Intraventricular flow patterns and stasis in the LVAD-assisted heart

K. Wong a, G. Samaroo a, I. Ling a, W. Dembitsky b, R. Adamson b, J.C. del Álamo c,d,K. May-Newman a,n

a Bioengineering Program, San Diego State University, Department of Mechanical Engineering, San Diego, CA 92182-1323, United Statesb Mechanical Circulatory Support, Cardiothoracic Surgery, Sharp Memorial Hospital, San Diego, CA 92182-1323, United Statesc Mechanical and Aerospace Engineering, U.C. San Diego, La Jolla, CA 92093-0411, United Statesd Institute for Engineering in Medicine, U.C. San Diego, La Jolla, CA 92093, United States

a r t i c l e i n f o

Article history:Accepted 21 December 2013

Keywords:Left ventricleMechanical circulatory assistPIVVortexThrombus

a b s t r a c t

Left ventricular assist device (LVAD) support disrupts the natural blood flow path through the heart,introducing flow patterns associated with thrombosis, especially in the presence of medical devices. Theaim of this study was to quantitatively evaluate the flow patterns in the left ventricle (LV) of the LVAD-assisted heart, with a focus on alterations in vortex development and stasis. Particle image velocimetry ofa LVAD-supported LV model was performed in a mock circulatory loop. In the Pre-LVAD flow condition, avortex ring initiating from the LV base migrated toward the apex during diastole and remained in the LVby the end of ejection. During LVAD support, vortex formation was relatively unchanged although vortexcirculation and kinetic energy increased with LVAD speed, particularly in systole. However, as pulsatilitydecreased and aortic valve opening ceased, a region of fluid stasis formed near the left ventricularoutflow tract. These findings suggest that LVAD support does not substantially alter vortex dynamicsunless cardiac function is minimal. The altered blood flow introduced by the LVAD results in stasisadjacent to the LV outflow tract, which increases the risk of thrombus formation in the heart.

& 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Left ventricular assist devices (LVADs) are mechanical pumpsthat are surgically connected to the left ventricle (LV) and aorta toincrease aortic flow and end-organ perfusion. Clinical studies havedemonstrated that LVADs improve patient health and quality oflife and significantly reduce the mortality of cardiac failure(Slaughter et al., 2010). A foreseeable limit of LVAD implantationis the native heart's capacity to endure the changes induced bylong-term LVAD use. During high LVAD support, blood flowoccurs entirely through the LVAD, the aortic valve is continuouslyclosed, and the heart operates in series with the pump. Withsufficient intrinsic contraction, the heart and the pump operate inparallel (John et al., 2010). Abnormal flow patterns such asstagnation areas have been previously correlated with thrombosis,especially in the presence of medical devices (Bluestein, 2004).Small thrombi can form in areas of blood flow stagnation andgrow to become emboli, which present a serious risk for systemicembolization or pump dysfunction. Previous studies of theatrial appendage have found that slow flow and reduced ejection

are strong predictors of clot formation (Goswami et al., 2004;Fyrenius et al., 2001). The changes in intraventricular flow thataccompany LVAD use can lead to areas of stasis in the subvalvularleft ventricular outflow tract (LVOT) (May-Newman et al., 2013)and in the LVAD apex near the LVAD inflow cannula (Estep et al.,2010), where thrombi have been observed (Miyake et al., 2004).

Previous studies have demonstrated that blood flow in the LV isasymmetric, unsteady, and three-dimensional with a common pat-tern consisting of a large diastolic vortex that channels blood towardsthe aortic valve (Rodevand et al., 1999; Kim et al., 1995; Pedrizzettiand Domenichini, 2005; Kheradvar et al., 2010). This vortex has beenshown to contribute to diastolic suction (Kheradvar et al., 2007), andto minimize kinetic energy (KE) losses and cardiac work (Pedrizzettiand Domenichini, 2005). Furthermore, the vortex has been shown tofacilitate the blood entering the LV to be washed out completely afterfew cardiac cycles thereby preventing intraventricular blood stagna-tion (Eriksson et al., 2010; Hendabadi et al., 2013).

Disruption in the vortex patterns has been associated withenergy losses leading to higher workload of the heart muscle(Faludi et al., 2010; Pierrakos and Vlachos, 2006) as well asincreased intraventricular blood stasis (Watanabe et al., 2008).Despite the potential impact of LV flow patterns in cardiacperformance and blood stasis, little is known about the alterationsinduced by LVAD support. A previous computational study by

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/jbiomechwww.JBiomech.com

Journal of Biomechanics

http://dx.doi.org/10.1016/j.jbiomech.2013.12.0310021-9290 & 2014 Elsevier Ltd. All rights reserved.

n Corresponding author. Tel.: þ1 619 594 5652; fax: þ1 619 594 3599.E-mail address: kmaynewm@mail.sdsu.edu (K. May-Newman).

Journal of Biomechanics 47 (2014) 1485–1494

Loerakker and colleagues found that LVAD support decreasesvortex strength and promotes their early decay but was limitedto a simplified axisymmetric tubular geometry (Loerakker et al.,2008). Our objective was to measure the intraventricular flowpatterns in a mock loop of the LVAD-assisted left ventricle usingDigital Particle Image Velocimetry (DPIV). Our hypothesis is thatthe addition of an LVAD to the circulation modifies the normalflow pattern of the LV by (i) increasing the extent of the stagnationregions and (ii) disrupting the normal diastolic vortex.

DPIV is an experimental technique for flow visualization andhas been widely used to study coherent structures in cardiovas-cular fluid flow in vitro (Willert and Gharib, 1991). This techniqueprovides instantaneous velocity measurement of a time-varyingflow field. Experimental results obtained from in-vitro PIV studieshave been used to validate clinical observations of LV fluid flow(Garcia et al., 2010; Abe et al., 2013).

2. Materials and methods

2.1. Experimental setup

Experimental studies were performed with the SDSU cardiac simulator (CS), amock loop of the LVAD-assisted heart (Zamarripa Garcia et al., 2008; May-Newmanet al., 2010). A Heartmate II LVAD (Thoratec Inc.; Pleasanton, CA) was attached to atransparent LV manufactured from platinum-cured silicone rubber (Young's mod-ulus of 620 kPa at 100% elongation, a tensile strength of 5.52 MPa, and elongationat break of 400%) to anatomically mimic the flow geometry of the LVAD-assistedventricle (see Fig. 1). For all flow conditions tested, left atrial pressure (LAP), LVpressure (LVP), aortic root pressure (AoP), total aortic flow (Q-total) and LVAD flowrate (Q-LVAD) were recorded at a frequency of 200 Hz using LabChart (ADinstruments) (May-Newman et al., 2010). The simulator settings were varied overa wide range to ensure coverage of physiological conditions. Seven different LVADrates (6500–12500 rpm) and simulated cardiac function (On/Off) were combined tocreate a matrix of 15 test conditions. Baseline measurements were made tosimulate a pre-LVAD patient, with the LVAD off and the LVAD conduit clamped.The design of the simulator is based on a 3-element Windkessel model in whichRp¼1.7�108 Pa-s/m3, C¼1.2�10�8 m3/Pa and Rs¼7.8�106 Pa-s/m3 following themethods of Yin and Liu (1989). By adjusting the systemic resistance clamp down-stream of the aorta, the pre-LVAD condition had a AoP of 73 mmHg and a Q-total of3.74 L/min, representative of HF patients with a severity of NYHA IV (Maurer et al.,2009; Travis et al., 2007). The resistance and compliance of the circuit were maintainedconstant for the duration of the study and heart rate maintained at 70 bpm.

The velocity field was measured in the CS using DPIV (Willert and Gharib,1991). A laser light sheet 1–2 mm in thickness was focused on the midplane of theLV, and a LaVision DPIV system (LaVision, Inc) used to capture and analyze theimages. The circulating fluid was a viscosity-matched blood analog consisting of40% glycerol (viscosity of 3.72 cP at 20C) and saline (Segur and Oberstar, 1951) andwas seeded through the left atrial chamber with 20-mm neutrally buoyantfluorescent particles (PMMA-RHb) during imaging. The camera (LaVision ImagerIntense) recorded 12-bit digital images with a time interval of 700–2000 ms basedon the average velocity of the flow field. Interrogation windows of 32�32 wereapplied to a field of 1376�1040 pixels to obtain sufficient spatial resolution.Triggered image pairs were acquired at 40 Hz for each condition and ten (N¼10)image sets collected for each time point and phase averaged. A sequentialacquisition technique that “phase-locked” specific phases of the cardiac cyclebased on predefined left ventricular and aortic pressure profiles was used(Falahatpisheh and Kheradvar, 2012; Abe et al., 2013).

2.2. Data analysis

Vortex identification enables visualization and quantification of swirlingcoherent structures. Vortices were identified in the recorded velocity fields andtheir dynamical properties were tracked in time using procedures based on the Qcriterion (Hunt, 1988) as described by Garcia et al. (2010). For an incompressiblefluid flow, Q is a measure of excessive rotation rate with respect to the strain rate,and is given by

Q ¼ 12ð Ωj jj2� Sj jj2Þ

���� ð1Þ

where S and Ω are the symmetric and anti-symmetric components of the velocitygradient tensor ∇v. In two dimensions, Q can be simply expressed as

Q ¼ �12

∂v2x∂y

þ2∂vx∂y

∂vy∂x

þ∂v2y∂y

!ð2Þ

Points with Q higher than a threshold Qth were identified as belonging to a vortex.The value of the threshold Qth was defined as the standard deviation of Q in spaceand time, as is customary in complex flows (Del Alamo et al., 2006). For each timeframe, contiguous pixels with Q4Qth were clustered in space and small (o10pixels) spurious objects were filtered out. The instantaneous position, radius,circulation and KE of the identified vortex cores were computed as a function oftime. Circulation was defined as

Γ ¼ZAωðx; yÞdA ð3Þ

where ω(x,y) is the vorticity and A is the area enclosed within the vortex core(Garcia et al., 2010). Vortices are defined as clockwise (CW) or counter-clockwise(CCW) depending on the sign of their circulation. KE was defined as

KE¼ 12ρ

ZAvðx; yÞ2dA ð4Þ

where ρ is the fluid density. Note that, because this quantity is computed in twodimensions, its units are J/m. The position of the vortex centroid was defined as

xCyC

!¼ 1Γ

ZAωðx; yÞ

x

y

!dA ð5Þ

The vortex core was fit to an ellipse with axes given by the second order momentsof the vorticity distribution

O¼ 1Γ

ZAωðx; yÞ ðx�xC Þ2 ðx�xC Þ ðy�yC Þ

ðx�xC Þ ðy�yC Þ ðy�yC Þ2" #

dA ð6Þ

such that the eigenvalues and eigenvectors of the matrix O provide the major andminor axes of the ellipse, a and b, and their orientation. The radius of the vortexwas defined as

R¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffia2þb2

2

s: ð7Þ

The results were made independent of the threshold Qth by recomputing the vortexproperties over the area defined by an ellipse centered at (xc, yc) and with majorand minor axes given by 2a and 2b. This procedure was repeated iteratively untilthe vortex radius varies by less than 1% between iterations, which was usuallyachieved in 4–5 steps.

We employed a variation of the Stagnation Index (SI) proposed by Quaini et al.(2011), to assess the amount of fluid stasis in the LV associated with different levelsof LVAD support. SI in our analysis is defined as

SIðx; yÞ ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

ALV TRT jvðx; yÞj2 dt

sð8Þ

where T is the period of interest (e.g. diastole, systole or the overall cardiac cycle)and ALV is the average area of the LV section considered in this study. This Eulerianindex has dimensions of time and is inversely proportional to the time-averaged KEat each point; a high value of SI indicates a low local velocity sustained over timeand, thus, the likelihood of fluid stagnation. We are not aware of any work that hascorrelated SI with intraventricular thrombus formation in patients with heartfailure. However, our previous work indicates that SI is typically o1 s in normalleft ventricles (Hendabadi et al., 2013). SI was averaged over a region of interestdefined in the left ventricular outflow tract (LVOT) to assess fluid stasis in this area.

3. Results

The test matrix of 15 flow settings resulted in four basicintraventricular flow patterns: pre-LVAD, which serves as a base-line, and three post-LVAD patterns: parallel pulsatile, seriespulsatile and nonpulsatile. Some of these patterns were producedby several different combinations of settings (e.g., series pulsatile)and are grouped together in the results.

3.1. Hemodynamics

Table 1 summarizes the hemodynamic data for all flow settings.Compared to the nonpulsatile flow, all three pulsatile flow condi-tions showed higher AoP, LVP, and Q-total for all LVAD speeds.Conversely, the pressure drop across the aortic valve (TVP) washigher for nonpulsatile flow. LVAD-generated flow contributed58% to the total aortic flow at the lowest level of LVAD support(6500 rpm), and reached 100% at LVAD speeds of 11500 rpm andabove. Except for LVP, all the hemodynamic parameters displayeda gradual increase with increasing LVAD support. Instantaneous

K. Wong et al. / Journal of Biomechanics 47 (2014) 1485–14941486

flow for pulsatile conditions at three different LVAD speeds(0, 9500, and 11500 rpm) are shown in Fig. 2. During parallelpulsatile flow, both the LVAD and the heart contributed to Q-total,but pulsatility decreased and the valve opened for less time eachcycle compared to Pre-LVAD flow. This trend continued as theLVAD flow was increased, eventually producing sufficient TVP tokeep the aortic valve closed throughout the cardiac cycle.

3.2. Pre-LVAD flow patterns

For the Pre-LVAD condition, the cardiac cycle began with atransmitral fluid jet that formed a vortex ring which appeared astwo counter-rotating vortex cores in the LV midplane. The mainantero-septal core rotated in the clockwise (CW) direction andchanneled blood towards the aortic tract, whereas the smaller

infero-lateral core rotated counter-clockwise (CCW) (Fig. 3a). TheCCW vortex was constrained due to its proximity to the ventricularwall (Fig. 3b) and eventually vanished from the flow field (Fig. 3c).The CW vortex propagated along the center axis of the LV,resulting in a large CW swirling motion during diastasis(Fig. 3d). A late diastolic jet during the A-wave generated ashort-lived CCW vortex core (Fig. 3e), while the CW vortexresulted from the late diastolic jet merged with the pre-existingCW vortex (Fig. 3e,f). The resulting swirling flow seemed tofacilitate fluid redirection towards the aortic valve (Fig. 3g) andwas still observed at mid-systole (Fig. 3h). These flow patternsand their time evolution agreed well with flow fields obtainedby echocardiography and magnetic resonance imaging in patientswith heart failure (Garcia et al., 2010; Hendabadi et al., 2013;Bermejo et al., 2014).

Fig. 1. (A) The SDSU cardiac simulator is a mock circulatory loop that reproduces the LVAD-supported heart. The left ventricle is modeled with a transparent, thin-walledrubber sac attached to a Heartmate II LVAD and suspended in a fluid filled chamber as shown in the schematic and (B) in a photo. (C) The velocity profile applied by thesimulator's piston pump to generate the pulsatile cardiac function of a heart failure patient.

K. Wong et al. / Journal of Biomechanics 47 (2014) 1485–1494 1487

3.3. Flow patterns during LVAD support (post-LVAD)

Fig. 4 shows the LV flow patterns at matched time points for allfour flow cases. For post-LVAD pulsatile flow conditions, the vortexring appeared, similar to the pre-LVAD case. However, the inflowjets and the vortices, particularly the CCW vortex, were stronger inthe post-LVAD cases. The differences were more prominent duringsystole. At a medium level of LVAD support, the flow wassimultaneously ejected through both the aortic valve and theLVAD conduit, establishing a parallel flow pattern. For series flow,the aortic valve remained closed during systole as the fluid wasejected through the apex of the LV into the LVAD only. In theabsence of cardiac pulsatility, the mitral valve was always open,the transmitral inflow was continuously directed towards theLVAD inflow conduit. As a consequence, ventricular outflowcompletely bypassed the aortic valve, leading to a large stagnationregion near the LVOT.

3.4. Vortex analysis

Fig. 5 displays the time evolution of LV vortices for each flowcondition. The circulation of both vortices (Fig. 5a) increased

during early diastole for all pulsatile conditions, but the CCWvortex rapidly decayed. The A-wave jet led to the reappearance ofthe CCW vortex and reinforcement of the existing CW vortexduring LVAD support as diastole progressed. By the onset ofsystole, the CCW vortex had disappeared but the CW vortexpersisted throughout systole. The evolution of vortex KE (Fig. 5b)closely followed that of circulation. A significant fraction of theenergy stored in the vortex ring was lost by end-diastole; by earlysystole, the CW retained E25% of its peak KE in both pre-LVADand post-LVAD pulsatile conditions. For nonpulsatile flow, thecirculation and KE of both vortices remained constant andrelatively low.

Average vortex circulation and KE increased moderately withLVAD support (Fig. 6a,b). However, for both parallel and seriespost-LVAD conditions, the circulation and KE of the CW vortex alsodecayed faster during systole (Fig. 5a,b). While the total aortic flowincreased by 76% between the pre-LVAD condition and the post-LVAD condition at maximum LVAD rate (12,500 rpm), the circula-tion of the main CW vortex increased by only 6% and KE increasedby 40%. The circulation and KE of the vortices show that the CWvortex was stronger and more persistent than the CCW regardlessof the level of LVAD support, which is consistent with the flow

Table 1Average hemodynamic data for pulsatile and continuous flow conditions.

Cardiac function LVAD (krpm) AoP (mmHg) LVP (mmHg) TVP (mmHg) CO (L/min) QLVAD (L/min) QLVAD/CO (%) Flow condition

On Off 73.370.2 27.870.0 45.570.2 3.770.1 0 0 Pre-LVADOn 6.5 86.570.1 26.870.0 59.770.1 4.370.1 2.570.0 58 ParallelOn 7.5 92.670.1 26.070.1 66.670.0 4.670.1 3.170.0 67 ParallelOn 8.5 101.470.1 27.070.0 74.470.1 4.970.1 3.970.0 80 ParallelOn 9.5 112.370.1 29.070.1 83.370.0 5.370.1 4.570.0 85 ParallelOn 10.5 121.570.1 28.970.0 92.670.1 5.770.1 5.370.0 93 ParallelOn 11.5 132.170.2 28.870.0 103.370.2 6.070.0 6.070.0 100 SeriesOn 12.5 146.770.1 30.570.1 116.270.0 6.570.1 6.570.0 100 SeriesOff 6.5 55.470.2 6.770.0 48.770.2 3.270.0 3.270.0 100 SeriesOff 7.5 64.470.2 6.470.0 58.070.2 3.870.0 3.870.0 100 SeriesOff 8.5 75.070.3 6.170.1 68.970.2 4.370.0 4.370.0 100 SeriesOff 9.5 86.970.3 5.770.0 81.270.3 4.970.1 4.970.1 100 SeriesOff 10.5 100.670.4 5.370.1 95.370.3 5.470.1 5.470.1 100 SeriesOff 11.5 115.370.5 4.870.1 110.570.4 5.970.1 5.970.1 100 SeriesOff 12.5 132.270.4 4.470.1 127.870.3 6.270.1 6.270.1 100 Series

LVAD, left ventricular assist device; AoP, aortic root pressure; LVP, left ventricular pressure; TVP¼AoP–LVP, trans-valvular pressure across the aortic valve. CO, total cardiacoutput; LVADQ, LVAD flow rate.

0

10

20

0 0.5 1 1.5 2 2.5 3 3.5

Q (L

/min

)

CO

QLVAD

0

10

20

0 0.5 1 1.5 2 2.5 3 3.5

Q (L

/min

)

CO

QLVAD

0

10

20

0 0.5 1 1.5 2 2.5 3 3.5

Q (L

/min

)

Time (s)

CO

QLVAD

Pre-LVAD

Series Flow(LVAD 11500rpm)

Parallel Flow(LVAD 9500rpm)

DiastoleDiastole

Systole

Fig. 2. Cardiac output (solid line) and LVAD flow rate (dashed line) for three different pulsatile flow conditions: pre-LVAD (top), parallel (middle), and series (bottom) showan increasing proportion of flow through the LVAD as the support level is increased.

K. Wong et al. / Journal of Biomechanics 47 (2014) 1485–14941488

patterns in Figs. 3 and 4. Fig. 5c displays the distance between thecenter of the CW vortex and the center of the mitral annulus, dv.The CW vortex traveled rapidly towards the center of the LVduring early E wave, but during the rest of the cardiac cycle movedaround a constant value corresponding to the fixed vortex position

during nonpulsatile flow (Fig. 5c). Interestingly, LVAD support ledto higher CW vortex mobility during diastole but lower mobilityduring systole. Overall, the vortex core in the pre-LVAD flowcondition shows the highest variation in position during thecardiac cycle among all flow conditions, (0.19odvo0.77).

A-wave

Systole

a

(m/s)

(m/s)

(m/s)

DiastasisE-wave

bc

de

f g

h

Fig. 3. Velocity field images at eight instants of the cardiac cycle for the pre-LVAD flow condition (labeled on LV flow rate waveform in the bottom panel) overlaid on imagesof the LV model. The arrows indicate the direction and magnitude (color) of the local velocity. The vortices are superimposed on the images with the local vorticity shown aswhite (CW) or magenta (CCW).

K. Wong et al. / Journal of Biomechanics 47 (2014) 1485–1494 1489

3.5. Stagnation index

Fig. 7 shows the spatial distribution of SI in the LV averagedover diastole, systole, and the entire cardiac cycle for the four flowpatterns. For all pulsatile flow conditions, SI was low duringdiastole, reaching its minimum value near the inflow jet and itsmaximum value around the CW vortex core. During systole, SIreached low values near the LVOT in the pre-LVAD condition butincreased with LVAD support as flow was diverted towardsthe apex. This led to an overall increase in the post-LVAD SI nearthe LVOT over the cardiac cycle, which increased with LVAD speed.A region of interest in which average SI in the LVOT was calculatedis illustrated in Fig. 7, and the results summarized in Table 2 forpre-LVAD, parallel pulsatile, and nonpulsatile flow. In the case of

pre-LVAD, the SI dropped by 7% from diastole to systole. Similarly,the parallel condition showed a relatively low SI during diastole,but increased 55% during systole. Overall, the SI for the parallelflow condition increased by 12% when compared to the pre-LVADcondition. The series nonpulsatile flow case had a substantiallyhigher SI than parallel flow and pre-LVAD conditions.

Altogether, the SI maps (Fig. 7) and vortex trajectory data(Fig. 5) reveal that fluid stasis was reduced when the CW vortexchanged its position, increasing the mixing in the LV. Note that thehigher mobility of the LV vortex observed during diastole in thepost-LVAD pulsatile cases (Fig. 7C) was associated with lowerdiastolic SI during these flow conditions, when compared with thepre-LVAD case (Fig. 5). Likewise, the higher vortex mobilityobserved in the pre-LVAD case during systole was associated with

Fig. 4. Velocity field images for several flow conditions: pre-LVAD (first row) and post-LVAD implantation (rows 2–4). Three instants of the cardiac cycle are represented: lateE-wave (left column), late A-wave (central column), and peak systole (right column). The vectors show the direction and magnitude (color) of local velocity, withsuperimposed vorticity. Parallel pulsatile flow is characterized by a bifurcated outflow during systole. Series flow is characterized by outflow occurring exclusively throughthe LVAD. For nonpulsatile flow, the pattern is steady.

K. Wong et al. / Journal of Biomechanics 47 (2014) 1485–14941490

lower systolic SI. Consistent with this idea, SI was highest when thesystem operated in nonpulsatile mode. It is important to note thata large region of elevated SI was observed near the LVOT, which

has been directly associated with the formation of a largethrombus in this region of the ventricle (May-Newman et al.,2013).

The results of this in vitro study demonstrate that, underpulsatile conditions, vortex formation in early diastole was rela-tively unchanged by LVAD support, but vortex decay followed adifferent courses. LVAD assistance boosted vortex circulation andKE during the A wave, which decayed more rapidly during systolethan under Pre-LVAD conditions. Flow stasis in the LVOT underpulsatile conditions did not vary during diastole with LVADsupport, but increased in systole and was high and constant withnonpulsatile flow conditions. These findings suggest that contin-uous flow LVAD support does not substantially disrupt normalvortex dynamics unless cardiac function is minimal and pulsatilityis absent. The altered blood flow introduced by the LVAD results inflow stasis adjacent to the LVOT, which increases the risk ofthrombus formation in the heart.

4. Discussion

For the pre-LVAD condition, our results compare favorably withthe flow patterns reported in previous clinical studies using bothphase contrast MRI (Kim et al., 1995; Kilner et al., 2000; Bolgeret al., 2007) and different echocardiography modalities (Del Alamoet al., 2009; Garcia et al., 2010; Hendabadi et al., 2013; Bermejoet al., 2014; Bazilevs et al., 2010; Rodriguez et al., 2013; Senguptaet al., 2012). These studies have suggested that swirling flow in theLV may increase the efficiency of the healthy heart by redirectingintraventricular flow, which reduces KE losses and cardiac work

Fig. 5. (A) Time-varying left and right vortex circulation for selected pulsatile flowcases: pre-LVAD, parallel pulsatile, series pulsatile and nonpulsatile conditions.(B) Kinetic energy during the cardiac cycle. (C) Instantaneous distance of the CW vortexcore relative to the base of the mitral valve plotted against dimensionless time (t/T).

Fig. 6. (A) Average circulation of the leading CW and CCW vortices increased inmagnitude with LVAD speed; linear trend lines are shown. (B) Average kineticenergy of the two main vortices as a function of LVAD speed.

K. Wong et al. / Journal of Biomechanics 47 (2014) 1485–1494 1491

(Kilner et al., 2000; Kim et al., 1995) and increases mixing(Eriksson et al., 2010). We postulate that LVAD design could beimproved by considering which LV flow patterns are beneficialwhen the blood flow path is rerouted. This study has, for the firsttime, characterized the intraventricular flow dynamics developedfollowing LVAD implantation using an in vitro model. Few

previous studies have examined the changes in ventricular flowwith LVAD use, and those that have are based on computationalmodels in simplified LV geometries (Loerakker et al., 2008).

Intraventricular flow architectures were generally not alteredwith LVAD support provided some cardiac function was present.Both CW and CCW vortices were observed and their time evolu-tion was similar to the pre-LVAD case. The strength of thesevortices increased with LVAD support, particularly during latefilling (A-wave). However, this LVAD-induced vortex strengtheningwas moderate considering that the total aortic flow was substan-tially increased by LVAD support. These results are in partialagreement with a previous computational study that showed thatthe strength of the LV vortex ring is lowered during cardiac failureand the use of a LVAD further decreases vortex strength, ultimatelycausing its early disappearance (Loerakker et al., 2008).

Several studies have proposed that the diastolic CW vortex mayact as a reservoir of KE that is consumed during systole to facilitatethe ejection of blood (Kilner et al., 2000; Abe et al., 2013;Kheradvar et al., 2007). LVAD support did not alter the overall

Fig. 7. Spatial distribution of the stagnation index, SI, over the ventricular midplane during different flow conditions. A higher value for SI indicates a low local velocitysustained over time and greater degree of flow stasis, averaged over diastole (left column), systole (middle column) and the entire cardiac cycle (right column). A region ofinterest defining the LVOT is shown superimposed on the nonpulsatile diastolic frame, within which SI was averaged for Table 2.

Table 2Average SI for selected flow conditions in the LVOT.

Pulsatile flow Pre-LVAD LVAD off Q-total: 3.7 L/min

Parallel LVAD9500 rpm Q-total:5.3 L/min

Nonpulsatile flow SeriesLVAD 10,500 rpm Q-total: 5.4 L/min

Diastole 4.3670.49 3.9970.5 11.5971.12Systole 4.0470.33 6.1971.19 11.5971.12Overall 4.2570.44 4.7370.76 11.5971.12

SI is the stagnation index and is defined in Eq. (8); a high SI indicates greaterlikelihood for thrombus formation.

K. Wong et al. / Journal of Biomechanics 47 (2014) 1485–14941492

efficiency of this process, as the same fraction of peak CW vortexKE was retained at the onset of systole both in the pre-LVAD andthe post-LVAD pulsatile flow conditions. Systolic flow patterns,however, were altered by LVAD support considerably morethan diastolic ones. As a larger fraction of the total flow exitsthrough the LVAD conduit at the LV apex, the aortic ejection jetweakened, until it eventually disappeared in series flow. In allpulsatile modes, the vortex cores experience a sharp decreasein left ventricular volume during systole, leading to a decay invortex circulation as predicted by Stewart (Stewart et al., 2012).The ventricular volume remained constant for each of thenonpulsatile modes.

During pulsatile flow conditions in the heart, the existence of adynamic vortex ring that travels from the mitral annulus towardsthe apex provides thorough washing of the LV cavity. In fact, ourresults suggest that LV vortex mobility may be an important factorin order to prevent flow stasis. During pulsatile conditions, LVADsupport did not substantially modify the distribution of fluid stasisoverall, although the stagnation index increased substantiallyduring systole. However, during nonpulsatile LVAD support, theaortic valve remained closed and a fixed flow pattern developed.The KE of the resulting vortex ring was lower than in the pulsatileflow cases, and the vortex cores were localized downstream of themitral valve and did not change position with time. However, aspulsatility decreased and aortic valve opening ceased, a region offluid stasis formed near the LVOT, as shown in Table 2. Thesefindings are in agreement with previous reports that patients withseverely impaired cardiac contractility and high LVAD support mayhave a greater risk of intraventricular thrombosis, particularlyalong the LVOT (Loerakker et al., 2008; Estep et al., 2010) andcan result in fatal consequences (May-Newman et al., 2013). Thus,the time evolving position of the CW vortex is a prominent featureof LV flow dynamics deserving special attention. We confirm thatintraventricular flow pulsatility is required to prevent flow stasisin the LVAD-supported heart, such that the risk of thrombusformation can be minimized. For end-stage heart failure patients,pulsatility may not be possible with the damaged myocardium,but must be provided by other sources, i.e., the LVAD.

5. Limitations

The CS has been used previously for measuring hemodynamicsand aortic valve biomechanics in the LVAD-assisted ventricle(May-Newman et al., 2010; Zamarripa Garcia et al., 2008).Although the design is based on the Windkessel model, it cannotadapt to changes in pressure and flow like the native cardiovas-cular system and maintains the same impedance until the settingsare changed. Furthermore, pulsatile cardiac function is achieved bycompressing the ventricle, neglecting the effect of myocardialtwist which has shown to be important for cardiac energetics.However, investigators have found reduced twist in heart failurepatients (Grosberg and Gharib, 2009; Tan et al., 2009), whichwould be compatible with our model. Although the mitral valvegeometry plays an important role in the initiation of vortices in thenormal heart and is not reproduced in our model, we and othershave observed the same fluid patterns in a mock loop with aporcine bioprosthetic valve (Pierrakos and Vlachos, 2006). Finally,LV size decreases and native LV contractility can be partiallyrecovered in patients after LVAD implantation (Ambardekar andButtrick, 2011), which was not simulated in this study.

6. Conclusion

Measurement of the LV flow field under different levels of LVADsupport was conducted using DPIV in a mock circulatory loop. The

results showed that the Pre-LVAD vortex formation in earlydiastole was relatively unchanged by LVAD support, but vortexdecay followed a different course. LVAD assistance boosted vortexcirculation and KE during the A wave, which decayed more rapidlyduring systole than under Pre-LVAD conditions. When pulsatilitywas absent, i.e., when the heart was not contracting, the vortexremained in a fixed position with a low and constant circulationand KE. These findings suggest that continuous flow LVAD supportdoes not substantially disrupt normal vortex dynamics unlesscardiac function is minimal and pulsatility is absent. Flow stasisin the LVOT did not change substantially with LVAD supportduring diastole, but increased in systole and with decreasedcardiac function. The flow stasis adjacent to the LVOT, introducedby the LVAD, increases the risk of thrombus formation in the heart,which can lead to emboli and stroke.

Conflict of interest statement

None of the authors have received funding from an entityperforming research or commercialization in this area.

Acknowledgments

The authors wish to thank Medtronic and Thoratec for theircontribution of equipment for this with this project. The technicalassistance of Fernando Olea and Michelle Jordan was greatlyappreciated. This work was supported by the San Diego StateUniversity and the Sharp Healthcare Foundation. JCA was partiallysupported by NIH Grant 1R21 HL108268-01 and the HellmanFamily Fund.

References

Abe, H., Caracciolo, G., Kheradvar, A., Pedrizzetti, G., Khandheria, B.K., Narula, J.,Sengupta, P.P., 2013. Contrast echocardiography for assessing left ventricularvortex strength in heart failure: a prospective cohort study. Eur. Heart J.Cardiovasc. Imaging 14, 1049–1060.

Ambardekar, A.V., Buttrick, P.M., 2011. Reverse remodeling with left ventricularassist devices: a review of clinical, cellular, and molecular effects. Circ. HeartFail. 4, 224–233.

Bazilevs, Y., Del Alamo, J.C., Humphrey, J.D., 2010. From imaging to prediction:emerging non-invasive methods in pediatric cardiology. Prog. Pediatr. Cardiol.30, 81–89.

Javier, Bermejo, Yolanda, Benito, Marta, Alhama, Raquel, Yotti, Pablo, Martínez-Legazpi,Candelas, Pérez del Villar, Esther, Pérez-David, Ana, González-Mansilla, Cristina,Santa-Marta, Alicia, Barrio, Francisco, Fernández-Avilés, del Alamo, Juan C., 2014.Intraventricular vortex properties in non-ischemic dilated cardiomyopathy. Am. J.Physiol. – Heart Circ. Physiol. http://dx.doi.org/10.1152/ajpheart.00697.2013.

Bluestein, D., 2004. Research approaches for studying flow-induced thromboem-bolic complications in blood recirculating devices. Expert Rev. Med. Devices 1,65–80.

Bolger, A.F., Heiberg, E., Karlsson, M., Wigstrom, L., Engvall, J., Sigfridsson, A.,Ebbers, T., Kvitting, J.P., Carlhall, C.J., Wranne, B., 2007. Transit of blood flowthrough the human left ventricle mapped by cardiovascular magnetic reso-nance. J. Cardiovasc. Magn. Reson. 9, 741–747.

Del Alamo, J.C., Jimenez, J., Zandonade, P., Moser, R.D., 2006. Self-similar vortexclusters in the turbulent logarighmic region. J. Fluid Mech. 561, 329–358.

Del Alamo, J.C., Marsden, A.L., Lasheras, J.C., 2009. Recent advances in theapplication of computational mechanics to the diagnosis and treatment ofcardiovascular disease. Rev. Esp. Cardiol. 62, 781–805.

Eriksson, J., Carlhall, C.J., Dyverfeldt, P., Engvall, J., Bolger, A.F., Ebbers, T., 2010. Semi-automatic quantification of 4D left ventricular blood flow. J. Cardiovasc. Magn.Reson. 12, 9.

Estep, J.D., Stainback, R.F., Little, S.H., Torre, G., Zoghbi, W.A., 2010. The role ofechocardiography and other imaging modalities in patients with left ventri-cular assist devices. JACC Cardiovasc. Imaging 3, 1049–1064.

Falahatpisheh, A., Kheradvar, A., 2012. High speed particle image velocimetry toassess cardiac fluid dynamics in vitro: from performance to validation. Eur.J. Mech.—B/Fluids 35, 2–8.

Faludi, R., Szulik, M., D'hooge, J., Herijgers, P., Rademakers, F., Pedrizzetti, G., Voigt, J.U., 2010. Left ventricular flow patterns in healthy subjects and patients withprosthetic mitral valves: an in vivo study using echocardiographic particleimage velocimetry. J. Thorac. Cardiovasc. Surg. 139, 1501–1510.

K. Wong et al. / Journal of Biomechanics 47 (2014) 1485–1494 1493

Fyrenius, A., Wigstrom, L., Ebbers, T., Karlsson, M., Engvall, J., Bolger, A.F., 2001.Three dimensional flow in the human left atrium. Heart 86, 448–455.

Garcia, D., Del Alamo, J.C., Tanne, D., Yotti, R., Cortina, C., Bertrand, E., Antoranz, J.C.,Perez-David, E., Rieu, R., Fernandez-Aviles, F., Bermejo, J., 2010. Two-dimensional intraventricular flow mapping by digital processing conventionalcolor-Doppler echocardiography images. IEEE Trans. Med. Imaging 29,1701–1713.

Goswami, K.C., Yadav, R., Bahl, V.K., 2004. Predictors of left atrial appendage clot: atransesophageal echocardiographic study of left atrial appendage function inpatients with severe mitral stenosis. Indian Heart J. 56, 628–635.

Grosberg, A., Gharib, M., 2009. Modeling the macro-structure of the heart: healthyand diseased. Med. Biol. Eng. Comput. 47, 301–311.

Hendabadi, S., Bermejo, J., Benito, Y., Yotti, R., Fernandez-Aviles, F., Del Alamo, J.C.,Shadden, S.C., 2013. Topology of blood transport in the human left ventricle bynovel processing of Doppler echocardiography. Ann. Biomed. Eng. (Epub aheadof print)

Hunt, J.C.R., 1988. Eddies, streams and convergence zones in turbulent flows, Centerfor Turbulence Research Report CTR-S88

John, R., Mantz, K., Eckman, P., Rose, A., May-Newman, K., 2010. Aortic valvepathophysiology during left ventricular assist device support. J. Heart LungTransplant. 29, 1321–1329.

Kheradvar, A., Houle, H., Pedrizzetti, G., Tonti, G., Belcik, T., Ashraf, M., Lindner, J.R.,Gharib, M., Sahn, D., 2010. Echocardiographic particle image velocimetry: anovel technique for quantification of left ventricular blood vorticity pattern. J.Am. Soc. Echocardiogr. 23, 86–94.

Kheradvar, A., Milano, M., Gharib, M., 2007. Correlation between vortex ringformation and mitral annulus dynamics during ventricular rapid filling. ASAIOJ. 53, 8–16.

Kilner, P.J., Yang, G.Z., Wilkes, A.J., Mohiaddin, R.H., Firmin, D.N., Yacoub, M.H., 2000.Asymmetric redirection of flow through the heart. Nature 404, 759–761.

Kim, W.Y., Walker, P.G., Pedersen, E.M., Poulsen, J.K., Oyre, S., Houlind, K.,Yoganathan, A.P., 1995. Left ventricular blood flow patterns in normal subjects:a quantitative analysis by three-dimensional magnetic resonance velocitymapping. J. Am. Coll. Cardiol. 26, 224–238.

Loerakker, S., Cox, L.G., van Heijst, G.J., de Mol, B.A., van de Vosse, V., 2008. Influenceof dilated cardiomyopathy and a left ventricular assist device on vortexdynamics in the left ventricle. Comput. Methods Biomech. Biomed. Eng. 11,649–660.

Maurer, M.M., Burkhoff, D., Maybaum, S., Franco, V., Vittorio, T.J., Williams, P.,White, L., Kamalakkannan, G., Myers, J., Mancini, D.M., 2009. A multicenterstudy of noninvasive cardiac output by bioreactance during symptom-limitedexercise. J. Card. Fail. 15, 689–699.

May-Newman, K., Enriquez-Almaguer, L., Posuwattanakul, P., Dembitsky, W., 2010.Biomechanics of the aortic valve in the continuous flow VAD-assisted heart.ASAIO J. 56, 301–308.

May-Newman, K., Wong, Y.K., Adamson, R., Hoagland, P., Vu, V., Dembitsky, W.,2013. Thromboembolism is linked to intraventricular flow stasis in a patientsupported with a left ventricle assist device. ASAIO J. 59, 452–455.

Miyake, Y., Sugioka, K., Bussey, C.D., Di, T.M., Homma, S., 2004. Left ventricularmobile thrombus associated with ventricular assist device: diagnosis bytransesophageal echocardiography. Circ. J. 68, 383–384.

Pedrizzetti, G., Domenichini, F., 2005. Nature optimizes the swirling flow in thehuman left ventricle. Phys. Rev. Lett. 95, 108101.

Pierrakos, O., Vlachos, P.P., 2006. The effect of vortex formation on left ventricularfilling and mitral valve efficiency. J. Biomech. Eng. 128, 527–539.

Quaini, A., Canic, S., Paniagua, D., 2011. Numerical characterization of hemody-namics conditions near aortic valve after implantation of left ventricular assistdevice. Math. Biosci. Eng. 8, 785–806.

Rodevand, O., Bjornerheim, R., Edvardsen, T., Smiseth, O.A., Ihlen, H., 1999. Diastolicflow pattern in the normal left ventricle. J. Am. Soc. Echocardiogr. 12, 500–507.

Rodriguez, M.D., Markl, M., Moya Mur, J.L., Barker, A., Fernandez-Golfin, C.,Lancellotti, P., Zamorano Gomez, J.L., 2013. Intracardiac flow visualization:current status and future directions. Eur. Heart J. Cardiovasc. Imaging 14,1029–1038.

Segur, J.B., Oberstar, H.E., 1951. Viscosity of glycerol and its aqueous solutions. Ind.Eng. Chem. 43, 2117–2120.

Sengupta, P.P., Pedrizzetti, G., Kilner, P.J., Kheradvar, A., Ebbers, T., Tonti, G., Fraser,A.G., Narula, J., 2012. Emerging trends in CV flow visualization. JACC Cardiovasc.Imaging 5, 305–316.

Slaughter, M.S., Pagani, F.D., Rogers, J.G., Miller, L.W., Sun, B., Russell, S.D., Starling,R.C., Chen, L., Boyle, A.J., Chillcott, S., Adamson, R.M., Blood, M.S., Camacho, M.T.,Idrissi, K.A., Petty, M., Sobieski, M., Wright, S., Myers, T.J., Farrar, D.J., 2010.Clinical management of continuous-flow left ventricular assist devices inadvanced heart failure. J. Heart Lung Transplant. 29, S1–39.

Stewart, K.C., Niebel, C.L., Jung, S., Vlachos, P.P., 2012. The decay of confined vortexrings. Exp. Fluids 53, 163–171.

Tan, Y.T., Wenzelburger, F., Lee, E., Heatlie, G., Leyva, F., Patel, K., Frenneaux, M.,Sanderson, J.E., 2009. The pathophysiology of heart failure with normal ejectionfraction: exercise echocardiography reveals complex abnormalities of bothsystolic and diastolic ventricular function involving torsion, untwist, andlongitudinal motion. J. Am. Coll. Cardiol. 54, 36–46.

Travis, A.R., Giridharan, G.A., Pantalos, G.M., Dowling, R.D., Prabhu, S.D., Slaughter,M., Sobieski, M., Undar, A., Farrar, D.J., Koenig, S.C., 2007. Vascular pulsatility inpatients with a pulsatile or continuous flow ventricular assist device. J. ThoracicCardiovasc. Surg. 133, 517–524.

Watanabe, H., Sugiura, S., Hisada, T., 2008. The looped heart does not save energyby maintaining the momentum of blood flowing in the ventricle. Am. J. Physiol.Heart Circ. Physiol. 294, H2191–H2196.

Willert, C.E., Gharib, M., 1991. Digital particle image velocimetry. Exp. Fluids 10,181–193.

Yin, F.C., Liu, Z.R., 1989. Estimating arterial resistance and compliance duringtransient conditions in humans. Am. J. Physiol. 257, H190–H197.

Zamarripa Garcia, M.A., Enriquez, L.A., Dembitsky, W., May-Newman, K., 2008. Theeffect of aortic valve incompetence on the hemodynamics of a continuous flowventricular assist device in a mock circulation. ASAIO J. 54, 237–244.

K. Wong et al. / Journal of Biomechanics 47 (2014) 1485–14941494