Research article Special Issues

Investigation on aortic hemodynamics based on physics-informed neural network


  • Received: 28 March 2023 Revised: 14 April 2023 Accepted: 16 April 2023 Published: 05 May 2023
  • Pressure in arteries is difficult to measure non-invasively. Although computational fluid dynamics (CFD) provides high-precision numerical solutions according to the basic physical equations of fluid mechanics, it relies on precise boundary conditions and complex preprocessing, which limits its real-time application. Machine learning algorithms have wide applications in hemodynamic research due to their powerful learning ability and fast calculation speed. Therefore, we proposed a novel method for pressure estimation based on physics-informed neural network (PINN). An ideal aortic arch model was established according to the geometric parameters from human aorta, and we performed CFD simulation with two-way fluid-solid coupling. The simulation results, including the space-time coordinates, the velocity and pressure field, were obtained as the dataset for the training and validation of PINN. Nondimensional Navier-Stokes equations and continuity equation were employed for the loss function of PINN, to calculate the velocity and relative pressure field. Post-processing was proposed to fit the absolute pressure of the aorta according to the linear relationship between relative pressure, elastic modulus and displacement of the vessel wall. Additionally, we explored the sensitivity of the PINN to the vascular elasticity, blood viscosity and blood velocity. The velocity and pressure field predicted by PINN yielded good consistency with the simulated values. In the interested region of the aorta, the relative errors of maximum and average absolute pressure were 7.33% and 5.71%, respectively. The relative pressure field was found most sensitive to blood velocity, followed by blood viscosity and vascular elasticity. This study has proposed a method for intra-vascular pressure estimation, which has potential significance in the diagnosis of cardiovascular diseases.

    Citation: Meiyuan Du, Chi Zhang, Sheng Xie, Fang Pu, Da Zhang, Deyu Li. Investigation on aortic hemodynamics based on physics-informed neural network[J]. Mathematical Biosciences and Engineering, 2023, 20(7): 11545-11567. doi: 10.3934/mbe.2023512

    Related Papers:

    [1] Yufei Wang, Huidong Cheng, Qingjian Li . Dynamic analysis of wild and sterile mosquito release model with Poincaré map. Mathematical Biosciences and Engineering, 2019, 16(6): 7688-7706. doi: 10.3934/mbe.2019385
    [2] Guirong Jiang, Qishao Lu, Linping Peng . Impulsive Ecological Control Of A Stage-Structured Pest Management System. Mathematical Biosciences and Engineering, 2005, 2(2): 329-344. doi: 10.3934/mbe.2005.2.329
    [3] Xiaoxiao Yan, Zhong Zhao, Yuanxian Hui, Jingen Yang . Dynamic analysis of a bacterial resistance model with impulsive state feedback control. Mathematical Biosciences and Engineering, 2023, 20(12): 20422-20436. doi: 10.3934/mbe.2023903
    [4] Huidong Cheng, Hui Xu, Jingli Fu . Dynamic analysis of a phytoplankton-fish model with the impulsive feedback control depending on the fish density and its changing rate. Mathematical Biosciences and Engineering, 2023, 20(5): 8103-8123. doi: 10.3934/mbe.2023352
    [5] Zhenzhen Shi, Huidong Cheng, Yu Liu, Yanhui Wang . Optimization of an integrated feedback control for a pest management predator-prey model. Mathematical Biosciences and Engineering, 2019, 16(6): 7963-7981. doi: 10.3934/mbe.2019401
    [6] Yuan Tian, Sanyi Tang . Dynamics of a density-dependent predator-prey biological system with nonlinear impulsive control. Mathematical Biosciences and Engineering, 2021, 18(6): 7318-7343. doi: 10.3934/mbe.2021362
    [7] Yilin Tu, Jin-E Zhang . Event-triggered impulsive control for input-to-state stability of nonlinear time-delay system with delayed impulse. Mathematical Biosciences and Engineering, 2025, 22(4): 876-896. doi: 10.3934/mbe.2025031
    [8] Biwen Li, Qiaoping Huang . Synchronization of time-delay systems with impulsive delay via an average impulsive estimation approach. Mathematical Biosciences and Engineering, 2024, 21(3): 4501-4520. doi: 10.3934/mbe.2024199
    [9] Yazhi Wu, Guangyao Tang, Changcheng Xiang . Dynamic analysis of a predator-prey state-dependent impulsive model with fear effect in which action threshold depending on the prey density and its changing rate. Mathematical Biosciences and Engineering, 2022, 19(12): 13152-13171. doi: 10.3934/mbe.2022615
    [10] Yang Wang, Yongyang Liu, Yansheng Liu . Total controllability of non-autonomous second-order measure evolution systems with state-dependent delay and non-instantaneous impulses. Mathematical Biosciences and Engineering, 2023, 20(2): 2061-2080. doi: 10.3934/mbe.2023095
  • Pressure in arteries is difficult to measure non-invasively. Although computational fluid dynamics (CFD) provides high-precision numerical solutions according to the basic physical equations of fluid mechanics, it relies on precise boundary conditions and complex preprocessing, which limits its real-time application. Machine learning algorithms have wide applications in hemodynamic research due to their powerful learning ability and fast calculation speed. Therefore, we proposed a novel method for pressure estimation based on physics-informed neural network (PINN). An ideal aortic arch model was established according to the geometric parameters from human aorta, and we performed CFD simulation with two-way fluid-solid coupling. The simulation results, including the space-time coordinates, the velocity and pressure field, were obtained as the dataset for the training and validation of PINN. Nondimensional Navier-Stokes equations and continuity equation were employed for the loss function of PINN, to calculate the velocity and relative pressure field. Post-processing was proposed to fit the absolute pressure of the aorta according to the linear relationship between relative pressure, elastic modulus and displacement of the vessel wall. Additionally, we explored the sensitivity of the PINN to the vascular elasticity, blood viscosity and blood velocity. The velocity and pressure field predicted by PINN yielded good consistency with the simulated values. In the interested region of the aorta, the relative errors of maximum and average absolute pressure were 7.33% and 5.71%, respectively. The relative pressure field was found most sensitive to blood velocity, followed by blood viscosity and vascular elasticity. This study has proposed a method for intra-vascular pressure estimation, which has potential significance in the diagnosis of cardiovascular diseases.



    1. Introduction

    The definition of impulsive semi-dynamical system and its properties including the limit sets of orbits have been investigated [1,9]. The generalized planar impulsive dynamical semi-dynamical system can be described as follows

    {dxdt=P(x,y),dydt=Q(x,y),(x,y)M,x=a(x,y),y=b(x,y),(x,y)M, (1)

    where (x,y)R2, x=x+x and y=y+y. P,Q,a,b are continuous functions from R2 into R, MR2 denotes the impulsive set. For each point z(x,y)M, the map or impulsive function I:R2R2 is defined as

    I(z)=z+=(x+,y+)R2, x+=x+a(x,y), y+=y+b(x,y)

    and z+ is called as an impulsive point of z.

    Let N=I(M) be the phase set (i.e. for any zM,I(z)=z+N), and NM=. Let (X,Π,R+) or (X,Π) be a semi-dynamical system, where X=R2 is a metric space, R+ is the set of all non-negative reals. For any zX, the function Πz:R+X defined by Πz(t)=Π(z,t) is clearly continuous such that Π(z,0)=z for all zX, and Π(Π(z,t),s)=Π(z,t+s) for all zX and t,sR+. The set

    C+(z)={Π(z,t)|tR+}

    is called the positive orbit of z. For all t0 and zX, we define F(z,t)={w:Π(w,t)=z}, and further for any set MX, let

    M+(z)=C+(z)M{z}.

    Based on above notations, the definition of impulsive semi-dynamical system is defined as follows [1,9,23].

    Definition 1.1. An planar impulsive semi-dynamic system (R2,Π;M,I) consists of a continuous semi-dynamic system (R2,Π) together with a nonempty closed subset M (or impulsive set) of R2 and a continuous function I:MR2 such that for every zM, there exists a ϵz>0 such that

    F(z,(0,ϵz))M= and Π(z,(0,ϵz))M=.

    Definition 1.2. Let (R2,Π) be a dynamical system. An open set V in R2 is called a tube if there exist an s>0 and a subset SV such that

    1. Π(z,t)V for all zS and s<t<s, and

    2. for each zV there is a unique s(z), |s(z)|<s such that Π(z,s(z))S.

    It is clear that V=[Π(z,t)|zS,s<t<s]. The set S is called an (s,V)-section of the tube V. It is well known that if z0R2 is not a rest point, that is, Π(z0,t)z0 for all t, then there exists a tube containing z0.

    Definition 1.3. Let (R2,Π) be a dynamical system. A subset MR2 is said to be well placed in R2 if every point zM lies in some tube whose section lies in M.

    Denote the points of discontinuity of Πz by {z+n} and call z+n an impulsive point of zn. A function Φ from R2 into the extended positive reals R+{} is defined as follows: let zX, if M+(z)= then Φ(z)=, otherwise if M+(z) then Φ(z)=s, where Π(z,t)M for 0<t<s but Π(z,s)M. Obviously, this function describes the times of meeting impulsive sets, and its continuity plays an important role in impulsive semi-dynamical system. In 1990 [9], Kual defined the time function (denoted as Φ)-the time of reaching impulsive points in the given impulsive semi-dynamical system (X,π;M,I). And he also provided a theorem that the function Φ is continuous on X when the impulsive set M is well placed. That is, the following main results on the continuity of function Φ have been provided in [1,9] (i.e. Theorem 2.1 in [1] and Theorem 4 in [9]).

    Theorem 1.4. Let (R2,Π) be a dynamical system. Let M be a closed and well placed subset of R2. Suppose that (R2,Π,M,I) is an impulsive semi-dynamical system. Then Φ is continuous on R2.


    2. Example and main results

    In 2004 [2], the author pointed out some errors on Theorem 1.4, that is, it need not be continuous under the assumptions. And the main aspect concerned in the paper [2] is the continuality of Φ on the impulsive set M. In another word, the author has proved that the function Φ sometimes is not continuous (actually not low semicontinuous) on some points in the impulsive set M and more conditions are needed to guarantee its continuity. They concluded that Φ just continuous on XM under the assumptions in [9]. And in paper [1], the authors have adopted this result. In this note, we have make further efforts that the function would not be continuous on XM under the assumption in the paper [9]. We first address this in the following by a realistic example in pest control and discuss generally what extra condition should be added to guarantee the continuity of Φ on XM.

    In the following we will provide an example to show this Theorem is not true for some special cases. Considering the following model with state-dependent feedback control

    {dx(t)dt=ax(t)[1x(t)K]βx(t)y(t)1+ωx(t),dy(t)dt=ηβx(t)y(t)1+ωx(t)δy(t),}x<ET,x(t+)=(1θ)x(t),y(t+)=y(t)+τ,}x=ET. (2)

    where x(t+)x+ and y(t+)y+ denote the numbers of pests and natural enemies after an integrated control strategy is applied at time t [11,17]. The definitions of all parameters and their biological meanings please refer to literatures [11,17]. For convenience, we assume that the initial density of the pest population is always less than the economic threshold (ET) (i.e. x(0+)x+0<ET,y(0+)y+0>0) and ET<K. In model (2), 0θ<1 is the proportion by which the pest density is reduced by killing and/or trapping once the number of pests reaches ET, while τ (τ0) is the constant number of natural enemies released at this time t.

    Define four curves as follows

    L0:x=δηβδω;   L1:y=rβ[1xK](1+ωx);
    L2:x=ET; and   L3:x=(1θ)ET.

    The intersection points of two lines L1 and L2, L1 and L3 are QET=(ET,yET) and QθET=((1θ)ET,yθET), respectively, where

    yET=rβ[1ETK](1+ωET), yθET=rβ[1(1θ)ETK](1+ω(1θ)ET).

    Define the open set in R2+ as follows

    Ω={(x,y)|x>0,y>0,x<ET}R2+={(x,y)|x0,y0}. (3)

    In the following we assume that model (2) without impulsive effects exists an unstable focus E with

    E=(xe,ye)=(δηβδω,rη(KηβKδωδ)K(ηβδω)2),

    which means that model (2) without impulsive effects has a unique stable limit cycle (denoted by Γ)[11,17], as shown in Fig. 1. Note that the Γ intersects with the isocline L1 at two points EΓ1(xΓ1,yΓ1) and EΓ2(xΓ2,yΓ2) with xΓ1<xΓ2.

    Figure 1. Illustrations of impulsive set, phase set and definition of impulsive semi-dynamical system for model (2). The parameter values are fixed as: r=1,K=52,β=0.19,η=0.45,ω=0.19,δ=0.36,τ=5,ET=35,θ=0.7.

    In the following we show that model (2) defines an impulsive semi-dynamical system. From a biological point of view, we focus on the space R2+ to discuss model (2). As an example, we only focus on the special case, i.e. xΓ1<(1θ)ET<ET<xΓ2, as shown in Fig. 1, and the trajectory initiating from the point QθET (denoted by Γ1) plays a key role in defining the impulsive set and determining the continuity of the function Φ. There are three cases: (A) Γ1 reaches the section SET={(x,y)|x=ET,y0} without crossing the curve L1; (B) Γ1 reaches the section SET={(x,y)|x=ET,y0} just on the point QET; and (C) Γ1 will cross the curve L1 several times before reaching the section SET, shown in Fig. 2(A), (B) and (C), respectively.

    Figure 2. Three possible trajectories of model (2) with xΓ1<(1θ)ET<ET<xΓ2 for model (2). The parameter values are fixed as: r=1,K=52,β=0.19,η=0.45,ω=0.19,δ=0.36,τ=5,ET=35,θ=0.7(A),0.659(B) and 0.5(C).

    Further, we define the section SθET={(x,y)|x=(1θ)ET,y0} and assume that it does not involve the rest point E in it. Choosing the section SθET as a Poincaré section. Assume that the point P+k=((1θ)ET,y+k) lies in the section SθET (here ESθET), and the trajectory Ψ(t,t0,(1θ)ET,y+k)=(x(t,t0,(1θ)ET,y+k),y(t,t0,(1θ)ET,y+k)) initiating from P+k will reach at the section SET in a finite time t1 (i.e. x(t1,t0,(1θ)ET,y+k)=ET and Φ(y+k)=t1), denote the intersection point as Pk+1=(ET,yk+1). This indicates that yk+1 is determined by y+k, i.e. we have yk+1=y(t1,t0,(1θ)ET,y+k)P(y+k). One time state dependent feedback control action is implemented at point Pk+1 such that it jumps to point P+k+1=((1θ)ET,y+k+1) with y+k+1=yk+1+τ on SθET. Therefore, we can define the Poincaré map PM as

    y+k+1=P(y+k)+τ=y(t1,t0,(1θ)ET,y+k)+τPM(y+k), and Φ(y+k)=t1. (4)

    Now define the impulsive set M for three cases shown in Fig. 1 as

    M={(x,y)| x=ET,0yYM}, (5)

    which is a closed subset of R2+, where YM=P(yθET) for cases (A) and (B), and YM=yET for case (C). It is easy to see that any solution of model (2) without impulsive effects initiating from Ω will reach in the impulsive set M in a finite time except for the rest point E. Define the continuous function I:(ET,y)M(x+,y+)=((1θ)ET,y+τ)Ω. Thus, the phase set N can be defined as follows

    N=I(M)={(x+,y+)Ω| x+=(1θ)ET,τy+P(yθET)+τ}. (6)

    Therefore, (R2+,Π;Ω,M,I) or (R2+,Π;M,I) with respect to model (2) defines an impulsive semi-dynamical system.

    According to the Definition 1.3 and topological structure of orbits of model (2) without impulsive effects, it is easy to see that M defined in (5) is a closed and well placed subset of R2+, and further no initial point in (R2,Π) belongs to the impulsive set M. This indicates that according to Theorem 1 the function Φ is continuous for the three cases shown in Fig. 2. For the cases shown in Fig. 2(A) and (B) it is easy to see that the results are true.

    However, this is not true for case (C) shown in Fig. 2(C). In fact, for case (C) there exists a trajectory (denoted by Γ2) initiating from Ω will tangent to the section SET at point QET, and the orbit Γ2 will intersect with the section SθET at several points, for example, point A=((1θ)ET,yA) as shown in Fig. 2(C). Now we consider this point A and choose an open set U which contains the point A. Taking any two points AiUSθET(i=1,2) with yA1<yA<yA2, we see that the trajectory starting from the point A2 will cross the curve L1 at least one time before hitting the impulsive set M, and the trajectory starting from the point A1 will reach the impulsive set M directly. This confirms that both trajectories initiating from the set USθET can not simultaneously reach the tube V of the impulsive set M for small s, and consequently the function Φ is not continuous at the point A.

    If we fixed all the parameter values as those shown in Fig. 3, then we can see that the continuities of the Poincaré map and the function Φ depend on the parameter θ. For larger θ values both the Poincaré map and the function Φ are continuous, and for smaller θ values both the Poincaré map and the function Φ exist several discontinuous points, as shown in Fig. 3. Moreover, we can show that the function Φ is continuous when the E is stable equilibrium or E does not exist. Thus, this counter example shows that more exact conditions are needed to show the continuity of the function Φ[1,9]. Based on above discussion, we can see that if the impulsive set M is transversal, i.e. there is no trajectory of model (2) without impulsive effect which tangents to the set M, then the function Φ is continuous. Therefore, we have the following result on the continuity of the function Φ which can be proved by using the similar method as those in[1,2,9].

    Figure 3. Continuity of Poincaré map and time function without impulse of model (2) for different θ. The other parameter values are fixed as: r=1,K=52,β=0.19,η=0.45,ω=0.19,δ=0.36,τ=5,ET=35.

    Theorem 2.1. Let (R2,Π) be a dynamical system and consider an impulsive semi-dynamical system (R2,Π,M,I). Let M be a closed and well placed subset of R2 which is transversal. Then Φ is continuous on R2 if and only if zM.

    Note that the transversality condition in Theorem 2.1 may exclude the case (B) in Fig. 2(B). In fact, based on our example we can conclude that the function Φ is continuous on phase set provided that the Poincaré map reaches its maximal value at yθET. This confirms that the conditions shown in Theorem 2.1 could be further improved.


    3. Dicsussion

    Recently, impulsive semi-dynamical systems or state dependent feedback control systems arise from many important applications in life sciences including biological resource management programmes and chemostat cultures [5,6,10,12,17,18,19,20,21,22,24], diabetes mellitus and tumor control [8,13], vaccination strategies and epidemiological control [14,15], and neuroscience [3,4,7]. In those fields, the threshold policies such as ET and threshold glucose level for closed-loop treatment play the pivotal role in the pest control and medications to treat diabetes. For example, control actions such as spraying pesticides and releasing natural enemies for pest control and drug administrations (such as insulin and glucose-lowering pills) for disease treatment depend on the threshold levels including density of pest population and glucose level [8,16,17,24].

    The above state-dependent feedback control strategies can be defined in broad terms in real biological problems, which are usually modeled by the impulsive semi-dynamical systems. The continuity of the function Φ describing the times of meeting impulsive set which is equivalent to the continuity of the Poincaré map determined by the impulsive points in phase set (as shown in Fig. 3) plays an important role in investigating the qualitative theories of the impulsive semi-dynamical system. In particular, it is very useful for determining the existence and stability of order one periodic solution or order one limit cycle [21,23]. However, in order to guarantee the continuity of this function, only some special cases for the proposed models have been discussed in literatures [11,12,16,23]. For example, the proposed models without impulsive effects do not have any positive equilibrium in the first quadrant, or the threshold level must be less than the values of the steady state once the models exist the interior equilibrium. Moreover, the Poincaré map can also help us to study the dynamic complexity of impulsive semi-dynamical system and its application in reality. Therefore, determining the continuity of the function describing the times of meeting impulsive set and then the Poincaré map are crucial for proposed models arising from different fields.




    [1] M. Kadem, L. Garber, M. Abdelkhalek, B. K. Al-Khazraji, Z. Keshavarz-Motamed, Hemodynamic modeling, medical imaging, and machine learning and their applications to cardiovascular interventions, IEEE Rev. Biomed. Eng., 16 (2023), 403–423. https://doi.org/10.1109/RBME.2022.3142058 doi: 10.1109/RBME.2022.3142058
    [2] M. F. O'Rourke, A. Adji, W. W. Nichols, C. Vlachopoulos, E. R. Edelman, Application of arterial hemodynamics to clinical practice: a testament to medical science in London, Artery Res., 18 (2017), 81–86. https://doi.org/10.1016/j.artres.2017.03.003 doi: 10.1016/j.artres.2017.03.003
    [3] K. Chatterjee, The swan-ganz catheters: past, present, and future: a viewpoint, Circulation, 119 (2009), E548. https://doi.org/10.1161/CIRCULATIONAHA.109.192583 doi: 10.1161/CIRCULATIONAHA.109.192583
    [4] R. Kett-White, P. J. Hutchinson, P. G. Al-Rawi, A. K. Gupta, J. D. Pickard, P. J. Kirkpatrick, Adverse cerebral events detected after subarachnoid hemorrhage using brain oxygen and microdialysis probes, Neurosurgery, 50 (2002), 1212–1221. https://doi.org/10.1097/00006123-200206000-00008 doi: 10.1097/00006123-200206000-00008
    [5] P. van Ooij, W. V. Potters, J. Collins, M. Carr, J. Carr, S. C. Malasrie, et al., Characterization of abnormal wall shear stress using 4D flow MRI in human bicuspid aortopathy, Ann. Biomed. Eng., 43 (2015), 1385–1397. https://doi.org/10.1007/s10439-014-1092-7 doi: 10.1007/s10439-014-1092-7
    [6] Y. Qin, J. H. Wu, Q. M. Hu, D. N. Ghista, K. K. L. Wong, Computational evaluation of smoothed particle hydrodynamics for implementing blood flow modelling through CT reconstructed arteries, J. X-ray Sci. Technol., 25 (2017), 213–232. https://doi.org/10.3233/XST-17255 doi: 10.3233/XST-17255
    [7] V. M. Pereira, B. Delattre, O. Brina, P. Bouillot, M. I. Vargas, 4D flow MRI in neuroradiology: techniques and applications, Top. Magn. Reson. Imaging, 25 (2016), 81–87. https://doi.org/10.1097/RMR.0000000000000082 doi: 10.1097/RMR.0000000000000082
    [8] K. Y. Lin, T. C. Shih, S. H. Chou, Z. Y. Chen, C. H. Hsu, C. Y. Ho, Computational fluid dynamics with application of different theoretical flow models for the evaluation of coronary artery stenosis on ct angiography: Comparison with invasive fractional flow reserve, Biomed. Phys. Eng. Express, 2 (2016), 065011. https://doi.org/10.1088/2057-1976/2/6/065011 doi: 10.1088/2057-1976/2/6/065011
    [9] A. Dubey, B. Vasu, O. Anwar Beg, R. S. R. Gorla, A. Kadir, Computational fluid dynamic simulation of two-fluid non-newtonian nanohemodynamics through a diseased artery with a stenosis and aneurysm, Comput. Methods Biomech. Biomed. Eng., 23 (2020), 345–371. https://doi.org/10.1080/10255842.2020.1729755 doi: 10.1080/10255842.2020.1729755
    [10] E. C. Mason, S. McGhee, K. Zhao, T. Chiang, L. Matrka, The application of computational fluid dynamics in the evaluation of Laryngotracheal Pathology, Ann. Otol. Rhinol. Laryngol., 128 (2019), 453–459. https://doi.org/10.1177/0003489419826601 doi: 10.1177/0003489419826601
    [11] C. Zhang, B. Lin, D. Li, Y. Fan, Application of multiscale coupling models in the numerical study of circulation system, Med. Novel Technol. Devices, 14 (2022), 100117. https://doi.org/10.1016/j.medntd.2022.100117 doi: 10.1016/j.medntd.2022.100117
    [12] L. Liang, M. Liu, C. Martin, W. Sun, A deep learning approach to estimate stress distribution: a fast and accurate surrogate of finite-element analysis, J. R. Soc. Intrface, 15 (2018), 20170844. https://doi.org/10.1098/rsif.2017.0844 doi: 10.1098/rsif.2017.0844
    [13] G. Hajgato, B. Gyires-Toth, G. Paal, Accelerating convergence of fluid dynamics simulations with convolutional neural networks, Period. Polytech. Mech. Eng., 63 (2019), 230–239. https://doi.org/10.1098/rsif.2017.0844 doi: 10.1098/rsif.2017.0844
    [14] G. Y. Li, H. R. Wang, M. Z. Zhang, S. Tupin, A. K. Qiao, Y. J. Liu, et al., Prediction of 3D cardiovascular hemodynamics before and after coronary artery bypass surgery via deep learning, Commun. Biol., 4 (2021), 1–12. https://doi.org/10.1038/s42003-020-01638-1 doi: 10.1038/s42003-020-01638-1
    [15] L. Liang, W. Mao, W. Sun, A feasibility study of deep learning for predicting hemodynamics of human thoracic aorta, J. Biomech., 99 (2020), 109544. https://doi.org/10.1016/j.jbiomech.2019.109544 doi: 10.1016/j.jbiomech.2019.109544
    [16] G. Kissas, Y. B. Yang, E. Hwuang, W. R. Witschey, J. A. Detre, P. Perdikaris, Machine learning in cardiovascular flows modeling: predicting arterial blood pressure from non-invasive 4D flow MRI data using physics-informed neural networks, Comput. Methods Appl. Eng., 358 (2020), 112623. https://doi.org/10.1016/j.cma.2019.112623 doi: 10.1016/j.cma.2019.112623
    [17] G. E. Karniadakis, I. G. Kevrekidis, L. Lu, P. Perdikaris, S. F. Wang, L. Yang, Physics-informed machine learning, Nat. Rev. Phys., 3 (2021), 422–440. https://doi.org/10.1038/s42254-021-00314-5 doi: 10.1038/s42254-021-00314-5
    [18] M. Raissi, A. Yazdani, G. E. Karniadakis, Hidden fluid mechanics: learning velocity and pressure fields from flow visualizations, Science, 367 (2020), 1026–1030. https://doi.org/10.1126/science.aaw4741 doi: 10.1126/science.aaw4741
    [19] S. Z. Cai, Z. P. Mao, Z. C. Wang, M. L. Yin, G. E. Karniadakis, Physics-informed neural networks (pinns) for fluid mechanics: a review, Acta Mech. Sinica, 37 (2021), 1729–1740. https://doi.org/10.1007/s10409-021-01148-1 doi: 10.1007/s10409-021-01148-1
    [20] M. Alber, A. B. Tepole, W. R. Cannon, S. De, S. Dura-Bernal, K. Garikipati, et al., Integrating machine learning and multiscale modeling-perspectives, challenges, and opportunities in the biological, biomedical, and behavioral sciences, NPJ Digital Med., 2 (2019), 115. https://doi.org/10.1038/s41746-019-0193-y doi: 10.1038/s41746-019-0193-y
    [21] E. Weinan, Machine learning and computational mathematics, Commun. Comput. Phys., 28 (2020), 1639–1670. https://doi.org/10.4208/cicp.OA-2020-0185 doi: 10.4208/cicp.OA-2020-0185
    [22] L. Huan, W. Lei, E. Weinan, Machine-learning-based non-newtonian fluid model with molecular fidelity, Phys. Rev. E, 102 (2020), 043309. https://doi.org/10.1103/PhysRevE.102.043309 doi: 10.1103/PhysRevE.102.043309
    [23] E. Samaniego, C. Anitescu, S. Goswami, V. M. Nguyen-Thanh, H. Guo, K. Hamdia, et al., An engergy approach to the solution of partial differential equations in computational mechanics via machine learning: concepts, implementation, and applications, Comput. Methods Appl. Mech. Eng., 362 (2020), 112790. https://doi.org/10.1016/j.cma.2019.112790 doi: 10.1016/j.cma.2019.112790
    [24] N. Sukumar, A. Srivastava, Exact imposition of boundary conditions with distance functions in physics-informed deep neural networks, Comput. Methods Appl. Mech. Eng., 389 (2022), 114333. https://doi.org/10.1016/j.cma.2021.114333 doi: 10.1016/j.cma.2021.114333
    [25] X. W. Jin, S. Z. Cai, H. Li, G. E. Karniadakis, NSFnets (Navier-Stokes Flow Nets): physics-informed neural networks for the incompressible Navier-Stokes equations, J. Comput. Phys., 426 (2021), 109951. https://doi.org/10.1016/j.jcp.2020.109951 doi: 10.1016/j.jcp.2020.109951
    [26] A. Valencia, F. Solis, Blood flow dynamics and arterial wall interaction in a saccular aneurysm model of the basilar artery, Comput. Struct., 84 (2006), 1326–1337. https://doi.org/10.1016/j.compstruc.2006.03.008 doi: 10.1016/j.compstruc.2006.03.008
    [27] M. Tremmel, S. Dhar, E. I. Levy, J. Mocco, H. Meng, Influence of Intracranial aneurysm-to-parent vessel size ratio on hemodynamics and implication for rupture: results from a virtual experimental study, Neurosurgery, 64 (2009), 622–630. https://doi.org/10.1227/01.NEU.0000341529.11231.69 doi: 10.1227/01.NEU.0000341529.11231.69
    [28] K. M. Tse, R. Chang, H. P. Lee, S. P. Lim, S. K. Venkatesh, P. Ho, A computational fluid dynamics study on geometrical influence of the aorta on hemodynamics, Eur. J. Cardio-Thoracic Surg., 43 (2012), 829–838. https://doi.org/10.1093/ejcts/ezs388 doi: 10.1093/ejcts/ezs388
    [29] M. Simao, J. Ferreira, A. C. Tomas, J. Fragata, H. Ramos, Aorta ascending aneurysm analysis using CFD models towards possible anomalies, Fluid, 2 (2017), 31. https://doi.org/10.3390/fluids2020031 doi: 10.3390/fluids2020031
    [30] R. Savabi, M. Nabaei, S. Farajollahi, N. Fatouraee, Fluid structure interaction modeling of aortic arch and carotid bifurcation as the location of baroreceptors, Int. J. Mech. Sci., 165 (2020), 105222. https://doi.org/10.1016/j.ijmecsci.2019.105222 doi: 10.1016/j.ijmecsci.2019.105222
    [31] M. Raissi, Z. C. Wang, M. S. Triantafyllou, G. E. Karniadakis, Deep learning of vortex induced vibrations, J. Fluid Mech., 861 (2019), 119–137. https://doi.org/10.1017/jfm.2018.872 doi: 10.1017/jfm.2018.872
    [32] J. Lin, S. Zhou, H. Guo, A deep collocation method for heat transfer in porous media: verification from the finite element method, J. Energy Storage, 28 (2020), 101280. https://doi.org/10.1016/j.est.2020.101280 doi: 10.1016/j.est.2020.101280
    [33] K. M. Hamdia, X. Zhuang, T. Rabczuk, An efficient optimization approach for designing machine learning models based on genetic algorithm, Neural Comput. Appl., 33 (2021), 1923–1933. https://doi.org/10.1007/s00521-020-05035-x doi: 10.1007/s00521-020-05035-x
    [34] A. Kendall, Y. Gal, R. Cipolla, Multi-task learning using uncertainty to weigh losses for scene geometry and semantics, in Proceedings of the IEEE Computer Society, Conference on Computer Vision and Pattern Recognition, IEEE Computer Society, (2018), 7482–7491. https://doi.org/10.1109/CVPR.2018.00781
    [35] K. H. Thung, C. Y. Wee, A brief review on multi-task learning, Multimedia Tools Appl., 77 (2018), 29705–29725. https://doi.org/10.1007/s11042-018-6463-x doi: 10.1007/s11042-018-6463-x
    [36] M. L. Wang, H. X. Li, X. Chen, Y. Chen, Deep learning-based model reduction for distributed parameter systems, IEEE Trans. Syst. Man Cybern.-Syst., 46 (2016), 1664–1674. https://doi.org/10.1109/TSMC.2016.2605159 doi: 10.1109/TSMC.2016.2605159
  • This article has been cited by:

    1. Qianqian Zhang, Biao Tang, Sanyi Tang, Vaccination threshold size and backward bifurcation of SIR model with state-dependent pulse control, 2018, 455, 00225193, 75, 10.1016/j.jtbi.2018.07.010
    2. Juhua Liang, Qian Yan, Changcheng Xiang, Sanyi Tang, A reaction-diffusion population growth equation with multiple pulse perturbations, 2019, 74, 10075704, 122, 10.1016/j.cnsns.2019.02.015
    3. Sanyi Tang, Changtong Li, Biao Tang, Xia Wang, Global dynamics of a nonlinear state-dependent feedback control ecological model with a multiple-hump discrete map, 2019, 79, 10075704, 104900, 10.1016/j.cnsns.2019.104900
    4. Sanyi Tang, Xuewen Tan, Jin Yang, Juhua Liang, Periodic Solution Bifurcation and Spiking Dynamics of Impacting Predator–Prey Dynamical Model, 2018, 28, 0218-1274, 1850147, 10.1142/S021812741850147X
    5. Qian Li, Yanni Xiao, Dynamical Behavior and Bifurcation Analysis of the SIR Model with Continuous Treatment and State-Dependent Impulsive Control, 2019, 29, 0218-1274, 1950131, 10.1142/S0218127419501311
    6. Xiyin Liang, Yongzhen Pei, Jianguo Tan, Yunfei Lv, Optimal parameter selection problem of the state dependent impulsive differential equations, 2019, 34, 1751570X, 238, 10.1016/j.nahs.2019.07.001
    7. Qianqian Zhang, Biao Tang, Tianyu Cheng, Sanyi Tang, Bifurcation Analysis of a Generalized Impulsive Kolmogorov Model With Applications to Pest and Disease Control, 2020, 80, 0036-1399, 1796, 10.1137/19M1279320
    8. Tianyu Cheng, Sanyi Tang, Robert A. Cheke, Threshold Dynamics and Bifurcation of a State-Dependent Feedback Nonlinear Control Susceptible–Infected–Recovered Model1, 2019, 14, 1555-1415, 10.1115/1.4043001
    9. Qian Li, Yao Bai, Biao Tang, Modelling the pulse population-wide nucleic acid screening in mitigating and stopping COVID-19 outbreaks in China, 2023, 23, 1471-2334, 10.1186/s12879-023-08265-1
    10. Qian Li, Yanni Xiao, Analysis of a hybrid SIR model combining the fixed-moments pulse interventions with susceptibles-triggered threshold policy, 2023, 453, 00963003, 128082, 10.1016/j.amc.2023.128082
    11. Yongfeng Li, Song Huang, Xinyu Song, Global dynamic analysis of a nonlinear state-dependent feedback control SIR model with saturation incidence, 2023, 138, 2190-5444, 10.1140/epjp/s13360-023-04277-7
    12. Chenxi Huang, Qianqian Zhang, Sanyi Tang, Non-smooth dynamics of a SIR model with nonlinear state-dependent impulsive control, 2023, 20, 1551-0018, 18861, 10.3934/mbe.2023835
    13. Shuai Chen, Wenjie Qin, Antipredator behavior of a nonsmooth ecological model with a state threshold control strategy, 2024, 9, 2473-6988, 7426, 10.3934/math.2024360
    14. Wenjie Qin, Zhengjun Dong, Lidong Huang, Impulsive Effects and Complexity Dynamics in the Anti-Predator Model with IPM Strategies, 2024, 12, 2227-7390, 1043, 10.3390/math12071043
    15. Zhanhao Zhang, Yuan Tian, Dynamics of a nonlinear state-dependent feedback control ecological model with fear effect, 2024, 9, 2473-6988, 24271, 10.3934/math.20241181
    16. Chenxi Huang, Qianqian Zhang, Sanyi Tang, Global analyses of a ratio-dependent state feedback control system for infectious disease, 2025, 151, 10075704, 109068, 10.1016/j.cnsns.2025.109068
  • Reader Comments
  • © 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)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

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

Metrics

Article views(4284) PDF downloads(512) Cited by(10)

Article outline

Figures and Tables

Figures(12)  /  Tables(6)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog