-i\tau_1^*\omega_1^*} \quad &{0} \quad &-n_{14}e^{-i\tau_2^*\omega_1^*}\\ -m_{21}e^{-i\tau_1^*\omega_1^*} \quad & l_{22}^* \quad &{0} \quad &{0}\\ {0} \quad &-l_{32} \quad & l_{33}^* \quad &{0}\\ {0} \quad &-l_{42} \quad &-l_{43} \quad &l_{44}^* \end{array} \right)^{-1}\times\left( \begin{array}{llll} P_1^{(1)}\\ P_1^{(2)}\\ \quad 0\\ \quad {0} \end{array} \right),$

$P_2 = -\left( \begin{array}{cccc} l_{11}+m_{11} \quad &m_{12} \quad &{0} \quad &n_{14}\\ m_{21} \quad &l_{22}+m_{22} \quad &{0} \quad &{0}\\ {0} \quad &l_{32} \quad &l_{33} \quad &{0}\\ {0} \quad &l_{42} \quad &l_{43} \quad &l_{44}+n_{44} \end{array} \right)^{-1}\times\left( \begin{array}{llll} P_2^{(1)}\\ P_2^{(2)}\\ \quad {0}\\ \quad {0} \end{array} \right).$

where

$\begin{eqnarray} l_{11}^*& = &2i\omega_1^*-l_{11}-m_{11}e^{-i\tau_1^*\omega_1^*}, \\ l_{22}^*& = &2i\omega_1^*-l_{22}-m_{22}e^{-i\tau_1^*\omega_1^*}, \\ l_{33}^*& = &2i\omega_1^*-l_{33}, \\ l_{44}^*& = &2i\omega_1^*-l_{44}-n_{44}e^{-i\tau_2^*\omega_1^*}, \end{eqnarray}$

and

$\begin{eqnarray} P_1^{(1)}& = &h_{11}u_2+h_{12}, \\ P_1^{(2)}& = &h_{21}u_2+h_{22}, \\ P_2^{(1)}& = &h_{11}(u_2+\bar{u}_2)+2h_{12}, \\ P_2^{(2)}& = &h_{21}(u_2+\bar{u}_2)+2h_{22}. \end{eqnarray}$

Then, we can obtain

$\begin{equation} \left. \begin{array}{llll} C_1(0) = \frac{i}{2\tau_1^*\omega_1^*}\bigg(g_{11}g_{20}-2|g_{11}|^2-\frac{|g_{02}|^2}{3}\bigg)+\frac{g_{21}}{2}\\ \mu_2 = -\frac{\text{Re}\{C_1(0)\}}{\text{Re}\{\lambda^\prime(\tau_1^*)\}}, \\ \beta_2 = 2{\text{Re}\{C_1(0)\}}, \\ T_2 = -\frac{\text{Im}\{C_1(0)\}+\mu_2\text{Im}\{\lambda^\prime(\tau_1^*)\}}{\tau_1^*\omega_1^*}, \end{array} \right. \end{equation}$

(4.8)

Thus, we have Theorem 6 about the Hopf bifurcation at $\tau_1^*$ .

Theorem 6. For system (1.2), the following results hold. If $\mu_2 > 0$ ( $\mu_2 < 0$ ), then the Hopf bifurcation is supercritical (subcritical); If $\beta_2 < 0$ ( $\beta_2 > 0$ ), then the bifurcating periodic solutions are stable (unstable); If $T_2 > 0$ ( $T_2 < 0$ ), then the period of the bifurcating periodic solutions increase (decrease).

5. Numerical simulations

In order to identify the correctness of above results, some parameters are used to numerical simulations. The values of all parameters are shown in Table 1.

Table 1. Estimation for values of the parameters.

Parameter	Value	Source
$b$	1	^[13]
$p$	0.9	^[13]
$\beta$	0.65	^[13]
$\sigma$	0.4	^[13]
$d$	0.1	assumed
$\delta$	0.7	^[13]
$\varepsilon$	0.14	assumed
$\gamma$	0.3	^[13]
$\eta$	0.1	assumed
$\alpha_1$	0.1	assumed
$\alpha_2$	0.18	assumed

| Show Table

DownLoad: CSV

Then, system (1.2) takes the form

$\begin{equation} \left\{ \begin{array}{llll} \frac{dS(t)}{dt}& = &0.1-\frac{0.65 S(t-\tau_1)I(t-\tau_1)}{1+0.4 S(t-\tau_1)}-0.1S(t)+0.7R(t-\tau_2), \\ \frac{dI(t)}{dt}& = &\frac{0.65 S(t-\tau_1)I(t-\tau_1)}{1+0.4 S(t-\tau_1)}-0.64I(t), \\ \frac{dQ(t)}{dt}& = &0.14I(t)-0.48Q(t), \\ \frac{dR(t)}{dt}& = &0.9+0.1 Q(t)+0.3 I(t)-0.7 R(t-\tau_2)-0.1 R(t), \end{array} \right. \end{equation}$

(5.1)

from which we can obtain $R_0 = 1.98 > 1$ and $E^*(1.6244, 2.0598, 0.7739, 2.0011)$ .

To verify Theorem 1, we use Matlab software and obtain $\omega_{10} = 0.0786$ and $\tau_{10} = 9.3985$ . shows that system (5.1) is locally asymptotically stable when $\tau_1\in[0, \tau_{10})$ , $\tau_2 = 0$ and a Hopf bifurcation arises when $\tau_1 = \tau_{10}$ . After that, from , system (5.1) becomes unstable when $\tau_1 > \tau_{10}$ .

Figure 1. Evolutions of S, I, Q, R for

$\tau_1 = 7.2010 < \tau_{10}$ of system (5.1) versus time t.

DownLoad: Full-Size Img PowerPoint

Figure 2. Evolutions of S, I, Q, R for

$\tau_1 = 11.9806 > \tau_{10}$ of system (5.1) versus time t.

DownLoad: Full-Size Img PowerPoint

In the same way, we apply Matlab software to verify Theorem 2. Then, we obtain $\omega_{20} = 0.3307$ and $\tau_{20} = 2.2352$ . From , system (5.1) is locally asymptotically stable when $\tau_1 = 0$ , $\tau_2\in[0, \tau_{20})$ , and a Hopf bifurcation arises when $\tau_2 = \tau_{20}$ . Otherwise, system (5.1) becomes unstable when $\tau_2 > \tau_{20}$ in Figure 4.

Figure 3. Evolutions of S, I, Q, R for

$\tau_2 = 2.0803 < \tau_{20}$ of system (5.1) versus time t.

DownLoad: Full-Size Img PowerPoint

Figure 4. Evolutions of S, I, Q, R for

$\tau_2 = 2.4012 > \tau_{20}$ of system (5.1) versus time t.

DownLoad: Full-Size Img PowerPoint

A short calculation revealed that $\omega_{30} = 0.3833$ and $\tau_{*0} = 1.9282$ . Afterwards, shows that system (5.1) is locally asymptotically stable when $\tau_1 = \tau_2\in[0, \tau_{*0})$ , and it can be seen a Hopf bifurcation when $\tau_1 = \tau_2 = \tau_{*0}$ . shows that system (5.1) becomes unstable when $\tau_1 = \tau_2 > \tau_{*0}$ .

Figure 5. Evolutions of S, I, Q, R for

$\tau_1 = \tau_2 = 1.8457 < \tau_{*0}$ of system (5.1) versus time t.

DownLoad: Full-Size Img PowerPoint

Figure 6. Evolutions of S, I, Q, R for

$\tau_1 = \tau_2 = 1.9796 > \tau_{*0}$ of system (5.1) versus time t.

DownLoad: Full-Size Img PowerPoint

It is easy to obtain $\omega_{40} = 0.0710$ and $\tau_{1*0} = 10.4056$ . When $\tau_1\in[0, \tau_{1*0})$ , $\tau_2\in[0, \tau_{20})$ , system (1.2) is locally asymptotically stable, and system (5.1) undergoes a Hopf bifurcation when $\tau_1 = \tau_{1*0}$ . Once $\tau_1 > \tau_{1*0}$ , system (5.1) becomes unstable. The corresponding simulations are shown in Figure 7 and Figure 8.

Figure 7. Evolutions of S, I, Q, R for

$\tau_1 = 8.8464 < \tau_{1*0}$ ,

$\tau_2 = 1.5764\in[0, \tau_{20})$ of system (5.1) versus time t.

DownLoad: Full-Size Img PowerPoint

Figure 8. Evolutions of S, I, Q, R for

$\tau_1 = 12.7054 > \tau_{1*0}$ ,

$\tau_2 = 1.5764\in[0, \tau_{20})$ of system (5.1) versus time t.

DownLoad: Full-Size Img PowerPoint

Through simple calculation, $\omega_{50} = 0.3971$ and $\tau_{2*0} = 1.8612$ can be got. As shows, system (5.1) is locally asymptotically stable when $\tau_1\in[0, \tau_{10})$ , $\tau_2\in[0, \tau_{2*0})$ and a Hopf bifurcation arises when $\tau_2 = \tau_{2*0}$ . And we can see that system (5.1) becomes unstable when $\tau_2 > \tau_{2*0}$ in Figure 10.

Figure 9. Evolutions of S, I, Q, R for

$\tau_1 = 8.7421\in[0, \tau_{10})$ ,

$\tau_2 = 1.3413 < \tau_{2*0}$ of system (5.1) versus time t.

DownLoad: Full-Size Img PowerPoint

Figure 10. Evolutions of S, I, Q, R for

$\tau_1 = 8.7421\in[0, \tau_{10})$ ,

$\tau_2 = 1.9735 > \tau_{2*0}$ of system (5.1) versus time t.

DownLoad: Full-Size Img PowerPoint

6. Conclusions

In this paper, we propose a novel Susceptible-Infected-Quarantined-Recovered (SIQRS) computer virus propagation model with quarantine strategy based on the model formulated in ^[13]. We consider not only incubation period delay, but also temporary immunization time delay when we observe the dynamics of the SIQRS model. The local stability of the virus-free equilibrium and the virus-existent equilibrium also has been discussed. Furthermore, we analyze the local stability and Hopf bifurcation under another five cases with different delays. If $\tau$ is less than the key value, system (1.2) is local stable; otherwise, there is a Hopf bifurcation. Then, we determine the direction of Hopf bifurcation and the stability of bifurcating periodic solutions by using the normal form and center manifold theorem. Ultimately, some numerical simulations are used to prove the validity of the theoretical results.

Compared with the model in ^[13], our novel model consider quarantine strategy, which is used to the prevention and cure of computer virus, so our model is closer to the actual situation. Furthermore, incubation period is one of the significant characteristics of computer virus, and it is very important to take the latency delay into account. Nowadays, antivirus software, which enable computers to gain temporary immunity, plays a very important role in the defense of computer virus. Temporary immunity delay is a common phenomena in real life. In our SIQRS model, the above cases are taken into account at the same time, and our model has more reference value over the existing ones. Global asymptotic stability is as important as local asymptotic stability, and it will be studied in the future.

Acknowledgments

This research was supported by the Natural Science Foundation of the Higher Education Institutions of Anhui Province (No.KJ2020A0002), the Project of Natural Science Foundation of Anhui Province (No.2008085QA09) and National Natural Science Research Foundation Project (No.12061033).

Conflict of interest

The authors declare that there are no conflicts of interest regarding the publication of this paper.

References

[1]	O. A. Ataguba, J. E. Ataguba, Social determinants of health: the role of effective communication in the COVID-19 pandemic in developing countries, Global Health Action, 1 (2020), 1788263. https://doi.org/10.1080/16549716.2020.1788263 doi: 10.1080/16549716.2020.1788263
[2]	World Health Organization, WHO coronavirus disease (COVID-19) Dashboard, 2022. Available from: https://covid19.who.int/?gclid = Cj0KCQjwtZH7BRDzARIsAGjbK2ZXWRpJROEl97HGmSOx0_ydkVbc02Ka1FlcysGjEI7hnaIeR6xWhr4aAu57EALw_wcB.
[3]	Q. Huang, X. Huang, Z. Kong, X. Li, D. Tao, Bi-phase evolutionary searching for biclusters in gene expression data, IEEE Trans. Evol. Comput., 5 (2018), 803–814. https://doi.org/10.1109/TEVC.2018.2884521 doi: 10.1109/TEVC.2018.2884521
[4]	Q. Huang, J. Yao, J. Li, M. Li, M. R. Pickering, X. Li, Measurement of quasi-static 3-D knee joint movement based on the registration from CT to US, IEEE Trans. Ultrason. Free, 6 (2020), 1141–50. https://doi.org/10.1109/TUFFC.2020.2965149 doi: 10.1109/TUFFC.2020.2965149
[5]	J. Xi, Z. Miao, L. Liu, X. Yang, W. Zhang, Q. Huang, et al., Knowledge tensor embedding framework with association enhancement for breast ultrasound diagnosis of limited labeled samples, Neurocomputing, 468 (2022), 60–70. https://doi.org/10.1016/j.neucom.2021.10.013 doi: 10.1016/j.neucom.2021.10.013
[6]	Q. Huang, F. Pan, W. Li, F. Yuan, H. Hu, J. Huang, et al., Differential diagnosis of atypical hepatocellular carcinoma in contrast-enhanced ultrasound using spatio-temporal diagnostic semantics, IEEE J. Biomed. Health, 10 (2020), 2860–2869. https://doi.org/10.1109/JBHI.2020.2977937 doi: 10.1109/JBHI.2020.2977937
[7]	J. Xi, D. Wang, X. Yang, W. Zhang, Q. Huang, Cancer omic data based explainable AI drug recommendation inference: A traceability perspective for explainability, Biomed. Signal Process., 79 (2023), 104144. https://doi.org/10.1016/j.bspc.2022.104144 doi: 10.1016/j.bspc.2022.104144
[8]	W. Shi, W. N. Chen, S. Kwong, J. Zhang, H. Wang, T. Gu, et al., A coevolutionary estimation of distribution algorithm for group insurance portfolio, IEEE Trans. Syst. Man CY-S, 11 (2021), 6714–28. https://doi.org/10.1109/TSMC.2021.3096013 doi: 10.1109/TSMC.2021.3096013
[9]	Y. Yuan, M. Chao, Y.C. Lo, Automatic skin lesion segmentation using deep fully convolutional networks with jaccard distance, IEEE Trans. Med. Imaging, 9 (2017), 1876–1886. https://doi.org/10.1109/TMI.2017.2695227 doi: 10.1109/TMI.2017.2695227
[10]	P. Liskowski, K. Krawiec, Segmenting retinal blood vessels with deep neural networks, IEEE Trans. Med. Imaging, 11 (2016), 2369–2380. https://doi.org/10.1109/TMI.2016.2546227 doi: 10.1109/TMI.2016.2546227
[11]	H. Fu, J. Cheng, Y. Xu, D. W. K. Wong, J. Liu, X. Cao, Joint optic disc and cup segmentation based on multi-label deep network and polar transformation, IEEE Trans. Med. Imaging, 7 (2018), 1597–1605. https://doi.org/10.1109/TMI.2018.2791488 doi: 10.1109/TMI.2018.2791488
[12]	H. Fu, Y. Xu, S. Lin, X. Zhang, D. W. K. Wong, J. Liu, et al., Segmentation and quantification for angle-closure glaucoma assessment in anterior segment OCT, IEEE Trans. Med. Imaging, 9 (2017), 1930–1938. https://doi.org/10.1109/TMI.2017.2703147 doi: 10.1109/TMI.2017.2703147
[13]	M. Anthimopoulos, S. Christodoulidis, L. Ebner, A. Christe, S. Mougiakakou, Lung pattern classification for interstitial lung diseases using a deep convolutional neural network, IEEE Trans. Med. Imaging, 5 (2016), 1207–1216. https://doi.org/10.1109/TMI.2016.2535865 doi: 10.1109/TMI.2016.2535865
[14]	J. M. Wolterink, T. Leiner, M. A. Viergever, I. Išgum, Generative adversarial networks for noise reduction in low-dose CT, IEEE Trans. Med. Imaging, 12 (2017), 2536–2345. https://doi.org/10.1109/TMI.2017.2708987 doi: 10.1109/TMI.2017.2708987
[15]	S. Albarqouni, C. Baur, F. Achilles, V. Belagiannis, S. Demirci, N. Navab, Aggnet: Deep learning from crowds for mitosis detection in breast cancer histology images, IEEE Trans. Med. Imaging, 5 (2016), 1313–1321. https://doi.org/10.1109/TMI.2016.2528120 doi: 10.1109/TMI.2016.2528120
[16]	X. Zhang, G. Wang, S. G. Zhao, CapsNet-COVID19: Lung CT image classification method based on CapsNet model, Math. Biosci. Eng., 19 (2022), 5055–5074. https://doi.org/10.3934/mbe.2022236 doi: 10.3934/mbe.2022236
[17]	M. J. Horry, S. Chakraborty, B. Pradhan, M. Fallahpoor, H. Chegeni, M. Paul, Factors determining generalization in deep learning models for scoring COVID-CT images, Math. Biosci. Eng., 18 (2021), 9264–9293. https://doi.org/10.3934/mbe.2021456 doi: 10.3934/mbe.2021456
[18]	A. Singh, K. K. Singh, M. Greguš, I. Izonin, CNGOD-An improved convolution neural network with grasshopper optimization for detection of COVID-19, Math. Biosci. Eng., 19 (2022), 12518–12531. https://doi.org/10.3934/mbe.2022584 doi: 10.3934/mbe.2022584
[19]	S. H. Wang, S. C. Satapathy, M. X. Xie, Y. D. Zhang, ELUCNN for explainable COVID-19 diagnosis, Soft Comput., 2023 (2023), 1–17. https://doi.org/10.1007/s00500-023-07813-w doi: 10.1007/s00500-023-07813-w
[20]	Y. Zhang, M. A. Khan, Z. Zhu, S. Wang, SNELM: SqueezeNet-guided ELM for COVID-19 recognition, Comput. Syst. Sci. Eng., 1 (2023), 13–26. https://doi.org/10.32604/csse.2023.034172 doi: 10.32604/csse.2023.034172
[21]	Y. Zhang, X. Zhang, W. Zhu, ANC: Attention network for COVID-19 explainable diagnosis based on convolutional block attention module, CMES COMP Model. Eng., 3 (2021), 1037–1058. https://doi.org/10.32604/cmes.2021.015807 doi: 10.32604/cmes.2021.015807
[22]	C. C. Lai, T. P. Shih, W. C. Ko, H. J. Tang, P. R. Hsueh, Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and coronavirus disease-2019 (COVID-19): The epidemic and the challenges, Int. J. Antimicrob. Ag., 3 (2020), 105924. https://doi.org/10.1016/j.ijantimicag.2020.105924 doi: 10.1016/j.ijantimicag.2020.105924
[23]	E. Irmak, Implementation of convolutional neural network approach for COVID-19 disease detection, Physiol. Genomics, 12 (2020), 590–601. https://doi.org/10.1152/physiolgenomics.00084.2020 doi: 10.1152/physiolgenomics.00084.2020
[24]	F. Xie, J. Xi, Q. Duan, Driver attribute filling for genes in interaction network via modularity subspace-based concept learning from small samples, Complexity, 2020 (2020), 1–12. https://doi.org/10.1155/2020/6643551 doi: 10.1155/2020/6643551
[25]	Q. Huang, F. Zhang, X. Li, A new breast tumor ultrasonography CAD system based on decision tree and BI-RADS features, World Wide Web, 21 (2018), 1491–1504. https://doi.org/10.1007/s11280-017-0522-5 doi: 10.1007/s11280-017-0522-5
[26]	G. Dong, Z. C. Zhang, J. Feng, X. M. Zhao, MorbidGCN: Prediction of multimorbidity with a graph convolutional network based on integration of population phenotypes and disease network, Brief Bioinform., 4 (2022). https://doi.org/10.1093/bib/bbac255 doi: 10.1093/bib/bbac255
[27]	S. Wang, P. Li, P. Chen, P. Phillips, G. Liu, S. Du, et al., Pathological brain detection via wavelet packet tsallis entropy and real-coded biogeography-based optimization, Fund. Inform., 4 (2017), 275–291. https://doi.org/10.3233/FI-2017-1492 doi: 10.3233/FI-2017-1492
[28]	L. V. Fulton, D. Dolezel, J. Harrop, Y. Yan, C. P. Fulton, Classification of Alzheimer's disease with and without imagery using gradient boosted machines and ResNet-50, Brain Sci., 9 (2019), 212. https://doi.org/10.3390/brainsci9090212 doi: 10.3390/brainsci9090212
[29]	Q. Guan, Y. Huang, Z. Zhong, Z. Zheng, L. Zheng, Y. Yang, Diagnose like a radiologist: Attention guided convolutional neural network for thorax disease classification, preprint, arXiv: 1801.09927.
[30]	P. P. Ypsilantis, G. Montana, Learning what to look in chest X-rays with a recurrent visual attention model, preprint, arXiv: 1701.06452.
[31]	E. Pesce, S. J. Withey, P. P. Ypsilantis, R. Bakewell, V. Goh, G. Montana, Learning to detect chest radiographs containing pulmonary lesions using visual attention networks, Med. Image Anal., 53 (2019), 26–38. https://doi.org/10.1016/j.media.2018.12.007 doi: 10.1016/j.media.2018.12.007
[32]	M. Toğaçar, B. Ergen, Z. Cömert, COVID-19 detection using deep learning models to exploit Social Mimic Optimization and structured chest X-ray images using fuzzy color and stacking approaches, Comput. Biol. Med., 121 (2020), 103805.https://doi.org/10.1016/j.compbiomed.2020.103805 doi: 10.1016/j.compbiomed.2020.103805
[33]	J. P. Cohen, L. Dao, K. Roth, P. Morrison, Y. Bengio, A. F. Abbasi, et al., Predicting covid-19 pneumonia severity on chest x-ray with deep learning, Cureus J. Med. Sci., 12 (2020), e9448. https://doi.org/10.7759/cureus.9448 doi: 10.7759/cureus.9448
[34]	Q. Ni, Z. Y. Sun, L. Qi, W. Chen, Y. Yang, L. Wang, et al., A deep learning approach to characterize 2019 coronavirus disease (COVID-19) pneumonia in chest CT images, Eur. Radiol., 30 (2020), 6517–6527. https://doi.org/10.1007/s00330-020-07044-9 doi: 10.1007/s00330-020-07044-9
[35]	H. Ko, H. Chung, W. S. Kang, K. W. Kim, Y. Shin, S. J. Kang, et al., COVID-19 pneumonia diagnosis using a simple 2D deep learning framework with a single chest CT image: model development and validation, J. Med. Int. Res., 6 (2020), e19569. https://doi.org/10.2196/19569 doi: 10.2196/19569
[36]	X. Wang, X. Deng, Q. Fu, Q. Zhou, J. Feng, H. Ma, et al., A weakly-supervised framework for COVID-19 classification and lesion localization from chest CT, IEEE Trans. Med. Imaging, 8 (2020), 2615–2625. https://doi.org/10.1109/TMI.2020.2995965 doi: 10.1109/TMI.2020.2995965
[37]	A. I. Khan, J. L. Shah, M. M. Bhat, CoroNet: A deep neural network for detection and diagnosis of COVID-19 from chest x-ray images, Comput. Method Prog. Bio., 196 (2020), 105581. https://doi.org/10.1016/j.cmpb.2020.105581 doi: 10.1016/j.cmpb.2020.105581
[38]	L. Hussain, T. Nguyen, H. Li, A. A. Abbasi, K. J. Lone, Z. Zhao, et al., Machine-learning classification of texture features of portable chest X-ray accurately classifies COVID-19 lung infection, Biomed. Eng. Online, 19 (2020), 1–18. https://doi.org/10.1186/s12938-020-00831-x doi: 10.1186/s12938-020-00831-x
[39]	J. Zhao, Y. Zhang, X. He, P. Xie, Covid-ct-dataset: a ct scan dataset about covid-19, 2020.
[40]	COVID-19 CT segmentation dataset. Available from: https://medicalsegmentation.com/covid19/.
[41]	D. P. Fan, T. Zhou, G. P. Ji, Y. Zhou, G. Chen, H. Fu, et al., Inf-net: Automatic covid-19 lung infection segmentation from ct images, IEEE Trans. Med. Imaging, 8 (2020), 2626–2637. https://doi.org/10.1109/TMI.2020.2996645 doi: 10.1109/TMI.2020.2996645
[42]	S. H. Wang, V. V. Govindaraj, J. M. Górriz, X. Zhang, Y. D. Zhang, Covid-19 classification by FGCNet with deep feature fusion from graph convolutional network and convolutional neural network, Inform. Fusion, 67 (2021), 208–229. https://doi.org/10.1016/j.inffus.2020.10.004 doi: 10.1016/j.inffus.2020.10.004
[43]	R. R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, D. Batra, et al., Grad-cam: Visual explanations from deep networks via gradient-based localization, Int. J. Comput. Vision, 2 (2020), 336–359. https://doi.org/10.1007/s11263-019-01228-7 doi: 10.1007/s11263-019-01228-7
[44]	A. Chattopadhay, A. Sarkar, P. Howlader, V. N. Balasubramanian, Grad-cam++: Generalized gradient-based visual explanations for deep convolutional networks, in 2018 IEEE winter conference on applications of computer vision (WACV), (2018), 839–847. https://doi.org/10.1109/WACV.2018.00097
[45]	D. Chicco, N. Tötsch, G. Jurman, The Matthews correlation coefficient (MCC) is more reliable than balanced accuracy, bookmaker informedness, and markedness in two-class confusion matrix evaluation, Biodata Min., 1 (2021), 1–22. https://doi.org/10.1186/s13040-021-00244-z doi: 10.1186/s13040-021-00244-z
[46]	Q. Huang, Y. Lei, W. Xing, C. He, G. Wei, Z. Miao, et al., Evaluation of pulmonary edema using ultrasound imaging in patients with COVID-19 pneumonia based on a non-local Channel attention ResNet, Ultrasound Med. Biol., 5 (2022), 945–953. https://doi.org/10.1016/j.ultrasmedbio.2022.01.023 doi: 10.1016/j.ultrasmedbio.2022.01.023
[47]	F. J. P. Montalbo, Diagnosing Covid-19 chest x-rays with a lightweight truncated DenseNet with partial layer freezing and feature fusion, Biomed. Signal Process., 68 (2021), 102583. https://doi.org/10.1016/j.bspc.2021.102583 doi: 10.1016/j.bspc.2021.102583
[48]	R. Mastouri, N. Khlifa, H. Neji, S. Hantous-Zannad, A bilinear convolutional neural network for lung nodules classification on CT images, Int. J. Comput. Ass Rad, 16 (2021), 91–101. https://doi.org/10.1007/s11548-020-02283-z doi: 10.1007/s11548-020-02283-z
[49]	A. Garg, S. Salehi, M. La Rocca, R. Garner, D. Duncan, Efficient and visualizable convolutional neural networks for COVID-19 classification using Chest CT, Expert Syst. Appl., 195 (2022), 116540. https://doi.org/10.1016/j.eswa.2022.116540 doi: 10.1016/j.eswa.2022.116540
[50]	F. Chollet, Xception: Deep learning with depthwise separable convolutions, in Proceedings of the IEEE conference on computer vision and pattern recognition (CVPR), (2017), 1251–1258. https://doi.org/10.48550/arXiv.1610.02357
[51]	H. Panwar, P. Gupta, M. K. Siddiqui, R. Morales-Menendez, P. Bhardwaj, V. Singh, A deep learning and grad-CAM based color visualization approach for fast detection of COVID-19 cases using chest X-ray and CT-Scan images, Chaos Soliton Fract., 140 (2020), 110190. https://doi.org/10.1016/j.chaos.2020.110190 doi: 10.1016/j.chaos.2020.110190

This article has been cited by:

1.	Ge Zhang, Zhiming Li, Anwarud Din, A stochastic SIQR epidemic model with Lévy jumps and three-time delays, 2022, 431, 00963003, 127329, 10.1016/j.amc.2022.127329
2.	Jiaxing Chen, Chengyi Xia, Matjaž Perc, The SIQRS Propagation Model With Quarantine on Simplicial Complexes, 2024, 11, 2329-924X, 4267, 10.1109/TCSS.2024.3351173
3.	Haokuan Cheng, Min Xiao, Yunxiang Lu, Haibo Bao, Leszek Rutkowski, Jinde Cao, Complex pattern evolution of a two-dimensional space diffusion model of malware spread, 2024, 99, 0031-8949, 045237, 10.1088/1402-4896/ad30ee
4.	Linji Yang, Qiankun Song, Yurong Liu, Stability and Hopf bifurcation analysis for fractional-order SVEIR computer virus propagation model with nonlinear incident rate and two delays, 2023, 547, 09252312, 126397, 10.1016/j.neucom.2023.126397
5.	Jie 杰 Gao 高, Jianfeng 建锋 Luo 罗, Xing 星 Li 李, Yihong 毅红 Li 李, Zunguang 尊光 Guo 郭, Xiaofeng 晓峰 Luo 罗, Triadic percolation in computer virus spreading dynamics, 2025, 34, 1674-1056, 028701, 10.1088/1674-1056/ad9ff8
6.	JunLing Wang, Xinxin Chang, Lei Zhong, A SEIQRS Computer Virus Propagation Model and Impulse Control With Two Delays, 2025, 0170-4214, 10.1002/mma.10721
7.	Junling Wang, Lei Zhong, Xinxin Chang, Sandip V. George, SEIQRS model analysis and optimal control with two delays, 2025, 20, 1932-6203, e0319417, 10.1371/journal.pone.0319417

Reader Comments

Your name:*

Email:*
© 2023 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)