A High Performance Pulsatile Pump For Aortic Flow Experiments in 3-Dimensional Models

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Cardiovascular Engineering and Technology, Vol. 7, No. 2, June 2016 (Ó 2016) pp.

148–158
DOI: 10.1007/s13239-016-0260-3

A High Performance Pulsatile Pump for Aortic Flow Experiments


in 3-Dimensional Models
RAFEED A. CHAUDHURY ,1 VICTOR ATLASMAN,2 GIRISH PATHANGEY,1 NICHOLAS PRACHT,1
RONALD J. ADRIAN,3 and DAVID H. FRAKES1,2
1
School of Biological and Health Systems Engineering, Arizona State University, Tempe, AZ, USA; 2School of Electrical,
Computer, and Energy Engineering, Arizona State University, Tempe, AZ, USA; and 3School for Engineering of Matter,
Transport and Energy, Arizona State University, Tempe, AZ, USA
(Received 10 September 2015; accepted 3 March 2016; published online 16 March 2016)

Abstract—Aortic pathologies such as coarctation, dissection, morbidity and mortality worldwide.19 The most com-
and aneurysm represent a particularly emergent class of mon pathologies affected include atheromatous aortic
cardiovascular diseases. Computational simulations of aortic disease, aortic coarctation, dissection, and abdominal
flows are growing increasingly important as tools for gaining
understanding of these pathologies, as well as for planning and thoracic aneurysm. Available treatment options
their surgical repair. In vitro experiments are required to include deployment of medical devices and surgical
validate the simulations against real world data, and the alteration to modify blood flow. However, the out-
experiments require a pulsatile flow pump system that can comes of these procedures are often uncertain a priori.
provide physiologic flow conditions characteristic of the Effective tools for predicting post-operative outcomes
aorta. We designed a newly capable piston-based pulsatile
flow pump system that can generate high volume flow rates on a pre-operative basis would thus have great
(850 mL/s), replicate physiologic waveforms, and pump high potential to increase interventional success rates. Gi-
viscosity fluids against large impedances. The system is also ven the clinical timescale associated with treating
compatible with a broad range of fluid types, and is operable aortic pathologies, the most suitable foundation upon
in magnetic resonance imaging environments. Performance which to build such tools is blood flow simulation.
of the system was validated using image processing-based
analysis of piston motion as well as particle image velocime- Although blood flow simulation is far from ubiqui-
try. The new system represents a more capable pumping tous in the clinic, work in the field is maturing at a
solution for aortic flow experiments than other available rapid and accelerating pace.2,3,8,37,39
designs, and can be manufactured at a relatively low cost. Hemodynamic flow simulations are often performed
through computational fluid dynamics and/or direct
Keywords—Piston pump, Flow loop, Pulsatile flow, Blood numerical simulations. The approaches can make sim-
flow, Aorta, Physiological waveform, Heart valves. plifying assumptions on the system dynamics and/or
parameters that do not represent in vivo or even in vitro
conditions well. This is particularly true for simulations
INTRODUCTION of high Reynolds numbers (between Re ¼ 9400 and
10,000) flows such as those found in the aorta. Experi-
The American Heart Association recently reported
mental validation against real world in vitro data is thus a
that an estimated 85.6 million Americans (>1 in 3
critical step toward the development of more capable
adults) suffer from cardiovascular disease (CVD).29
simulation tools. Real world experimental measurements
CVD accounts for 17.3 million deaths per year glob-
of in vitro aortic flows can be made using particle image
ally, a number that is expected to grow to >23.6
velocimetry (PIV), laser Doppler anemometry (LDA),
million by 2030.1 Further, the total direct medical costs
phase contrast magnetic resonance imaging (PC-MRI),
of CVD in the United States are projected to triple
and echocardiography, among others.20 Of course taking
from $272.5 to $818.1 billion over the same period.16
such measurements requires a pulsatile pump system
Aortic diseases account for significant cardiovascular
capable of generating flows characteristic of the aorta.
Several groups have developed flow pump systems to
Address correspondence to Rafeed A. Chaudhury, School of
replicate physiologic flows, but these are limited in terms
Biological and Health Systems Engineering, Arizona State University, of accuracy, power, practicality, and/or cost in the con-
Tempe, AZ, USA. Electronic mail: [email protected] text of replicating aortic waveforms.12,14,15,17,26,32,35,38
148
1869-408X/16/0600-0148/0 Ó 2016 Biomedical Engineering Society
A High Performance Pulsatile Pump 149

The dual-chamber pump that uses a reversing cylinder- pump to drive the steady component and then using a
piston to generate pulsatile flow by Frayne and Holds- T-connector or similar to combine the flows), which
worth et al. was capable flow rates up to 30 mL/s and had creates bias when studying aortic flows. Here, ‘‘exter-
a usable stroke volume of 180 mL.14,17 The computer- nal turbulence’’ refers to turbulence generated by finite
controlled progressive cavity pump by Eriksson et al. had disturbances in the system that would not otherwise be
a 250 mL/s peak flow rate.12 Tsai and Savaş developed a present in a system with a clean outlet chamber design.
computer-controlled pump that combined aspects of The commercially available pumps mentioned pre-
both a gear pump to generate the mean component and a viously use valves and flow connector orientations
piston pump to generate the oscillatory component.35 which fall short of designs needed to achieve low tur-
This design allowed for a 300 mL/s throughput, but the bulence intensity in fluid dynamic experiments. For
combination of the two types of pumps created multiple example, in the ViVitro SuperPump and Harvard
sources of disturbance in the flow loop.35 Yilmaz et al. Apparatus 1400 Series blood flow pumps, the non-
used a centrifugal pump with pneumatic and solenoid streamlined designs of the outlet chambers would
valves to generate flow.38 Although the pump was able to clearly produce flow disturbances of greater intensity
output 900 mL/s, it took 10 s to accelerate to that flow than the streamlined design of our outlet chamber, as
rate. Mock circulatory loops, such as the Donovan mock compared in Fig. 1. Specifically, a ball valve in the flow
circulatory system and other in-house built systems, at- connector region is used in the pump head outlet
tempt to recreate the systemic and pulmonary circula- chamber of the Harvard Apparatus pump. The foreign
tions of the normal human for in vitro evaluation and body, in this case a ball, in the main flow channel
testing of artificial heart systems, including total artificial would redirect the flow path around the ball instead of
hearts and ventricular assist devices.9–11,34 achieving a streamlined fluid entry with no barriers to
Commercially available piston driven pumps, such as overcome. The ViVitro SuperPump has a large cham-
the ViVitro SuperPump (ViVitro Labs, Inc., Victoria, ber, but a small non-streamlined outlet orifice, which
BC), CardioFlow 5000 MR (Shelley Medical Imaging would produce more biased valve turbulence mea-
Technologies, London, ON), and the 1400 Series Har- surements downstream. The sketch of the SuperPump
vard Apparatus (Harvard Apparatus, Holliston, MA) presented in Fig. 1 is a typical output interface con-
pulsatile blood pump, among others, have been used to figuration used during experimentation. Other pump
create pulsatile flow in arterial flow experiments in head orientations and/or valve combinations may be
studies several studies.3,18,21,27,30,32 The ViVitro Super- also used. Additionally, the CardioFlow 5000 MR has
Pump combines a linear actuator and piston to replicate similar valve and interface features as the SuperPump
an input waveform. While the pump mimics a compliant and 1400 Series blood flow pumps that are not optimal
elastic model of the heart, replicating a specific arterial for achieving low turbulence intensity in fluid dynamic
waveform requires adjusting the downstream im- experiments.
pedance and is limited by a maximum stroke volume of Thus, there is a need for high performance piston
150 mL.36 The CardioFlow 5000 MR is a programmable pumps that produce high Reynolds number flows and
pump used to generate periodic, reproducible flow have low inlet disturbances. Such pumps are particu-
profiles at anatomically relevant velocities up to 300 mL/ larly important for high-resolution heart valve turbu-
s while using 5/8 in inner diameter tubing connectors.31 lence measurements. With a low disturbance system
The 1400 Series Harvard Apparatus pulsatile blood and heart valve adapters placed in the pump head, the
pump is a single ventricle simulator designed for a turbulence contributions made by the valve can be
maximum output of 167 mL/s and was not fully auto- further isolated. The disturbance level refers to un-
mated, requiring the operator to manually adjust each wanted fluctuations in time and/or distortions of the
parameter at the pump.15,32 mean velocity profile from axisymmetric profile near
Notably, these pumps cannot generate flows at the the inlet or the parabolic profile downstream in a
high end of the physiologic Reynolds number range, straight section. Temporal fluctuations are caused by
which is between Re ¼ 9400 and 10,000 found in the fluctuations in the flow rate produced by the pump,
aorta. Most physiological pump systems are unable to such as periodic variations caused by the teeth in a gear
approach Re ¼ 5000 without exacerbating practicality. motor; fluctuating piston velocities caused by caused
Commercial pumps only output a maximum flow rate by the mechanical driver or slip/stick motion caused by
between 100 and 300 mL/s, perhaps due to their the friction between the piston and the cylinder wall;
inability to ensure safety during use. Furthermore, the and/or unsteady flow disturbances produced by the
pumps that are able to generate over 300 mL/s intro- exit flow design, such as complex outlet flow geometry
duce external turbulence into their flow by combining or turbulence caused by mixing two streams. The
flow sources (e.g., simultaneously using a piston pump present pump attempts to minimize the effects by using
to drive the pulsatile component while using a gear an electric cylinder that provides very smooth motion;
150 R. A. CHAUDHURY et al.

FIGURE 1. Simplified comparison of outlet chamber designs and flow patterns between the ViVitro SuperPump (left), Harvard
Apparatus 1400 Series Blood Pumps (center), and the newly proposed pump (right).

low friction between the pump wall and the piston; and METHODS AND MATERIALS
a smoothly contracting, axisymmetric exit from the
pump. The value of the head loss coefficient is Pump Design
indicative of the smoothness of the exit flow (the Force Requirements
smaller the value of the loss coefficient, the lower the
disturbance in the inlet region). Typically, a flow sys- A force analysis of a piston based flow loop, de-
tem is considered to be ‘‘low noise’’ if the turbulence picted in Fig. 2a, was performed to obtain force and
intensity (root mean square (RMS) of the velocity energy balance equations, which were used to deter-
fluctuation at the test section divided by the mean) is mine the total piston force output required. Applying
small. In the present pump the measured turbulence Newton’s second law yields
intensity is <2 %. The most compelling evidence of Pa Ap þ Fp  Pp Ap  Ff ¼ mp V_ p ; ð1Þ
low noise is the high value of the Reynolds number
(8000) at which the downstream flow transitions from where Pa is the atmospheric pressure, Ap is the area of
laminar to turbulent flow compared to transition for the piston, Fp is the force applied on the piston, Pp is
Reynolds number between 1800 and 2300 that char- the pressure applied on the piston, Ff is the friction
acterizes ‘normal’ pipe flows. This ‘delayed’ transition force, mp is the mass of the piston, and V_ p is the
is a known effect in systems that have inlet flow dis- acceleration of the piston. The energy equation for
turbances lower than ‘normal’. These advancements in incompressible flow is
heart valve research would also make the replication of
physiological turbulence in in vitro systems more pos- V22 P2 V2 P3
þ þ z2 ¼ 3 þ þ z3 þ hL þ V_ 2 L12 : ð2Þ
sible. Here, we propose a new computer programmable 2g qg 2g qg
piston pump that produces physiologically realistic,
accurate, and reproducible aortic waveforms at a low
Here, V2 and V3 are the velocities at locations 2 and
cost.
3 in Fig. 2a, respectively, g is the acceleration due to
We designed the new piston-based pulsatile flow
gravity, P2 and P3 are the pressures at locations 2 and
pump system to meet several application needs. Those
3, q is the density of the working fluid, z2 and z3 are the
were to: (1) generate high volume flow rates like those
height at locations (L) 2 and 3, while hL is the head loss
in the human aorta; (2) allow for custom input of and
replicate physiologic waveforms; (3) generate very of the system. V_ 2 L12 is the acceleration over the
small inlet disturbances so as to enable experiments on streamline from location 1 to 2. Location 1 is area of
transition from laminar to turbulent flow; (4) be the pump chamber where the piston travels, location 2
operable in MRI environments; (5) be capable of is the area where the fluid travels through the smaller
pumping fluids with high viscosities (e.g., 4 cP) against diameter tube, and location 3 is the area where the
large impedances depending on tubing length; (6) be fluid enters the reservoir. Equation (2) then simplifies
chemically compatible with different fluids and corro- to
sion resistant. In this technical note, we describe the  
P2  Pa L V22 V22 _ Ap
design and production of a pulsatile flow pump system þ ðz2  z3 Þ ¼ f þ Kv þ Vp L12 ;
qg D 2g 2g A2
that meets all of these application needs. Validation
was accomplished by testing system performance with ð3Þ
respect to each need, including image processing-based given that z2 ¼ z1 , P3 ¼ Pa , and V3 ¼ V2 . The head
analysis of piston motion and PIV-based measurement loss term is expanded to include the friction coefficient,
of flow waveforms. f, total length of tubing, L, inner diameter of tubing, D,
A High Performance Pulsatile Pump 151

(a)

(b)

FIGURE 2. (a) Simplified drawing of a piston-based flow loop system. (b) Free body diagram of a piston with all forces acting on it.
P denotes pressure, F denotes force, and A denotes area. The subscripts a, p, and f denote atmosphere, piston, and friction,
respectively.

and the head loss coefficient, Km . For the purposes of Simplifying Eq. (7) further and solving for Fp , the
the pump system described in this paper, the friction force required for the piston to push all of the fluid
coefficient between plastic O-rings and an acrylic through the system, we find that
cylinder was estimated to be 0.04 while the loss coef-
ficient for a well rounded entrance was also estimated Fp ¼ ðmp þ mc ÞV_ p þ qQðV2  Vp Þ
   2 
to be 0.04. Thus, Eq. (3) becomes L V2
  þ A2 Pa þ f þ Kv q  qgðz2  z3 Þ
D 2 ð8Þ
L V22 q Ap  
P2  Pa ¼ f  qgðz2  z3 Þ þ V_ p L12 ; ð4Þ Ap
D 2 A2  Pa Ap þ Ff þ V_ p L12 :
A2
where P2  Pa is the gauge pressure and q is the den-
sity of the working fluid. A free body diagram of the Here, mp is the mass of the piston, mc is the mass of the
piston is shown in Fig. 2b. The net force acting on the fluid in the cylinder, V_ p is the acceleration of the pis-
piston, FNet , is ton, where Vp is the piston velocity, and Q is the
maximum rate of flow in the cylinder. Substituting for
FNet ¼ Fp þ Pa Ap  Pp Ap
ð5Þ V2  Vp , Q, and V22 using conservation of mass, the
¼ Fp þ ðPa  Pp ÞAp : piston force required is
 
By treating the area in which the piston moves as a Ap
Fp ¼ ðmp þ mc ÞV_ p þ q  1 Ap V2p þ
control volume and applying conservation of A2
2   3
momentum, we arrive at   V2 Ap 2
Z Z 6 L p A 2 7
X @ A2 4Pa þ f þ Kv q  qgðz2  z3 Þ5 ð9Þ
F¼ qVdV þ VqVdA; D 2
ð6Þ
C:V:
@t C:V: C:S:
 
_ Ap
where t is time, V is the volume, A is the area, and the  Pa Ap þ Ff þ Vp L12 :
A2
subscripts C.V. and C.S. denote the control volume
and control surface, respectively. Solving for Pp Ap in
Eqs. (1) and (5), we can show that System Design

Pp Ap  Ff  mp V_ p ¼ Pp Ap ¼ qQðV2  Vp Þ The computer controlled piston pump system we


  designed is shown in Fig. 3a. It consists of a cylindrical
Ap ð7Þ
_
þ P2 A2 þ q _ Vp þ Vp_ L12 : acrylic chamber and pump head, an electric cylinder,
A2
152 R. A. CHAUDHURY et al.

FIGURE 3. (a) Photograph of pump with major components labeled. (b) Design of the pump head entrance to the tube.

and a servo amplifier/drive mounted on an aluminum determined by the system response to achieve the de-
base plate. The chamber houses a 57.0 mm diameter sired volumetric flow and velocity.
piston having a 420 mm stroke actuated by the electric The pump head, shown in Fig. 3b, was connected to
cylinder (N2-AKM23D-BNC102B-18, Danaher Mo- the long pump chamber. The smooth exit from the
tion, Washington, DC) controlled by an analog signal pump is designed to produce smooth (low disturbance)
input to the servo amplifier/drive (AKD-B00306- flow in a smaller diameter tubing. The 6.35 mm radius
NAAN, Danaher Motion, Washington, DC). The entrance of the pump head provides for low distur-
electric cylinder has a ballscrew, which utilizes ballnuts bances in the entry flow, resulting in a smaller head
riding on recirculating ball bearings resulting in higher loss coefficient. A cylindrical 3.175 mm thick rubber
speeds (76 cm/s), loads (1220 N), and cycle rates gasket fitted between the head and chamber prevents
(100 % duty). The electric cylinder reaches maximum air pockets and leaks. The head houses a 25.4 mm
output when running the servo drive on three phase outer diameter (O.D.) glass pipe used as the inlet to the
power, at 240 VAC. test section. It was sealed with three 2.4 mm width,
Analog signal inputs were generated by computer 21.8 mm inner diameter (I.D.) nitrile rubber (NBR) o-
using LabVIEW software (National Instruments, rings spaced 25.4 mm apart.
Austin, TX) and sent to a multiple channel analog If low disturbance level entry flow is not needed, the
output data acquisition (DAQ) board (NI PCIe-6351, mounting arrangement or the tube can be replaced by
National Instruments, Austin, TX) to generate syn- a simple barbed hose connector, enabling the use of
chronized waveforms for analog position control and a flexible tubing such as Tygon. Custom pump head
trigger signal for MRI or PIV image acquisition. The adapters can be inserted before the pump head to
signals for a single cycle were discretized into 1000 change characteristic input flow. Experimentation with
samples per second and output at the necessary rate heart valves, such as the bileaflet valve shown in
A High Performance Pulsatile Pump 153

FIGURE 4. (a) Front view of bileaflet heart valve adapter used between pump head and chamber. (b) Side view of bileaflet heart
valve adapter.

Figs. 4a and 4b, is a prime use case that the custom tank. A one-way flow valve capable of withstanding
pump adapters can address. In the example from 250 lbs of force was placed in the circuit between the
Fig. 4a, a model of a 22 mm St. Jude Regent bileaflet reservoir and the pump chamber to prevent flow into
mechanical heart valve was fitted into the adapter. The the reservoir during forward piston motion. Following
adapter and other components of the pulsatile pump forward pumping, solution in the reservoir was drawn
system were built using corrosion resistant materials back into the pump cylinder by reversing the piston
(i.e., acrylic, stainless steel, and glass), which allows motion. A second one-way flow valve was installed in
water, glycerin, sodium iodide, and other chemicals the test section-reservoir line to prevent back-flow
and mixtures to be used without damaging the system. during reverse piston movement.

Piston Motion Validation


Experimental Flow Loop
Image processing based estimation of piston veloc-
Figure 5a shows a schematic of the flow loop sys-
ity and output flow rate measurements were used to
tem, which consists of three major components: the
evaluate pump system performance for various un-
piston pump, a urethane block phantom/test section,
steady and steady flow conditions. Two different
and a reservoir tank to receive the flow, while Fig. 5b
ascending aortic waveforms were specified to validate
depicts the control circuit logic. The pump, the test
the pump input and output velocities. The first wave-
section, and the 8 L reservoir plus various connecting
form, henceforth referred to as Waveform 1, was based
tubes form a closed-loop system. The fluid circulated
on phase-contrast MRI data and had a cardiac cycle of
through the loop was a sodium iodide-based solution
0.7 s.33 The second waveform, henceforth referred to
(m ¼ 2:05  106 m2 s1 ) with a refractive index mat-
as Waveform 2, was based on Fourier harmonics from
ched to the urethane block model (n ¼ 1:49). The vis-
literature and had a cardiac cycle of 0.8 s.28
cosity of the solution was measured using a Cannon-
Fenske routine viscometer. The sodium iodide-based
Estimation of Piston Velocity
solution rendered the interior urethane walls invisible
and eliminated optical distortion, allowing optical ac- A high-speed camera, triggered using a frame
cess for flow visualization and PIV measurements. grabber, was used to monitor piston movements. The
The test section was a 19.0 mm diameter 304 mm images were collected at 260 frames per second at a
long cylindrical hole through a 355 mm block of cast resolution of 640  480 pixels, and were then written
polyurethane created using methods previously pub- into a National Instruments technical data manage-
lished.5–7 The inlet end of the test section was con- ment streaming (NI TDMS) file. The images were later
nected to the pump by a 52.5D long, 19.05 mm I.D. extracted from the TDMS file into separate time or-
glass tube, and the outlet end was connected to a 16D dered portable networks graphics (PNG) images and
long 19.05 mm I.D. glass tube that transitioned to 32D imported into MATLAB for processing. A single
of 19 mm I.D. Tygon tubing leading to the reservoir image frame cropped to the piston area was used as the
154 R. A. CHAUDHURY et al.

(a)

(b)

FIGURE 5. (a) Flow loop schematic. Arrows represent flow direction. (b) Diagram of control circuit logic.

and total flow rate output were calculated based on


conservation of mass. The total output volumetric flow
error between the prescribed input and measured output
was obtained for flow rates up to 850 mL/s with stroke
times up to 120 s and an uncertainty of 0.5 mL.

RESULTS AND DISCUSSION

While the main purpose in constructing this pump


was to produce a low disturbance inlet flow, the cost of
materials and invoiced machine shop time used in
building the pump was less than 1/4 that of a typical
commercial pump. The pump system achieved a
FIGURE 6. Piston motion estimation is accomplished by maximum instantaneous peak flow rate and velocity of
tracking a cropped template image from frame to frame. The
box around the piston indicates the template image used for 850 mL/s and 300 cm/s, respectively. Such high peak
tracking. flow rates will enable studies for exercise situations
where the cardiac output can be as high as 40 L/min in
template for piston tracking, as shown in Fig. 6.13 athletes. At a Reynolds number of 10,000 the RMS of
Cross correlation was used to find the displacement of turbulence fluctuation at x=D ¼ 55 was less than 2 %,
the template between image frames. The product of the measured using stereo PIV at the downstream location
pixel displacement between frames and the image scale in a straight urethane phantom model. Figure 7 shows
(inches/pixel), divided by the time between each frame, a linear relationship between the input voltage and the
produced the motion estimated instantaneous piston output piston velocity and volumetric flow rate. The
velocity over time. R2 value for each linear fit to the data points is 1.00.
Though not shown in Fig. 7, the linear trend continues
Output Flow Measurements beyond a 4 V input up to 10 V. Velocities can be fur-
For steady flows, the volume flow rate was checked by ther scaled by adjusting the revolutions per minute per
integrating the velocity profile at a downstream location volt in the pump software. This linearity is useful as it
(55D) and comparing the results to output flow mea- simplifies rescaling. The measurements were obtained
surements obtained with a bucket and stopwatch, using the bucket and stopwatch technique and image
wherein a bucket is used in place of the reservoir at the processing method described earlier. The total output
outlet and flow into the bucket is timed with a stop- volumetric flow error between the prescribed input and
watch. The input voltage to the piston was converted measured output was less than 0.5 % for flow rates up
into revolutions per minute per volt, and using a revo- to 850 mL/s. Error bars are not shown because they
lutions per inch scaling, theoretical piston displacement are too small to be appreciated.
A High Performance Pulsatile Pump 155

(a)
8

Velocity (cm/s)
6

0
0.0 0.5 1.0 1.5
Time (s)

FIGURE 7. Steady state response calibration curves for in- (b)


put voltage versus piston velocity output (circles) and volu- 0.08
metric flow rate output (triangles).

RMS Error
0.06
As shown in Figs. 8a and 9, the major features of
the pump’s output (including peak flow velocity) 0.04
matched well with the features present in the two
physiologic input waveforms that we tested (Waveform 0.02
1 and Waveform 2 as described in the methods sec-
tion). The piston output for Waveform 1 was almost 0.00
identical to the input. The R2 value, or coefficient of 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
determination, for the entirety of Waveform 1 was Time (s)
greater than 0.999. Figure 8b depicts the cycle-to-cycle
piston output variation as a function of time charac- FIGURE 8. (a) Comparison of input (solid) and output (dot-
ted) piston velocities for an ascending aortic waveform
terized by the RMS error between each cycle and their (Waveform 1) with time relative to start of the cardiac cycle.
mean using six sample cycles. The numerical value of Waveform 1 is representative of an aortic waveform with a
the RMS error integrated over the cycle is 2.68 % of similar mean output flow rate found in healthy and unhealthy
individuals under resting conditions and is obtained through
the time averaged mean velocity of the pulse history. validating the piston motion using image processing tech-
Shown in Fig. 9 is an extreme aortic waveform with a niques. The piston output for Waveform 1 was almost identi-
velocity scale 3.75 times higher than that shown in cal to the input. The R 2 value for the entirety of Waveform 1
was greater than 0.999. (b) Cycle-to-cycle piston output vari-
Fig. 8a, which stretches the pump system to its limits. ation as a function of time (dashed) characterized by the root
Although there is a lag between the input and output, mean square (RMS) error between each cycle using six sam-
ple cycles. The numerical value of the RMS error integrated
the R2 values for the time-shifted systolic and diastolic over the cycle (solid) is 2.68 % of the time averaged mean
curves for Waveform 2 were 0.966 and 0.974, respec- velocity of the pulse history.
tively.
Figure 10 compares the first piston input waveform
to the centerline velocity 55D downstream in an aortic multiscale model, which has been used and referenced
phantom model measured with PIV. The measured many times.22–25 In such a case, the heart function is
downstream pump output closely followed the piston provided by the pump and the downstream impedance
input waveform during systole, but deviated during the of the circulation is provided by the Windkessel model.
diastolic phase of the cardiac cycle. The discrepancy in Similar to most piston-based pumps, there are limi-
pump input and output is mostly due to impedance tations. Our design does not allow for continuous non-
from the system and flow loop. There is a need for zero flow output as the piston must be pulled back at the
improved waveform matching and propose a solution end of each stroke. However, the number of beats per
based on linear and time invariant (LTI) theory to stroke varies based on the input waveform used and is
predict and adjust the input waveform as necessary to between 20 and 30 beats for most aortic flows. Wave-
obtain the desired downstream flow output.4,7 forms with a significant backflow component cannot be
In our experiments, which use rigid models for PIV replicated because a negative pressure is accumulated
studies, the characteristic impedance is not the same as per stroke due to the current design, which includes one-
for a mock circulatory loop that uses elastic tubing and way valves. This limitation is easily circumvented by
downstream resistances. We note that it is possible to removing the one-way valves and implementing auto-
combine both a piston pump and a Windkessel-based matic flow control valves that open and close with the
156 R. A. CHAUDHURY et al.

35 the MRI and inside the control room. This experience,


while limited, supports compatibility with MRI tech-
nology and indicates that the pump motor, controller,
Velocity (cm/s)

25 and wires are properly shielded.

15
CONCLUSIONS

Aortic flows are of great interest to researchers and


5
clinicians who need to understand cardiovascular
pathophysiology. Computational simulations of aortic
0.0 0.3 0.5 0.8 1.0 1.3 1.5 flows compatible with the interventional planning
-5
Time (s) timescales require in vitro experiments for validation,
which in turn require a pulsatile flow pump system
FIGURE 9. Comparison of input (solid) and output (dotted)
piston velocities for an ascending aortic waveform (Waveform
capable of reproducing high Reynolds number flows.
2) with time relative to start of the cardiac cycle. Waveform 2 is Design and production of such a computer-controlled
representative of an aortic waveform in an extreme case with a piston pump system was presented here.
mean output flow rate that is almost four times higher (3.753)
than Waveform 1 and is used to stretch the pump system to its
The total output volumetric flow error between the
limits. The output curve is obtained through image process- prescribed input and measured output was <0.5 % for
ing of the piston motion. Here, the time-shifted systolic and flow rates up to 850 mL/s. The R2 value for a common
diastolic curves were compared, which resulted in R 2 values
of 0.966 and 0.974, respectively. aortic waveform was 0.999, and the RMS error of the
cycle-to-cycle piston output variation was 2.68 % of the
time averaged mean velocity of the pulse history. Al-
though the piston output waveform lagged when the
pump system was stretched to its limits, the time-shifted
systolic and diastolic curves both still had R2 values
above 0.966. The slopes of the systolic and diastolic
curves between the input and output waveforms varied
by <2 % in such cases. Piston motion was validated
using image processing techniques and by comparing
the pump’s output to specified inputs using physiologic
waveforms as a testbed. The system is economical and its
components can be acquired commercially or machined
in-house. Most importantly, the pump system can real-
istically generate aortic flows into and beyond the
transitional regime, which is a capability that other
available pump systems do not provide. Specifically, our
FIGURE 10. Comparison of piston input (solid) and mea-
sured downstream centerline velocity (dotted) obtained with pump achieved Reynolds numbers above 10,000 in an
phase-averaged stereographic PIV for an ascending aortic aortic diameter tube 55D downstream whereas the most
waveform (Waveform 1) with time relative to start of the car- capable commercially available pump can only achieve
diac cycle. The measured downstream pump output closely
followed the piston input waveform during systole, but devi- no more than Re ¼ 5000 based on its specifications
ated during the diastolic phase of the cardiac cycle. The dif- (CompuFlow 5000 MR, 300 mL/s maximum flow rate,
ferences are considerable, as they should be. A means to 5/8 in inner tubing diameter, water as working fluid).
determine the input waveform to achieve the desired time-
dependent output velocities is provided in separate refer- The pump successfully met all of the application
ences.4,7 needs we established: it actuated high Reynolds num-
ber flows like those in the human aorta, allowed for
piston motion. The pump is able to produce physio- custom input of and replicated physiologic waveforms,
logical pressure in the downstream test section. To generated transitional and turbulent flows without
maintain the physiological pressure, the gas release valve introducing external turbulence, triggered MRI and
must be opened periodically while reversing the piston to PIV image acquisition at user-defined points in the
eliminate any negative pressure build up. flow waveform, pumped high viscosity fluids against
The pump did not interfere with the operation of large impedances, and demonstrated compatibility
MRI, the triggering of MRI, or the images captured with corrosive fluids often used in with flow imaging.
for experiments conducted in a 1.5 T Phillips MRI Further, the pump can enable high-resolution heart
scanner with the pump located 8 m from the head of valve turbulence measurements and studies on repli-
A High Performance Pulsatile Pump 157
6
cating physiological turbulence in in vitro systems (as Chaudhury, R. A., M. Herrmann, D. H. Frakes, and R. J.
in wind tunnels). With a low disturbance flow inlet, Adrian. Length and time for development of laminar flow
in tubes following a step increase of volume flux. Exp.
various heart valve designs can be studied to obtain
Fluids. 56(1):22–10, 2015.
unbiased high-resolution heart valve turbulence mea- 7
Chaudhury, R. A., J. Ryan, D. H. Frakes, and R. J.
surements. The pump system may also be adapted to- Adrian. Prediction of downstream velocity waveforms for
wards pediatric applications. The only necessary design in vitro flow experiments. In: Summer Biomechanics,
changes would relate to adjusting the acrylic cylinder, Bioengineering, and Biotransport Conference. Snowbird
Resort, Utah, 2015.
piston, and tubing diameters to achieve the appropriately 8
Chung, B., and J. R Cebral. CFD for evaluation and
reduced flow rates. Future work will include isolating treatment planning of aneurysms: review of proposed
sources of variation within the system, examining other clinical uses and their challenges. Ann. Biomed. Eng. p. 1–
flow waveforms, and applying the system to investigate 17, 2015.
9
clinically significant aortic pathologies. Crosby, J. R., K. J. DeCook, P. L. Tran, R. G. Smith, D.
F. Larson, Z. I. Khalpey, et al. Physiological characteri-
zation of the SynCardia total artificial heart in a mock
circulation system. ASAIO J. 61(3):274–281, 2015.
10
Donovan, F. Design of a hydraulic analog of the circula-
ACKNOWLEDGMENTS tory system for evaluating artificial hearts. Artif. Cells
Blood Substit. Biotechnol. 3(4):439–449, 1975.
11
The authors gratefully acknowledge financial sup- Dur, O., M. Yoshida, P. Manor, A. Mayfield, P. D.
port from the Arizona Biomedical Research Commis- Wearden, V. O. Morell, et al. In vitro evaluation of right
sion (ABRC Grant #11028752), National Science ventricular outflow tract reconstruction with bicuspid
valved polytetrafluoroethylene conduit. Artif. Organs.
Foundation (NSF CAREER Award #1151232), the 34(11):1010–1016, 2010.
Ira A. Fulton Endowment, and ARCS Foundation. 12
Eriksson, A., H. W. Persson, and K. Lindström. A com-
puter-controlled arbitrary flow wave form generator for
physiological studies. Rev. Sci. Instrum. 71(1):235, 2000.
13
CONFLICT OF INTEREST Frakes, D., C. Zwart, and W. Singhose. Extracting motion
data from video using optical flow with physically-based
Rafeed A. Chaudhury, Victor Atlasman, Girish constraints. Int. J. Control Autom. Syst. 11(1):48–57, 2013.
14
Pathangey, Nicholas Pracht, Ronald J. Adrian, and Frayne, R., and D. Holdsworth. Computer-controlled flow
simulator for MR flow studies. J. Magn. Reson. Imaging.
David H. Frakes declare that they have no conflict of 2(5):605–612, 1992.
interest. 15
Harvard Apparatus. Series 1400 Pulsatile Blood Pumps
User’s Manual. Holliston, MA, 2004.
16
Heidenreich, P.A., J. G. Trogdon, O. A. Khavjou, J. But-
HUMAN AND ANIMAL STUDIES ler, K. Dracup, M. D. Ezekowitz, et al. Forecasting the
future of cardiovascular disease in the United States a
No human or animal studies were carried out by the policy statement from the American heart association.
authors for this article. Circulation. 123(8):933–944, 2011.
17
Holdsworth, D. W., D. W. Rickey, M. Drangova, D. J. M.
Miller, and A. Fenster. Computer-controlled positive dis-
placement pump for physiological flow simulation. Med.
Biol. Eng. Comput. 29(6):565–570, 1991.
REFERENCES 18
Hoskins, P. R. Simulation and validation of arterial
ultrasound imaging and blood flow. Ultrasound Med. Biol.
1
Alwan, A. Global Status Report on Noncommunicable 34(5):693–717, 2008.
19
Diseases 2010. Geneva: World Health Organization, 2011. Isselbacher, E. M. Thoracic and abdominal aortic aneur-
2
Arbia, G., C. Corsini, M. E. Moghadam, A. L. Marsden, ysms. Circulation 111(6):816–828, 2005.
20
F. Migliavacca, G. Pennati, et al. Numerical blood flow Ku, D. Blood flow in arteries. Ann. Rev. Fluid Mech.
simulation in surgical corrections: what do we need for an 29(1):399–434, 1997.
21
accurate analysis? J. Surg. Res.186(1):44–55, 2014. Ku, J. P., C. J. Elkins, and C. A. Taylor. Comparison of
3
Babiker, M. H., L. F. Gonzalez, J. Ryan, F. Albuquerque, CFD and MRI flow and velocities in an in vitro large artery
D. Collins, A. Elvikis, et al. Influence of stent configura- bypass graft model. Ann. Biomed. Eng. 33(3):257–269,
tion on cerebral aneurysm fluid dynamics. J. Biomech. 2005.
22
45(3):440–447, 2012. Kung, E. O. In-vitro experimental validation of finite ele-
4
Chaudhury, R. A. Improved Techniques for Cardiovas- ment analysis of blood flow and vessel wall dynamics.
cular Flow Experiments. Tempe, AZ: Arizona State Stanford: Stanford University, 2010.
23
University, 2015. Kung, E. O., A. S. Les, C. A. Figueroa, F. Medina, K.
5
Chaudhury, R. A., M. Herrmann, D. H. Frakes, and R. J. Arcaute, R. B. Wicker, et al. In vitro validation of finite
Adrian. Impact of development time on boundary condi- element analysis of blood flow in deformable models. Ann.
tions for constant volume flux start-up flow in arterial flow Biomed. Eng. 39(7):1947–1960, 2011.
24
piston pumps. In: 7th World Congress of Biomechanics. Kung, E. O., A. S. Les, F. Medina, R. B. Wicker, M. V.
Boston, MA, 2014. McConnell, and C. A. Taylor. In vitro validation of finite-element
158 R. A. CHAUDHURY et al.
33
model of AAA hemodynamics incorporating realistic outlet The Medical Image Computing and Computer Assisted
boundary conditions. J. Biomech. Eng. 133(4):041003, 2011. Intervention Society. CFD Challenge: Simulation of
25
Kung, E. O., and C. A. Taylor. Development of a physical Hemodynamics in a Patient-Specific Aortic Coarctation
Windkessel module to re-create in vivo vascular flow impedance Model, 2012. http://vascularmodel.org/miccai2012.
34
for in vitro experiments. Cardiovasc. Eng. Technol. 2(1):2–14, 2010. Timms, D., M. Hayne, K. McNeil, and A. Galbraith. A
26
Law, Y.F., R. S. Cobbold, K. W. Johnston, and P. A. complete mock circulation loop for the evaluation of left,
Bascom. Computer-controlled pulsatile pump system for right, and biventricular assist devices. Artif. Organs.
physiological flow simulation. Med. Biol. Eng. Comput. 29(7):564–572, 2005.
35
25(5):590–595, 1987. Tsai, W., and Ö. Savaş. Flow pumping system for physiological
27
Lieber, B.B., V. Livescu, and L. N. Hopkins. Particle image waveforms. Med. Biol. Eng. Comput. 48(2):197–201, 2010.
36
velocimetry assessment of stent design influence on intra- ViVitro Labs Inc., SuperPump System User Manual. Vic-
aneurysmal flow. Ann. Biomed. Eng. 30(6):768–777, 2002. toria, BC, 2014.
28 37
Milnor, W. R. Hemodynamics. Baltimore: Williams & Xiang, J., A. Siddiqui, and H. Meng. The effect of inlet
Wilkins, 1982. waveforms on computational hemodynamics of patient-
29
Mozaffarian, D., E. J. Benjamin, A. S. Go, D. K. Arnett, specific intracranial aneurysms. J. Biomech. 47(16):3882–
M. J. Blaha, M. Cushman, et al. Heart disease and stroke 3890, 2014.
38
statistics-2015 update: a report from the American Heart Yilmaz, S., O. Toker, N. Arslan, and H. Sedef. Optimal
Association. Circulation. 2014. in vitro realization of pulsatile coronary artery flow wave-
30
Pahlevan, N. M., and M. Gharib. In-vitro investigation of forms using closed-loop feedback algorithms with multiple
a potential wave pumping effect in human aorta. J. Bio- flow control devices. Turk. J. Electr. Eng. Comput. Sci.
mech. 46(13):2122–2129, 2013. 20(6):1006–1030, 2012.
31 39
Shelley Medical Imaging Technologies. CardioFlow 5000 MR Yun, B. M., D. B. McElhinney, S. Arjunon, L. Mirabella,
Computer-Controlled Flow Pump System. London, ON, 2015. C..K. Aidun, and A. P. Yoganathan. Computational sim-
32
Taylor, C. E., Z. W. Dziczkowski, and G. E. Miller. ulations of flow dynamics and blood damage through a
Automation of the harvard apparatus pulsatile blood bileaflet mechanical heart valve scaled to pediatric size and
pump. J. Med. Devices. 6(4):045002, 2012. flow. J. Biomech. 47(12):3169–3177, 2014.

You might also like