Finite Element Modelling of Pulsatile Blood Flow in Idealized Model of Human Aortic Arch: Study of Hypotension and Hypertension

A three-dimensional computer model of human aortic arch with three branches is reproduced to study the pulsatile blood flow with Finite Element Method. In specific, the focus is on variation of wall shear stress, which plays an important role in the localization and development of atherosclerotic plaques. Pulsatile pressure pulse is used as boundary condition to avoid flow entry development, and the aorta walls are considered rigid. The aorta model along with boundary conditions is altered to study the effect of hypotension and hypertension. The results illustrated low and fluctuating shear stress at outer and inner wall of aortic arch, proximal wall of branches, and entry region. Despite the simplification of aorta model, rigid walls and other assumptions results displayed that hypertension causes lowered local wall shear stresses. It is the sign of an increased risk of atherosclerosis. The assessment of hemodynamics shows that under the flow regimes of hypotension and hypertension, the risk of atherosclerosis localization in human aorta may increase.


Introduction
Atherosclerosis is the disease of large arteries (carotid, aorta, and other proximal arteries) and tends to localize in regions of curvature and branching in arteries. Caro et al. [1,2] and DeBakey et al. [3] categorized aortic arch, major branches of aortic arch, and abdominal aorta as susceptible sites for creation and development of atherosclerosis. The complex anatomies of arteries are often associated with abnormal flow dynamics and stress distributions. Ku et al. [4], Nerem [5], Tarbell [6], and Zarins et al. [7] have reported that atherosclerosis is more prone to occur in regions of low shear stress and oscillating shear stress increases the risk of atherosclerosis localization. Fry [8] demonstrated that under acute elevation of shear stress the endothelial layer of arterial wall may damage and increases its permeability for lipids. These studies prove that there is a positive correlation between low and fluctuating wall shear stress and risk of development of atherosclerosis.
The human aorta has complex anatomy with curvature, branching, and distal tapering. Caro [2] and Fry [8] have shown that the human aortic arch is vulnerable to localization of atherosclerosis due to the complex anatomy of aorta. In his work, Utepov [10] has demonstrated the tapering of arteries as one of the risk factors for manifestation of atherosclerosis. Kilner et al. [11] detected, in vivo, the complicated helical and retrograde flow caused by the aortic arch and pulsatile nature of the inflow of blood from left ventricle of heart. Thus the complex arterial flow mechanics in aorta may promote the early development atherosclerotic plaque. The blood flow in thoracic aorta models has been also studied numerically [12][13][14][15][16][17][18][19]. However, some of studies neglected the three major of branches of aortic arch. Towfiq [9] and Dabagh et al. [20] have shown that the aorta size is subjected to change with blood pressure. But no attention has been made on studying the corresponding influence on the blood flow features. Moreover, the effect of blood pressure on changing the aorta geometry has also been ignored.   28.0 Diameter (A 3 ) 9 . 9 Distance between BA and LCA (d 1 ) 3 . 6 Distance between LCA artery and LSA (d 2 ) 4 . 7 In the present study, we have reconstructed a threedimensional (3D) aorta model adopted from the literature. The aorta model includes three major branches (brachiocephalic artery, left carotid artery and left subclavian artery) in the aortic arch. The pulsatile blood flow through three 3D model of aortic arch was simulated with incompressible Navier-Stokes equations. The governing equations along with boundary conditions are solved with Finite element method-(FEM-) based code Comsol Multiphysics V3.4. The idealized aorta model along with inlet pressure profile was further modified to simulate aorta under hypotension and hypertension conditions. Though the transient variation of blood pressure within a cardiac cycle causes instantaneous deformation of the aorta wall, which in turn may affect the blood flow features (two-way coupling between wall deformation and blood flow), such effects are ignored in this study. Because of the critical role played by the shear stress in arterial wall diseases, the wall shear stress (WSS) is studied extensively within branches and through the arch of aorta. Moreover, the velocity profiles across various cross sections of aortic arch and branches are investigated. The first objective of present study is to find out how the aortic geometry is connected with the blood flow development and the shear stress distribution in the aorta wall. The second objective is to realize how blood pressure and its associated wall deformation affect the distribution of WSS, the flow profiles, and the volumetric outflow across branches. Therefore, this study seeks the role of aortic arch geometry (with branches) and the mean blood pressure on the resulting transient blood flow only when the wall of aorta is rigid. However, one should note that the corresponding results might be different than what we observe in reality due to the flexibility of real wall and the complexity of the real aorta geometry.

Materials and Methods
The idealized model of human aorta was reconstructed based on aorta model used by Shahcheranhi et al. [18]. The model is constructed in six parts, namely, ascending aorta, aortic arch, descending aorta, and three branches (brachiocephalic artery, left common carotid artery, and left subclavian artery). The simplification of the geometry is basically associated with neglecting the undetermined curvature of the abdominal part of aorta and torsion of ascending aorta, as the original aorta models were reconstructed from the computed tomographic (CT) images. Moreover, there may be minor differences in original aorta models and aorta models used in this study due to different design module used for reconstruction purpose. Figure 1(a) illustrates the schematic diagram of aorta with the details of geometric measures given in Table 1. The aorta model was modified for two working pressures inside the aorta mimicking the hypotension (65-105 mmHg) and hypertension (100-140 mmHg) conditions. The variation of the inlet crosssectional area versus pressure is demonstrated in Figure 1(b), which has been used in the construction of the geometry under the effect of hypotension and hypertension conditions.
The blood is assumed to be a Newtonian fluid. The assumption of Newtonian fluid for blood with a constant viscosity is feasible in large arteries. Although some works such as Khanfar et al. [21] showed that the non-Newtonian assumption of blood affects the blood flow in aorta aneurysms, their simulations did not display significant differences in shear stress calculated from Newtonian and non-Newtonian simulations. The driving force for the blood flow in an artery is the pressure gradient along the artery. Thus the pulsatile pressure-inlet boundary condition (with zero viscous stress) was used at inlet and outlets. The outflow Table 2: Values of coefficients for polynomial used as pressure waveform (×10 5 ). from left ventricle of heart is not always uniform as flow disturbance is induced by aortic valve. Although an exact effect of tricuspid aortic valve cannot be exactly mimicked, with the deployment of pressure-inlet boundary condition we can avoid uniform inflow at inlet of aorta which has been typically used in several works. Thus the pressureinlet boundary condition in the present study will be more realistic. In the present study, a pulsatile pressure is deployed as inlet and outlet boundary conditions. An eighth degree polynomial was used to reproduce the inlet pressure pulse in mmHg from the data given by Conlon et al. [22] as where n is the cardiac cycle number varying from 1 to 4.
The polynomial coefficients C i are introduced in Table 2.
The period of systole and diastole in the pressure pulse took 0.35 s and 0.5 s, respectively. The pressure pulse at the inlet of ascending aorta is shown under normal pressure condition in Figure 1(c). For boundary conditions at the outlets, the pressure pulse is multiplied by a certain coefficient that represents pressure drop between inlet and respective outlet. The coefficients for respective branch outlets are obtained by a separate series of steady-state simulations with flat velocity at inlet and target mass flow at outlet. Li [23] has suggested that in large arteries it is reasonable to assume that the blood close to artery moves with same speed as that of wall. Thus over the walls of aorta, no-slip boundary condition is applied. Under transient condition and the assumption of incompressible flow with Newtonian rheology, the flow of blood in the aorta is governed by the continuity and Navier-Stokes equations where u denotes fluid velocity vector and p represents hydrostatic pressure. Also, ρ is the density of blood taken as 1060 kg/m 3 and μ is the dynamic viscosity of blood that is a constant as 0.005 Pa.s.
To ensure periodic nature of the flow, the simulations were performed for four cardiac cycles (pressure pulse) where each cycle is 0.85 s. The results from fourth cardiac cycles are discussed in the results section. The governing equations were discretized with Backwards Difference Method, which is known to be a very stable method for discretization. The set of governing (2)-(3) along with boundary conditions were solved by means of the FEM method provided in COMSOL Multiphysics, v. 3.4. The mesh grid for each aorta case was refined subsequently to obtain mesh-independent-results. The computational mesh utilized in primary simulations consisted of nearly 35000 to 42000 tetrahedral elements. The mesh grid was refined till the resulting distributions of the flow and WSS were qualitatively identical: spatially and temporally. The final mesh grid consisted of nearly 140000 grid elements. To further increase accuracy and reliability of the solutions, the local variations of flow variable within the grid elements were predicted by quadratic piecewise functions. Memory friendly iterative solvers GMRES and FGMRES were used for solving the discretized governing equations. The residual for solution was kept at 1 × 10 −4 and the simulations progressed with a time step of 0.001 s. A steady-state simulation with 80 mmHg pressure was performed with direct solver UMFPACK and was used as an initial guess for the transient simulations. The transient simulations were performed with personal computer with 3 GHz Core 2 Duo processor with 3 GB of RAM.
The aorta model used in the current study was adopted from the literature and thus lacked some features of realistic aorta, for example, curvature in descending aorta. The study was focused on flow dynamics in aortic arch and three major branches, thus the branches of thoracic aorta were neglected. In real aorta, the branch entry regions have blunt corners; however, this feature of aorta model was not included due to lack of exact measurements. This feature of aorta model was also found missing in the original works [13,18]. Due to limited memory available for computation, the fluidstructure interaction could not be included in the current set of simulation. The reported study is mainly focused on understanding the effects of hypotension and hypertension on the flow and WSS distributions. The reported study is mainly focused on understanding the effects of hypotension and hypertension on the flow and WSS distributions. Although it is indicative that assumption of rigid wall in blood flow simulation may underestimate temporal and spatial flow and wall motion, the aorta walls were assumed to be rigid for flow simulations under all three pressure regimes. Long-term hypertension can cause thickening of arterial wall and loss of elasticity of arterial wall [24]. Thus, aorta walls can be assumed rigid under hypertensive flow regime.
The assumption of rigid artery wall in other two cases is acceptable as earlier works [21,[25][26][27]

Results and Discussion
The axial velocity distributions and peripheral WSS at peak systolic time instance t = 0.18 s are compared. The distribution of velocity in vicinity of the walls is crucial for the calculating WSS, as low and fluctuating WSS are known to be the key parameters of the development of atherosclerotic plaques. Higher WSS is also known to be closely related to aorta dissection. Ku [24] has reported that skewness in axial profiles can cause alteration and oscillation of local WSS, which can cause localization of atherosclerosis. The distributions of WSS at peak systole for aorta model under normal pressure regime are presented in Figure 4. The distributions are shown in two view angles with full and low range of WSS. The anterior and posterior views in Figure 4(a) and Figure 4  on the proximal walls of branches and the higher WSS distributed on distal walls of branches. On lowering down the scale we also realized extremely low WSS distributed at the outer wall or ascending aorta and inner wall of the aortic arch (Figure 4(c) and Figure 4(d)). The demonstrated distribution of flow and WSS were in qualitative agreement with the earlier numerical [13,18,28] and experimental [5] works.
To further analyze the WSS distributions quantitatively, circumferential WSS captured at axial cross-sections D-I is plotted as polar plots in Figure 5. In sections D, F, and I of branches, the low values of WSS are distributed in the angular domain from 90 • to 270 • counterclockwise, while the higher values of WSS are distributed on the opposite domain, that is, from 270 • to 90 • counterclockwise. The distributions of circumferential WSS at axial sections close to branch outlets follows cardioids-like pattern.
In order to get stable values, WSS is averaged peripherally for brachiocephalic left carotid, and left subclavian arteries. An area weighted average is obtained for 10 arc sections over each branch. The arch sections are essentially of arc length of 2.6 mm. The peripherally average WSS for brachiocephalic, left carotid and left subclavian arteries is presented in Figure 5(g). To avoid errors in averaging due to extreme WSS values close to branch entry and exit, WSS values close to the branch entry and exit regions are excluded from averaging. The values of mean WSS display limited variation along each branch which is about 3.5 Pa for brachiocephalic, 2.5 Pa for left carotid, and 1.5 Pa for left subclavian branches. As the lower WSS is blamed for the development of atherosclerotic plaques, the brachiocephalic artery displays more vulnerability to the development of plaques than the two others arteries. This conclusion is in agreement with earlier works [2,3], where the brachiocephalic artery is determined as one of the predominant sites for localization of atherosclerosis plaque. However, Figure 5  large for the brachiocephalic artery. It should be noted that there are several other factors not studied here such as the geometry of aortic arch (patient-specific data associated with the shape of arch and branches with local details of the wall surface) and the flexibility of the aorta wall. Such factors will be studied through our next papers. The mass flow rate across branch outlets followed distribution similar to outlet boundary condition. The flow rates were fairly the same in all three branches as it was observed for WSS, too. The reason behind similarity of mass flow rates is also the fact that the actual boundary condition was determined from equal mass flow rate condition.

Effect of Hypotension and Hypertension on the Distributions of Flow Field and WSS.
The effect of pressure on flow parameters is studied by changing the dimensions of the aorta model based on the pressure-lumen area relation in aorta from Towfiq et al. [9]. Moreover, the boundary conditions are also modified to mimic hypotension and hypertension pressure regimes. The flow distributions in coronal plane are shown in Figures 2(b) and 2(d), while axial flow distributions across sections A-I (as shown in Figure 3 Within the branches, the flow is skewed toward the proximal walls, and the effect of hypotension and hypertension affects both the profile and the magnitude of velocity. Again the velocity profiles for hypertension case are more pronounced and amplified. Also amongst all the axial sections, maximum velocity magnitude is observed in section E (Figures 7(g)-6(i)), which lies in vicinity of outlet of brachiocephalic artery, for hypertension case.
The effect of pressure regimes on WSS is closely investigated. The peripheral WSS for hypotension, normal pressure and hypertension is presented in form of polar plots in Figure 8. It can be noticed in Figure 8 that the magnitude of peripheral WSS does not change proportional to the pressure regimes. The WSS magnitude under hypertension condition is lower than that observed for normal cases while qualitatively the distributions are similar. Interestingly, for hypertension case, the values of WSS shown in Figure 8  Relatively lower WSS at the branch entry regions and inner aortic arch may play a role in altering vascular biology and the localization of atherosclerosis [4,6]. The aorta is under greater peripheral stresses during hypertension. Dabagh et al. [29] have reported that due to the change in the morphology of endothelial cells higher rates of mitosis or apoptosis are expected, which increases the number of LDL pathways over endothelium. Thus an elevated risk of developing atherosclerotic plaques is anticipated under hypertensive conditions. It is also instructive to look at the variation of mean WSS through different sections along the branches. In order to get stable values, WSS is averaged along the length of the brachiocephalic, left carotid, and left subclavian branches for aorta model in Figure 9. It can be seen that the levels of WSS corresponding to hypertension do not considerably surpass those of normal pressure along each branch, but they may even go lower. As mentioned earlier, this can potentially boost the penetration of LDL macromolecules across the arterial wall. Figure 9 indicates that the average peripheral WSS is almost invariant along the branch though its peripheral distribution is not uniform. Note that the hypertensive wall is under higher strain while its shear stress is either unchanged or lower than what normally exist. According to Dabagh et al. [29], this can potentially boost the penetration of LDL macromolecules across the arterial wall.

Conclusions
In the present study, an idealized model of human aorta was reconstructed with measurements from the literature. Pulsatile blood flow in aorta model was simulated with FEM-based numerical solver. The aorta model along with boundary condition was modified to mimic the effect of hypotension and hypertension. The flow dynamic in the work displayed a reasonable qualitative agreement with other earlier works. Figures 4, 5, and 8 illustrate extremely low WSS at outer wall of aortic arch, proximal wall of branches, entry region of branches, and inner aortic arch. Low and oscillating shear stress are associated with risk of developing atherosclerosis, thus the listed sites are susceptible to localization of atherosclerosis. This conclusive observation is also consistent with earlier studies. The study has provided the fundamental understanding of the flow dynamics dependence on the artery geometry and the blood pressure in human aorta. Results reveal that the hemodynamics is significantly affected by the anatomy of aorta, which emphasizes on the patient-specific factor of the aorta geometry in developing atherosclerosis. The anatomy and orientation of branches have asignificant role in flow distribution and flow dynamics. On the other hand, the blood pressure affects the site of low WSS. It also appeared that the magnitude of WSS decreases under hypertension. It is associated with the deformation of aorta under high pressure. Our simulations were based on pressure-inlet boundary condition in the inlet and pressureoutlet boundary condition at the outlets, which reproduced developed flow profiles from the inlet section. The results show the fact that flow dynamics may be more complex when the geometry of aorta varies under pressure. A very important feature for our future study is the coupling the blood flow in aorta with wall deformation under elevated pressures to analyze the blood hemodynamics more precisely. To overcome limitations offered by idealized aorta models, computed-tomography-(CT-) scans-based realistic aorta will be used in future simulations.

Conflict of Interests
There are no conflicts of interests.