Artificial Organs
33(4):334–345, Wiley Periodicals, Inc.
© 2009, Copyright the Authors
Journal compilation © 2009, International Center for Artificial Organs and Transplantation and Wiley Periodicals, Inc.
Multimodal Flow Visualization and Optimization of
Pneumatic Blood Pump for Sorbent Hemodialysis System
*Fangjun Shu, †Robert Parks, ‡John Maholtz, §Steven Ash, and *James F. Antaki
*Department of Biomedical Engineering, Carnegie Mellon University; †Daedalus Excel, Pittsburgh, PA; ‡Renal Solutions,
Inc., Warrendale, PA; and §Hemocleanse, Inc., Lafayette, IN, USA
Abstract: Renal Solutions Allient Sorbent Hemodialysis
System utilizes a two-chambered pneumatic pump (Pulsar
Blood Pump, Renal Solutions, Inc., Warrendale, PA, USA)
to avoid limitations associated with peristaltic pumping
systems. Single-needle access is enabled by counter-pulsing
the two pump chambers, thereby obviating compliance
chambers or blood reservoirs. Each chamber propels 20 cc
per pulse of 3 s (dual access) or 6 s (single access) duration,
corresponding to a peak Reynolds number of approxi-
mately 8000 (based on inlet velocity and chamber diameter).
A multimodal series of flow visualization studies (tracer
particle, dye washout, and dye erosion) was conducted on a
sequence of pump designs with varying port locations and
diaphragms to improve the geometry with respect to risk
of thrombogenesis. Experiments were conducted in a
The Allient Sorbent Hemodialysis System is
intended for renal replacement therapy for acute and
chronic renal failure (Fig. 1a). One of its unique fea-
tures is the use of a Pulsar dual-chambered pneumatic
pump (Renal Solutions, Inc. [RSI], Warrendale, PA,
USA) to withdraw and infuse blood from/to the
patient (Fig. 1b). This configuration enables single-
lumen vascular access without the need for a blood
reservoir or compliance chamber, and allows dual-
lumen access when desired, for faster flow. The pump
provides expanded safety and versatility advantages
over conventional peristaltic pumps through its
unique, pressure-limited, two-chamber design. A
baseline design of the Pulsar pump is comprised of a
blood chamber featuring coaxial inlet and outlet
doi:10.1111/j.1525-1594.2009.00724.x
Received September 2007; revised March 2008.
Address correspondence and reprint requests to Dr. Fangjun
Shu, Department of Biomedical Engineering, Carnegie Mellon
University, 700 Technology Drive, Pittsburgh, PA 15219, USA.
E-mail: fangjun@gmail.com
334
simplified flow loop using occluders to simulate flow resist-
ance induced by tubing and dialyzer. Tracer visualization
revealed flow patterns and qualitatively indicated turbu-
lence intensity. Dye washout identified dwell volume and
areas of flow stagnation for each design. Dye erosion results
indicated the effectiveness and homogeneity of surface
washing. Compared to a centered inlet which resulted in a
fluid jet that produced two counter-rotating vortices, a tan-
gential inlet introduced a single vortex, and kept the flow
laminar. It also provided better surface washing on the
pump inner surface. However, a tangential outlet did not
present as much benefit as expected. On the contrary, it
created a sharp defection to the flow when transiting
from filling to ejection. Key Words: Thrombogenicity—
Surface wash—Dye washout—Dye erosion—Blood pump.
ports, separated from a pneumatically driven air
chamber by a flexible, elastic diaphragm.The specified
flow is 100–400 mL/min for dual access flow, 100–
200 mL/min for single lumen flow, requiring pneu-
matic drive pressure of 350 mm Hg and vacuum of
-250 mm Hg to augment filling.
Previous clinical experience with the baseline RSI
Pulsar blood pump revealed minimal hemolysis in
this system (1,2), thus was not a concern of this study.
However, deposition of thrombus was commonly
found in the pump and/or dialyzer after 3–8 h of
treatment, despite steady infusion of heparin (3).
Therefore, this study was undertaken to improve the
baseline design to supply the required flow and pres-
sure while avoiding deleterious fluid dynamics that
could promote thrombosis and cellular trauma.
The visualization of the flow field inside a
diaphragm-type blood pump is generally confounded
by several factors. The curved surfaces of virtually
every component of the blood path prevent optical
access of the full region of interest. Additionally, the
continuously flexing diaphragm cyclically blocks the
 SORBENT HEMODIALYSIS SYSTEM
BLOOD OUTLET
(TO DIALYZER)
BLOOD CHAMBER
OCCLUDERS
DIALYZER
CONNECTION TO
PATIENT
optical path, and limits the focal depth. Therefore,
three flow visualization techniques were applied in
this study to perform a comprehensive assessment of
the relevant flow characteristics. The geometric fea-
tures of the pump under consideration included the
position and orientation of the blood ports as well as
the contour of both housing and diaphragm. Due to
the coupling between these design features, various
alternatives were considered in combination.The can-
didate designs were evaluated on the basis of minimiz-
ing stagnation and optimizing surface washing to limit
thrombus deposition while also avoiding flow distur-
bances such as jets and chaotic flow to reduce risk of
cellular trauma and platelet activation.
MATERIALS AND METHODS
Dialyzer system and experimental setup
The complete blood flow path within the dialysis
system depicted in Fig. 1a includes the blood pump as
well as vascular access catheters, a dialyzer, and inter-
connecting tubing. The pump is driven by a pneu-
matic control unit providing both positive pressure
and vacuum in a cyclical fashion at 10 cycles per
minute (in dual access mode). Pneumatically actu-
ated tubing occluders upstream and downstream of
the pump are synchronized with the drive pulse to
provide unidirectional flow.
The fluid loop was simplified for the flow visualiza-
tion studies by replacing the dialyzer and cannula
with an equivalent fluid resistance provided by an
335
FIG. 1. Allient Sorbent Dialysis System.
(a) Dual pneumatic pumps allow single-
needle access to patient. (b) Construction
BLOOD INLET of pump comprised of a pair of snap-fit
(FROM PATIENT) housings with elastomeric diaphragm sep-
arating pneumatic and blood chambers.
DIAPHRAGM
PNEUMATIC PORT
AIR CHAMBER
occluder valve. Because the two pump chambers
shown in Fig. 1b are symmetric, and do not commu-
nicate with one another, only one was studied for
the purpose of these experiments. The resistance
was adjusted to achieve a minimum inlet pressure
of approximately -40 mm Hg during filling, and a
maximum outlet pressure during ejection of approxi-
mately 125 mm Hg—corresponding to clinical opera-
tion of the system in the field. A solution of glycerin
and water (1:2 volume, approximately 3.7 cP, and
1.08 g/mL) was used as blood analog.
The parametric study of alternative pump caps and
diaphragms is summarized in Fig. 2. All the caps have
a base diameter of 45 mm and the height of 15 mm
with inlet and outlet inner diameter of 4 mm. These
designs were generated by solid-modeling software
(SolidWorks 2006, SolidWorks, Inc., Madison, WI,
USA) and rapid-prototyped by stereolithography
(3D Systems, Valencia, CA, USA) with an optically
clear resin (RenShap SL 7565, Vantico A&T US, Inc.,
E. Lansing, MI, USA).The parts were sanded with 320
grit paper to remove any surface irregularities and
then sprayed with a clear lacquer gloss coat.The parts
were then wet sanded at 600 grit to provide smooth
surface for optical access. The baseline pump design
(A0) features a flat diaphragm and coaxial inlet and
outlet ports, centered within the body of the cap.
The baseline flat diaphragm is nominally centered
between upper and lower housings, and is inflated into
the blood chamber during ejection and vacuumed into
the air chamber during filling. The prototype domed
Artif Organs, Vol. 33, No. 4, 2009
336 F. SHU ET AL.
Diaphragm
Cap
A (BASELINE)
0 (BASELINE) A0 (BASELINE)
1 (RIBBED) A1
2 (Tangential inlet) A2
3 (Tangential inlet/outlet) A3
4
(PARALLEL A4
INLET/OUTLET)
diaphragms are molded of silicone rubber and are
flexed rather than stretched during operation. Conse-
quently, their time-varying contour is not as uniform
(congruent) as are the inflated diaphragms. However
the advantage, and motivation, for this design was to
reduce elastic stresses and risk of pull-out at the sealed
edges. Caps 1, 2, 3, and 4 are modified variations of the
baseline cap, intended to improve the hemodynamics.
Cap 1 incorporates two parallel ribs axially aligned
with the inlet and outlet ports, intended to guide the
flow and limit the contact of the diaphragm with
the housing. Cap 2 features a tangential inlet port,
intended to establish a primary vortex upon filling.
Caps 3 and 4 have both tangential inlets and outlets.
The ports of Cap 3 are diametrically opposed, while
the ports of Cap 4 are aligned on the same side of the
housing. These design options resulted in 10 potential
combinations that were studied; however, for practi-
cal purposes, only four representative combinations
are presented here.
Artif Organs, Vol. 33, No. 4, 2009
B (DOMED)
B0
FIG. 2. Matrix of cap/diaphragm comb-
B1 inations.
B2
B3
B4
Flow visualization techniques
Three flow visualization methods were used in this
investigation. Tracer particle visualization was used
to investigate the path lines of the flow field inside
the pump housing throughout the pump cycle. A dye
washout (also known as dye dilution) technique was
used to assess the overall residence, hence clearance
of blood through the pump chamber. Dye erosion
(also known as dye wash-off) was performed to vali-
date the surface washing caused by surface shear, not
easily visualized by the previous two modalities.
Tracer particle visualization employed neutrally
buoyant spherical fluorescent polystyrene micro-
spheres (Duke Science Corp., Palo Alto, CA, USA)
with a mean diameter of 33 mm to mark the flow field.
A monochromatic light sheet (approximately 1 mm
thick) generated by a solid state diode-pumped laser
(532 nm, Millennia II, Spectra-Physics Lasers, Inc.,
Mountain View, CA, USA), operated at 0.6 Watts, was
aimed through the pump housing to illuminate the
 SORBENT HEMODIALYSIS SYSTEM
area of interest. A high-resolution digital camera
(D70, Nikon Corp., Tokyo, Japan) was used to record
the trajectory of the particles. The exposure time was
between 10–50 ms, adjusted to optimize the image
quality and length of streaks. A long pass optical filter
(550 nm cut-on wavelength, Edmunds Optics, Bar-
rington, NJ, USA) was used to eliminate the boundary
glare and thus improve the contrast.
A dye washout technique was used to reveal
regions of stagnation or recirculation that may
potentially lead to blood deposition. An aqueous
solution of fluorescein disodium salt (C20H10O5Na2)
with a concentration of 5 ppm was used as fluores-
cent dye. Due to Stokes shift phenomenon, the dye
absorbs irradiated light at 532 nm and emits a
broadband longer wavelength. The same laser sheet
was used to illuminate the area of interest as the
tracer particles visualization method. A 2-mL bolus
of dye was injected into the inlet tubing of the pump
during the beginning of the filling phase. The same
long pass filter was used to eliminate background
laser illumination and boundary glare. A standard
DVD camcorder (VDR-D100, Panasonic, Secaucus,
NJ, USA) was used to record the flow pattern for
further analysis.
FIG. 3. Flow pattern in the baseline pump (A0) at various phases of the cycle. The flow direction is from bottom (inlet) to the top (outlet).
Figures (a), (b), (c), and (d) (top row) designate the filling phase, and (e), (f), (g), and (h) (bottom row) the ejection phase. Inset illustrates
approximate phase of the pump cycle for each figure.
337
The dye erosion experiments employed a paint
dot flow visualization technique, modified from the
method described by Merzkirch (4) and described by
our group previously (5) to evaluate the efficacy of
surface washing. Highly viscous droplets of pig-
mented paint (Alizarin Crimson, Winsor & Newton,
London, England) were applied on the interior of
pump caps using a quill-type spotter. The diameter of
the droplets was approximately 1 mm. After drying
overnight at ambient temperature, the pump was
gently assembled, connected into the flow loop, oper-
ated for 3.5 min, and disassembled immediately
thereafter. The duration was selected to assure thor-
ough, yet incomplete, dot erosion to facilitate com-
parison between cases. A high resolution digital
camera (D70, Nikon Corp.) was used to photograph
the residual droplets within the cap.
RESULTS
Tracer particle visualization
The flow patterns illustrated by the streaks of
tracer particles in the baseline pump are shown in
Fig. 3. Eight phases of the pump cycle are provided,
with the respective point in the volume–time graphs
Artif Organs, Vol. 33, No. 4, 2009
338
indicated in the inset of each figure. At the beginning
of filling (Fig. 3a), pneumatic vacuum draws fluid
into the pump chamber through the inlet and was
observed to immediately form a pair of counter-
rotating vortices. The vortices were initially confined
to a region adjacent to the inlet, and migrated down-
stream toward the center of the pump (Fig. 3b).
During the remainder of the filling phase, the vortex
grew to occupy the entire chamber. This resulted in a
sharp deviation at the distal wall causing retrograde
flow along the sidewalls of the chamber. Two regions
of disturbed flow were apparent near the outlet and
inlet, characterized by a sharp deflection of the par-
ticle path lines. The vortex pair also caused the retro-
grade flow along the sidewalls to collide with the
inflow jet. These collisions produced small-scale tur-
bulent structures (Fig. 3c,d), hence dissipating turbu-
lent kinetic energy.The loss of kinetic energy resulted
in decreased flow velocity at the end of filling. The
small turbulent structures dissipated gradually at the
beginning of ejection (Fig. 3e,f). Therefore, there
were fewer small-scale structures observed through-
out ejection. Near the end of ejection, the flow inside
the pump was almost laminar (Fig. 3g,h).
Pump B0 (baseline with domed diaphragm), A1,
and B1 (ribbed housings) exhibited similar flow
FIG. 4. Flow pattern in the tangential inlet pump (A2) at various phases of the cycle. Images are arranged as in Fig. 3.
Artif Organs, Vol. 33, No. 4, 2009
F. SHU ET AL.
behavior, thus not presented. The effect of the ribs
and domed diaphragm was overwhelmed by the flow
disturbance caused by the colliding flow streams.
To eliminate the disturbed regions induced by col-
lisions in the baseline pump, and improve confluence
of the flow during filling, a tangential inlet was
evaluated. The results corresponding to pump A2 are
presented in Fig. 4. At the beginning of filling, the
fluid velocity was relatively higher than observed in
B0—indicated by the length of the streaks. This was
attributed to the thin gap between the pump housing
and the diaphragm (Fig. 4a). After a very short
period of chaotic flow (Fig. 4b), the fluid in the pump
was observed to rotate clockwise. No small-scale tur-
bulent structures were observed. Consequently, the
accumulated kinetic energy of filling maintained
the rotation throughout the beginning of ejection
(Fig. 4e,f). Another beneficial property of the tangen-
tial inlet of pump A2 was found to be the smooth
transition of the flow velocity from filling to ejection
(Fig. 4e,f). At the end of ejection, the vortex was
rapidly interrupted, due to minimal residual volume.
(Fig. 4h).
Pumps A3 and A4 have identical tangential inlets
to A2, and therefore, the flow patterns during the
filling phase were virtually identical. Contrasted with
 SORBENT HEMODIALYSIS SYSTEM
FIG. 5. Flow pattern in the tangential inlet/outlet pump (A3) at various phases of the cycle. Images are arranged as in Fig. 3.
the coaxial outlet of A1 and A2, the outlets of A3 and
A4 are tangential, although oriented 180° from one
another. The ejection flow fields of A3 and A4 were
quite different from A2, however, very similar to one
another. Therefore, only results for pump A3 are pre-
sented (Fig. 5). Unlike A2, in which the flow con-
verged symmetrically through the outlet collector,
the tangential outlet caused an asymmetric fluid tra-
jectory (from upper right quadrant toward the outlet
in the figure). Because the main vortex was arrested
suddenly at the end of filling, the ejected volume was
initiated from rest; hence, there was no benefit
observed by locating the outlet tangentially. Overall,
the flow patterns throughout ejection in A3 and A4
were more chaotic than A2.
Figure 6 presents the flow patterns for pump B2,
featuring the domed diaphragm. The deformation of
the domed diaphragm was asymmetrical when com-
pared to the deformation of the flat one, and more
influenced by the flow. However, no apparent differ-
ence in the flow patterns was observed as compared
to the baseline (flat) diaphragm (e.g., pump A2).
Dye washout
Figures 7 and 8 provide representative images of
dye washout results for pumps A0 (baseline) and A2
339
(tangential inlet), respectively. The baseline pump
exhibited nearly instantaneous mixing of fluorescent
dye after injection—evidenced by the uniform
brightness of the illuminated region (Fig. 7f,g). This
demonstrates a high level of turbulent mixing as evi-
denced by the particle tracer results.
For tangential inlet pumps (A2–B4), there was less
mixing observed during filling. This was evidenced by
the non-uniformity of the dye concentration at the
end of filling and even at the beginning of ejection.
Bright streaks were observed throughout the cycle,
indicating a low level of turbulence intensity and
incomplete. This also agreed with the particle tracer
results.
For all the pumps tested, virtually all of the dye was
expelled from the pump at the end of ejection, with
the exception of the inlet region in which there
remained a small residual volume. This residual
volume mixed with the incoming fluid on the next
cycle and was almost imperceptible at the end of the
second cycle.
Dye erosion
Dye erosion results of the baseline pump are
shown in Fig. 9. The original paint dots are shown in
Fig. 9a, and Fig. 9b shows residual paint after the
Artif Organs, Vol. 33, No. 4, 2009
340
FIG. 6. Flow pattern in pump B2, with tangential inlet and domed diaphragm at various phases of the cycle. Images are arranged as in
Fig. 3. Images appear darker due to black coloration of the diaphragm.
pump was operated for 3.5 min. The lateral walls and
central keel of the pump chamber (regions 1, 2, and 3)
demonstrated greater surface washing than the inlet
diffuser region (region 6) and the regions corre-
sponding to the eyes of the two counter-rotating vor-
tices (regions 4 and 5) (Fig. 3b). The schematic of the
flow structure in this region is shown in Fig. 9c.
Results of pump A2 are presented in Fig. 10 to
illustrate the contrast between the pumps displaying
the greatest and least turbulence in the previous
studies. After 3.5 min of operation, almost all the
paint dots within pump A2 were washed off. An
exception was a small spot located in the inner side of
the inlet (circled in Fig. 10b) corresponding to the
likely zone of separation due to the excessive diffu-
sion angle of the inlet.
DISCUSSION
It is understood that blood trauma and thrombosis
are multifactorial phenomena, and are affected by
the fluid dynamics as well as surface chemistry and
condition of the blood. The current study presumed
that best practices will be utilized for material selec-
tion and anticoagulation. Therefore, the focus was
Artif Organs, Vol. 33, No. 4, 2009
F. SHU ET AL.
primarily on hemodynamic factors—both to limit or
eliminate the incidence of thrombosis and to mini-
mize the dosage requirement for heparin, thereby
reducing the associated adverse effects.
Tracer particle visualization has been used effec-
tively for obtaining qualitative appraisal of the flow
fields within a variety of blood-wetted devices such as
pulsatile pumps (6,7), heart valves (8–10), and rotary
pumps (5,11–13) using a variety of tracer particles
and imaging modalities. Several previous visual-
ization studies of pulsatile blood pumps have indi-
cated that unfavorable disturbed flow, recirculation,
separation, and insufficient “surface washing” can
potentially be responsible for thrombus formation
(7,10,12,14–20). Therefore, a common objective is to
assure that all surfaces are “washed” to prevent accu-
mulation of platelet agonists and reduce the resi-
dence time in which platelets may come in contact
with the biomaterial surface. Conversely, excessive
velocity may elevate shear stress, and/or cause turbu-
lence or flow disturbances that can also contribute to
thrombogenesis by activating platelets (9,21). The
three flow visualization techniques used in this study
aimed to evaluate, and minimize, these deleterious
flow features. The multimodal approach to flow visu-
 SORBENT HEMODIALYSIS SYSTEM
FIG. 7. Dye washout results of the baseline pump (A0) at various times in the pump cycle. Fluorescent dye was injected in the beginning
of filling. Figures (a), (b), (c), (d), (e), and (f) correspond to filling, and ejection corresponds to figures (g), (h), (i), (j), (k), and (l).
alization proved to be beneficial, both for cross-
validation and for providing more comprehensive
information about the hemodynamics.
The coaxial inlet design of the baseline pump
caused a great deal of chaotic flow due to the dis-
charge of the jet into the initially stagnant blood in
the chamber. As the jet propagated downstream, the
flow impinged on the outlet port, reversing direction
to form a pair of counter-rotating vortices. This ret-
rograde flow eventually collided with the incoming
jet. Overall, this was a rather suboptimal design.
The tangential inlet was therefore introduced to
minimize fluid dynamic disturbance during filling.
Fluid entering the pump chamber through the tan-
gential port established a single vortex, and rejoined
the inlet jet without disturbing the vortex. Elimina-
tion of flow disturbance or turbulence during the
filling phase allowed the accumulated kinetic energy
to maintain the velocity of the circumferential
341
motion (like a “fluid flywheel”) and thereby prolong-
ing the surface “washing” effect.
A tangential outlet was introduced with the inten-
tion of maintaining the vortex initiated during filling
throughout ejection. It was anticipated that aligning
the outlet port with the (anticipated) circumferen-
tial velocity field would thereby preserve the inertia.
Other blood pumps, such as the Novacor LVAS
have reported such a behavior (8). However, we
found that the vortex in pumps A3 and A4 was
damped abruptly at the end of filling, and therefore,
the fluid was accelerated from rest. Hence, the
tangential port presented more of a detriment by
arresting the vortex, and causing the fluid volume to
exit the chamber asymmetrically. The attenuation
of the vortex in pumps A3 and A4 with tangential
outlet compared to A2 with an axial outlet was
counterintuitive. This may be attributed to the rela-
tively spherical shape of the current device relative
Artif Organs, Vol. 33, No. 4, 2009
342
FIG. 8. Dye washout results of pump A2 at various times in the pump cycle. Fluorescent dye was injected in the beginning of filling.
Figures (a), (b), (c), (d), (e), and (f) correspond to filling, and ejection corresponds to figures (g), (h), (i), (j), (k), and (l).
to the flat “pancake” shape of the Novacor LVAS
pumping sac. The current results more closely
resemble the flow described within the Penn State
device (22) which also features a relatively spherical
fluid chamber. Another likely factor is the relatively
long pump cycle provided by the pneumatic drive
system, and the correspondingly long dwell time
between filling and ejection. Future attention to the
timing, phasing, and rates of filling and ejection are
likely to further improve the hemodynamics.
Improvements to the diaphragm were considered
due to the physical stretching that occurred in the flat
diaphragm of the baseline pump.A domed design was
introduced to replace stretching with bending, and to
alleviate the tension at the anchor locations.Although
the choice of diaphragm did not present any percep-
tible hemodynamic benefits, the domed shape of the
Artif Organs, Vol. 33, No. 4, 2009
F. SHU ET AL.
diaphragm B was determined to be preferable based
on its superior mechanical characteristics.
Dye washout (dilution) studies were conducted to
reveal regions of flow stasis, which would affect the
dwell time of blood within the pump. We had antici-
pated an incremental dilution and clearance of dye
over several cycles. However, there was no percep-
tible dye in any of the pumps after two cycles. This is
understandable due to the very small residual volume
at the end of ejection, and the absence of crevices
which could trap stagnant blood.
Dye erosion (erosion) studies were undertaken
to indirectly evaluate the shear stress on the
blood-contacting surfaces, hence evaluate “surface
washing.” This is a rather effective and efficient
method of evaluation because it does not require
elaborate interpretation of complex flow patterns,
 SORBENT HEMODIALYSIS SYSTEM
and most closely represents the potential of deposi-
tion due to inadequate washing. These results
also proved to be very sensitive at distinguishing
the differences between designs. The presence of
residual dye indicators was found to correspond to
regions in which low velocity was observed in the
particle tracking studies, such as the center of the
vortices. Likewise, regions in which the dye was
completely eroded corresponded to high surface
velocity, such as the periphery of the vortices. The
residual dye within the inlet diffuser also revealed
diminished surface shear, caused by boundary layer
separation, which was not readily observable by the
other techniques.
343
FIG. 9. Dye erosion results of the base-
line pump (A0). (a) Dye droplets deposited
prior to pumping. (b) Residual dye droplets
after 3.5 min of operation. Well-washed
regions (e.g., 1, 2, and 3) found inter-
spersed with poorly washed regions (e.g.,
5 and 6). (c) Separation zones responsible
for poor surfaced washing.
CONCLUSION
Rational design of blood-wetted devices requires
careful consideration of shear-induced trauma
(hemolysis) and activation of blood elements (throm-
bus) (23).
This has proven to be a difficult undertaking,
requiring several decades of evolution.
The safety and efficacy of the devices currently in
clinical use have evolved over several decades, ben-
efiting from a great amount of research and incre-
mental improvements. Indeed, the design of pulsatile
pumps dates to the origins of artificial heart develop-
ment (24). In this study, we attempted to optimize
Artif Organs, Vol. 33, No. 4, 2009
344 F. SHU ET AL.
a pump design in a relatively short time frame.
Geometric modifications to the baseline design
were introduced primarily based on intuition and
experience.
Due to the immature design of the initial baseline
pump, it was relatively easy to introduce dramatic
improvements. Based on the results, a pump with
tangential inlet, coaxial outlet, and a domed dia-
phragm (Pump B2) offered the best performance.
However, others have shown that there are many
minor nuances to the design that are important for
optimizing flow, and hence reducing risk of thrombo-
sis (22). Due to the vast selection of design param-
eters, it is impractical to explore all possible design
variations and combinations. Accordingly, it is
unlikely that the final pump in this series is in fact the
“optimal” design, and additional design iterations
would most probably yield improved results. Numeri-
cal optimization may facilitate this process by replac-
ing time-consuming experiments with computational
fluid dynamics (CFD) simulation and “shape optimi-
zation” techniques (25–29). CFD can also provide a
more quantitative and elegant tool to optimize the
pump geometry. However, CFD simulation of pulsa-
tile blood pumps is particularly challenging due to
dynamic interaction between the fluid and the
moving, flexible diaphragm (30,31). Therefore, flow
visualization experiments, combined with rapid pro-
Artif Organs, Vol. 33, No. 4, 2009
FIG. 10. Dye erosion results of pump A2.
(a) Dye droplets deposited prior to
pumping. (b) Residual dye droplets after
3.5 min of operation. Residue found in two
small regions: near the inlet diffuser and
outlet collector.
totyping, offers a pragmatic approach to rapidly
identify potential problems and gain dramatic
improvements to the hemodynamics.
REFERENCES
1. Ash SR. The allient dialysis system. Semin Dial 2004;17:164–6.
2. Tate LJ, Kameneva MV, Snyder T, Marascalco P, Marhefka J.
An innovative blood pumping device for use in hemodialysis.
J Am Soc Nephrol 2004;15:150A.
3. Ash SR, Carr DJ. Coagulation test method for extracorporeal
blood circuits. ASAIO J 2005;51:44A.
4. Merzkirch W. Flow Visualization, 2nd Edition. New York: Aca-
demic Press, 1987.
5. Wu ZJ, Antaki JF, Burgreen GW, Butler KC, Thomas DC,
Griffith BP. Fluid dynamic characterization of operating con-
ditions for continuous flow blood pumps. ASAIO J 1999;45:
442–9.
6. Hochareon P, Manning KB, Fontaine AA, Deutsch S, Tarbell
JM. Diaphragm motion affects flow patterns in an artificial
heart. Artif Organs 2003;27:1102–9.
7. Hochareon P, Manning KB, Fontaine AA, Tarbell JM,
Deutsch S. Fluid dynamic analysis of the 50 cc Penn State
artificial heart under physiological operating conditions using
particle image velocimetry. J Biomech Eng 2004;126:585–93.
8. Woodard JC, Shaffer FD, Schaub RD, Lund LW, Borovetz HS.
Optimal management of a ventricular assist system. Contribu-
tion of flow visualization studies. ASAIO J 1992;38:M216–9.
9. Liu JS, Lu PC, Chu SH. Turbulence characteristics down-
stream of bileaflet aortic valve prostheses. J Biomech Eng
2000;122:118–24.
10. Manning KB, Przybysz TM, Fontaine AA, Tarbell JM, Deutsch
S. Near field flow characteristics of the Bjork–Shiley Monos-
trut valve in a modified single shot valve chamber. ASAIO J
2005;51:133–8.
 SORBENT HEMODIALYSIS SYSTEM
11. Kerrigan JP, Yamazaki K, Meyer RK, et al. High-resolution
fluorescent particle-tracking flow visualization within an intra-
ventricular axial flow left ventricular assist device. Artif
Organs 1996;20:534–40.
12. Manning KB, Miller GE. Flow through an outlet cannula of
a rotary ventricular assist device. Artif Organs 2002;26:714–
23.
13. Song X, Wood HG, Allaire PE, Antaki JF, Olsen DB. Inlet and
outlet devices for rotary blood pumps. Artif Organs 2004;28:
911–5.
14. Hochareon P, Manning KB, Fontaine AA, Tarbell JM,
Deutsch S. Wall shear-rate estimation within the 50 cc Penn
State artificial heart using particle image velocimetry. J
Biomech Eng 2004;126:430–7.
15. Karino T, Goldsmith HL. Adhesion of human platelets to
collagen on the walls distal to a tubular expansion. Microvasc
Res 1979;17:238–62.
16. Karino T, Goldsmith HL. Aggregation of human platelets in
an annular vortex distal to a tubular expansion. Microvasc Res
1979;17:217–37.
17. Mandrusov E. Blood plasma and platelet reactions in shear
and separated flows: mechanism for thrombogenesis on
prewetted artificial surfaces [Doctor of Philosophy]: Columbia
University, 1997.
18. Schultz JS, Lindenauer SM, Penner JA, Barenberg S. Determi-
nants of thrombus formation on surfaces. Trans Am Soc Artif
Intern Organs 1980;26:279–83.
19. Mandrusov E, Puszkin E, Vroman L, Leonard EF. Separated
flows in artificial organs. A cause of early thrombogenesis?
ASAIO J 1996;42:M506–13.
20. Manning KB, Kini V, Fontaine AA, Deutsch S, Tarbell JM.
Regurgitant flow field characteristics of the St. Jude bileaflet
mechanical heart valve under physiologic pulsatile flow using
particle image velocimetry. Artif Organs 2003;27:840–6.
345
21. Tamagawa M, Akamatsu T, Saitoh K. Prodiction of hemolysis
in turbulent shear orifice flow. Artif Organs 1996;20:553–9.
22. Deutsch S, Tarbell JM, Manning KB, Resenburg G, Fontaine
AA. Experimental fluid mechanics of pulsatile artificial blood
pumps. Annu Rev Fluid Mech 2006;38:65–86.
23. Leonard EF. Principles of cardiovascular device design.
Cardiovasc Pathol 1993;2(Suppl):3–10.
24. Liotta D, Crawford ES, Cooley DA, DeBarkey ME, de Urquia
M, Feldman L. Prolonged partial left ventricular bypass by
means of an intrathoracic pump implanted in the left chest.
Trans Am Soc Artif Intern Organs 1962;8:90–9.
25. Antaki JF, Ghattas O, Burgreen GW, He B. Computational
flow optimization of rotary blood pump components. Artif
Organs 1995;19:608–15.
26. Burgreen GW, Antaki JF, Griffith BP. A design improvement
strategy for axial blood pumps using computational fluid
dynamics. ASAIO J 1996;42:M354–60.
27. Burgreen GW, Antaki JF, Wu J, le Blanc P, Butler KC. A
computational and experimental comparison of two outlet
stators for the Nimbus LVAD. Left ventricular assist device.
ASAIO J 1999;45:328–33.
28. Hund S, Antaki JF. Improving the methodologies for shape
optimization in the medical device industry. 1st International
Conference on e-Design. Atlanta, GA, May 14, 2005.
29. Hund S, Antaki JF. Improved methodologies for the e-design
of medical devices. 1st Internation Conference on e-Design.
Atlanta, GA, May 14, 2005.
30. Avrahami I, Rosenfeld M, Raz S, Einav S. Numerical model of
flow in a sac-type ventricular assist device. Artif Organs
2006;30:529–38.
31. Medvitz RB, Kreider JW, Manning KB, Fontaine AA, Deutsch
S, Paterson EG. Development and validation of a computa-
tional fluid dynamics methodology for simulation of pulsatile
left ventricular assist devices. ASAIO J 2007;53:122–31.
Artif Organs, Vol. 33, No. 4, 2009