bs_bs_banner
© 2011, Copyright the Authors
Artificial Organs © 2011, International Center for Artificial Organs and Transplantation and Wiley Periodicals, Inc.
Validation of an Axial Flow Blood Pump: Computational
Fluid Dynamics Results Using Particle Image Velocimetry
*Boyang Su, †Leok Poh Chua, and ‡Xikun Wang
*DHI-NTU Water & Environment Centre and Education Hub, Nanyang Technological University; †School of Mechanical
and Aerospace Engineering, Nanyang Technological University; and ‡Maritime Research Centre, Nanyang Technological
University, Singapore
Abstract: A magnetically suspended axial flow blood
pump is studied experimentally in this article. The pump
casing enclosed a three-blade straightener, a two-blade
impeller shrouded by a permanent magnet-embedded cyl-
inder, and a three-blade diffuser. The internal flow fields
were simulated earlier using computational fluid dyna-
mics (CFD), and the pump characteristic curves were
determined.The simulation results showed that the internal
flow field was basically streamlined, except the diffuser
region. Particle image velocimetry (PIV) measurement of
the 1:1 pump model was conducted to validate the CFD
result. In order to ensure the optical access, an acrylic pro-
totype was fabricated with the impeller driven by a servo-
motor instead, as the magnet is opaque. In addition to the
Each year, approximately 30 000 patients at risk of
dying from severe heart failure are awaiting heart
transplant, but only around 3500 cases are performed
(1) mainly constrained by the limited heart donors
available. Ventricular assist devices (VADs) have
been developed for decades as an alternative
therapeutic modality and can be produced in the
quantity needed. VADs are used for bridge-to-
transplantation, bridge-to-recovery, or even long-
term cardiac support (2,3). The continuous-flow
VADs usually include a rotating impeller to generate
adequate flow rate and pressure head, and so are also
named as rotary blood pumps. Typically, the rotating
speed of a centrifugal blood pump is 2000~4000
rpm (4–7), and that of an axial blood pump is
doi:10.1111/j.1525-1594.2011.01362.x
Received November 2010; revised July 2011.
Address correspondence and reprint requests to Mr. Boyang Su,
School of Mechanical and Aerospace Engineering, Nanyang Tech-
nological University, Nanyang Drive, Singapore 639798, Singapore.
E-mail: bysu@ntu.edu.sg
aor_1362 359..367
transparent model, the blood analog fluid with the refrac-
tive index close to that of acrylic was used to avoid
refraction. According to the CFD results, the axial flow
blood pump could generate adequate pressure head at the
rotating speed of 9500 rpm and flow rate of 5 L/min, and
the same flow condition was applied during the PIV
measurement. Through the comparisons, it was found that
the experimental results were close to those obtained by
CFD and had thus validated the CFD model, which could
complement the limitation of the measurement in assessing
the more detailed flow fields of the axial flow pump. Key
Words: Computational fluid dynamics—Axial flow blood
pump—Particle image velocimetry.
6000~12 000 rpm (8–11). The hemolysis induced by a
rotary blood pump is the main concern besides the
hydraulic performance (12). It is found that the
hemolysis is significantly related to the flow behavior,
and the scalar shear stress and exposure time are two
main parameters used to estimate the hemolysis (13).
Therefore, the rotary blood pump poses a challenge,
especially for axial flow blood pump with higher
rotating speed.
Computational fluid dynamics (CFD) has been
widely used as a powerful tool to investigate many
blood-contacting medical devices including blood
pumps (14,15). With the aid of CFD, the design phase
can be shortened significantly from several years to
several months (16). It can also support and explain
experimental results. Compared with the traditional
design, CFD can predict the detailed flow fields,
pump characteristics curves, and other parameters
such as hemolysis (12,17). However, the numerical
results still need validating using the nonintrusive
measurements, although they are usually difficult and
limited by the optical access. For the whole flow field,
Artificial Organs 2012, 36(4):359–367
360 B. SU ET AL.
particle image velocimetry (PIV) technique is widely
adopted (18–20), as compared with the point-wise
hot wire (21) or laser Doppler anemometer (22,23).
The internal flow seeded with particles is exposed by
double-frame laser sheets, and the velocity vector
map is calculated using a cross-correlation technique
in PIV measurement.
Due to the complexity of the structure of an axial
flow blood pump, which usually includes a straight-
ener, impeller, and diffuser, and its high rotating
speed, there are only few studies using PIV to
measure its internal flow (14,20) as compared with
the centrifugal blood pump (18,19,24–26). Therefore,
in this study, the axial flow blood pump was analyzed
by both CFD and PIV techniques, and both numeri-
cal and experimental results were compared for
validation.
MATERIALS AND METHODS
Physical model
The axial blood pump comprises of three compo-
nents: a three-blade straightener, an enclosed two-
blade impeller, and a three-blade diffuser as shown in
Fig. 1. Its overall axial length and outer diameter of
the inner components are 60 and 19 mm, respectively.
The impeller was the only rotating component
enclosed inside a cylinder, which was driven by the
Lorentz force. As a result, the impeller and the cylin-
der rotated simultaneously as a single component
(27). As the magnet was installed in the rotor cylin-
der, there was a clearance gap between the rotor
cylinder and the pump casing. The current diffuser
has only three blades for the ease of fabrication and
PIV measurement.
CFD method
The axial flow blood pump was generated using
computer aided design software Pro/Engineer Wild-
FIG. 1. The axial flow VAD.
Artif Organs, Vol. 36, No. 4, 2012
FIG. 2. Grid independence test.
fire 4.0 (Parametric Technology Corporation,
Needham, MA, USA), and was then exported into
commercial grid generator software ICEM (ANSYS,
Inc., Canonsburg, PA, USA). The numerical model
consisted of three domains, that is, straightener,
impeller, and diffuser, with both straightener and dif-
fuser prolonged to make the flow develop. The grid
independence test was conducted based on the pres-
sure head across the blood pump as shown in Fig. 2. It
can be found that the pressure head difference with
respect to the finest model was less than 1% as the
grid number of the numerical model increased to 1.3
million. Therefore, the model with 1.3 million grids
was selected as a compromise between the accuracy
and computational resource. In between the clear-
ance gap between the pump casing and the impeller
shroud, 10 layers of grid cells were generated in the
selected model. The skew angles of these grids were
between 20 and 160° with the y+ value of grids at
near-wall region maintained around 1, in order to
ensure the quality of grids. Figure 3a shows the
surface grids on the inner components and the longi-
tudinal cross section with the complete computa-
tional domain is shown in Fig. 3b.
The numerical simulation was conducted using the
finite-volume CFD code FLUENT 6.3.26. The blood
was assumed to be Newtonian with a viscosity of
0.0035 Pa·s and a density of 1050 kg/m3 (28). At the
rotating speed of 9500 rpm, the Reynolds number
based on the impeller tip was about 3.4 ¥ 104, so the
internal flow was turbulent.The shear stress transport
(SST) k-w turbulence model developed by Menter
(29) was selected because it combines the advantage
of k-w and k-e models in the near-wall regions and
outside the boundary layers, respectively (26,30,31).
In addition, it predicts better results in adverse pres-
sure gradients and separating flow (32). The sliding
mesh model was applied to accurately predict the
 VALIDATION OF BLOOD PUMP USING PARTICLE IMAGE VELOCIMETRY
(a) Detailed surface grids
(b) Complete numerical domain
rotor–stator interaction with 300 time steps in one
period (33). Velocity inlet and pressure outlet were
applied at the pump inlet and outlet, respectively. The
static pressure head across the numerical model was
used as the macro parameter to monitor the conver-
gence, and the iteration was stopped when the con-
vergence criterion of 10-4 was satisfied (34).
Experimental method
Figure 4a shows the schematic drawing of the con-
structed experimental test rig. The pump model con-
structed is the exact size of the prototype in order to
avoid any scaling problem that alters the results.
When the pump operates, the flow rate is controlled
by the throttle valve located downstream of the flow-
meter, and the pressure head across the pump model
and the flow rate is obtained from the pressure trans-
ducer and the flowmeter, respectively.
In order to reduce the possibility of hemolysis,
magnetic suspension technique is widely adopted in
rotary blood pumps, including the present study.
However, it is impossible to perform PIV measure-
ments due to the opaque magnets. Therefore, the
impeller is driven by a servomotor through a shaft.
361
FIG. 3. Surface grids on the axial flow
blood pump. (a) Detailed surface grids, (b)
complete numerical domain.
The test model is partially fabricated from acrylic for
the illumination of the laser sheet and image pairs
capturing as demonstrated in Fig. 4b. In order to
minimize the vibration of the operating impeller, a
bearing holder, which is a hollow step cylinder, is
designed as shown in Fig. 4c. Its front end is designed
to place the hollow diffuser in position with the diam-
eter of the diffuser hub equal to that of the bearing
holder at Section B. Note that the influences of the
vibrations of the inner components such as impeller
and diffuser are not considered in this study. The
straightener, impeller, and diffuser are all made of
acrylic as well as the impeller shroud that is glued
together with the impeller. The impeller is driven by
a steel shaft, which is inserted through the bearings
and bearing holder. These components are enclosed
in the chamber, which is capped by two components
connected to the test rig.
The PIV system used in the present study was the
Dantec FlowMap PIV system (Dantec Dynamics
A/S, Skovlunde, Denmark). The Gemini PIV 200-15
(New Wave Research, Fremont, CA, USA), a double-
cavity pulsed neodymium yttrium aluminum garnet
(Nd:YAG) laser system, was used as the illumination
Artif Organs, Vol. 36, No. 4, 2012
362 B. SU ET AL.

(a)

(b)

FIG. 4. Schematic view of (a) the test rig

and (b) test model mounted on the
traverse stage in PIV measurement.
(c) CAD drawing of the bearing holder.

(c)

Artif Organs, Vol. 36, No. 4, 2012

 VALIDATION OF BLOOD PUMP USING PARTICLE IMAGE VELOCIMETRY
source. The two pulsing beams from a double-cavity
Q-switched Nd:YAG laser are combined and formed
into pulsing light sheets using optics. Q-switched
pulse can effectively freeze the motion of particle,
and two cavities ensure the widest possible ranges of
time between pulses. The CCD camera used in
the FlowMap PIV system is HiSense MkII
camera (Hamamatsu C8484-05 digital CCD chip
(Hamamatsu City, Shizuoka, Japan); resolution
1280 ¥ 1024) equipped with Nikon AF Micro-Nikkor
60 mm f/2.8D (Tokyo, Japan). FlowManager software
(Dantec Dynamics A/S) controlled the synchroniza-
tion sequence to acquire data, and stored the record-
ing data. The “once-per-revolution” signal from the
shaft encoder was sent to the processor, which then
fires the laser and acquires the images from the PIV
camera as shown in Fig. 3a, so images could be taken
at a certain angular position of the impeller. More-
over, the timing signals are used to control the
camera to make sure that the CCD sensor is only
exposed to the light from the illumination system, as
the ambient light will reduce the signal : noise ratio.
In order to measure the flow field at a different
angular position, of the impeller a delay was
set between the external signal and the image
acquisition. Two successive images are then trans-
ferred to the FlowMap processor and stored finally.
As a result, a pair of seeding particle images with
time delay was captured at a certain impeller angular
position, and the velocity vector map was derived
from the movement of particles during the time
delay.
In order to avoid refraction, the working fluid
involved mixing 40% of glycerin and 60% of aqueous
ammonium thiocyanate (NH4SCN) solution by
363
volume (35). Its refractive index was 1.488 to match
with that of the acrylic model with a density of
1155 kg/m3 and a dynamic viscosity of 3.8 mPa·s. The
kinetic viscosities of the working fluid and blood
were 3.29 and 3.33 mm2/s, respectively, so as to main-
tain Reynolds number similarity. The solution was
seeded with neutral buoyant hollow glass spheres
with the mean diameter of 10 mm. In the present PIV
measurement, 100 image pairs were captured at each
measured area. The measurement area was 10.76 ¥
8.26 mm2 with the scale factor, S, of 1.25. The record-
ing aperture was f/5.6~8, and thus, the camera image
of the seeding particle had a diameter 2~3 pixels
under the wavelength of 532 nm. The interrogation
area of 32 ¥ 32 pixels was used with 50% overlap. The
pulse duration of this PIV is 10 ns, so the particle
movement during this time duration was negligible.
In the straightener, the pulse separation varied
10~25 ms at different measurement areas according to
the local flow velocity. In order to minimize the
effects of the out-of-plane velocity component, the
thickness of the laser sheet was 0.5 mm, and the time
between light pulses was 10 ms so as to satisfy the
equation below (36):
vz ⋅ Δt
≤ 0.25
Zl
where vz is the out-of-plane velocity, Dt is the time
between pulses and Zl is the thickness of the laser
sheet. Postprocessing validation methods including
peak height validation and velocity range validation
were applied on the raw vector. Finally, the mean
phase-averaged velocity profiles over 100 recordings
were obtained in each region.
FIG. 5. Experimental and numerical
pump characteristic curves.
Artif Organs, Vol. 36, No. 4, 2012
364 B. SU ET AL.

RESULTS AND DISCUSSION

Figure 5 compares the pump performance curves

of both CFD and experimental results at different
rotating speeds. Note that each point corresponds to
a specific flow rate and rotating speed under steady
flow condition. Both the experimental and CFD
simulation results show that the static pressure head
decreases with the increasing flow rate. It can be
found that CFD simulations overestimate the perfor-
mance, mainly due to the obstruction of the bars as
indicated by the black color in Fig. 4c. However, they (a) straightener
follow the trends of the experimental results. In order
to match the operating conditions used in the
numerical simulation, the impeller rotating speed and
the pump flow rate were 9500 rpm and 5 L/min,
respectively, in the PIV experiment.
In order to show the flow field between the hub
(R = 5.5 mm) and pump casing (R = 7.5 mm), Fig. 6
shows the longitudinal cross-sectional planes through
the Z-axis in each region. Figure 7 shows the com-
parisons of velocity profiles between the CFD and
PIV results in the straightener, impeller, and diffuser
regions, respectively. In these figures, the solid line
and dotted line denote PIV and CFD results,
(b) impeller
respectively. In Fig. 7a, the flow in the straightener
region is smooth in the positive axial direction. At the
region close to the walls of the hub and casing of the
pump, the fluid has lower speed due to the boundary
layer effect as compared to that at the middle of the
channel. The axial velocity profile is almost parabolic
in shape. Due to the reduced cross-sectional area,
the fluid is accelerated as it flows through the
straightener.
As the fluid flows into the impeller channel, the
dominant flow is in the positive Z direction, and
the high axial velocity is located slightly close to the
impeller hub (R = 5.5 mm), especially further down- (c) diffuser
stream until Z = 18 mm, as shown in Fig. 7b. As a
result, the fluid velocity close to the impeller shroud FIG. 6. Demonstration of the cross-sectional planes in each
region. (a) Straightener, (b) impeller, (c) diffuser.
(R = 7.5 mm) is much smaller than that close to the
hub (R = 5.5 mm). From Z = 18 mm onward, it can be
observed that the forward velocity at the impeller
shroud has increased in magnitude, while near diffuser is the most irregular with strong reverse flow
the impeller exit, reverse flow is detected close to the close to the diffuser hub (R = 5.5 mm). It is because
impeller hub. Because the diffuser placed at the the fluid there has lower velocity and cannot over-
impeller downstream induces an adverse static come the adverse pressure gradient. However, the
pressure gradient at the impeller exit, the fluid near forward flow becomes more dominant from Z =
the rotating impeller hub, which has lower momen- 26.5 mm onward where the blade ends. A similar flow
tum than that near the impeller shroud, cannot pattern is predicted by the numerical simulation.
overcome it and flow is reversed. A qualitative agreement between the numerical
Figure 7c shows the velocity vector distributions in and experimental results can be observed, especially
the diffuser. Compared with the flow patterns in both in the straightener followed by the impeller and the
straightener and the impeller, the flow pattern in the diffuser. The discrepancy becomes relatively obvious

Artif Organs, Vol. 36, No. 4, 2012

 VALIDATION OF BLOOD PUMP USING PARTICLE IMAGE VELOCIMETRY
(a) Straightener
(b) Impeller
(c) Diffuser
at the regions close to blades, hub, and inner casing
surfaces due to a number of reasons. First, the
trapped seeding particles on surfaces could induce
relatively strong reflection, resulting in error in PIV
measurements. Second, the fluid close to the inter-
faces usually has higher velocity gradient in radial
direction, and it could cause error due to the difficulty
in PIV measurements. In order to calculate the esti-
mated velocity accurately, the seeding particles
within an interrogation area, where the camera image
is usually divided into 32 pixels by 32 pixels in size,
should be homogenous, and have exactly the same
displacement. However, the displacement of indi-
vidual particles is always different from each other
under high velocity gradient. Third, the out-of-plane
motion of the flow field will introduce signal dropout.
However, it has been minimized by selecting proper
365
FIG. 7. Comparison of the velocity pro-
files obtained from PIV (solid line) and
CFD (dotted line) results in (a) straight-
ener, (b) impeller, and (c) diffuser.
laser thickness and the time between pulses as men-
tioned earlier in the experimental method section.
The L2 relative error norm of the differences in axial
velocity between CFD and PIV with respect to the
PIV in the straightener, impeller, and diffuser is 11,
19, and 27%, respectively. Note that the L2 relative
error norm is defined as
N
∑ ( fi E − fi N )2
i
N
∑ ( fi E )2
i
E N
where fi and fi represent the values at node
i for PIV experiment and numerical simulation,
respectively. These values are consistent with the
observation earlier.
Artif Organs, Vol. 36, No. 4, 2012
366 B. SU ET AL.
CONCLUSION
Computational fluid dynamics was initially used to
study the internal flow field of the axial flow blood
pump.The experimental test showed that the numeri-
cal simulation could predict the pump characteristic
curves with similar trends. Furthermore, the particle
image velocimetry measurement was conducted for
the validation at the flow conditions of 9500 rpm
rotating speed and 5 L/min flow rate. Generally, the
CFD results agree fairly well with the PIV results,
especially in the straightener region. It is also found
that the matching between numerical and measured
results is related to the flow pattern, that is, a better
agreement is always obtained in a smooth flow field
such as in the straightener region. The major differ-
ences between the experimental and simulation
results are usually found near interfaces, due to the
high velocity gradients, strong reflections from the
seeding particles trapped there, and the high out-
of-plane velocity component. Basically, a fair agree-
ment of the simulations and measurements can be
observed, and it has reasonably validated the simula-
tion results. Therefore, SST k-w turbulence model is
able to predict the flow phenomenon in the axial flow
blood pump.
REFERENCES
1. Mielniczuk L, Mussivand T, Davies R, et al. Patient selection
for left ventricular assist devices. Artif Organs 2004;28:
152–7.
2. Borovetz HS, Ramasamy N, Zerbe TR, Portner PM. Evalua-
tion of an implantable ventricular assist system for humans
with chronic refractory heart failure. ASAIO J 1995;41:42–
8.
3. Tayama E, Olsen DB, Ohashi Y, et al. The DeBakey ventricu-
lar assist device: current status in 1997. Artif Organs 1999;23:
1113–6.
4. Anderson JB, Wood HG, Allaire PE, Olsen DB. Numerical
analysis of blood flow in the clearance regions of a continuous
flow artificial heart pump. Artif Organs 2000;24:492–500.
5. Song X, Throckmorton AL, Wood HG, Antaki JF, Olsen DB.
Quantitative evaluation of blood damage in a centrifugal
VAD by computational fluid dynamics. J Fluids Eng 2004;126:
410–8.
6. Chua LP, Song G, Yu SCM, Lim TM. Computational fluid
dynamics of gap flow in a biocentrifugal blood pump. Artif
Organs 2005;29:620–8.
7. Legendre D, Antunes P, Bock E, Andrade A, Biscegli JF, Ortiz
JP. Computational fluid dynamics investigation of a centrifu-
gal blood pump. Artif Organs 2008;32:342–8.
8. Mitamura Y, Nakamura H, Okamoto E, Yozu R, Kawada S,
Kim D-W. Development of the Valvo pump: an axial flow
pump implanted at the heart valve position. Artif Organs
1999;23:566–71.
9. Zhu X, Zhang M, Zhang G, Liu H. Numerical investigation on
hydrodynamics and biocompatibility of a magnetically sus-
pended axial blood pump. ASAIO J 2006;52:624–9.
10. Throckmorton AL, Untaroiu A, Allaire PE, et al. Numerical
design and experimental hydraulic testing of an axial flow
Artif Organs, Vol. 36, No. 4, 2012
ventricular assist device for infants and children. ASAIO J
2007;53:754–61.
11. Zhang Y, Zhan Z, Gui X, et al. Design optimization of an axial
blood pump with computational fluid dynamics. ASAIO J
2008;54:150–5.
12. Arvand A, Hormes M, Reul H. A validated computational
fluid dynamics model to estimate hemolysis in a rotary blood
pump. Artif Organs 2005;29:531–40.
13. Paul R, Apel J, Schugner S, Shwindke P, Reul H. Shear stress
related blood damage in laminar couette flow. Artif Organs
2003;27:517–29.
14. Triep M, Brucker C, Schroder W, Siess T. Computational fluid
dynamics and digital particle image velocimetry study of the
flow through an optimized micro-axial blood pump. Artif
Organs 2006;30:384–91.
15. Zhang J-M, Chua LP, Ghista DN, Zhou T-M, Tan YS. Valida-
tion of numerical simulation with PIV measurements for two
anastomosis models. Med Eng Phys 2008;30:226–47.
16. Burgreen GW, Antaki JF, Wu ZJ, Holmes AJ. Computational
fluid dynamics as a development tool for rotary blood pumps.
Artif Organs 2001;25:336–40.
17. Goubergrits L, Affeld K. Numerical estimation of blood
damage in artificial organs. Artif Organs 2004;28:499–
507.
18. Sankovic JM, Kadambi JR, Mehta M, Smith WA, Wernet MP.
PIV investigations of the flow field in the volute of a rotary
blood pump. J Fluids Eng 2004;126:730–4.
19. Day SW, McDaniel JC. PIV measurements of flow in a cen-
trifugal blood pump: steady flow. J Biomech Eng 2005;127:244–
53.
20. Triep M, Brucker C, Kerkhoffs W, Schumacher O, Marseille O.
Investigation of the washout effect in a magnetically driven
axial blood pump. Artif Organs 2008;32:778–84.
21. Chua LP, Ong KS, Yu CMS, Chan WK, Wong YW. Gap veloc-
ity measurements of a blood pump model. Artif Organs
2002;26:682–94.
22. Chua LP, Ong KS, Song G. Study of velocity and shear stress
distributions in the impeller passages and the volute of a bio-
centrifugal ventricular assist device. Artif Organs 2008;32:376–
87.
23. Chua LP, Ong KS, Song G, Ji W. Measurements by laser
doppler velocimetry in the casing/impeller clearance gap of a
biocentrifugal ventricular assist device model. Artif Organs
2009;33:360–72.
24. Day SW, McDaniel JC, Wood HG, Allaire PE, Landrot N,
Curtas A. Particle image velocimetry measurements of blood
velocity in a continuous flow ventricular assist device. ASAIO
J 2001;47:406–11.
25. Day SW, McDaniel JC, Wood HG, et al. A prototype
HeartQuest ventricular assist device for particle image veloci-
metry measurements. Artif Organs 2002;26:1002–5.
26. Song X, Wood HG, Day SW, Olsen DB. Studies of turbulence
models in a computational fluid dynamics model of a blood
pump. Artif Organs 2003;27:935–7.
27. Lim TM, Zhang D. Development of Lorentz force-type self-
bearing motor for an alternative axial flow blood pump design.
Artif Organs 2006;30:347–53.
28. Chua LP, Su BY, Lim TM, Zhou TM. Numerical simulation of
an axial blood pump. Artif Organs 2007;31:560–70.
29. Menter FR. Two-equation eddy-viscosity turbulence
models for engineering applications. AIAA J 1994;32:1598–
605.
30. Zhu L, Zhang X, Yao Z. Shape optimization of the diffuser
blade of an axial blood pump by computational fluid dynamics.
Artif Organs 2010;34:185–92.
31. Kido K, Hoshi H, Watanabe N, et al. Computational fluid
dynamics analysis of the pediatric tiny centrifugal blood pump
(TinyPump). Artif Organs 2006;30:392–9.
32. Fraser KH, Taskin ME, Griffith BP, Wu ZJ. The use of com-
putational fluid dynamics in the development of ventricular
assist devices. Med Eng Phys 2011;33:263–80.
 VALIDATION OF BLOOD PUMP USING PARTICLE IMAGE VELOCIMETRY 367

33. Chua LP, Song GL, Zhou TM. Numerical analysis of the inner
flow field of a biocentrifugal blood pump. Artif Organs
2006;30:467–77.
34. Throckmorton AL, Untaroiu A, Allaire PE, et al. Computa-
tional analysis of an axial flow pediatric ventricular assist
device. Artif Organs 2004;28:881–91.
35. Doyle MG, Vergniaud JB, Tavoularis S, Bourgault Y.
Numerical simulations of blood flow in artificial and natural
hearts with fluid–structure interaction. Artif Organs 2008;
32:870–9.
36. Dantec-Dynamics. Flowmanager software and introduction to
PIV instrumentation (software user’s guide). 2002.

Artif Organs, Vol. 36, No. 4, 2012