- Email: [email protected]
The hemodynamic effects of the LVAD outflow cannula location on the thrombi distribution in the aorta: A primary numerical study
computer methods and programs in biomedicine 133 (2016) 217–227
j o u r n a l h o m e p a g e : w w w. i n t l . e l s e v i e r h e a l t h . c o m / j o u r n a l s / c m p b
The hemodynamic effects of the LVAD outflow cannula location on the thrombi distribution in the aorta: A primary numerical study Yage Zhang, Bin Gao, Chang Yu * School of Life Science and BioEngineering, Beijing University of Technology, Beijing 100124, China
A R T I C L E
I N F O
A B S T R A C T
Article history:
Although a growing number of patients undergo LVAD implantation for heart failure treat-
Received 19 October 2015
ment, thrombi are still the devastating complication for patients who used LVAD. LVAD outflow
Received in revised form
cannula location and thrombi generation sources were hypothesized to affect the thrombi
6 May 2016
distribution in the aorta. To test this hypothesis, numerical studies were conducted by using
Accepted 31 May 2016
computational fluid dynamic (CFD) theory. Two anastomotic configurations, in which the LVAD outflow cannula is anastomosed to the anterior and lateral ascending aortic wall (named
Keywords:
as anterior configurations and lateral configurations, respectively), are designed. The par-
Heart failure
ticles, whose sized are same as those of thrombi, are released at the LVAD output cannula
LVAD
and the aortic valve (named as thrombiP and thrombiL, respectively) to calculate the dis-
CFD
tribution of thrombi. The simulation results demonstrate that the thrombi distribution in
Stroke
the aorta is significantly affected by the LVAD outflow cannula location. In anterior con-
Thrombi distribution
figuration, the thrombi probability of entering into the three branches is 23.60%, while that in lateral configuration is 36.68%. Similarly, in anterior configuration, the thrombi probabilities of entering into brachiocephalic artery, left common carotid artery and left subclavian artery, is 8.51%, 9.64%, 5.45%, respectively, while that in lateral configuration it is 11.39%, 3.09%, 22.20% respectively. Moreover, the origins of thrombi could affect their distributions in the aorta. In anterior configuration, the thrombiP has a lower probability to enter into the three branches than thrombiL (12% vs. 25%). In contrast, in lateral configuration, the thrombiP has a higher probability to enter into the three branches than thrombiL (47% vs. 35%). In brief, the LVAD outflow cannula location significantly affects the distribution of thrombi in the aorta. Thus, in the clinical practice, the selection of outflow location of LVAD and the risk of thrombi formed in the left ventricle should be paid more attention than before. © 2016 Elsevier Ireland Ltd. All rights reserved.
1.
Introduction
With prevalence increasing, heart failure has become a severe threat to human health [1]. Heart transplant is the effective way for the patient treatment with end-stage failure [2].
However, the number of donor hearts is far less than the number of awaiting heart transplant patients. Hence, the left ventricular assist device (LVAD) has become an important method for heart failure patients to improve quality of life and to prolong survival rate [3]. Along with the application of LVAD, the thrombosis has become the most devastating complication
* Corresponding author. School of Life Science and BioEngineering, Beijing University of Technology, Beijing 100124, China. Tel.: +86 010 67391685; fax: +86 010 67391685. E-mail address: [email protected] (C. Yu). http://dx.doi.org/10.1016/j.cmpb.2016.05.017 0169-2607/© 2016 Elsevier Ireland Ltd. All rights reserved.
218
computer methods and programs in biomedicine 133 (2016) 217–227
for patients supported by this device [4,5]. Holman et al. [6] reported that the thrombosis issue is an independent predictor of the survival rate of patients supported by LVAD. John et al. [7] demonstrated that the incidence of adverse events caused by LVAD could be reduced by solving the thrombosis issue. Consequently, Hasin et al. [8] also pointed that the thrombosis is one of the leading causes of readmission for patients with LVAD support. In order to solve this problem, researchers tried to prevent thrombosis by improving the hemodynamic structure of LVAD [9] and by refining the strategies of anticoagulation [10]. However, the following studies demonstrate that the incidence of stroke caused by LVAD is still high. For instance, Starling et al. [4] reported that the pump thrombosis related to the use of the HeartMate II has been abruptly increasing, which is associated with substantial morbidity and mortality. Similarly, Inci et al. [11] found that HeartWare ventricular assist system reported the same complications. Currently, the incidences of adverse events, related to thrombosis, range from 14% to 47% over a period of 6–12 months [12–14]. In order to further reduce the incidence of adverse events caused by thrombi, researchers attempt to regulate the hemodynamic effects in the aorta to refine the distribution of thrombi. For instance, Ricardo et al. study the difference in percentage of particles entering the cerebral vessels when out-graft of LVAD is anastomosed to the ascending aorta and descending aorta [15]. They found that the percentage of particles entering the cerebral vessels ranged from 6% for the descending aorta VADOG anastomosis, to 14% for the ascending aorta. Similarly, Osorio et al. found that the angle of out-graft of LVAD is also a factor that affects the distribution of thrombi in the aorta [16]. Although these studies show that the hemodynamic states could significantly affect the distribution of thrombi, none of these studies concern the effects of outflow cannula location at the ascending aorta on the distribution of thrombi in the aorta. Moreover, the relationship between the thrombi generation sources and the risk of thrombi related to stroke is also unclear. Along with the development of LVAD, two kinds of support modes, named as fully support and partial support, are widely studied. Under fully support mode, the aortic valve will keep closing throughout the whole cardiac period. And there is no blood jetting through the aortic valve. Hence, the distribution of thrombi, generated in the left ventricular chamber and in the aortic root, was not studied by previous studies. However, the partial support mode, currently, has attracted more and more interesting, because of its advantages on improving cardiac function recovery [17], on prevention from aortic valve commissural fusion [18] and on preventing from aortic insufficiency [19]. However, the blood, flowing through the aortic valve, makes the thrombosis movement become more complex than that under full support mode. The thrombi, formed in the left ventricular chamber and in the aortic root, may have a chance to flow into the three bifurcation vessels. Thus, the study on the relation between the distribution of thrombi and their generation source also has important meaningful to the application of partial support mode. According to literatures, the LVAD outflow cannula location is an important factor affecting the aortic hemodynamic states. For instance, Kar [20] et al. first studied the effect of the LVAD graft position on the hemodynamic states in the
cardiovascular system. The study shows that the LVAD graft position could significantly change the aortic hemodynamic status. Similarly, May-Newman [21] et al. studied the effect of LVAD aortic outflow conduit location on the 3-D flow in the native aorta over a range of boundary conditions. After that, Karmonik [22] et al. proposed that a large region of disturbed flow was observed surrounding the LVAD outflow, under lateral configuration. And then, Yang et al. [23,24] demonstrated that the LVAD support causes an acute increase in flow splitting and turbulence in the major branch vessels. Currently, Caruso [25] et al. reported that the distance between the LVAD outflow cannular graft and ST junction would significantly change the hemodynamics in the aortic arch. The above-mentioned studies demonstrate that the outflow cannula location would significantly change the aortic hemodynamics. Thus, we hypothesize that the outflow cannula location of LVAD could affect the distribution of thrombi in the aorta and then affect the risk of stroke. In order to test this hypothesis, numerical studies are conducted based on patient-specific aortic model with two different outflow cannula locations on the ascending aorta. The computational fluid dynamic (CFD) theory and Lagrangian particle tracking theory were adopted for calculating the aortic hemodynamic states and the thrombi distribution. The blood flow velocity vector, wall shear stress (WSS), the probability of thrombi distribution (Pi) and the risk index (RIi) of thrombi with different sources are used to evaluate the relationship between the thrombi distribution in the aorta, the outflow cannula location and thrombi generation sources.
2.
Material and methods
2.1.
Aortic model generation
In order to clarify the hemodynamic state and distribution of thrombi, a patient-specific aortic model of a heart failure patient without LVAD support was reconstructed based on a series of computed tomography angiography (CTA) images. Written informed consent was obtained from the patient for using medical images in the study. The 3D patient-specific anatomical model was constructed by using these image segmentations and by using commercial 3D reconstruction software MIMICS (Materialise, Belgium) to provide the stereolithographic files (STL format). And then, the model was send into software Geomagic (Geomagic, USA) to obtain a better surface quality, smoothing the aorta (Table 1). The LVAD, used in this study, was designed by Beijing University of technology, whose outflow cannula was 12-mm
Table 1 – The geometry size of the aorta. Location a b c d e f
Diameter(mm) 29.91 12.05 6.04 9.22 22.48 12
computer methods and programs in biomedicine 133 (2016) 217–227
219
Fig. 1 – The configurations of the LVAD cannula. a, ascending aorta; b, brachiocephalic artery; c, left common carotid artery; d, left subclavian artery; e, descending aorta; f, LVAD cannula.
polyester outflow graft. Hence, in this study, the outflow cannular diameters are typically set to be 12 mm. For CFD analysis, the outflow cannula was modeled as a rigid cannula and was virtually added to the anterior and lateral ascending aortic wall (named as the anterior configuration and the lateral configuration), respectively (Fig. 1), by using commercial software FreeForm (Geomagic, USA).
2.2.
Meshing generation
Both geometric models were meshed by using the grid generator ICEM (ANSYS, Canonsburg, PA, USA), and then transported into CFD software FLUENT (ANSYS). In order to determine the appropriate numbers of grids for this work, grid independency tests, targeting the flow rate in the three branches, are conducted. The results demonstrate that 2.1 million tetrahedral elements are sufficient for this study (Fig. 2, Table 2).
of the blood. The Navier–Stokes equations were solved in ANSYS/CFX 14.0 utilizing a finite volume-based pressurecorrection algorithm and a second-order upwinding scheme for the convective derivatives. In order to analyze the distribution of the thrombi, the Lagrangian particle tracking theory, which is through the Laplace coordinates of the particles by integral differential equation of force to solve the trajectory of discrete phase particles [27], was utilized in this work, as Eq. (3):
mp
dup = ∑ Fdrag + ∑ Fvirtual + ∑ Fpressure dt
where mp is the mass of particles, Fvirtual represents the additional virtual force, Fpressure represents the pressure force. µp is the particle velocity, Fdrag represents the drag force per unit mass. Then, the drag force, caused by the relative velocity of the particles with respect to the blood flow, is given by Eq. (4) [27].
Fdrag = 0.5 ∗ Cd ρp Ap u − up (u − up )
2.3.
Numerical approaches
The flow simulation was based on the mass and momentum conservation for incompressible fluid, known as the Continuity and Navier–Stokes Eqs. [26] (1) and (2).
∇⋅u = 0
ρ
∂u + ρ (u ⋅ ∇ ) u = −∇p + μ∇2u ∂t
(1)
(2)
where u is velocity vector, t represents the time, and p denotes the pressure, ρ and µ are the density and the dynamic viscosity
(3)
(4)
where C d represents the drag coefficient, A p is the crosssection of particle, ρp denotes the particle’s density, u is the velocity of blood, and the velocity of particle is denoted by up. The Fvirtual and Fpressure were given by Eqs. (5) and (6).
Fvirtual =
ρ d (u − up ) 2ρp dt
Fpressure =
ρ ∂u μ pi ρp ∂x i
where ρ is the density of blood.
(5)
(6)
220
computer methods and programs in biomedicine 133 (2016) 217–227
Fig. 2 – The optimal mesh, nearly 2.1 million elements. (A) Mesh of anterior configuration; (B) Mesh of lateral configuration.
Table 2 – The results of the grid independence test. Number of cells 1658950 2083905 2292967
2.4.
Flow rate in brachiocephalic artery
Relative error
Flow rate in the left common carotid artery
Relative error
Left subclavian artery
Relative error
12.286 ml/s 13.333 ml/s 13.583 ml/s
— 8.52% 1.87%
3.083 ml/s 3.047 ml/s 3.107 ml/s
— 1.16% 1.96%
6.738 ml/s 5.738 ml/s 5.464 ml/s
— 14.84% 4.77%
The determination of thrombi numbers
In order to determine the optimal thrombi number, released from aortic valve and LVAD outlet cannula, a partial number independency test was conducted. In this test, the thrombi were released alone at aortic valve and LVAD outlet cannula, respectively. The percentage of thrombi entering descending aorta was chosen as the target to determine the appropriate thrombi number, released at aortic valve and LVAD outlet cannula. The test demonstrated that the 1825 thrombi, released at aortic valve, and 167 thrombi, released at LVAD outlet cannula, were sufficient for this study (Tables 3 and 4).
2.5.
Boundary conditions
According to clinical practice, the hemodynamic states of partial support, in which the rotational speed of LVAD was regulated to maintain the aortic valve periodically opens and closes during the whole cardiac cycle, were studied (Fig. 3). The physiological pressure and velocity data were used as the boundary conditions for CFD calculation. The boundary conditions were derived from a lumped parameter model [28] (Fig. 3A), of which
Table 3 – The results of the partial number independence test for the thrombi released at aortic valve. The thrombi number, released at aortic valve 167 334
Percentage of thrombi, entering into descending aorta 86.16% 85.51%
the accuracy was validated by clinical data. In this study, the rotational speed of LVAD was set to 8000 rpm, in which the total flow rate, cardiac output and LVAD output were about 5.3 L/ min, 0.8 L/min and 4.5 L/min, respectively (Fig. 3C). And the ratio of cardiac output to the LVAD output was consistent with the clinical practice. This lumped parameter model consists of the left atrial, left ventricle, LVAD, systemic circulatory system, right atrial, right ventricle and pulmonary circulatory system. In this study, the blood velocity at the aortic valve and outflow cannula of LVAD were used as the inlet boundary conditions. Moreover, the blood velocity and pressure waveform, used as the
computer methods and programs in biomedicine 133 (2016) 217–227
Table 4 – The results of the partial number independence test for the thrombi released at outlet cannula. The thrombi number, released at LVAD outlet cannula 1825 3650
Percentage of thrombi, entering into descending aorta 76.40% 75.17%
2.6.
Calculation settings
In order to evaluate the thrombi distribution, the particles with a diameter of 2 mm [29], which was similar with the size of thrombi in the LVAD and left ventricle, were used to evaluate the thrombi distribution. The partial density [30] was set to be 1117 kg/m3, which was similar to the density of thrombi in the aorta. In addition, the collision between the aortic wall and the thrombi was assumed as the elastic collision, in which the energy lost was ignored. The particles were released both at the aortic valve and LVAD outflow cannula (named as thrombiL and thrombiP, respectively) at the beginning of the simulation. The numbers of thrombiL and thrombiP were 1825 and 167, respectively. The blood is assumed to be the homogeneous, incompressible and Newtonian fluid, of which the density and viscosity is set to 1050 kg/m3 [31] and 0.0035 Pa•s [32], respectively. The cardiac cycle was set up 0.8 s and the step length was 0.01 s. According to the geometric size of aorta, the peak systolic blood flow velocity and the blood density, the peak systolic Reynolds number Re is larger than 5000. Hence, the k-e turbulence model, proven to be appropriate for illustrating the turbulence flow states in the aorta [33], was used in this study. The backward Euler implicit time integration scheme was implemented with a fixed time increment (time step of 0.01 s). Fourteen cardiac cycles have been calculated to obtain stable results independent from initialization, and the hemodynamic parameters in the 14th cardiac cycles have been extracted for hemodynamic analysis and for evaluating the thrombi distribution. The convergence precision in this study is set to be 10−3.
2.7.
Hemodynamic analysis
In order to evaluate the distributions of thrombi, the thrombi probability of passing through varied outlets of the aorta is calculated, as Eq. (7):
Pi =
Ni Ntotal
artery, the left common carotid artery, the left subclavian artery and the descending aorta. Ni denotes the numbers of thrombi, passing through different outlets of the aorta. Ntotal is the total numbers of thrombi, released in the aorta. To assess the risk of stroke caused by thrombi with different origins, the risk index (RI) was defined as Eq. (8).
RIi =
boundary conditions, were imposed at the outlet of three branch vessels and the descending aorta (Fig. 3B and C), in which the heart beat was set to 75 bpm and the duration of systolic period was set to 0.3 s.
(7)
where Pi represents the probability of thrombi, passing through different outlets of the aorta. The subscript i is used to mark the different outlets of the aorta, including the brachiocephalic
221
PiP PiL
(8)
where RIi represents the risk index, in which the subscript i indicates the different outlets of the aorta, including the brachiocephalic artery, the left common carotid artery, the left subclavian artery and the descending aorta. PiP represents the probability of thrombiP. PiL denotes the probability of thrombiL. For each outlets, RIi = 1 means the risk, caused by the thrombiL, is equal to that caused by thrombiP. Similarly, RIi > 1 means that the thrombiP is more likely to enter the vessels than thrombiL, and vice versa.
3.
Results
In order to evaluate the aortic hemodynamic states and the thrombi distribution, the probability entering each vessel, the thrombi risk index (RI), WSS and blood flow pattern were illustrated from Figs. 4–7. And three special time points (0.21 s, 0.38 s and 0.44 s) were chosen in this study to show hemodynamic states in the aorta. 0.21 s is the time point when the velocity of blood, jetting from left ventricle, reaches its maximum value. 0.38 s is the middle point of the blood velocity of LVAD. Similarly, 0.44 s is the time point when the velocity of blood, jetting from left ventricle, reaches its minimum values. Fig. 4 shows the probability of thrombi passing through each vessel. In anterior configuration, the probability of thrombi entering into the three branch vessels is 23.60%, while that in lateral configuration it has been increased to 36.68%. Among them, for anterior configuration, the probability of thrombi entering into the brachiocephalic artery, the left common carotid artery, and the left subclavian artery is 8.51%, 9.64% and 5.45%, respectively, while that in the lateral configuration, it is 11.39%, 3.09% and 22.20%, respectively. Fig. 5 compares the changes in the risks of thrombi with two origins, entering into three branches in both anterior configuration and lateral configuration. In anterior configuration, the probability of thrombi generated in LVAD entering into three branch vessels is lower than that generated in the left ventricle (probability: 12% vs. 25%). In contrast, for the lateral configuration, the thrombi from LVAD have higher probability to enter into the three branch vessels compared with that from left ventricle (probability: 47% vs. 35%). Fig. 6 shows the risk index of thrombi under both types. In anterior configurations, the RIdescendingaorta , RIbrachiocephalicartery , RIleftcommoncarotidartery and RIleftsubclavianartery are 1.18, 0.1, 0.09 and 2.01, respectively, while those for lateral configuration they are 0.83, 0.08, 0.09 and 2.46, respectively. Fig. 7 shows the flow pattern at aorta arch and three bifurcations in both types. In anterior configuration, a recirculation
222 computer methods and programs in biomedicine 133 (2016) 217–227
Fig. 3 – Inlet and outlet boundary conditions. A is a lumped parameter model (Ref. 25). B is the pressure waveform at the three epiaortic vessels. C reports the inlet and outlet velocity waveforms, in which the dotted line represents the inlet velocity waveform at the ascending aorta; the dashed line represents the velocity waveform at the descending aorta; the solid line represents the velocity waveform at the LVAD cannula.
computer methods and programs in biomedicine 133 (2016) 217–227
223
Fig. 4 – The histogram of thrombi probability entering into all vessels. The thrombi probability, entering into descending aorta, brachiocephalic artery, left common carotid artery and left subclavian artery under anterior configuration and lateral configuration, were illustrated in this figure.
zone is observed at 0.21 s (region a), while that is not observed in lateral configuration. In contrast, for both types, the recirculation zones are observed at 0.38 s (region b). Moreover, in lateral configuration, recirculation zones are formed at the region c.
4.
Discussion
Along with the increase of application of LVAD, the adverse events, resulting from arterial thrombi, have become a severe complication of LVAD [4,34,35]. Besides refining the hemodynamic structure and the strategies of anticoagulation, the optimization of hemodynamic states in the aorta has become an alternative approach to reduce the incidence of thrombi relating complications. Although several studies on the relationship between the distributions of thrombi and the
geometric characteristics of outflow graft [15,16], none of these studies focus on the relationship between distribution of thrombi, outflow cannula location and the generation sources of thrombi. This study is the first time to clarify the relationship between LVAD outflow cannula location, thrombi generation sources and the distribution of thrombi in the aorta. The main objective of this study is to determine whether changing the LVAD outflow cannula location significantly affected the trajectory of thrombi emanating from both the LVAD conduit and from aortic valve. In our work, we utilized CFD to investigate the trajectories of representative 2 mm diameter randomly released particles at the LVAD cannula inlet plane and aortic valve inlet plane. We then computed the percentage of released particles reaching the critical arteries as well as the pooled statistics. The results obtained in this study indicated that there is indeed a significant relation between the LVAD outflow cannula location and the number of thrombi flowing through the three bifurcation vessels and descending
Fig. 5 – Thrombi probability for all outlets, considering the two configurations and the two inlets.
224
computer methods and programs in biomedicine 133 (2016) 217–227
Fig. 6 – The numerical value of RI.
aorta. Although the incidence of thrombi to three bifurcation vessels was never completely eliminated by the changes in the LVAD outflow cannula location, the primary study confirmed the hypothesis that there is an optimal LVAD outflow cannula location, which could help to reduce the incidence of thrombi flowing into the three bifurcation vessels. These results are consisted with the findings of Kato’s study [36], in which the hemispheric distribution of stroke in LVAD patients was confirmed to affect by anatomic configuration of LVAD outflow cannula. Similarly, Osorio et al. [16] reported that it is possible to minimize the number of thrombi flowing into the carotid and vertebral arteries by regulating the aortic hemodynamic states. The above-mentioned studies and the obtained results in this study all demonstrated that the distribution of thrombi could be optimized by regulating the anastomosis characteristics. According to the study results, the LVAD outflow cannula location could significantly change the distribution of thrombi in the aorta (Fig. 4). And the changes in the aortic flow pattern, resulting from varied graft locations, might be the key reasons (Fig. 7). From Fig. 7, a recirculation zone (region a) is obviously observed for anterior configuration, where thrombi have a tendency to get captured by the large recirculation zones, reducing their potential to flow into three bifurcation vessels. In contrast, the recirculation zone does not continually exist for lateral configuration. Similarly, the vortex (region b) for the lateral configuration is closer to the three branch vessels than that for anterior configuration, which may prevent thrombi from flowing into the descending aorta and increase the probability of thrombi flowing into the three bifurcation vessels. Moreover, there is a vortex at the region e for lateral configuration, which is not observed for anterior configuration. This vortex may lead the thrombi more easily to enter into the three bifurcation vessels. Moreover, the obtained results demonstrated that the source of thrombi have significant effects on their probability, flowing into three bifurcation vessels (Fig. 5). In lateral configuration, thrombiL may more easily enter into the three branch vessels than thrombiP. Moreover, thrombiP, in lateral configuration, easily flow into the left subclavian artery, while thrombiL easily enter more into the left subclavian artery and brachiocephalic artery. In contrast, the probability of thrombi P entering into the
brachiocephalic artery and left common carotid artery is approximately zero. These results could provide very useful information for surgeons and LVAD operator. In anterior configuration, the surgeons and operators should pay more attention to the left ventricular flow pattern and avoid the thrombosis. And for lateral configuration, the thrombi generated by LVAD should be paid more attention.
4.1.
Limitation
In this study, the flow condition was chosen according to the clinical practice and the anastomotic location was also the common configurations used in clinical practice. Hence, the obtained results could provide useful information to surgeons. However, the results of numerical study might be affected by the flow condition (including the rotational speed of LVAD and support level) and anastomotic location, hence, if the flow condition and anastomotic location are quite different from those reported in this study, the conclusions may be altered. In this study, the deformation of the aortic wall has been neglected to reduce the computational cost. Hence, in the future, the elasticity of aortic wall will be considered by utilizing fluid– structure interaction (FSI). In addition, thrombi with same diameter are considered in this work. According to clinical practice, the thrombi in the aorta mainly ranges from 2 to 5 mm [29]. In the future, thrombi with multi-diameters will be used to more accurately clarify this mechanism. Moreover, only two anastomotic locations were considered. According to simulation results, the anastomotic location could significantly change the distribution of thrombi, but maybe both are not the optimal design. Hence, more anastomotic locations will be studied to determine the optimal position.
5.
Conclusion
Numerical studies were conducted to evaluate the effects of varied LVAD outflow cannula location and thrombi generation sources on the possible thromboembolic events leading to stroke. Furthermore, this study confirmed the hypothesis
computer methods and programs in biomedicine 133 (2016) 217–227
Fig. 7 – The velocity vector at different time of the two configurations. A is the velocity vector of anterior configuration; B is the velocity vector of lateral configuration. Symbol a aims to the position of recirculation zone occurred at the ascending aorta. Symbol b reflects the position of vortex at aortic arch. Symbol c denotes to the position of recirculation zones occurred at the aortic arch. Symbol d and e reflects the vortex at the descending aorta.
225
226
computer methods and programs in biomedicine 133 (2016) 217–227
that an optimal LVAD outflow cannula location that reduces the number of thrombi flowing into three bifurcation vessels is a viable engineering solution to the present problem posed by thrombus formation in LVAD supported patients. The simulation results demonstrate that the thrombi distribution in the aorta can be significantly affected by the LVAD outflow cannula location. In anterior configuration, the probability of thrombi, entering into the descending aorta, is up to 76.40%, while that in lateral configuration is reduced to 63.32%. Similarly, in anterior configuration, the probability of thrombi, entering into the brachiocephalic artery, the left common carotid artery, and the left subclavian artery is 8.51%, 9.64%, 5.45%, respectively, while that in the lateral configuration, it is 11.39%, 3.09%, 22.20%, respectively. In addition, the study demonstrates that the thrombi with varied origin reflect different distributions in the aorta. In the anterior configuration, the thrombiP has a lower probability to enter into the three branches than thrombiL (probability 11.59% vs. 23.60%). In contrast, in lateral configuration, the thrombiP has a higher probability to enter into the descending aorta than thrombiL (probability 46.36% vs. 35.02%). In brief, the LVAD outflow cannula location could significantly change the hemodynamic states and the distribution of thrombi in the aorta. In the anterior configuration, the thrombi have lower probability to cause stroke than that in the lateral configuration. Hence, besides improving the structure of LVAD and anticoagulant strategies, optimize anastomosis characteristics should be considered as another useful method to reduce the incidence of stroke caused by LVAD support. Moreover, the thrombiL is easier to cause stroke than thrombiP. Hence, the thrombi generated in the left ventricle should be paid more attention in the clinical practice.
Acknowledgment This work was partly sponsored by the National Natural Science Foundation of China (Grant No. 11572014, 11272022, 91430215). This work was also sponsored by the BJUT Foundation Fund (KM201510005028), BJUT Talent Found (2014000020124G045).
REFERENCES
[1] M. Loebe, E. Soltro, V. Thohan, J.A. Lafuente, G.P. Noon, New surgical therapies for heart failure, Curr. Opin. Cardiol. 18 (3) (2003) 194–198. [2] J. Garbade, H.B. Bittner, M.J. Barten, A. Rastan, S. Lehmann, F.W. Mohr, et al., Combined surgical left ventricular reconstruction and left ventricular assist device implantation for destination therapy in end-stage heart failure, Circ. Heart Fail. 4 (2011) el, 4-5. [3] S.J. Park, C.A. Milano, A.G. Tatooles, J.G. Rogers, R.M. Adamson, D.E. Steidley, et al., Outcomes in advanced heart failure patients with left ventricular assist devices for destination therapy, Circ. Heart Fail. 5 (2012) 241–248. [4] R.C. Starling, N. Moazami, S.C. Silvestry, G. Ewald, J.G. Rogers, C.A. Milano, et al., Unexpected abrupt increase in left ventricular assist device thrombosis, NEJM 370 (1) (2014) 33–40.
[5] M.R. Mehra, G.C. Stewart, P.A. Uber, The vexing problem of thrombosis in long-term mechanical circulatory support, J. Heart Lung Transplant. 33 (1) (2014) 1–11. [6] W.L. Holman, R.L. Kormos, D.C. Naftel, M.A. Miller, F.D. Pagani, E. Blume, et al., Predictors of death and transplant in patients with a mechanical circulatory support device: a multi-institutional study, J. Heart Lung Transplant. 28 (2009) 44–50. [7] R. John, F. Kamdar, K. Liao, M. Colvin-Adams, A. Boyle, L. Joyce, Improved survival and decreasing incidence of adverse events with the HeartMate II left ventricular assist device as bridge-to-transplant therapy, Ann. Thorac. Surg. 86 (4) (2008) 1227–1234. [8] T. Hasin, Y. Marmor, W. Kremers, Y. Topilsky, C.J. Severson, J.A. Schirger, et al., Readmissions after implantation of axial flow left ventricular assist device, J. Am. Coll. Cardiol. 61 (2) (2013) 153–163. [9] R. Radovancevic, N. Matijevic, A.W. Bracey, B. Radovancevic, M. Elayda, L.D. Gregoric, et al., Increased leukocyte-platelet interactions during circulatory support with left ventricular assist devices, ASAIO J. 55 (5) (2009) 459–464. [10] B. Meuris, J. Arnout, D. Vlasselaers, M. Schetz, B. Meyns, Long-term management of an implantable left ventricular assist device using low molecular weight heparin and antiplatelet therapy: a possible alternative to oral anticoagulants, Artif. Organs 31 (5) (2007) 402–405. [11] G. Inci, E. Sorguven, Effect of LVAD outlet graft anastomosis angle on the aortic valve, wall, and flow, ASAIO J. 58 (4) (2012) 373–381. [12] M. Deng, L.B. Edwards, M.I. Hertz, A.W. Rowe, B.M. Keck, R.L. Kormos, et al., Mechanical circulatory support device database of the international society for heart and lung transplantation: third annual report, J. Heart Lung Transplant. 24 (2005) 1182–1187. [13] C. Schmid, M. Weyand, D.G. Nabavi, D. Hammel, M.C. Deng, E.B. Ringelstein, et al., Cerebral and systemic embolization during left ventricular support with the Novacor N100 device, Ann. Thorac. Surg. 65 (1998) 1703–1710. [14] C.E. Thomas, D. Jichici, R. Petrucci, V.C. Urrutia, R.J. Schwartzman, Neurologic complications of the Novacor left ventricular assist device, Ann. Thorac. Surg. 72 (4) (2001) 1311–1315. [15] I.R. Argueta-Morales, R. Tran, A. Ceballos, W. Clark, R. Osorio, E.A. Divo, et al., Mathematical modeling of patient-specific ventricular assist device implantation to reduce particulate embolization rate to cerebral vessels, J. Biomech. Eng. 136 (7) (2014). [16] A.F. Osorio, R. Osorio, A. Ceballos, R. Tran, W. Clark, E.A. Divo, I.R. Argueta-Morales, et al., Computational fluid dynamics analysis of surgical adjustment of left ventricular assist device implantation to minimize stroke risk, Comput. Methods Biomech. Biomed. Engin. 16 (6) (2013) 622–638. [17] W. Oriyanhan, H. Tsuneyoshi, T. Nishina, S. Matsuoka, T. Ikeda, K. Komeda, Determination of optimal duration of mechanical unloading for failing hearts to achieve bridge to recovery in a rat heterotopic heart transplantation model, J. Heart Lung Transplant. 26 (1) (2007) 16–23. [18] J.R. Martina, M.E.I. Schipper, N. de Jonge, F. Ramjankhan, R.A. Weger, J.R. Lahpor, et al., Analysis of aortic valve commissural fusion after support with continuous-flow left ventricular assist device, Interact. Cardiovasc. Thorac. Surg. 17 (4) (2013) 616–624. [19] S.D. Gregory, M.C. Stevens, E. Wu, J. Fraser, D. Timms, In vitro evaluation of aortic insufficiency with a rotary left ventricular assist device, Artif. Organs 37 (9) (2013) 802–809. [20] B. Kar, R.M. Delgado, O.H. Frazier, I.D. Gregoric, M.T. Harting, Y. Wadia, et al., The effect of LVAD aortic outflow-graft placement on hemodynamics and flow: implantation
computer methods and programs in biomedicine 133 (2016) 217–227
[21]
[22]
[23]
[24]
[25]
[26]
[27] [28]
technique and computer flow modeling, Tex. Heart Inst. J. 32 (3) (2005) 294–298. K. May-Newman, B. Hillen, W. Dembitsky, Effect of left ventricular assist device outflow conduit anastomosis location on flow patterns in the native aorta, ASAIO J. 52 (2) (2006) 132–139. C. Karmonik, S. Partovi, M. Loebe, B. Schmacjk, A. Ghodsizad, M.R. Robbin, et al., Influence of LVAD cannula outflow tract location on hemodynamics in the ascending aorta: a patient-specific computational fluid dynamics approach, ASAIO J. 58 (6) (2012) 562–567. N. Yang, S. Deutsch, E.G. Paterson, K. Manning, Numerical study of blood flow at the end-to-side anastomosis of a left ventricular assist device for adult patients, J. Biomech. Eng. 131 (11) (2009) 111005. N. Yang, S. Deutsch, E.G. Paterson, K. Manning, Hemodynamics of an end-to-side anastomotic graft for a pulsatile pediatric ventricular assist device, J. Biomech. Eng. 132 (3) (2010) 031009. M.V. Caruso, V. Gramigna, M. Rossi, G.F. Serraino, A. Renzulli, G. Fragomeni, A computational fluid dynamics comparison between different outflow graft anastomosis locations of left ventricular assist device (LVAD) in a patient-specific aortic model, Int. J. Numer. Method. Biomed. Eng. 31 (2) (2015). http://dx.doi.org/10.1002/cnm.2700. F. Kabinejadian, L.P. Chua, D.N. Ghista, M. Sankaranarayanan, Y.S. Tan, A novel coronary artery bypass graft design of sequential anastomoses, Ann. Biomed. Eng. 38 (10) (2010) 3135–3150. Fluent Inc, Fluent V6.1 User Guide, Fluent Inc, Lebanon, NH, 2003. Z.M. Song, K.Y. Gu, B. Gao, F. Wan, Y. Chang, Y. Zeng, Hemodynamic effect of various support modes of
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
227
continuous flow LVADs on the cardiovascular system: a numerical study, Med. Sci. Monit. 20 (2014) 733–741. V.J. Marder, D.J. Chute, S. Starkman, A.M. Abolian, C. Kidwell, D. Liebeskind, et al., Analysis of thrombi retrieved from cerebral arteries of patients with acute ischemic stroke, Stroke 37 (8) (2006) 2086–2093. V.M. Nahimyak, S.W. Yoon, C.K. Holland, Acoustomechanical and thermal properties of clotted blood, J. Acoust. Soc. Am. 119 (6) (2006) 3766–3772. S.S. Khalafvand, E.Y.K. Ng, L. Zhong, T.K. Hunghung, Fluid-dynamics modelling of the human left ventricle with dynamic mesh for normal and myocardial infarction: preliminary study, Comput. Biol. Med. 42 (8) (2012) 863– 870. X. Zhao, Y.J. Liu, F. Bai, X.C. Ren, J.L. Ding, F. Wang, Numerical study on bilateral bidirectional Glenn shunt, J. Med. Biomechan. 27 (5) (2012) 488–494. Z. Cheng, F.P. Tan, C.V. Riga, C.D. Bicknell, M.S. Hamady, R.G.J. Gibbs, et al., Analysis of flow patterns in a patient-specific aortic dissection model, J. Biomech. Eng. 132 (5) (2010) 051007. D.S. Likosky, C.A.S. Marrin, L.R. Caplan, Y.R. Baribeau, J.R. Morton, R.M. Weintraub, et al., Determination of etiologic mechanisms of strokes secondary to coronary artery bypass graft surgery, Stroke 34 (12) (2003) 2830–2834. C.G. Lyndsey, Q. Nidia, J.M. Eric, L.M. David, B. Karsten, Neurocognitive outcomes after extracorporeal membrane oxygenation, Stroke 29 (2) (2015) 123–135. T.S. Kato, T. Ota, P.C. Schulze, M. Farr, U. Jorde, H. Takayama, et al., Asymmetric pattern of cerebrovascular lesions in patients after left ventricular assist device implantation, Stroke 43 (3) (2012) 872–874.
j o u r n a l h o m e p a g e : w w w. i n t l . e l s e v i e r h e a l t h . c o m / j o u r n a l s / c m p b
The hemodynamic effects of the LVAD outflow cannula location on the thrombi distribution in the aorta: A primary numerical study Yage Zhang, Bin Gao, Chang Yu * School of Life Science and BioEngineering, Beijing University of Technology, Beijing 100124, China
A R T I C L E
I N F O
A B S T R A C T
Article history:
Although a growing number of patients undergo LVAD implantation for heart failure treat-
Received 19 October 2015
ment, thrombi are still the devastating complication for patients who used LVAD. LVAD outflow
Received in revised form
cannula location and thrombi generation sources were hypothesized to affect the thrombi
6 May 2016
distribution in the aorta. To test this hypothesis, numerical studies were conducted by using
Accepted 31 May 2016
computational fluid dynamic (CFD) theory. Two anastomotic configurations, in which the LVAD outflow cannula is anastomosed to the anterior and lateral ascending aortic wall (named
Keywords:
as anterior configurations and lateral configurations, respectively), are designed. The par-
Heart failure
ticles, whose sized are same as those of thrombi, are released at the LVAD output cannula
LVAD
and the aortic valve (named as thrombiP and thrombiL, respectively) to calculate the dis-
CFD
tribution of thrombi. The simulation results demonstrate that the thrombi distribution in
Stroke
the aorta is significantly affected by the LVAD outflow cannula location. In anterior con-
Thrombi distribution
figuration, the thrombi probability of entering into the three branches is 23.60%, while that in lateral configuration is 36.68%. Similarly, in anterior configuration, the thrombi probabilities of entering into brachiocephalic artery, left common carotid artery and left subclavian artery, is 8.51%, 9.64%, 5.45%, respectively, while that in lateral configuration it is 11.39%, 3.09%, 22.20% respectively. Moreover, the origins of thrombi could affect their distributions in the aorta. In anterior configuration, the thrombiP has a lower probability to enter into the three branches than thrombiL (12% vs. 25%). In contrast, in lateral configuration, the thrombiP has a higher probability to enter into the three branches than thrombiL (47% vs. 35%). In brief, the LVAD outflow cannula location significantly affects the distribution of thrombi in the aorta. Thus, in the clinical practice, the selection of outflow location of LVAD and the risk of thrombi formed in the left ventricle should be paid more attention than before. © 2016 Elsevier Ireland Ltd. All rights reserved.
1.
Introduction
With prevalence increasing, heart failure has become a severe threat to human health [1]. Heart transplant is the effective way for the patient treatment with end-stage failure [2].
However, the number of donor hearts is far less than the number of awaiting heart transplant patients. Hence, the left ventricular assist device (LVAD) has become an important method for heart failure patients to improve quality of life and to prolong survival rate [3]. Along with the application of LVAD, the thrombosis has become the most devastating complication
* Corresponding author. School of Life Science and BioEngineering, Beijing University of Technology, Beijing 100124, China. Tel.: +86 010 67391685; fax: +86 010 67391685. E-mail address: [email protected] (C. Yu). http://dx.doi.org/10.1016/j.cmpb.2016.05.017 0169-2607/© 2016 Elsevier Ireland Ltd. All rights reserved.
218
computer methods and programs in biomedicine 133 (2016) 217–227
for patients supported by this device [4,5]. Holman et al. [6] reported that the thrombosis issue is an independent predictor of the survival rate of patients supported by LVAD. John et al. [7] demonstrated that the incidence of adverse events caused by LVAD could be reduced by solving the thrombosis issue. Consequently, Hasin et al. [8] also pointed that the thrombosis is one of the leading causes of readmission for patients with LVAD support. In order to solve this problem, researchers tried to prevent thrombosis by improving the hemodynamic structure of LVAD [9] and by refining the strategies of anticoagulation [10]. However, the following studies demonstrate that the incidence of stroke caused by LVAD is still high. For instance, Starling et al. [4] reported that the pump thrombosis related to the use of the HeartMate II has been abruptly increasing, which is associated with substantial morbidity and mortality. Similarly, Inci et al. [11] found that HeartWare ventricular assist system reported the same complications. Currently, the incidences of adverse events, related to thrombosis, range from 14% to 47% over a period of 6–12 months [12–14]. In order to further reduce the incidence of adverse events caused by thrombi, researchers attempt to regulate the hemodynamic effects in the aorta to refine the distribution of thrombi. For instance, Ricardo et al. study the difference in percentage of particles entering the cerebral vessels when out-graft of LVAD is anastomosed to the ascending aorta and descending aorta [15]. They found that the percentage of particles entering the cerebral vessels ranged from 6% for the descending aorta VADOG anastomosis, to 14% for the ascending aorta. Similarly, Osorio et al. found that the angle of out-graft of LVAD is also a factor that affects the distribution of thrombi in the aorta [16]. Although these studies show that the hemodynamic states could significantly affect the distribution of thrombi, none of these studies concern the effects of outflow cannula location at the ascending aorta on the distribution of thrombi in the aorta. Moreover, the relationship between the thrombi generation sources and the risk of thrombi related to stroke is also unclear. Along with the development of LVAD, two kinds of support modes, named as fully support and partial support, are widely studied. Under fully support mode, the aortic valve will keep closing throughout the whole cardiac period. And there is no blood jetting through the aortic valve. Hence, the distribution of thrombi, generated in the left ventricular chamber and in the aortic root, was not studied by previous studies. However, the partial support mode, currently, has attracted more and more interesting, because of its advantages on improving cardiac function recovery [17], on prevention from aortic valve commissural fusion [18] and on preventing from aortic insufficiency [19]. However, the blood, flowing through the aortic valve, makes the thrombosis movement become more complex than that under full support mode. The thrombi, formed in the left ventricular chamber and in the aortic root, may have a chance to flow into the three bifurcation vessels. Thus, the study on the relation between the distribution of thrombi and their generation source also has important meaningful to the application of partial support mode. According to literatures, the LVAD outflow cannula location is an important factor affecting the aortic hemodynamic states. For instance, Kar [20] et al. first studied the effect of the LVAD graft position on the hemodynamic states in the
cardiovascular system. The study shows that the LVAD graft position could significantly change the aortic hemodynamic status. Similarly, May-Newman [21] et al. studied the effect of LVAD aortic outflow conduit location on the 3-D flow in the native aorta over a range of boundary conditions. After that, Karmonik [22] et al. proposed that a large region of disturbed flow was observed surrounding the LVAD outflow, under lateral configuration. And then, Yang et al. [23,24] demonstrated that the LVAD support causes an acute increase in flow splitting and turbulence in the major branch vessels. Currently, Caruso [25] et al. reported that the distance between the LVAD outflow cannular graft and ST junction would significantly change the hemodynamics in the aortic arch. The above-mentioned studies demonstrate that the outflow cannula location would significantly change the aortic hemodynamics. Thus, we hypothesize that the outflow cannula location of LVAD could affect the distribution of thrombi in the aorta and then affect the risk of stroke. In order to test this hypothesis, numerical studies are conducted based on patient-specific aortic model with two different outflow cannula locations on the ascending aorta. The computational fluid dynamic (CFD) theory and Lagrangian particle tracking theory were adopted for calculating the aortic hemodynamic states and the thrombi distribution. The blood flow velocity vector, wall shear stress (WSS), the probability of thrombi distribution (Pi) and the risk index (RIi) of thrombi with different sources are used to evaluate the relationship between the thrombi distribution in the aorta, the outflow cannula location and thrombi generation sources.
2.
Material and methods
2.1.
Aortic model generation
In order to clarify the hemodynamic state and distribution of thrombi, a patient-specific aortic model of a heart failure patient without LVAD support was reconstructed based on a series of computed tomography angiography (CTA) images. Written informed consent was obtained from the patient for using medical images in the study. The 3D patient-specific anatomical model was constructed by using these image segmentations and by using commercial 3D reconstruction software MIMICS (Materialise, Belgium) to provide the stereolithographic files (STL format). And then, the model was send into software Geomagic (Geomagic, USA) to obtain a better surface quality, smoothing the aorta (Table 1). The LVAD, used in this study, was designed by Beijing University of technology, whose outflow cannula was 12-mm
Table 1 – The geometry size of the aorta. Location a b c d e f
Diameter(mm) 29.91 12.05 6.04 9.22 22.48 12
computer methods and programs in biomedicine 133 (2016) 217–227
219
Fig. 1 – The configurations of the LVAD cannula. a, ascending aorta; b, brachiocephalic artery; c, left common carotid artery; d, left subclavian artery; e, descending aorta; f, LVAD cannula.
polyester outflow graft. Hence, in this study, the outflow cannular diameters are typically set to be 12 mm. For CFD analysis, the outflow cannula was modeled as a rigid cannula and was virtually added to the anterior and lateral ascending aortic wall (named as the anterior configuration and the lateral configuration), respectively (Fig. 1), by using commercial software FreeForm (Geomagic, USA).
2.2.
Meshing generation
Both geometric models were meshed by using the grid generator ICEM (ANSYS, Canonsburg, PA, USA), and then transported into CFD software FLUENT (ANSYS). In order to determine the appropriate numbers of grids for this work, grid independency tests, targeting the flow rate in the three branches, are conducted. The results demonstrate that 2.1 million tetrahedral elements are sufficient for this study (Fig. 2, Table 2).
of the blood. The Navier–Stokes equations were solved in ANSYS/CFX 14.0 utilizing a finite volume-based pressurecorrection algorithm and a second-order upwinding scheme for the convective derivatives. In order to analyze the distribution of the thrombi, the Lagrangian particle tracking theory, which is through the Laplace coordinates of the particles by integral differential equation of force to solve the trajectory of discrete phase particles [27], was utilized in this work, as Eq. (3):
mp
dup = ∑ Fdrag + ∑ Fvirtual + ∑ Fpressure dt
where mp is the mass of particles, Fvirtual represents the additional virtual force, Fpressure represents the pressure force. µp is the particle velocity, Fdrag represents the drag force per unit mass. Then, the drag force, caused by the relative velocity of the particles with respect to the blood flow, is given by Eq. (4) [27].
Fdrag = 0.5 ∗ Cd ρp Ap u − up (u − up )
2.3.
Numerical approaches
The flow simulation was based on the mass and momentum conservation for incompressible fluid, known as the Continuity and Navier–Stokes Eqs. [26] (1) and (2).
∇⋅u = 0
ρ
∂u + ρ (u ⋅ ∇ ) u = −∇p + μ∇2u ∂t
(1)
(2)
where u is velocity vector, t represents the time, and p denotes the pressure, ρ and µ are the density and the dynamic viscosity
(3)
(4)
where C d represents the drag coefficient, A p is the crosssection of particle, ρp denotes the particle’s density, u is the velocity of blood, and the velocity of particle is denoted by up. The Fvirtual and Fpressure were given by Eqs. (5) and (6).
Fvirtual =
ρ d (u − up ) 2ρp dt
Fpressure =
ρ ∂u μ pi ρp ∂x i
where ρ is the density of blood.
(5)
(6)
220
computer methods and programs in biomedicine 133 (2016) 217–227
Fig. 2 – The optimal mesh, nearly 2.1 million elements. (A) Mesh of anterior configuration; (B) Mesh of lateral configuration.
Table 2 – The results of the grid independence test. Number of cells 1658950 2083905 2292967
2.4.
Flow rate in brachiocephalic artery
Relative error
Flow rate in the left common carotid artery
Relative error
Left subclavian artery
Relative error
12.286 ml/s 13.333 ml/s 13.583 ml/s
— 8.52% 1.87%
3.083 ml/s 3.047 ml/s 3.107 ml/s
— 1.16% 1.96%
6.738 ml/s 5.738 ml/s 5.464 ml/s
— 14.84% 4.77%
The determination of thrombi numbers
In order to determine the optimal thrombi number, released from aortic valve and LVAD outlet cannula, a partial number independency test was conducted. In this test, the thrombi were released alone at aortic valve and LVAD outlet cannula, respectively. The percentage of thrombi entering descending aorta was chosen as the target to determine the appropriate thrombi number, released at aortic valve and LVAD outlet cannula. The test demonstrated that the 1825 thrombi, released at aortic valve, and 167 thrombi, released at LVAD outlet cannula, were sufficient for this study (Tables 3 and 4).
2.5.
Boundary conditions
According to clinical practice, the hemodynamic states of partial support, in which the rotational speed of LVAD was regulated to maintain the aortic valve periodically opens and closes during the whole cardiac cycle, were studied (Fig. 3). The physiological pressure and velocity data were used as the boundary conditions for CFD calculation. The boundary conditions were derived from a lumped parameter model [28] (Fig. 3A), of which
Table 3 – The results of the partial number independence test for the thrombi released at aortic valve. The thrombi number, released at aortic valve 167 334
Percentage of thrombi, entering into descending aorta 86.16% 85.51%
the accuracy was validated by clinical data. In this study, the rotational speed of LVAD was set to 8000 rpm, in which the total flow rate, cardiac output and LVAD output were about 5.3 L/ min, 0.8 L/min and 4.5 L/min, respectively (Fig. 3C). And the ratio of cardiac output to the LVAD output was consistent with the clinical practice. This lumped parameter model consists of the left atrial, left ventricle, LVAD, systemic circulatory system, right atrial, right ventricle and pulmonary circulatory system. In this study, the blood velocity at the aortic valve and outflow cannula of LVAD were used as the inlet boundary conditions. Moreover, the blood velocity and pressure waveform, used as the
computer methods and programs in biomedicine 133 (2016) 217–227
Table 4 – The results of the partial number independence test for the thrombi released at outlet cannula. The thrombi number, released at LVAD outlet cannula 1825 3650
Percentage of thrombi, entering into descending aorta 76.40% 75.17%
2.6.
Calculation settings
In order to evaluate the thrombi distribution, the particles with a diameter of 2 mm [29], which was similar with the size of thrombi in the LVAD and left ventricle, were used to evaluate the thrombi distribution. The partial density [30] was set to be 1117 kg/m3, which was similar to the density of thrombi in the aorta. In addition, the collision between the aortic wall and the thrombi was assumed as the elastic collision, in which the energy lost was ignored. The particles were released both at the aortic valve and LVAD outflow cannula (named as thrombiL and thrombiP, respectively) at the beginning of the simulation. The numbers of thrombiL and thrombiP were 1825 and 167, respectively. The blood is assumed to be the homogeneous, incompressible and Newtonian fluid, of which the density and viscosity is set to 1050 kg/m3 [31] and 0.0035 Pa•s [32], respectively. The cardiac cycle was set up 0.8 s and the step length was 0.01 s. According to the geometric size of aorta, the peak systolic blood flow velocity and the blood density, the peak systolic Reynolds number Re is larger than 5000. Hence, the k-e turbulence model, proven to be appropriate for illustrating the turbulence flow states in the aorta [33], was used in this study. The backward Euler implicit time integration scheme was implemented with a fixed time increment (time step of 0.01 s). Fourteen cardiac cycles have been calculated to obtain stable results independent from initialization, and the hemodynamic parameters in the 14th cardiac cycles have been extracted for hemodynamic analysis and for evaluating the thrombi distribution. The convergence precision in this study is set to be 10−3.
2.7.
Hemodynamic analysis
In order to evaluate the distributions of thrombi, the thrombi probability of passing through varied outlets of the aorta is calculated, as Eq. (7):
Pi =
Ni Ntotal
artery, the left common carotid artery, the left subclavian artery and the descending aorta. Ni denotes the numbers of thrombi, passing through different outlets of the aorta. Ntotal is the total numbers of thrombi, released in the aorta. To assess the risk of stroke caused by thrombi with different origins, the risk index (RI) was defined as Eq. (8).
RIi =
boundary conditions, were imposed at the outlet of three branch vessels and the descending aorta (Fig. 3B and C), in which the heart beat was set to 75 bpm and the duration of systolic period was set to 0.3 s.
(7)
where Pi represents the probability of thrombi, passing through different outlets of the aorta. The subscript i is used to mark the different outlets of the aorta, including the brachiocephalic
221
PiP PiL
(8)
where RIi represents the risk index, in which the subscript i indicates the different outlets of the aorta, including the brachiocephalic artery, the left common carotid artery, the left subclavian artery and the descending aorta. PiP represents the probability of thrombiP. PiL denotes the probability of thrombiL. For each outlets, RIi = 1 means the risk, caused by the thrombiL, is equal to that caused by thrombiP. Similarly, RIi > 1 means that the thrombiP is more likely to enter the vessels than thrombiL, and vice versa.
3.
Results
In order to evaluate the aortic hemodynamic states and the thrombi distribution, the probability entering each vessel, the thrombi risk index (RI), WSS and blood flow pattern were illustrated from Figs. 4–7. And three special time points (0.21 s, 0.38 s and 0.44 s) were chosen in this study to show hemodynamic states in the aorta. 0.21 s is the time point when the velocity of blood, jetting from left ventricle, reaches its maximum value. 0.38 s is the middle point of the blood velocity of LVAD. Similarly, 0.44 s is the time point when the velocity of blood, jetting from left ventricle, reaches its minimum values. Fig. 4 shows the probability of thrombi passing through each vessel. In anterior configuration, the probability of thrombi entering into the three branch vessels is 23.60%, while that in lateral configuration it has been increased to 36.68%. Among them, for anterior configuration, the probability of thrombi entering into the brachiocephalic artery, the left common carotid artery, and the left subclavian artery is 8.51%, 9.64% and 5.45%, respectively, while that in the lateral configuration, it is 11.39%, 3.09% and 22.20%, respectively. Fig. 5 compares the changes in the risks of thrombi with two origins, entering into three branches in both anterior configuration and lateral configuration. In anterior configuration, the probability of thrombi generated in LVAD entering into three branch vessels is lower than that generated in the left ventricle (probability: 12% vs. 25%). In contrast, for the lateral configuration, the thrombi from LVAD have higher probability to enter into the three branch vessels compared with that from left ventricle (probability: 47% vs. 35%). Fig. 6 shows the risk index of thrombi under both types. In anterior configurations, the RIdescendingaorta , RIbrachiocephalicartery , RIleftcommoncarotidartery and RIleftsubclavianartery are 1.18, 0.1, 0.09 and 2.01, respectively, while those for lateral configuration they are 0.83, 0.08, 0.09 and 2.46, respectively. Fig. 7 shows the flow pattern at aorta arch and three bifurcations in both types. In anterior configuration, a recirculation
222 computer methods and programs in biomedicine 133 (2016) 217–227
Fig. 3 – Inlet and outlet boundary conditions. A is a lumped parameter model (Ref. 25). B is the pressure waveform at the three epiaortic vessels. C reports the inlet and outlet velocity waveforms, in which the dotted line represents the inlet velocity waveform at the ascending aorta; the dashed line represents the velocity waveform at the descending aorta; the solid line represents the velocity waveform at the LVAD cannula.
computer methods and programs in biomedicine 133 (2016) 217–227
223
Fig. 4 – The histogram of thrombi probability entering into all vessels. The thrombi probability, entering into descending aorta, brachiocephalic artery, left common carotid artery and left subclavian artery under anterior configuration and lateral configuration, were illustrated in this figure.
zone is observed at 0.21 s (region a), while that is not observed in lateral configuration. In contrast, for both types, the recirculation zones are observed at 0.38 s (region b). Moreover, in lateral configuration, recirculation zones are formed at the region c.
4.
Discussion
Along with the increase of application of LVAD, the adverse events, resulting from arterial thrombi, have become a severe complication of LVAD [4,34,35]. Besides refining the hemodynamic structure and the strategies of anticoagulation, the optimization of hemodynamic states in the aorta has become an alternative approach to reduce the incidence of thrombi relating complications. Although several studies on the relationship between the distributions of thrombi and the
geometric characteristics of outflow graft [15,16], none of these studies focus on the relationship between distribution of thrombi, outflow cannula location and the generation sources of thrombi. This study is the first time to clarify the relationship between LVAD outflow cannula location, thrombi generation sources and the distribution of thrombi in the aorta. The main objective of this study is to determine whether changing the LVAD outflow cannula location significantly affected the trajectory of thrombi emanating from both the LVAD conduit and from aortic valve. In our work, we utilized CFD to investigate the trajectories of representative 2 mm diameter randomly released particles at the LVAD cannula inlet plane and aortic valve inlet plane. We then computed the percentage of released particles reaching the critical arteries as well as the pooled statistics. The results obtained in this study indicated that there is indeed a significant relation between the LVAD outflow cannula location and the number of thrombi flowing through the three bifurcation vessels and descending
Fig. 5 – Thrombi probability for all outlets, considering the two configurations and the two inlets.
224
computer methods and programs in biomedicine 133 (2016) 217–227
Fig. 6 – The numerical value of RI.
aorta. Although the incidence of thrombi to three bifurcation vessels was never completely eliminated by the changes in the LVAD outflow cannula location, the primary study confirmed the hypothesis that there is an optimal LVAD outflow cannula location, which could help to reduce the incidence of thrombi flowing into the three bifurcation vessels. These results are consisted with the findings of Kato’s study [36], in which the hemispheric distribution of stroke in LVAD patients was confirmed to affect by anatomic configuration of LVAD outflow cannula. Similarly, Osorio et al. [16] reported that it is possible to minimize the number of thrombi flowing into the carotid and vertebral arteries by regulating the aortic hemodynamic states. The above-mentioned studies and the obtained results in this study all demonstrated that the distribution of thrombi could be optimized by regulating the anastomosis characteristics. According to the study results, the LVAD outflow cannula location could significantly change the distribution of thrombi in the aorta (Fig. 4). And the changes in the aortic flow pattern, resulting from varied graft locations, might be the key reasons (Fig. 7). From Fig. 7, a recirculation zone (region a) is obviously observed for anterior configuration, where thrombi have a tendency to get captured by the large recirculation zones, reducing their potential to flow into three bifurcation vessels. In contrast, the recirculation zone does not continually exist for lateral configuration. Similarly, the vortex (region b) for the lateral configuration is closer to the three branch vessels than that for anterior configuration, which may prevent thrombi from flowing into the descending aorta and increase the probability of thrombi flowing into the three bifurcation vessels. Moreover, there is a vortex at the region e for lateral configuration, which is not observed for anterior configuration. This vortex may lead the thrombi more easily to enter into the three bifurcation vessels. Moreover, the obtained results demonstrated that the source of thrombi have significant effects on their probability, flowing into three bifurcation vessels (Fig. 5). In lateral configuration, thrombiL may more easily enter into the three branch vessels than thrombiP. Moreover, thrombiP, in lateral configuration, easily flow into the left subclavian artery, while thrombiL easily enter more into the left subclavian artery and brachiocephalic artery. In contrast, the probability of thrombi P entering into the
brachiocephalic artery and left common carotid artery is approximately zero. These results could provide very useful information for surgeons and LVAD operator. In anterior configuration, the surgeons and operators should pay more attention to the left ventricular flow pattern and avoid the thrombosis. And for lateral configuration, the thrombi generated by LVAD should be paid more attention.
4.1.
Limitation
In this study, the flow condition was chosen according to the clinical practice and the anastomotic location was also the common configurations used in clinical practice. Hence, the obtained results could provide useful information to surgeons. However, the results of numerical study might be affected by the flow condition (including the rotational speed of LVAD and support level) and anastomotic location, hence, if the flow condition and anastomotic location are quite different from those reported in this study, the conclusions may be altered. In this study, the deformation of the aortic wall has been neglected to reduce the computational cost. Hence, in the future, the elasticity of aortic wall will be considered by utilizing fluid– structure interaction (FSI). In addition, thrombi with same diameter are considered in this work. According to clinical practice, the thrombi in the aorta mainly ranges from 2 to 5 mm [29]. In the future, thrombi with multi-diameters will be used to more accurately clarify this mechanism. Moreover, only two anastomotic locations were considered. According to simulation results, the anastomotic location could significantly change the distribution of thrombi, but maybe both are not the optimal design. Hence, more anastomotic locations will be studied to determine the optimal position.
5.
Conclusion
Numerical studies were conducted to evaluate the effects of varied LVAD outflow cannula location and thrombi generation sources on the possible thromboembolic events leading to stroke. Furthermore, this study confirmed the hypothesis
computer methods and programs in biomedicine 133 (2016) 217–227
Fig. 7 – The velocity vector at different time of the two configurations. A is the velocity vector of anterior configuration; B is the velocity vector of lateral configuration. Symbol a aims to the position of recirculation zone occurred at the ascending aorta. Symbol b reflects the position of vortex at aortic arch. Symbol c denotes to the position of recirculation zones occurred at the aortic arch. Symbol d and e reflects the vortex at the descending aorta.
225
226
computer methods and programs in biomedicine 133 (2016) 217–227
that an optimal LVAD outflow cannula location that reduces the number of thrombi flowing into three bifurcation vessels is a viable engineering solution to the present problem posed by thrombus formation in LVAD supported patients. The simulation results demonstrate that the thrombi distribution in the aorta can be significantly affected by the LVAD outflow cannula location. In anterior configuration, the probability of thrombi, entering into the descending aorta, is up to 76.40%, while that in lateral configuration is reduced to 63.32%. Similarly, in anterior configuration, the probability of thrombi, entering into the brachiocephalic artery, the left common carotid artery, and the left subclavian artery is 8.51%, 9.64%, 5.45%, respectively, while that in the lateral configuration, it is 11.39%, 3.09%, 22.20%, respectively. In addition, the study demonstrates that the thrombi with varied origin reflect different distributions in the aorta. In the anterior configuration, the thrombiP has a lower probability to enter into the three branches than thrombiL (probability 11.59% vs. 23.60%). In contrast, in lateral configuration, the thrombiP has a higher probability to enter into the descending aorta than thrombiL (probability 46.36% vs. 35.02%). In brief, the LVAD outflow cannula location could significantly change the hemodynamic states and the distribution of thrombi in the aorta. In the anterior configuration, the thrombi have lower probability to cause stroke than that in the lateral configuration. Hence, besides improving the structure of LVAD and anticoagulant strategies, optimize anastomosis characteristics should be considered as another useful method to reduce the incidence of stroke caused by LVAD support. Moreover, the thrombiL is easier to cause stroke than thrombiP. Hence, the thrombi generated in the left ventricle should be paid more attention in the clinical practice.
Acknowledgment This work was partly sponsored by the National Natural Science Foundation of China (Grant No. 11572014, 11272022, 91430215). This work was also sponsored by the BJUT Foundation Fund (KM201510005028), BJUT Talent Found (2014000020124G045).
REFERENCES
[1] M. Loebe, E. Soltro, V. Thohan, J.A. Lafuente, G.P. Noon, New surgical therapies for heart failure, Curr. Opin. Cardiol. 18 (3) (2003) 194–198. [2] J. Garbade, H.B. Bittner, M.J. Barten, A. Rastan, S. Lehmann, F.W. Mohr, et al., Combined surgical left ventricular reconstruction and left ventricular assist device implantation for destination therapy in end-stage heart failure, Circ. Heart Fail. 4 (2011) el, 4-5. [3] S.J. Park, C.A. Milano, A.G. Tatooles, J.G. Rogers, R.M. Adamson, D.E. Steidley, et al., Outcomes in advanced heart failure patients with left ventricular assist devices for destination therapy, Circ. Heart Fail. 5 (2012) 241–248. [4] R.C. Starling, N. Moazami, S.C. Silvestry, G. Ewald, J.G. Rogers, C.A. Milano, et al., Unexpected abrupt increase in left ventricular assist device thrombosis, NEJM 370 (1) (2014) 33–40.
[5] M.R. Mehra, G.C. Stewart, P.A. Uber, The vexing problem of thrombosis in long-term mechanical circulatory support, J. Heart Lung Transplant. 33 (1) (2014) 1–11. [6] W.L. Holman, R.L. Kormos, D.C. Naftel, M.A. Miller, F.D. Pagani, E. Blume, et al., Predictors of death and transplant in patients with a mechanical circulatory support device: a multi-institutional study, J. Heart Lung Transplant. 28 (2009) 44–50. [7] R. John, F. Kamdar, K. Liao, M. Colvin-Adams, A. Boyle, L. Joyce, Improved survival and decreasing incidence of adverse events with the HeartMate II left ventricular assist device as bridge-to-transplant therapy, Ann. Thorac. Surg. 86 (4) (2008) 1227–1234. [8] T. Hasin, Y. Marmor, W. Kremers, Y. Topilsky, C.J. Severson, J.A. Schirger, et al., Readmissions after implantation of axial flow left ventricular assist device, J. Am. Coll. Cardiol. 61 (2) (2013) 153–163. [9] R. Radovancevic, N. Matijevic, A.W. Bracey, B. Radovancevic, M. Elayda, L.D. Gregoric, et al., Increased leukocyte-platelet interactions during circulatory support with left ventricular assist devices, ASAIO J. 55 (5) (2009) 459–464. [10] B. Meuris, J. Arnout, D. Vlasselaers, M. Schetz, B. Meyns, Long-term management of an implantable left ventricular assist device using low molecular weight heparin and antiplatelet therapy: a possible alternative to oral anticoagulants, Artif. Organs 31 (5) (2007) 402–405. [11] G. Inci, E. Sorguven, Effect of LVAD outlet graft anastomosis angle on the aortic valve, wall, and flow, ASAIO J. 58 (4) (2012) 373–381. [12] M. Deng, L.B. Edwards, M.I. Hertz, A.W. Rowe, B.M. Keck, R.L. Kormos, et al., Mechanical circulatory support device database of the international society for heart and lung transplantation: third annual report, J. Heart Lung Transplant. 24 (2005) 1182–1187. [13] C. Schmid, M. Weyand, D.G. Nabavi, D. Hammel, M.C. Deng, E.B. Ringelstein, et al., Cerebral and systemic embolization during left ventricular support with the Novacor N100 device, Ann. Thorac. Surg. 65 (1998) 1703–1710. [14] C.E. Thomas, D. Jichici, R. Petrucci, V.C. Urrutia, R.J. Schwartzman, Neurologic complications of the Novacor left ventricular assist device, Ann. Thorac. Surg. 72 (4) (2001) 1311–1315. [15] I.R. Argueta-Morales, R. Tran, A. Ceballos, W. Clark, R. Osorio, E.A. Divo, et al., Mathematical modeling of patient-specific ventricular assist device implantation to reduce particulate embolization rate to cerebral vessels, J. Biomech. Eng. 136 (7) (2014). [16] A.F. Osorio, R. Osorio, A. Ceballos, R. Tran, W. Clark, E.A. Divo, I.R. Argueta-Morales, et al., Computational fluid dynamics analysis of surgical adjustment of left ventricular assist device implantation to minimize stroke risk, Comput. Methods Biomech. Biomed. Engin. 16 (6) (2013) 622–638. [17] W. Oriyanhan, H. Tsuneyoshi, T. Nishina, S. Matsuoka, T. Ikeda, K. Komeda, Determination of optimal duration of mechanical unloading for failing hearts to achieve bridge to recovery in a rat heterotopic heart transplantation model, J. Heart Lung Transplant. 26 (1) (2007) 16–23. [18] J.R. Martina, M.E.I. Schipper, N. de Jonge, F. Ramjankhan, R.A. Weger, J.R. Lahpor, et al., Analysis of aortic valve commissural fusion after support with continuous-flow left ventricular assist device, Interact. Cardiovasc. Thorac. Surg. 17 (4) (2013) 616–624. [19] S.D. Gregory, M.C. Stevens, E. Wu, J. Fraser, D. Timms, In vitro evaluation of aortic insufficiency with a rotary left ventricular assist device, Artif. Organs 37 (9) (2013) 802–809. [20] B. Kar, R.M. Delgado, O.H. Frazier, I.D. Gregoric, M.T. Harting, Y. Wadia, et al., The effect of LVAD aortic outflow-graft placement on hemodynamics and flow: implantation
computer methods and programs in biomedicine 133 (2016) 217–227
[21]
[22]
[23]
[24]
[25]
[26]
[27] [28]
technique and computer flow modeling, Tex. Heart Inst. J. 32 (3) (2005) 294–298. K. May-Newman, B. Hillen, W. Dembitsky, Effect of left ventricular assist device outflow conduit anastomosis location on flow patterns in the native aorta, ASAIO J. 52 (2) (2006) 132–139. C. Karmonik, S. Partovi, M. Loebe, B. Schmacjk, A. Ghodsizad, M.R. Robbin, et al., Influence of LVAD cannula outflow tract location on hemodynamics in the ascending aorta: a patient-specific computational fluid dynamics approach, ASAIO J. 58 (6) (2012) 562–567. N. Yang, S. Deutsch, E.G. Paterson, K. Manning, Numerical study of blood flow at the end-to-side anastomosis of a left ventricular assist device for adult patients, J. Biomech. Eng. 131 (11) (2009) 111005. N. Yang, S. Deutsch, E.G. Paterson, K. Manning, Hemodynamics of an end-to-side anastomotic graft for a pulsatile pediatric ventricular assist device, J. Biomech. Eng. 132 (3) (2010) 031009. M.V. Caruso, V. Gramigna, M. Rossi, G.F. Serraino, A. Renzulli, G. Fragomeni, A computational fluid dynamics comparison between different outflow graft anastomosis locations of left ventricular assist device (LVAD) in a patient-specific aortic model, Int. J. Numer. Method. Biomed. Eng. 31 (2) (2015). http://dx.doi.org/10.1002/cnm.2700. F. Kabinejadian, L.P. Chua, D.N. Ghista, M. Sankaranarayanan, Y.S. Tan, A novel coronary artery bypass graft design of sequential anastomoses, Ann. Biomed. Eng. 38 (10) (2010) 3135–3150. Fluent Inc, Fluent V6.1 User Guide, Fluent Inc, Lebanon, NH, 2003. Z.M. Song, K.Y. Gu, B. Gao, F. Wan, Y. Chang, Y. Zeng, Hemodynamic effect of various support modes of
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
227
continuous flow LVADs on the cardiovascular system: a numerical study, Med. Sci. Monit. 20 (2014) 733–741. V.J. Marder, D.J. Chute, S. Starkman, A.M. Abolian, C. Kidwell, D. Liebeskind, et al., Analysis of thrombi retrieved from cerebral arteries of patients with acute ischemic stroke, Stroke 37 (8) (2006) 2086–2093. V.M. Nahimyak, S.W. Yoon, C.K. Holland, Acoustomechanical and thermal properties of clotted blood, J. Acoust. Soc. Am. 119 (6) (2006) 3766–3772. S.S. Khalafvand, E.Y.K. Ng, L. Zhong, T.K. Hunghung, Fluid-dynamics modelling of the human left ventricle with dynamic mesh for normal and myocardial infarction: preliminary study, Comput. Biol. Med. 42 (8) (2012) 863– 870. X. Zhao, Y.J. Liu, F. Bai, X.C. Ren, J.L. Ding, F. Wang, Numerical study on bilateral bidirectional Glenn shunt, J. Med. Biomechan. 27 (5) (2012) 488–494. Z. Cheng, F.P. Tan, C.V. Riga, C.D. Bicknell, M.S. Hamady, R.G.J. Gibbs, et al., Analysis of flow patterns in a patient-specific aortic dissection model, J. Biomech. Eng. 132 (5) (2010) 051007. D.S. Likosky, C.A.S. Marrin, L.R. Caplan, Y.R. Baribeau, J.R. Morton, R.M. Weintraub, et al., Determination of etiologic mechanisms of strokes secondary to coronary artery bypass graft surgery, Stroke 34 (12) (2003) 2830–2834. C.G. Lyndsey, Q. Nidia, J.M. Eric, L.M. David, B. Karsten, Neurocognitive outcomes after extracorporeal membrane oxygenation, Stroke 29 (2) (2015) 123–135. T.S. Kato, T. Ota, P.C. Schulze, M. Farr, U. Jorde, H. Takayama, et al., Asymmetric pattern of cerebrovascular lesions in patients after left ventricular assist device implantation, Stroke 43 (3) (2012) 872–874.