Computational analysis of flow through a two-dimensional model of a mechanical bileaflet aortic heart valve

Mashhour Bani Amer*, L. Al-ibbini, H. Al-issa

Biomedical Engineering Departme , Jordan University of Science & Technology P.O.Box 3030, 22110 Irbid, JORDAN. *E-mail: m-b-amer@just.edu.jo Abstract Flow behavior through prosthetic heart valves capable of providing first-line information about valve performance. Computational fluid dynamic (CFD) models of the mechanical heart valves are presented in this paper describing steady and laminar flow characteristics in a bileaflet aortic heart valves models that are investigated over a range of systolic flow rates. The purpose of this study is to evaluate the performance of the bileaflet heart valve computationally. Velocity, stream function, pressure drop and shear stress distributions obtained from CFD were used to give an overall picture of the flow field. Some comparisons with experimental and invasive direct measurements are also included. In conclusion, the approach described herein might offer a way towards the alterations of blood flow associated with heart valve prostheses. I. Introduction The natural heart valves in the human heart (tricuspid, mitral, pulmonary, and aortic) are mechanical devices that permit the flow of blood in one direction only. Any one of these four valves may become diseased, due in part to rheumatic fever, calcification, and congenital defect. As a result, back flow of blood known as regurgitation (insufficiency) or restriction of the forward flow of blood known as stenosis may result, thereby the blood flow through the valve results in increased pressure drop across the valve orifice. However, the mitral and aortic valves are the ones most often afflicted by disease. Heart valve disease in prolonged and advanced forms severely affects the performance of a natural heart valve by causing permanent alteration to the valve tissue leading to valve dysfunction. So far, the need for valvular replacement arises when an improper functioning heart valve induces considerable disability and may ultimately lead to death if the problem is not treated. Biological and mechanical valve substitutes have been employed since the early 1960s [1]. On the other hand, the presence of the prosthesis disturbs the flow of blood producing areas of high fluid stress and separated flow regions leading to platelet activation and related thrombus formation [1]. Hence, it is extremely important to examine the likelihood of these problems by investigating as fully as possible the fluid dynamics of prosthetic heart valves, because there is a direct link between the fluid dynamics of heart valves and pathological problems encountered. Most of the previous research works on mechanical heart valves (MHV) were conducted using sophisticated and advanced measurement techniques for flow visualization. Such as Particle Image Velocimetry (PIV) or Laser Doppler Anemometry (LDA) that has been recorded by Grigioni et al. [2], Kadir et al. [3], and Lim et al. [4] to study flow patterns characteristics around these prostheses. However, the experimental studies have several drawbacks: they are not able to provide complete description of the complex flow patterns, expensive techniques, and it is tedious to apply geometrical changes in the proposed valves structure. Furthermore, a high degree of inaccuracies were found between PIV and LDA measurements presented by Browne et al. [5]. Thus, CFD considered an expedient method before detailed testing and used somehow to tackle the above experimental problems. The viability of the numerical simulation reported by Lai et al. [6] presented on the details flow dynamics in the vicinity of the leaflets during valve closure with variations in the leaflet tip geometry, which is difficult to be endured experimentally. In addition, vortex shedding that induced by valve design characteristics was also recorded computationally by Bluestein et al. [7]. It is postulated that these shed vortices provide the optimal conditions for the generation of microemboli that carry with them the main risk associated with mechanical heart valves, namely, cardioembolic stroke.

A comparative evaluation between experimental and finite-element computational studies on unsteady and laminar flows has also been reported in several papers as that presented by Stijnen et al. [8]. The measured and computed flow patterns match very well with slight differences. More recently Leuprecht et al. [9] employed a realistic flow simulation by the combination of Magnetic Resonance Imaging (MRI) and Computational Fluid Dynamic technique. The comparison of flow patterns predicted from simulations with data measured by MRI indicates the validity of the methods suggested and illustrates the potential for detailed studies of blood flow in and around cardiovascular implants in vivo. This paper is a continuation for Lim et al. [4] where a detailed study of the steady laminar flow field in the aortic root region immediately downstream of different prosthetic heart valves (except a bileaflet one) were used at the systolic peak flow phase under experimental evaluation. The case study here investigates a bileaflet MHV, which is the most commonly implanted type among other prostheses valves, because of its high durability, low transvalvular pressure drop, and centralized flow. This study illustrates the use of a numerical simulation to optimize prosthetic heart valve design and encourage the use of this method to examine valve performance when subjected to several flow conditions. The obvious discrepancy between our simulation and the work presented in [4] lies in the pressure drop results, where our computed gradients are satisfactorily in agreement with the in-vivo range in [10]. However, these predicted values exceed that observed experimentally around the St. Vincent Meditech Tilting Disc in [4], which discriminate the fact i.e. the bileaflet valve exhibited the lowest pressure drop among the previous models of MHVs. A possible explanation for this discrepancy is that the pressure gradients calculated from in-vivo techniques, such as echocardiography and direct measurements are commonly much higher than would be received from in-vitro measurements [10]. This is an advantage of this work in comparison with the work presented in [4] from the validity and more realistic simulation points of view. Although, the simulated bileaflet valve generates the lowest shear stress (450 dyn/cm2) in comparison with the reported stresses of the other models [4]. We consider in this work a planer CFD model of a bileaflet valve housed in an aortic chamber that presents somehow the realistic morphology of the aortic root from a physiological point of view [11]. Also, it is considered as a faithful replication of the in-vitro set-ups described in the previous experimental evaluations. Thus, the flow domain behavior is more honest to interpret and the agreements with the in-vitro and in-vivo studies are allowed. This provides informative qualitative measurements of the flow distribution and also influences the sensitive regions of stagnation or recirculation where thromboembolism events may occur. However, this study represents a first step towards a comprehensive predictive simulation of the MHV operation during the entire cardiac cycle, since it is restricted on steady and laminar flow conditions. Although, the used numerical model has been extensively validated where the present results are in good agreement with those of the analytic solutions of King et al. [12] in the reported velocity profiles under laminar flow. However, the great challenge lies in the development of an independent grid to attain results of high resolution, where a well match between laminar and turbulent results for small Reynolds numbers (Re<10000) as reported in [1]. Full mapping of the velocity contour or vector field, static pressure drop measurements and axial shear stress distributions at the aortic root wall and on the valve itself are also reported in this paper. The significant features of the results presented here are: () the velocity vectors field enabling the determination of a small wake region that suggests the bileaflet valve to be less prone to thrombus formation, () the pressure drop around the valve was mapped and compared with in-vivo results performed by Laske et al. [10], and () shear stresses on the lower surface of the valve indicating the lower propensity of the valve to damage blood elements and are also compared with the other prosthetic valves tested experimentally by Lim et al. [4]. The present investigation differs from previous ones in introducing the flow stream function, which gives a focus to local disturbances due to the formation of cavitation caused by a transient reduction in local pressure below the vapor pressure of blood seen at the tip of the valve leaflet. Furthermore, some details of flow patterns seen within the valve housing will be discussed in this publication. II. Valve model definition The CFD simulation presented in this paper is implemented on a bileaflet MHV having a travelling angle of 53 and an opening angle respect the symmetry axis of 12 at different sizes 20, 25, and 27mm with various systolic flow rates. These parameters are representative of a realistic aortic valve normally implanted. In such a problem the complex geometry of the valve must be simplified so that it can be studied numerically. This simplification still contains the essential features of the valve.

According to angiographic data, the valve was inserted in a geometrical model representing the shape of the aortic root, drawn on the basis of the anatomical findings of other researches [11]. For a physiological subject the height (H) of the aortic root is equal to the nominal diameter of the inlet section. The geometrical model of this device is shown in Fig.1.a. The flow is assumed to be symmetric about the centerline, and hence only one half of the actual geometry is used for the analysis as shown in Fig.1.b.

(a)

12 53

_________

2H

_________

(b)

Figure 1. (a) Model of the bileaflet aortic prototype, (b) Geometrical model for CFD investigation.

The length of the inlet wall is 4 H, while the length of the outlet wall is 10 times the height of the channel. Also, as performed by Stijnen et al., the length of the leaflet is 1.07 H and the thickness of the leaflet is 1mm. Based upon the foregoing assumptions, the governing equations for tow-dimensional steady incompressible flow of a Newtonian fluid are the continuity equation and the Navier-Stokes equations derived from [1]. u/x + v/y =0, (1) u u/x + v u/y + p/x= (2u/x2 + 2u/y2)/ Re (2) (3) u v/x + v v/y + p/y= (2v/x2 + 2v/y2)/ Re where u and v are the horizontal and vertical components, respectively, and p is the pressure. The Reynolds number Re is defined as [1]: Re= va* H / (* va *H)/ (4) va = 4Q/ (*H) ....(5) Where va is the average inflow velocity, H is the channel width, is the kinematic viscosity of the fluid, is the density, and is the fluid viscosity.

III. Numerical Simulation Numerical simulation was conducted using the commercial CFD software package FLUENT 6.0 in conjunction with GAMBIT 2.0 as a preprocessor where an efficient mesh generation procedure was developed by testing several grid densities. At this point, the comparison between the numerical

solutions indicated some ambiguous results occurring at lower grid densities. These grids were discarded until we reach the case in which the results showed independence of grid dimensions, since a simulation with a finer mesh indicated that the mesh density is fine enough to be regarded as grid independent, which has the benefit of maintaining high-resolution results. The computational grids 20, 25 (Fig.2) and 27mm are 2168 cell, 3700 cell, and 4062 cell in a quadrilateral form, respectively.

Figure 2. Computational grid used at 25mm diameter with boundary adaptation. In processing for solutions, segregated and implicit model was used in which the discrete, non-linear governing equations are linearized implicitly to produce a system of equations for the dependent variables of the velocity components and pressure in every computational cell. The resultant linear system is then solved to yield an updated flow-field solution. The assumptions for the physical model used are two-dimensional axisymmetric laminar flow of a Newtonian, incompressible, and steady fluid with a density of 1056 kg /m3 and a viscosity of 0.0035 kg /(m.s). Then the flow variables are specified on the boundaries appropriately of the physical model as a critical component of FLUENT simulations. A uniform velocity was assumed at the inflow boundary, varies according to the applied systolic flow rate, while the outflow boundary employed a constant pressure outlet set to zero. In addition, symmetry boundary condition -centerline is axisymmetry- used to model zero shear slip walls in viscous flows since the physical geometry of interest has mirror symmetry while the remaining boundaries are solid walls. Convergent numerical solutions are obtained at various systolic flow rates Q = 2, 5, 10, 12, 15, and 20 L/min applied on the used three valve sizes. It is best for both accuracy and convergence to have a mesh in which the changes in cell volume are gradual under certain iterations that serve to enhance the solution.

IV. Results and Discussion Simulation has been carried out with the valve leaflet fixed at the full open position. The computation was performed for a systolic flow rate ranging from 2 to 20 L/min. A total of three configurations are simulated and the results are discussed below. a. Velocity distribution The velocity magnitudes have been represented as contour plots with a color scale shown in Fig. 3. The flow patterns show a central velocity core and there is a large recirculation region where the velocities are very low near the wall of the aortic root. The field shows a maximum velocity at the center, which drops to zero at the edges. At the average physiological flow rate 5 L/min the highest velocity was found to be approximately 1.5 m/s at the leaflet tip in the lower valve size, which is the worst case to be studied.

(a)

(b)

(c)

Figure 3. Contours of Velocity Magnitudes (a) 20mm, (b) 25mm, and (c) 27mm.

Figure 4 shows simulated vectors plots at 20-mm valve diameter with two physiological flow rates, resting and exercise conditions (5 and 20 L/min, respectively). By examining the vector plots note the recirculation regions at the aortic root portion, and at the region below the valve leaflet which is called the wake region. (a)

(b)

Figure 4. Vectors of Velocity Magnitude, at (a) 5 L/min and (b) 20 L/min both for 20mm-valve size. In hydrodynamic terms, the presence of valve leaflets give additional flow structures not seen in the natural valve, where the main flow is essentially unobstructed. It is obvious from the previous velocity mappings (taking the mirror) that the flow is divided into three jets; two jets are between the valve leaflets and the aortic wall while the third one past the central orifice. As presented in the vector fields the wake regions reveal recirculating zones characterized by negative velocities are created between the leaflets. Those effectively confine the central jet to a smaller cross section; thus the acceleration of the central jet is maintained reducing the diverging effects that could have resulted between the leaflets in their fully open position. These effects are best seen in Fig. 5, where several axial velocity profiles in different cross sections (13 mm before the leaflet with 13 mm and 15 mm after the end tip of the leaflet) are superimposed.

Figure 5. The development of axial velocity profiles past the valve and in the wake. (25mm, 5 L/min) The velocity profile has a typical top distribution that precedes the pressure buildup at the stagnation region of the leaflets leading edge. Accordingly, the flow is accelerated, and the velocity gradients falling sharply. At x=15 mm downstream from the tip of the leaflet, the velocity of the flow is approximately 0.85 m/s, as compared to approximately 0.8 m/s proximal to the leaflet. The velocity gradients are getting steeper, translating into higher shear stresses. The acceleration of the flow along the upper and lower surfaces of the leaflet entails pronounced inertial effects; the higher velocity layer slides over the lower velocity layer, and the shear layer formed between the two layers rolls up into vortex trains. In general these velocity profiles (taking the mirror) seem to correlate with the findings of others [12] investigated the two-dimensional flow of 25 mm Carbomedics valve under laminar flow conditions. Moreover, the results obtained here are very similar in shape to those obtained in [7] with turbulent flow. Let us now turn our attention to the concept of stream function that is considered to be a convenient way to describe the flow patterns, in order to study to what extent the stasis i.e. reduction in blood flow velocity, recirculating, or separated flow will favor thrombus formation. The stream function formulated as a relation between the streamlines and the statement of conservation of mass. Stream function (x,y), is defined such that [1]: u=/y and = - /x .(6)

Streamlines contour plots are presented in Fig.6 for Re=1024, and 5122. The extent of recirculating regions along the aortic root is clearly seen and grew rapidly with higher Reynolds numbers. The location of the separation point on the valve remained at the downstream surface for the tested Reynolds numbers and the fluid layers adjacent to this solid surface are brought to rest and the flow separates from the surface while no separated regions are predicted upstream of the valve. Boundarylayer separation results in the formation of a relatively low-pressure region (wake region) which is deficient in momentum.

(a)

(b)

Figure 6. Contours of Stream Function (kg/s) applied on 25 mm (a) Re=1024, (b) Re=5122. Physically, the flow velocity inside the separated zone decreases relative to the surrounding flow as the Reynolds numbers increases. It is interested to note that the regions of high-speed flow occur where the streamlines are closed together while a lower speed flow occurs where streamline spacing is wider. b. Pressure drop The pressure gradients across valves of the different sizes are measured at low and high systolic flow rates. In our model the simulation data are taken under approximately the same conditions conducted by the direct invasive pressure gradient measurements studied by Laske et al. [10]. Where the tip of the aortic needle was placed 3 to 4 cm above the aortic valve. While in our computations two lines are superimposed in a cross-section around the created valve, one before the valve by 2 cm corresponds to the left ventricle and the other line was 3 cm after the valve. The computationally predicted valve pressure gradients (Fig.7) show that the high pressure (11.4 mmHg at flow rate 15 L/min) can be attributed to the effective reduction in orifice area of the lowest proposed valve size (20-mm). With higher systolic flow across the valves the gradients increase and at a given systolic flow, the gradients decrease from small to larger valve sizes. The obtained simulation results agree exactly with the invasive pressure gradients measurements presented in [10]. For clarity, the mean gradients are less than 10 mmHg under resting conditions and increase with exercise, but remain less than 13mmHg even in small valves with high systolic flow as reported in [10].

12

10

20 mm

) g H m m ( p o r D e r u s s e r P

25 mm
6

27 mm
4

8 10 12 Systolic Flow ( L/min )

14

16

18

20

Figure 7. Mean Pressure Gradients versus Systolic Flow Rates at different valve sizes.

c. Shear stress distribution The wall shear stress distribution along the aortic wall is depicted in Fig.8. The level of shear stress is generally found to be higher at higher flow rates. The flow separation at the entrance of the aortic root causes the wall shear stress to decrease rapidly. The values continue to be small and stay around zero over the first half of the aortic root. Then it increases negatively towards the end of the root where the flow is reversed. Its obvious that the wall shear stress reverse back to positive signs with a sharp increase due to a sudden change in the flow direction at the downstream end.

Figure 8. Axial wall shear stress (Pascal) distributed on the aortic wall for 25mm-valve size at various flow rates.

By examining the shear stresses on the valve itself we found that the maximum shear takes place on the lower surface due to high velocity gradients. At higher flow rates the wall shear stress increases with maximum value at the entry tip of the valve, the resultant simulation is shown in Fig. 9. Zones of high fluid stress indicate the possibility of thrombosis and haemolysis formation. By platelet activation at the entry tip where the high shear stress is generated then a depression of the platelets occurs in the wake region, which is the stasis and a low-pressure region.

Figure 9. Wall Shear Stress (Pascal) on the lower surface of the valve for 20 mm. At 5, 10 and 15 L/min. The maximum shear stress 45 Pascal i.e. (450 dyne/cm2) demonstrated in our simulation was at high flow rate and the smallest valve size (15 L/min, 20mm). By comparing the results of the analysis with those observed in the experimental investigation by Browne et al. [5], one can see that the LDV maximum turbulent shear stress plot (at 25.4 mm flow inlet under 30 L/min) show regions of high stress (500 dyne/cm2) located between the sinus wall and the main jet, which corresponds in our laminar study to the shear stress on the lower surface of the valve. In particular, the steady laminar flow results tend to underestimate the level of the shear stress in relation to those found in the pulsatile flow situation. V. Conclusions The flow characteristics of aortic bileaflet MHV models have been investigated computationally. This study demonstrated that the Computational Fluid Dynamic (CFD) simulation can be used to study the complex fluid dynamics in detailed and can be a useful tool for valve design before machining prototypes for detailed testing. Such method circumvents the experimental drawbacks that may encounter. The results presented in this study only refer to the systolic peak flow phase of the cardiac cycle. The following conclusions can be drawn from this study: 1. 2. The results show that the general flow features under various systolic flow rates i.e. the resting and exercise conditions are qualitatively similar but the latter yields a higher order of magnitude in terms of wall shear stress and static pressure drop. At the lower surface of the valve there is a local low-pressure spot for the MHV studied. The current bileaflet MHV designs enhance free emboli formation through the wake region generated by the valve, the challenge facing MHV design optimization would be to delineate those design parameters that would reduce or ideally eliminate the formation of shed vortices. The simulation also demonstrated that the mean gradients are less than 10 mmHg under resting conditions and increase with exercise, but remain less than 13 mmHg even in small valves with high systolic flow. An agreement between the current simulation results and the invasive

3.

10

4.

pressure direct measurements presented in [10] serve as further validation of the results of simulation. The computational shear stress distribution on the aortic wall as well as along the leaflet indicate that the maximum shear stress results using 20 mm valve size at 15 L/min is 450 dyn/cm2 at the tip of the valve, which has a good correlation with experimental results studied using Particle Image Velocimetry technique (PIV) [5].

Of course, the computational results correctly predict the main characteristics of the fluid. However, bileaflet MHVs designs are significantly have less flow impediment than other mechanical counterparts, a fact that encourage their widespread use throughout the world. It must be stated that, the conducted comparisons reveal in general good qualitative and quantitative agreements between the previous studied measurements and these simulated results.

References Jagan N. Mazumdar, 1998. Biofluid Mechanics. World Scientific Publishing Co. Pte. Ltd. 2. Grigioni, M., Daniele, C., Avenio, G., and Barbaro, V., 2001. The influence of the leaflets curvature on the flow field in two bileaflet prosthetic heart valves. Journal of Biomechanics, 613-621. 3. Kadir, I., Wan, Y. P., Walsh, C., Wilde, P., Bryan, A. J., and Angelini, G. D., 2001. Hemodynamic Performance of the 21-mm Sorin Bicarbon Mechanical Aortic Prostheses Using Dobutamine Doppler Echocardiography. The Society of Thoracic Surgeons, 49-53. 4. Lim, W. L., Chew, Y. T., Cew, H. T., and Low, H. T., 1998. Steady flow dynamics of prosthetic aortic heart valves: a comparative evaluation with PIV techniques. Journal of Biomechanics, 412, 411-421. 5. Browne, P., Ramuzat, A., Saxena, R., and Yoganathan, A. P., 2000. Experimental Investigation of the Steady Flow Downstream of the St. Jude Bileaflet Heart Valve: A Comparison Between Laser Doppler Velocimetry and Particle Image Velocimetry Techniques. Biomedical Engineering Society, 43-44, 39-47. 6. Lai, Y. G., Chandran, K. B., and Lemmon, J., 2002. A numerical simulation of mechanical heart valve closure fluid dynamics. Journal of Biomechanics, 881-892. 7. Bluestein, D., Rambod, E., and Gharib, M., 2000. Vortex Shedding as a Mechanism for Free Emboli Formation in Mechanical Heart Valves. Journal of ASME Biomechanical Engineering, 129, 125-134. 8. Stijnen, J.M.A., de Hart, J., Bovendeerd, P. H. M., and van de Vosse, F. N., 2004. Evaluation of a fictitous domain method for predicting dynamic response of mechanical heart valves. Journal of Fluid and Structures, 853-850. 9. Leuprecht, A., Kozerke, S., Boesiger, P., and Perktold, K., 2003. Blood flow in the human ascending aorta: a combined MRI and CFD study. Journal of Engineering Mathematics, 387 404. 10. Laske, A., Jenni, R., Maloigne, M., Vassallii, G., Bertel, O., and Turina, M. I., 1996. Pressure Gradients Across Bileaflet Aortic Valves by Direct Measurements and Echocardiography. The Society of Thoracic Surgeons, 48, 52, 48-57. 11. Grigioni, M., Daniele, C., Gaudio,C. D., Morbiducci, U., Balducci, A., Avenio, G. D., and Barbaro, V., 2003. Experimental and computational studies of flow through a bileaflet mechanical heart valve in a realistic aorta. Istituto Superiore di Sanit. 12. King, M. J., Corden, J., David, T., and Fisher, J., 1996. A three dimensional, time dependent analysis of flow through a bileaflet mechanical heart valve: comparison of experimental and numerical results. Journal of Biomechanics, 616, 609-618. 1.

11