Approximately 400,000 new cases of heart failure are diagnosed each year in the United States, including up to 15,000 births with congenital heart defects. Estimates are that anywhere from 35,000 to 200,000 patients could benefit from an artificial heart implant or from a ventricular assist device if one were available. 1 In light of this need, the National Institutes of Health has funded the development of both pulsatile and nonpulsatile artificial hearts for these patients. Although much of the early funding supported the development of diaphragm pulsatile type pumps, continuous flow pumps have been used in short duration circulatory assistance for many years. 2 Studies in animals have shown long-term survival with nonpulsatile blood flow with no adverse effects. 3–5 Conclusions drawn from these tests are that the nonpulsatile rotodynamic pump lends itself to application as an artificial heart of the complete implantation type. Other data reveal that human patients have been maintained with centrifugal pumps for extended periods before a heart transplant. 6 These results are significant, showing the rotodynamic pump’s potential for use as a bridge to transplant or implantable artificial heart. These promising clinical and animal study results have led to an increased emphasis on and development of centrifugal and axial flow pumps because of their inherent simplicity. With this interest, much attention has been focused on the pump bearings. 7–13
The optimal bearing support system appears to be totally noncontact magnetic bearings. A magnetic bearing support system has advantages from the viewpoints of power loss and blood damage. With novel designs, the power loss of a magnetic bearing can be extremely low, especially compared with a hydrodynamic bearing system. The noncontact nature of these bearings allows for fully washed flow paths to avoid stagnation points that might allow thrombi to form. Considerable flexibility in design of these flow paths can be obtained through variations in the magnetic bearing system design. With proper design, it becomes possible to both avoid stagnation points while at the same time maintaining large enough clearances throughout the flow path to also eliminate regions of locally high shear stress that would cause hemolysis. Equally important, the use of noncontact bearings totally eliminates the lifetime reducing wear and tribocompatibility issues that are present in rolling element or point contact bearing systems.
With an active bearing system, the issue of power consumption must be addressed. To operate, the spinning pump rotor must be constrained in five axes: one axial, two radial, and two tilt. Most active magnetic bearing supported pumps use a bearing system with five active axes to provide complete control of the pump rotor during operation. Clearly, a reduction in the number of actively controlled axes would reduce the power required to operate the bearings, thus increasing operating time before a battery change or recharge is required. For magnetic attractive type bearing configurations, no reduction in the number of controlled axes is possible. However, in theory, if magnetic repulsion is used, only one actively controlled axis is required. This approach will minimize the total bearing power draw.
The second generation pump under development by Mohawk Innovative Technology, Inc. (MiTi, Albany, NY) is based on a novel, patent-pending, hybrid passive–active bearing system that provides passive bearings to control the rotor in the two radial and two tilt axes and a unique active thrust bearing to control the rotor axial position. The magnetic assembly contains two pairs of permanent magnet rings positioned on both the rotor and stator in a repulsive radial bearing configuration. An electrically controlled coil is positioned around the stator magnet pairs and can modulate the force between stator and rotor magnet pairs to provide axial control of the rotor. To further reduce the power consumption of the magnetic bearings, current to the coils is controlled to position the rotor for maintaining a zero force balance (low current) position axially to the rotor. When additional axial forces are applied to the impeller, an increase in coil current is detected, and a different equilibrium rotor position is maintained for a new zero force balance position axially to the rotor.
These novel bearings are integrated into a pump assembly that provides a simple, direct, smooth blood recirculation flow path to minimize the potential for hemolysis. Flow passages are sized to maintain a minimum flow velocity, which also avoids the potential for formation of thrombi. The pump geometry has been configured to allow for conventional hydrodynamic backup bearings to be integrated into the pump. This is a key issue for any magnetically supported pump. If bearing power is lost, or a controller fault occurs (such as because of component failure or electromagnetic pulse), or some sudden overload shock transient exceeds the load capacity of the bearing system, there must be a means to provide continued support of the rotor during operation until the problem can be resolved. Hydrodynamic bearings, although they have much higher power loss and increased blood damage relative to magnetic bearings, are ideal as a backup because they are a totally passive bearing. As long as the rotor is spinning, the hydrodynamic bearing will have load capacity to support the rotor through a transient overload or compromised active control system scenario. For the MiTi pump, four of the five axes are supported by passive magnetic bearings, which require only a transient overload backup. This transient bearing is a conventional radial hydrodynamic bearing that is inherent in the recirculation flow path geometry. The fifth (axial) axis is actively controlled, thus requiring a backup bearing for both transient overload and controller failures.
In the current study, a series of pump performance and durability tests are reported. Additional results of a preliminary backup bearing design study are also discussed.
Materials and Methods
The prototype generation II blood pump, shown in Figure 1, is based on a modified first generation pump. The prototype pump is housed in an oversized polycarbonate outer housing for these tests. This outer housing is approximately 76 mm in diameter and 94 mm long (not including the end cover). The pump itself is based around a 34.5 mm diameter impeller that rotates at 2,500 to 3,500 rpm, depending on the operating point. The fully washed flow path recirculates blood around the rotor through a simple and direct flow path. The low power loss, hybrid passive–active bearing system described above is integrated into the rotor and stator along with a high efficiency permanent magnet motor. The low power requirements for the active bearing system at operating conditions of 3,000 rpm, 5.0 L/min, and pressure head of 100 mm Hg is only 0.007 watts, a significant reduction from 0.5 watts for the first generation pump. The total efficiency of this prototype pump is 14%, with the majority of losses from the impeller and viscus losses in the wash flow path. For the hemolysis tests, exposed metal surfaces were coated with Paralyne.
Water-glycerin blood analog fluid, with a nominal viscosity of 1.5 × 10−3 Pa-s at room temperature, was used as the test fluid for flow and long-term durability tests. The pump was installed in a simple flow loop and was instrumented for pressure rise, flow, inlet, outlet, housing and reservoir temperatures, shaft speed, and pump vibration. For transient orientation tests, the pump was installed in a fixture that could rapidly rotate the pump spin axis from horizontal to vertical. Figure 2 shows the test setup. The position was controlled with a timer, which was generally set for a orientation change every 15 minutes. In this test configuration, the pump was instrumented for pressure rise, inlet, outlet, housing and reservoir temperatures, shaft speed, and pump vibration.
For hemolysis testing, the pump was installed in a second flow loop configured on the basis of American Society for Testing and Materials (ASTM) practice F1841–7. For these tests, bovine blood was obtained by venipuncture and collected into a sterile 2 L blood bag containing heparin sulfate. The priming volume of the pump was approximately 36 ml, and the total volume of the circuit was 1.9 L. The tubing consisted of 9.5 mm medical grade polyvinylchloride. A nominal pump speed of 2,800 rpm, providing a flow rate of 5 L/min at a pressure head of 100 mm Hg, was used for these tests. The normalized index of hemolysis was calculated using the following equation:EQUATION
where ΔfreeHb is the increase in plasma free hemoglobin concentration (g/L) over the sampling time interval, V is the circuit volume in liters, Hct is the percent hematocrit, Q is the flow rate in L/min, and T is the sampling time interval in minutes.
Hematocrit was measured using a CritSpin micro hematocrit centrifuge from Stat-Spin Inc. A NesLab model RTE-211 constant temperature circulating bath was used to maintain the blood at 37°C for the duration of testing. An Eppendorf model 5402 centrifuge was used to spin the blood samples for plasma hemoglobin determinations. Plasma hemoglobin measurements were made at 30 minute intervals using a plasma hemoglobin kit (No. 527) from Sigma Diagnostics. Absorbance measurements were made using a Shimadzu model UV-160A UV-Vis spectrophotometer.
Figure 3 shows the flow versus pressure data for the pump at three operating speeds. Tests of three different impeller blade heights are shown on this figure (impellers 1, A, and B). Increasing flow corresponds to increasing blade height. These data show that the nominal operating point of 5 L/min at 100 mm Hg pressure rise can be achieved for all three blade heights by varying the pump speed. At the nominal operating point, the pump is expected to operate at approximately 3,000 rpm. Depending on the operating condition, flows from 2 to 7 L/min, with pressure rises ranging from 50 mm Hg to 150 mm Hg, were obtained. This range covers the projected range for the left ventricular assist device (LVAD) application.
Figures 4 through 9 present selected data recorded during a 28 day durability test. For this test, a nominal flow of approximately 2 L/min, with a nominal pressure rise of 102 mm Hg, was selected. Figure 4 presents the flow data. The flow remains fairly constant, with a slight downward trend that is related to a decrease in motor speed. Figure 5 presents the motor speed. Motor speed is determined by a DC voltage proportional to the speed provided by the motor controller. The drift in speed was not surprising because the motor was operating without a speed control system, and temperature drift of the motor controller will cause small changes in the indicated motor speed. Figure 6 presents the pressure data. The pressure is relatively constant, with some variations that are related to changes in outlet pressure. Figure 7 presents the temperature rise from inlet to outlet. The very small temperature rise reflects the low power dissipation of this pump and bearing system. Figure 8 presents typical horizontal vibration as acceleration versus frequency taken at several intervals over the course of the test. Figure 9 presents the corresponding axial data. The vibration recorded is quite low, is dominated by running speed, and remains consistent throughout the test.
Figures 10 through 14 present selected data from a 5 day horizontal-vertical transient orientation test. Figure 10 presents the pressure rise recorded during this test (flow data were not recorded continuously during this test). Relatively constant flow was recorded, with essentially no effect from position. Figure 11 presents the motor speed recorded during the test. The speed remains relatively constant in both orientations, even without a closed loop speed control system. Figure 12 presents the motor input power during this test. As expected from the constant speed and flow, power remains constant throughout the test. Figure 13 presents the averaged axial position during several vertical-horizontal transitions. The rotor position shifts slightly, as would be expected from the position control algorithm used. Figure 14 presents typical vibration spectra during the test. As with the 28 day test, the spectra are dominated by running speed. Very little change is seen between the horizontal and vertical orientations.
Figure 15 presents a typical hemolysis test result. The data indicate some initial damage during the startup of the loop, with an asymptotic decrease to a normalized index of hemolysis of 0.01 g/100 L.
Backup Bearing Sizing
The hydrodynamic thrust backup bearing is designed to provide an alternate load support mechanism in the event of a failed or compromised active magnetic thrust bearing. It is also designed to provide added load capacity in the event of a shock overload of the magnetic bearing (a high fall or vehicular accident, for example).
The surface of a hydrodynamic thrust bearing is designed in such a way that the space between the runner and the bearing forms a converging channel. As the runner is moving relative to the bearing, it drags the fluid (blood) into the converging channel, and the hydrodynamic pressure is developed. This hydrodynamic pressure can provide a substantial load carrying capacity, which strongly increases for decreasing gap size. During normal operation, the hydrodynamic bearing operates unloaded with a large gap, thus reducing power loss and minimizing blood damage. If the bearing function is needed, the gap closes because of rotor motion, allowing the bearing to generate load.
A thorough analysis of the fluid flow through a backup thrust bearing has been performed. The details of the analysis are beyond the scope of this article, and only a general description is provided here. The hydrodynamic performance of a thrust bearing has been modeled using the classical Reynolds equation of lubrication, which in polar coordinates assumes the form EQUATION
where r and θ are radial and angular coordinates, respectively, h is the gap between bearing surfaces, μ is fluid dynamic viscosity, ω is the rotational speed, and P is the unknown hydrodynamic pressure. When the appropriate flow geometry, fluid properties, and rotational speed are provided, equation 2 can be solved numerically to yield the generated hydrodynamic pressure field and the resulting hydrodynamic load.
Thrust bearings (which can carry axial load) typically consist of a number of pads (4–12) uniformly spaced around an angle of 360°. Because all pads are identical, the performance of a bearing can be analyzed by considering one pad only. Figure 16 shows a typical shape of one pad of a thrust bearing, whereas Figure 17 presents a typical solution to equation 2, which predicts the hydrodynamic pressure developed because of the rotation of the runner. Integration of the pressure profile yields the hydrodynamic load.
A challenge in designing a thrust bearing for the LVAD application is not just reaching the desired load capacity but also minimizing blood damage. A variety of models for hemolysis caused by blood flow can be found in literature. In general, they predict that the amount of blood hemolysis depends mainly on three factors: 1) level of shear stress in the fluid, 2) time duration for which blood is subject to the shear stress, and 3) the volume of blood subject to the shear stress. 14 By proper analysis of the solution to the Reynolds equation (equation 2), it is possible to determine the flow paths of blood particles, thus allowing for a determination of what volume of blood is subject to what level of shear stress and for how long. This information can then be used together with a model for hemolysis caused by blood flow to yield a prediction for the severity of blood damage in a bearing and an optimization metric for selecting the appropriate bearing configuration. With a set of bearing designs that provides adequate load, these designs may be thus ranked with regards to the potential for blood damage.
Figure 18 shows a typical streamline pattern over one pad of a thrust bearing. Fluid enters through the lower boundary of the bearing (Y = 0) and proceeds along the streamlines. It can be seen that, because of the converging gap between the runner and the bearing, the majority of fluid exits along the sides of the bearing.
The preliminary design study reveals that a number of bearing configurations can provide the required load with blood as the operating fluid. Typical power loss is on the order of a few watts or less. From the viewpoint of blood damage, it appears that bearings that have smaller regions of higher shear stress are preferred to bearings that reduce the peak shear stress but have larger regions of moderate stress. Testing is currently under way to confirm this result.
On the basis of the results obtained, it appears that the second generation LVAD pump is a viable candidate for use in this application. The range of flow and pressure are appropriate for this application. The pump durability under both steady state and transient conditions has been demonstrated. Low pump vibration has been observed in all cases. Reasonable performance with regard to hemolysis has been obtained. A backup bearing solution for the single active axis has been identified. Further design modifications are needed and will focus on reducing the overall size and weight of the pump and reducing the pumping losses from the impeller and wash flow path. The next step for this pump will be calf implant tests to confirm the promising results obtained to date.
The authors would like to acknowledge the support of the National Institutes of Health and Mohawk Innovative Technology, Inc. for funding this work. The efforts of Mr. Michael Albertini in fabrication and testing, as well as Mr. Thomas Russell for his assistance with testing, are also gratefully acknowledged.
1. Bernstein EF, Cosentino LC, Reich SM, et al: A compact, low-hemolysis, non-thrombogenic system for nonthoracotomy prolonged left ventricular bypass. Trans ASAIO 20: 643–653, 1974.
2. Saxton GA, Andrew CB: An ideal heart pump with hydrodynamic characteristics analogous to the mammalian heart. Trans ASAIO 6: 288–291, 1960.
3. Rosenbaum BJ, Kwak D, Benkowski RJ, Cetin K. NASA Tech Briefs, No. 7, p 86, July 1996.
4. Yada I, Golding LA, Harasaki H, et al: Physiopathological studies of nonpulsatile blood flow in chronic models. Trans ASAIO 29: 520-521, 1983.
5. Golding LA, Murakami G, Takatani S, et al: Chronic nonpulsatile blood flow. Trans ASAIO 28: 81–85, 1982.
6. Yozu R, Golding L, Yada I, et al: Do we really need pulse? Chronic nonpulsatile and pulsatile blood flow: from the exercise response viewpoints. Artif Organs
18: 638–642, 1994.
7. Anataki JF, Burgreen GW, Wu ZJ, et al: Development progress of the University of Pittsburgh streamliner: a mixed flow blood pump with magnetic bearings. ASAIO J 46: 194, 2000.
8. Qian KX, Zeng P, Ru WM, et al: A simple purge system enables the centrifugal pump absolutely free from thrombus. ASAIO J 46: 165, 2000.
9. Weiss WJ, Rosenberg G, Snyder AJ, et al: Steady state hemodynamic and energetic characterization of the Penn State/3M Health Care total artificial heart. ASAIO J 45: 189–193, 1999.
10. Allaire P, Hilton E, Baloh M, et al: Performance of a continuous flow ventricular assist device: magnetic bearing design, construction, and testing. Artif Organs 22: 475–480, 1998.
11. Masuzawa T, Kita T, Matsuda K, Okada Y: Magnetically suspended rotary blood pump with radial type combined motor-bearing. Artif Organs 24: 468–474, 2000.
12. Akamatsu T, Nakazeki T, Itoh H: Centrifugal blood pump with a magnetically suspended impeller. Artif Organs 16: 305–308, 1992.
13. Bearson GB, Olsen DB, Khanwilkar PS, et al: Implantable centrifugal pump with hybrid magnetic bearings. ASAIO J 44: 733–736, 1998.
14. Giersiepen M, Wurzinger LJ, Opitz R, Reul H: Estimation of shear stress-related blood damage in heart valve prostheses: in vitro comparison of 25 aortic valves. Int J Artif Organs 13: 300–306, 1990.