ASAIO Journal 2004

Bioengineering

Physiologic Control of Rotary Blood Pumps: An In Vitro Study

GURUPRASAD A. GIRIDHARAN,* GEORGE M. PANTALOS, KEVIN J. GILLARS, STEVEN C. KOENIG,

AND

MIKHAIL SKLIAR*

the flow rate generated by continuous flow VADs is selected

manually by physicians or trained support personnel. Mobile
patients operate the implanted continuous flow VADs by adjusting the pump rpm or flow rate using guidelines provided by
the doctor or until a perceived comfort level of perfusion is
achieved. No automatic feedback mechanism, directly based
upon physiologic indicators of cardiac demand, has been
successfully implemented.
Several physiologic indicators of cardiac demand, like atrial
pressure, blood oxygen saturation, lactic acid concentration in
blood, P wave activity of the atria, renal sympathetic nerve
activity, and aortic nerve activity, have been identified. Atrial
pressure (AP) is correlated to the preload in a straight forward
fashion, which makes AP the most widely used input control
parameter for the VAD.2 4 However, AP is not a direct indicator of physiologic demand and is influenced by many irrelevant factors. An excellent indication of tissue perfusion is
given by blood oxygen saturation.59 Unfortunately, the response time of blood oxygen saturation to the physiologic
demand is relatively slow, and implantable oxygen saturation
sensors capable of long-term, reliable operation are yet to be
developed. It was found that P wave activity decayed over
time10 once the natural heart was removed. With VAD, P wave
does not decay; however, the problem of decreasing sensitivity
of an electrical activity sensor over a long term still remains.
The renal sympathetic nerve activity and aortic nerve activity
suffers from a similar problem of signal degradation and longterm sensor reliability. The monitoring of lactic acid using
currently available sensors is not specific, reducing the value
of obtained results in on-line control of the total artificial heart
(TAH) and VAD. Moreover, physiologic indicators, which current control strategies (both in literature and in practice) rely
upon, change with variation in cardiac demand. For example,
if the control objective is to maintain a reference VAD rpm, an
increase in cardiac demand would necessitate an increase in
the desired rpm according to some expert rule or model prediction. Additionally, most of the discussed indicators are
affected by factors unrelated to cardiac demand. Consequently, it may be difficult to accurately correlate these indicators to the cardiac demand. In contrast, the average pressure
difference between pulmonary vein and aorta, Pa, is almost
constant for varying levels of cardiac demand, does not decay
with time, and can be reliably measured. Hence, we suggest
that maintaining an average reference Pa is an appropriate
strategy for controlling ventricular assist devices. The natural
regulatory system varies the vascular resistances to maintain
the required flow of blood11,12 with an almost constant average Pa. By using a VAD to assist the failing heart in maintaining the prescribed average Pa, we, in effect, synchronize

Rotary blood pumps (RBPs) are currently being used as a

bridge to transplantation as well as for myocardial recovery
and destination therapy for patients with heart failure. Physiologic control systems for RBPs that can automatically and
autonomously adjust the pump flow to match the physiologic
requirement of the patient are needed to reduce human
intervention and error, while improving the quality of life.
Physiologic control systems for RBPs should ensure adequate
perfusion while avoiding inflow occlusion via left ventricular
(LV) suction for varying clinical and physical activity conditions. For RBPs used as left ventricular assist devices (LVADs),
we hypothesize that maintaining a constant average pressure
difference between the pulmonary vein and the aorta (Pa)
would give rise to a physiologically adequate perfusion while
avoiding LV suction. Using a mock circulatory system, we
tested the performance of the control strategy of maintaining
a constant average Pa and compared it with the results
obtained when a constant average pump pressure head (P)
and constant rpm are maintained. The comparison was made
for normal, failing, and asystolic left heart during rest and at
light exercise. The Pa was maintained at 95 1 mm Hg for
all the scenarios. The results indicate that the Pa control
strategy maintained or restored the total flow rate to that of
the physiologically normal heart during rest (3.8 L/m) and
light exercise (5.4 L/m) conditions. The Pa approach
adapted to changing exercise and clinical conditions better
than the constant rpm and constant P control strategies. The
Pa control strategy requires the implantation of two pressure sensors, which may not be clinically feasible. Sensorless
RBP control using the Pa algorithm, which can eliminate the
failure prone pressure sensors, is being currently investigated.
ASAIO Journal 2004; 50:403409.

entricular assist devices (VADs) have been used successfully for many years as a bridge to transplantation1, and they
hold the potential to become long-term alternatives to donor
heart transplantation (destination therapy). However, a control
system that automatically responds to physiologic cardiac demand for the continuous flow VAD does not exist. In hospitals,
From the *Department of Chemical Engineering, University of Utah,
Salt Lake City, Utah; and the Jewish Hospital Heart and Lung Institute,
University of Louisville, Department of Surgery, Louisville, Kentucky.
Submitted for consideration June 2003; accepted for publication in
revised form May 2004.
Presented at the 49th ASAIO Conference, June 19 21, 2003, Washington, DC.
Correspondence: Dr. Guruprasad A. Giridharan, Department of
Chemical Engineering, 50 South Central Campus Dr., Rm 3290 MEB,
University of Utah Salt Lake City, UT 84112.
DOI: 10.1097/01.MAT.0000136652.78197.58

403

404

GIRIDHARAN ET AL.

Table 1. Baseline Parameters of the Mock Circulatory System Used in Simulating the Human Equivalence of a Normal, Failing,
and Asystolic LV
Normal LV
Parameter
a

Mean AoPr (mm Hg)

CO (L/m)b
DLP (mm Hg)c

Failing LV

Asystolic LV

Rest

Exercise

Rest

Exercise

96
3.8
210

100
5.4
300

68
2.0
100

75
3.1
150

Rest

Exercise

0
0

0
0

LV, left ventricle; AoPr, aortic root pressure; CO, cardiac output; DLP, drive line pressure.

the assisted and natural perfusion, thus indirectly incorporating

natural cardiovascular regulation into the VAD control.
In this study, the hypothesis of maintaining a constant average Pa to achieve physiologically motivated perfusion is
tested for a centrifugal blood pump using an in vitro mock
circulatory system13 for the following test conditions: normal,
failing, and asystolic left heart during rest and light exercise.
Materials and Methods
Experimental Design
An adult mock circulation (consisting of a mock left ventricle, ventricular apical inflow cannulation, and mock systemic
vasculature with aortic root outflow cannulation) along with a
centrifugal continuous flow blood pump (BioMedicus,
Medtronic, Eden Prairie, MN) was used to test the viability of
the Pa control strategy and compare it with constant rpm and
constant pump pressure head control strategies.
To study the range of applicability of the proposed approach, one normal and two different pathologic cases of the
VAD assisted perfusion were simulated using the mock circulatory system. In the first case, the left ventricular assist device
(LVAD) was attached to the human equivalent of a normal
healthy heart, which is a realistic case when testing with
animals or when natural left heart (LH) function has completely recovered after VAD implantation, as observed in several clinical cases.14,15 Equivalence of pathologic cases include LVAD assistance of failing and asystolic left heart (Table
1). Three scenarios were tested under rest and light exercise
conditions. Heart rate was 60 bpm during rest and 100 bpm
during light exercise. For all test conditions, 35% systole and
65% diastole was maintained. A lower value of heart rate and
the resulting lower cardiac output during rest were chosen to
increase the variability in the cardiac demand as the maximum
flow rate of the mock circulatory system is limited. The aortic
input impedance and vascular mechanical properties were
controlled to simulate the flow and impedance of the normal
human vasculature.16 The vascular resistance (total peripheral
resistance) and the driveline pressure (which controls the contractility of the LV) were adjusted to match the pressure and
flow waveform characteristics of the human circulatory system
under the described scenarios. Once the resistance and the
driveline pressure are determined, they are used consistently to
test different control strategies.
For different clinical and cardiac demand conditions, the
VAD rpm is adjusted manually until the setpoint for Pa, P,
or rpm is reached. The setpoint for Pa is selected as the Pa
value observed with normal unassisted heart (baseline case) at
rest and is equal to approximately 95 mm Hg. Based upon the

result of the previous simulation study,17 the setpoint for P is

selected as 75 mm Hg. The pump speed of approximately
1,440 rpm, which was needed to restore the cardiac output to
the physiologic level of 3.8 L/m for the case of failing heart at
rest, was selected as the rpm setpoint. Once the setpoint is
reached, the limit cycle hemodynamic waveforms were recorded with and without VAD assistance for each of the three
control strategies. Characterizing hemodynamic parameters,
waveform morphology, and ventricular pressure-volume loop
responses were calculated to identify differences in the performance with different control strategies for each test condition.
Mock Circulation
The adult mock circulation consists of atrium, ventricle, and
systemic and coronary vasculature components as illustrated
in Figure 1. In a previous study,13 the adult mock circulation
was shown to mimic human normal ventricle, failing ventricle,
and partial cardiac recovery physiologic responses as defined
by characterizing hemodynamic parameters, ventricular pressure-volume relationship, aortic input impedance, and vascular mechanical properties. An artificial atrium,18 made of a
flexible polymer sphere 50 mm in diameter, is connected
upstream of the inflow valve of a mock ventricle. The mock
ventricle consists of a flexing polymer sac inside a pressurization chamber.19 The ventricular sac is hemiellipsoid shaped
and is 70 mm wide at the base and 83 mm long from base to
apex. The base is covered by a semi-rigid polymer dome 20
mm high, with mounts for inflow (mitral) and outflow (aortic)
prosthetic valves. Metered pulses of compressed air (Utah
Heart Controller, CardioWest, Tuscon, AZ) are delivered to the
pressurization chamber during systole, compressing the ventricular sac to form coapting quadrants simulating contraction
of the normal and dysfunctional ventricle and the delivery of
cardiac stroke volume. An artificial aorta (polyurethane tube
segment 25 mm diameter) is connected downstream of the
outflow valve of the ventricular sac to the mock systemic
vasculature. The mock systemic vasculature consists of four
integrated chambers that represent lumped proximal resistance, systemic compliance, peripheral resistance, and venous
compliance.13 Introduction ports for the VAD uptake cannula
are incorporated into the ventricular sac apex and VAD output
flow cannula at the aortic root.
Instrumentation
A high fidelity, pressure-volume conductance catheter (Millar Instruments, Houston, TX) was inserted into an aortic introducer port and passed retrograde through the aortic valve

405

VAD COMPUTER SIMULATION

Figure 1. Schematic of the mock

circulatory system with the assist
device.

and down to the ventricular apex for simultaneous ventricular

pressure, root aortic pressure, and ventricular volume measurements. Single tip, high fidelity catheters (Millar Instruments, Houston, TX) were inserted into introducer ports for
measuring atrial pressure, distal aortic pressure, and driveline
pressures. Aortic root, aortic distal, and VAD output were
measured with inline transit time flow probes (Transonics,
Ithaca, NY). Pressure, flow, and volume transducers were preand postcalibrated, and transducer gains and offsets were calculated and applied to ensure measurement accuracy. Gains
were calculated for the LV volume data to match the stroke
volume of the LV, as sensed by the aortic root flow probe.
Offsets for the LV volume data were calculated taking into
consideration the total flow and left ventricular end diastolic
pressure (LVPed) data. Placement of instrumentation for hemodynamic measurements of pressures, flows, and volume is
shown in Figure 1. Signal conditioning was accomplished
using transducer amplifiers (Ectron, San Diego, CA), transittime flow meters (Transonics, Ithaca, NY), a volume conductance unit (Leycom, Sigma V, Netherlands), and other peripheral conditioners integrated in an instrumentation system
compliant with Good Laboratory Practice (GLP) guidelines.
Signal conditioned data were low pass filtered at 60 Hz,
analog to digitally converted (AT-MIO-16E-10 and LabVIEW,
National Instruments) at a sampling rate of 400 Hz, and stored
on a personal computer for postprocessing and analysis.20
Data Analysis
Differences in characterizing hemodynamic parameter values and ventricular pressure-volume loop response were calculated using a Hemodynamic Evaluation and Assessment
Research Tool (HEART) program21 and supporting m-files developed in Matlab (MathWorks, Natick, MA). Pressure, flow,
and volume waveforms were used to calculate the following
hemodynamic parameters: mean pulmonary vein pressure,
LVPed and VAD output flow, and the total flow. All hemodynamic parameters were calculated on a beat to beat basis, with
all beats in each data set averaged to obtain a single representative mean value for each parameter. Pressure-volume loops

were constructed by plotting ventricular pressure against ventricular volume, in which each loop represents one complete
cardiac cycle (one beat). Characterizing hemodynamic parameters and pressure-volume loops were calculated for all experimental conditions.
Results
The hemodynamic parameters for a normal, failing, and
asystolic LV with and without continuous assist for each of the
three control strategies during rest and light exercise are listed
in Table 2. Without VAD assistance, the values of the total flow
rate, P, and Pa decrease during ventricular failure at rest
and exercise in comparison with the normal LV at rest and
exercise. The left ventricular end diastolic pressure for all the
control strategies remain within 3 mm Hg of the baseline
normal LV value. Because the left ventricular pressure and
volume sensor is introduced through the aortic valve (Figure
1), there is a back flow through that valve for baseline and all
VAD assist scenarios.
Table 2 indicates that all of the tested control strategies
increase the total flow, P, and Pa with failing and asystolic
LV. The Pa control strategy maintains or restores the total
flow rate to that of the physiologically normal heart during rest
and exercise and adapts best to the need for support. For
example, in the case of the normal heart during rest and
exercise, the average net VAD flow rate with this strategy is
close to zero, as expected, because the native LV provides all
the required cardiac output.
Figure 2 shows the comparison between the total flow rates
(sum of VAD flow rate and cardiac output) at baseline and
during assistance using constant rpm, P, and Pa control
strategies during rest and exercise scenarios for each clinical
test condition. The baseline cardiac output of 3.8 L/m at rest
and 5.4 L/m during light exercise are considered to be physiologic flows for the corresponding physical activities. The
comparison of total flow rates produced with different control
strategies shows that the Pa approach best matched the physiologic flow rate. The comparison of rest and exercise cases
shows that the control strategy of maintaining an average Pa

406

GIRIDHARAN ET AL.
Table 2. Comparison of Assisted and Unassisted Perfusion Under Different Scenarios
Total Flow
(L/m)

P
(mm Hg)

Pa
(mm Hg)

0
0
0
0
0
0

3.8
2.0
0
5.4
3.1
0

54.3
30.5

54.4
32.9

94.4
51.5

98.4
58.7

0.0
0.0

0.0
0.0

0.0
17.2

6.7
16.4

800
1440
1490
650
1490
1600

3.6
3.9
3.9
5.5
5.2
5.7

55.3
71.9
96.4
51.7
68.6
98.3

95.0
95.2
94.9
95.8
94.4
95.2

0.2
4.5
4.5
0.4
5.6
6.3

1.0
0.8
1
5.3
5.1
2.8

1300
1450
1320
1280
1600
1400

4.6
4.0
3.3
6.4
5.7
4.8

75.3
78.2
75.6
74.6
76.6
74.1

115.4
101.2
74.5
117.5
101.5
71.6

3.2
4.5
3.9
3.5
6.2
5.4

1.3
1.3
0.0
7.6
6.2
0.1

1450
1440
1450
1440
1440
1440

4.7
3.9
3.7
6.8
4.8
5.0

80.1
71.9
89.3
80.0
70.5
76.6

120.2
95.2
87.8
122.4
97.0
74.0

3.8
4.5
4.3
4.4
5.3
5.5

1.8
0.8
1
7.9
4.8
0.6

VAD rate
(rpm)
Baseline values
Normal LV, rest
Failing LV, rest
Asystolic LV, rest
Normal LV, exercise
Failing LV, exercise
Asystolic LV, exercise
Centrifugal VAD with Pa control
Normal LV, rest
Failing LV, rest
Asystolic LV, rest
Normal LV, exercise
Failing LV, exercise
Asystolic LV, exercise
Centrifugal VAD with P control
Normal LV, rest
Failing LV, rest
Asystolic LV, rest
Normal LV, exercise
Failing LV, exercise
Asystolic LV, exercise
Centrifugal VAD with rpm control
Normal LV, rest
Failing LV, rest
Asystolic LV, rest
Normal LV, exercise
Failing LV, exercise
Asystolic LV, exercise

VAD Flow
(L/m)

LVPeda
(mm Hg)

VAD, ventricular assist device; LVPed, left ventricular end diastolic pressure, LV, left ventricle.
LVP for the asystolic LV is a constant value

leads to the correct adaptation to changing cardiac demand.

For a normal LV, the Pa control strategy best matches the
physiologic flow rates (Figure 2a). At the same time, the constant rpm and P control strategies result in higher than normal
flow rate. The overpumping (highest when constant rpm is
maintained) increases the risk of LV suction.
Figure 2b indicates that when the failing LV is assisted by a
VAD, all three control strategies restore the total cardiac output
to near physiologic level. The Pa approach best matches the
physiologic flow rate during rest and exercise. The P approach leads to an output that is slightly higher than the
physiologic flow rate. The constant rpm strategy results in a
lower than normal flow rate during exercise, increasing the
chances of under perfusion during higher cardiac demand.
With an asystolic LV (Figure 2c), the Pa strategy is the best
approach at restoring the flow rates to near physiologic values,
followed by constant rpm and P strategies. Overall, the Pa
strategy consistently produced a total flow rate that is the
closest to the physiologic flow rate for all heart conditions and
physical activity scenarios.
The left ventricular pressure-volume relationships for a normal and failing LV with and without assistance is shown in
Figure 3. All of the tested control strategies cause a leftward
shift in the pressure volume loop for a failing ventricle during
rest and exercise (Figure 3,c and d) and a lowering of left
ventricular end diastolic pressure, indicating a correct direction of adaptation for all the control strategies. Except for the
result with Pa control, a leftward shift in the PV loop and
lowering of LVPed was noticed for a normal heart assisted by

a VAD during rest and exercise (Figure 3, a and b), indicating

an increased likelihood for suction.
Discussion
The importance of adequate VAD control cannot be overstated. Although the design of the VAD itself is critical to the
long-term success of the electromechanical implant, the control of the VAD determines the confidence of doctors and
patients in the VAD as a permanent solution and an alternative
to donor heart transplantation. The key requirement of the
automatic control system is the adaptation of VAD generated
flow to the changing physiologic requirements of the patient
while reliably avoiding suction.2224
Our in vitro results show that maintaining a constant average
Pa is an effective way to the correct adaptation of the cardiac
output to changing requirements of the patient irrespective of
the type of rotary pump used to assist perfusion. The physiologic explanation of this conclusion rests with the fact that the
vascular bed resistance can increase or decrease by a factor of
two to five in response to the changing cardiac demand25 and
is the dominant factor in regulating perfusion. The blood flow
is inversely proportional to the vascular bed resistance so that
maintaining a constant Pa with changing resistance can increase or decrease the flow rate by the same factor of two to
five.
The desired (reference) Pa can be maintained by adjusting
the pump rpm within physiologically admissible limits despite
the changing vascular resistance, stroke volume, and heart

VAD COMPUTER SIMULATION

Figure 2. Comparison of total flow generated with Pa, P, and

constant rpm strategies with and without the VAD: (a) Normal LV, (b)
Failing LV, and (c) Asystolic LV. VAD, ventricular assist device; LV,
left ventricle.

rate, which represent the response of the natural regulatory

mechanisms to the changing physiologic cardiac output demand. The dominant role of the changing resistance in adaptation to physiologic demand11,12 implies that by maintaining,
on average, the prescribed Pa, we in effect synchronize the
assisted and natural perfusion, thus indirectly incorporating
natural cardiovascular regulation into VAD control.
The proposed approach to the control of RBP requires that
the natural regulatory mechanism functions properly in response to changing cardiac demand, which may not always be
the case. For example, medical intervention may be necessary
in the case of severe hypertension (often seen in the VAD
recipients after initial recovery), which could lead to higher
than normal arterial pressures, resulting in lung edema. Note
that neither the alternative VAD control strategies nor the
natural heart can directly mitigate arterial hypertension and the
resulting lung edema. Consequently, the long-term goal may
have to include the development of an automatic, autonomous, portable health monitoring and management system for
patients with the permanent VAD or TAH, which would combine real time control of the blood pump with the automatic
monitoring of the cardiac function and, if necessary, emergency drug administration and other advanced functionalities.
The primary advantage of the Pa control strategy is its
ability to autonomously adjust the total output, defined as the
sum of cardiac and pump outputs, to match the cardiac de-

407

mand better than any alternative strategies. The Pa, being the
difference between the aortic and the pulmonary venous pressure (equal to left atrial pressure), is sensitive to changes both
in preload and afterload. The current in vitro study shows that
the proposed strategy of maintaining the desired average Pa
leads to an adequate adaptation for widely changing cardiac
demand and clinical conditions in a completely autonomous
way. The results show that, although some degree of physiologic adaptation is achieved with constant rpm and constant
P, these alternatives are less effective for a wide range of
physical activities and rapidly changing status of cardiac function (such as a sudden transition from failing to asystolic heart).
With the constant rpm control strategy, a broad range of
physical and clinical conditions would require an external
intervention to change the rpm setpoint according to some
expert rule, model prediction, or operator input. The constant
P approach does not perform well for a broad range of
clinical conditions of the native heart, which changed from
normal to asystolic LV in this study, but adapts well to the
changing cardiac demand due to different exercise levels. The
in vitro study is consistent with the results of computer simulations,17,26 28 which showed that the P control strategy
adapts better to widely varying cardiac output requirements
when compared with the traditional constant rpm control
approach. Because of the limitations of the mock circulatory
system, we were unable to test the higher cardiac demand
conditions to make an in vitro comparison between the different control strategies. In the limited range of cardiac demands
that could be tested in vitro, the performance of the Pa
control strategy is superior to P and constant rpm control
alternatives.
Using the proposed approach, both the natural heart and the
assist device are contributing to the pumping action of maintaining an average Pa. If cardiac function improves, the
native heart will increase its contribution to maintaining the
reference pressure difference, with the VAD controller autonomously and automatically responding to the decreased need
for assisted perfusion, as evidenced by the near zero net VAD
flow rate with the normal heart during rest and exercise. When
the net flow rate through the VAD is close to zero, blood does
not stagnate inside the VAD, although the residence time of
blood in the pump is higher, increasing the probability of
hemolysis. Because this scenario occurs only with a normal or
near normal heart, the patient could be weaned from the pump
at this stage.
The ability of the proposed control strategy to automatically
adjust its contribution towards maintaining Pa may prove to
be well suited to the application of the ventricular assist devices in cardiac recovery therapy14,15,29 of end-stage heart
failure (alone or in combination therapy), as well as in the
destination therapy.
Although not directly addressed in this study, the overarching principle behind the proposed approach of maintaining
key pressure differences with mechanical blood pumps, while
relying on the natural regulation to adjust the resistances to the
blood flow to meet the physiologic demand, is also applicable
to the case of pulsatile ventricular assist devices, as well as the
total artificial heart. In the case of the TAH, the blood pump
should be controlled to maintain key pressure differences at
the average reference values, which, in the case of pulmonary

408

GIRIDHARAN ET AL.

Figure 3. PV loops with and without VAD assistance using constant

rpm, P, and Pa control strategies for the following test conditions: (a) Normal LV during rest; (b)
Normal LV during exercise; (c) Failing LV during rest; and (d) Failing LV
during exercise. VAD, ventricular
assist device; LV, left ventricle; PV,
pressure volume.

circulation, is the difference between pulmonary arterial and

vena cava pressures. The proposed approach for VAD control,
in its current form, requires the implantation of two pressure
sensors. Long-term clinical implantation of pressure sensors
may not be feasible because of decreasing sensitivity over
time, sensor drift, and the risk of thrombus formation. However, for different types of blood pumps, it may be possible to
estimate Pa using only the pump model and readily measurable intrinsic pump parameters (such as pump rpm, voltage,
and current), eliminating the need for implantable pressure
sensors. The approach that uses a blood pump as both the
actuator and the flow or pressure sensor can be viewed as a
sensorless control.27 Sensorless estimation of Pa is currently being pursued as a follow-up investigation.

Limitations
The performance of the mock circulation during normal,
failing, and asystolic heart test conditions is representative
of clinical observations from a purely hemodynamic/hydrodynamic viewpoint. Clearly, mock circulation experiments
are limited in replicating clinical conditions. For instance,
mock circulation cannot mimic neurohumoral responses,
tissue remodeling, or activation of regulatory proteins. The
particular mock circulatory system used in the experiment is
a single sided mock and can simulate the systemic (left side)
circulation only. In this mock system, the uptake cannula
inflow occlusion is physically impossible, and the reduced
LV pressure and volume are the only indications of suction.

VAD COMPUTER SIMULATION

The instrumentation used to record hemodynamic waveforms in the present study has inherent measurement errors
associated with each technique (i.e., pressure error 1
mm Hg, flow error 0.5 L/min), which we have attempted to minimize by using GLP compliant test equipment, calibration procedures, and documentation practices.
The gains and offsets assigned for the LV volume data may
have inherent errors associated with it, as it is based upon
the aortic root stroke volume, LVPed, and total flows
produced.
Despite these limitations, these results provide valuable insight into the differences of each control algorithm, which aids
in further development of control algorithms and experimental
protocols to identify optimal ventricular assist therapies that
can subsequently be validated in in vivo models.
Conclusions
The in vitro results show that maintaining an average pressure difference between the pulmonary vein and aorta (Pa)
provides an effective way to control a continuous flow LVAD
over a wide range of physiologic and cardiac demand conditions while reducing the probability of suction. Change in
vascular resistance is the dominant regulatory mechanism in
meeting the physiologic requirements for blood perfusion.
Maintaining the desired average Pa by adjusting the pump
rpm during changing cardiac demand, in effect, synchronizes
the assisted and natural perfusion. Therefore, the proposed
control strategy indirectly incorporates natural cardiovascular
regulation, which changes vascular resistance into VAD control. The comparison with the VAD control systems, which
maintain either constant reference pump rpm or constant
pump pressure head (P), shows that the proposed approach is
superior in autonomously maintaining an adequate perfusion
during changing cardiac demand for the test conditions simulated in vitro. Because the Pa control strategy automatically
adjusts its contribution to the total flow based upon the function of the native ventricle, the proposed approach may prove
to be well suited to the application of the ventricular assist
devices in recovery therapy.
Acknowledgment
This study was supported by the Established Investigator Award from
the American Heart Association.

References
1. Olsen D: Sixth international symposium for rotary blood pumps.
Artif Organs 26: 475 476, 1999.
2. Klute G, Tasch U, Geselowitz D: An optimal controller for an
electric ventricular assist device: Theory, implementation and
testing. IEEE Trans Biomed Eng 39: 394 403, 1992.
3. Boston J, Simaan M, Antaki J, Yu Y: Control issues in rotary heart
assist devices. Proceedings of the American Control Conference
34733477, 2000.
4. Schima H, Trubel W, Moritz A, et al: Noninvasive monitoring of
rotary blood pumps: Necessities, possibilities, and limitations.
Artif Organs 16: 195202, 1992.
5. Takatani S: Open-loop analysis of circulatory system in awake live
animals: Relations between mixed venous saturation (SvO2)
and hemodynamic parameters. Artif Organs 17: 79, 1992.
6. Takatani S, Noda H, Takano H, Akutsu T: Continuous in-line
monitoring of oxygen delivery to control artificial heart output.
Artif Organs 174: 458, 1990.

409

7. Ohno T, Masuzawa T, Nakamura M, et al: Characteristics of

mixed venous oxygen saturation and physical activity for artificial heart control. ASAIO J 43: M677M681, 1997.
8. Nakamura M, Homma A, Tatsumi E, et al: Mixed venous oxygen
saturation as a promising parameter for physiologic control of
total artificial heart. ASAIO J 46: 761766, 2000.
9. Makoto Nakamura, Toru Masuzawa, Eisuke Tatsumi, et al: Control
of total artificial heart using mixed venous oxygen saturation.
ASAIO J 45: 460 465, 1999.
10. Murray KD, Binkley PF, Watson KM, et al: Remnant atrial function
in human recipients of a total artificial heart. J Heart Lung Trans
11: 1066 1072, 1992.
11. Guyton A: Textbook of Medical Physiology, 6th Ed. Philadelphia:
Saunders, 1981.
12. Green J: Fundamental Cardiovascular and Pulmonary Physiology.
Philadelphia: Lea and Febiger, 1987.
13. Pantalos G, Gillars K, Ewert D, Giridharan G, Koenig S: Characterization of an adult mock circulation for testing cardiac support devices. ASAIO J 50: 37 46, 2004.
14. Nishimura M, Radovancevic B, Odegaard P, Myers T, Springer W,
Frazier O: Exercise capacity recovers slowly but fully in patients
with a left ventricular assist device. ASAIO J 42: M568 M570,
1996.
15. Nakatani T, Sasako Y, Kobayashi J, et al: Recovery of cardiac
function by long-term left ventricular support in patients with
end-stage cardiomyopathy. ASAIO J 44: M516 M520, 1998.
16. Nichols W, Conti C, Walker W, Milnor W: Input impedance of the
systemic circulation in man. Circ Res 40: 451 458, 1977.
17. Giridharan G, Skliar M: Control strategy for maintaining physiological perfusion with implantable rotary blood pumps. Artif
Organs 27: 639 648, 2003.
18. Rosenberg G, Phillips W, Landis D, Pierce W: Design and evaluation of the Pennsylvania State University mock circulatory
system. ASAIO J 4: 41 49, 1981.
19. Pantalos G, Hayes J, Khanwilkar P, Bearnson G, Olsen D: Left
ventricular imulator for cardiovascular device testing. ASAIO J
42: 46, 1996.
20. Drew G, Koenig S: Biomedical patient monitoring, data acquisition, and playback with labview. In: LabVIEW for Automotive,
Telecommunications, Semiconductor, Biomedical, and other
Applications. Hall T. Martin, Meg L. Martin (eds) Upper Saddle
River, NJ: Prentice Hall, 2000, pp. 9298.
21. Schroeder M, Perrrault B, Ewert D, Koenig S: HEART: An automated beat-to-beat cardiovascular analysis package using MatLab. Com Bio & Med 34: 371388, 2004 (in press, and available
on-line:
www.sciencedirect.com/science/journal/
0010104825).
22. Amin D, Antaki J, Litwak P, Thomas D, WuZ, Watach M: Induction of ventricular collapse by an axial flow blood pump.
ASAIO J 44: M685M690, 1998.
23. Araki K, Oshikawa M, Onitsuka T, Nakamura K, Anai H, Yoshihara H: Detection of total assist and sucking points based on
pulsatility of a continuous flow artificial heart: In vitro evaluation. ASAIO J 44: M708 M711, 1998.
24. Oshikawa M, Araki K, Nakamura K, Anai H, Onitsuka T: Detection of total assist and sucking points based on pulsatility of a
continuous flow artificial heart: In vivo evaluation. ASAIO J 44:
M704 M707, 1998.
25. Waters T, Allaire P, Adams M, et al: Motor feedback physiological
control for a continuous flow VAD. Artificial Organs 23: 480
486, 1999.
26. Giridharan G, Skliar M, Olsen D, Pantalos G: Modeling and
control of a brushless dc axial flow ventricular assist device.
ASAIO J 48: 272289, 2002.
27. Giridharan G, Skliar M: Physiological control of rotary blood
pumps using intrinsic pump parameters. 10th Congress of International Society for Rotary Blood Pumps, Osaka, Japan,
September 2002.
28. Giridharan G, Skliar M: Nonlinear controller for ventricular assist
devices. Artif Organs 26: 980 984, 2002.
29. Tansley P, Birks E, Hipkin M, et al: Early and mid-term results
of the Harefield left ventricular assist device combination
therapy for reversal of end stage heart failure. Eur Heart J 4:
618, 2002.