|Home | About | Journals | Submit | Contact Us | Français|
Cleveland Clinic’s continuous-flow total artificial heart has one motor and one rotating assembly supported by a hydrodynamic bearing. The right hydraulic output is self regulated by passive axial movement of the rotating assembly to balance itself with the left output. The purpose of this article is to present progress in four areas of development: the automatic speed control system, self-regulation to balance right/left inlet pressures and flows, hemolysis testing using calf blood, and coupled electromagnetics (EMAG) and computational fluid dynamics (CFD) analysis. The relationships between functions of motor power and speed, systemic flow, and systemic vascular resistance (SVR) were used for the sensorless speed control algorithm and demonstrated close correlations. Based on those empirical relationships, systemic flow and SVR were calculated in the system module and showed good correlation with measured pump flow and SVR. The automatic system adjusted the pump’s speed to obtain the target flow in response to the calculated SVR. Atrial pressure difference (left minus right atrial pressure) was maintained within ± 10 mm Hg for a wide range of SVR/PVR (systemic/pulmonary vascular resistance) ratios, demonstrating a wide margin of self-regulation under fixed-speed mode and 25% sinusoidally modulated speed mode. Hemolysis test results indicated acceptable values (normalized index of hemolysis <.01 mg/dL). The coupled EMAG/CFD model was validated for use in further device development.
We have developed a small continuous-flow total artificial heart (CFTAH) without valves or sensors (1–2). The CFTAH, 6 cm in diameter and 10 cm in length, has a priming volume of 37 ml and is implantable within the chest. The inflow cuffs are anastomosed to the atria, and left and right outlet grafts are anastomosed to the ascending aorta and pulmonary artery, respectively. The CFTAH features one motor and one rotating assembly supported by a hydrodynamic bearing (1–2). This device replaces the ventricles of the heart and delivers blood flow to both the systemic (left) and pulmonary (right) circulations. Based on previous in vitro and in vivo studies, the automatic speed control system has been advanced and evaluated through in vitro study. Also, the detailed phenomenon of the self-regulation, the results of hemolysis test, and coupled electromagnetics (EMAG) and computational fluid dynamics (CFD) analysis regarding this CFTAH have not been reported.
This purpose of this article is to present progress in four areas of our CFTAH program: the automatic speed control system, self-regulation to balance right/left inlet pressures and flows, hemolysis testing using calf blood, and coupled electromagnetics (EMAG) and computational fluid dynamics (CFD) analysis.
A primary pump design goal in this CFTAH project is simplicity, to avoid unreliable components and allow a more manufacturable and relatively low-cost device. This simple and unique design is derived from combining the left ventricular assist (LVAD) and right ventricular assist (RVAD) centrifugal pump features with one motor and one power cable (Fig. 1A). Impellers supporting the left and right circulation are mounted on opposing ends of the rotating assembly, allowing the left and right hydraulic environments to create opposing forces at opposite ends of the shaft (Fig. 1B). The nominal proportions of the rotating assembly are 2.8 cm in diameter by 6 cm in length.
The motor’s magnet assembly in the rotor is shorter than the motor’s steel laminations, allowing a degree of free axial movement in the direction of net axial force, primarily caused by an imbalance of pump inlet (atrial) pressures. This axial movement changes the pump geometry at the outer diameter of the right impeller, affecting relative left/right performance in a direction to correct the atrial pressure imbalance. Thus, the device is self-regulating, acting as an inlet pressure-balancing regulator while at the same time pumping both sides of the circulation. A new feature, a speed pulse, may be added to create a pulse in the patient and also provide an additional parameter for physiologic control. The rotating assembly is radially suspended using a blood-lubricated, hydrodynamic journal bearing, which is designed to minimize blood shear while maintaining stable operation. The stationary element – an yttria-stabilized polycrystalline tetragonal zirconia sleeve that lines the internal diameter of the stator – provides an excellent bearing surface and is nonconductive to avoid power losses from electrical eddy currents.
The self-regulating function is based on a balancing of hydraulic forces. When the left inlet pressure is higher than the right inlet pressure (caused by right overpumping, left underpumping, or right atrial suction), the rotating assembly is shifted by hydraulic forces to the right, thereby narrowing the right pump aperture and decreasing the right pump’s performance. When the right inlet pressure is higher than the left inlet pressure (caused by left overpumping, right underpumping, or left atrial suction), the rotating assembly moves to the left, thereby opening the aperture and increasing the right pump’s performance. This response is instantaneous in helping to relieve a possible suction condition while speed auto-regulation incrementally lowers pump speed in response to suction detection.
The resulting device is a simple total artificial heart (TAH), which has only one moving part and only one electromechanical component. There is no need for problematic components found in other TAH devices such as valves, sensors, actuation mechanisms, or flexible blood-pumping elements.
The automatic speed control system is based on two empirical correlations determined by bench characterization testing: the first correlation is between systemic flow and a function of motor power and speed (MPS), and a second correlation exists between systemic vascular resistance (SVR) and MPS. Using these correlations, real-time calculations of the systemic flow and SVR are performed in the control module and used to execute the speed control algorithm. Correlations were improved by subtracting the bearing power, which is derived from a bench calibration of power consumption vs. speed with impellers removed, from the total pump power.
System self-regulation and performance was evaluated under fixed-speed mode, 25% sinusoidally modulated speed mode, and automatic mode using a mock circulatory loop (3). The sensorless automatic speed control system adjusts the pump speed to maintain a target flow, depending on calculated SVR. The response to this automatic speed control system was evaluated over a range of hemodynamic conditions.
The mock loop was reconfigured as two separate loops to show the range in right pump output caused by imposed atrial pressure differences (1). To assess the relationship between right and left flows and pressure rise, the pump speed was fixed at 2,700 rpm under same left and right flow. Atrial pressure differences were set as 0, +10 mm Hg, and −10 mm Hg. The rotor position was maintained at neutral with 0 mm Hg of the atrial pressure difference. The rotor position shifted to the right or left with +10 mm Hg or −10 mm Hg of the atrial pressure difference [left atrial pressure minus right atrial pressure], respectively.
Using the full mock circulatory loop, the relationship between atrial pressure difference and SVR/pulmonary vascular resistance (PVR) ratio was determined to evaluate the range of self-regulation when coupled with the circulatory system. Normal human ranges of SVR and PVR are 900–1,400 dyne·s·cm−5 and 150–250 dyne·s·cm−5, respectively(4). Fixed-speed mode, and 25% sinusoidally modulated speed mode were used for this evaluation.
Pump flow and pressure rise are normalized with respect to speed to allow the plotting of a wide range of pump operating conditions on the same plot. Flow coefficient is expressed as Q/ωr3 (Q, the volume flow rate; ω, rotation speed; r, impeller radius), and pressure coefficient is expressed as Δp/(ρ(ωr)2) (ρ, fluid density; Δp, pressure rise).
To evaluate hemolysis characteristics of the developed CFTAH pump, an in vitro hemolysis test was performed. Two reservoirs containing bovine blood, representing the right and left circuits, were placed in a warm water bath set to 37°C. The pump was initially started at a fixed speed of 2,700 rpm, and the flow restrictor was adjusted to create a pump flow of 5 L/min. Hemolysis was also evaluated at a fixed mean speed of 3,400 rpm with a flow of 6.5L/min and a fixed mean speed of 2,000 rpm with a flow 3.6 L/min and also at a mean speed 2,700 rpm and flow of 5 L/min with a 25% sinusoidally modulated speed mode to produce pulsatile blood flow.
The blood was sampled every 30 min until the 4-hr end point was reached, then was centrifuged prior to measurement of the plasma free hemoglobin. As a control, 60 ml of blood was preserved in the warm water bath and sampled at 0, 60, 120, and 240 min. A normalized index of hemolysis (NIH) was adopted to compare the hemolysis properties of each pump:
where ΔPf Hb/Δt is the change in plasma free hemoglobin per unit time (g/L/min), V is the blood volume in the loop, Q is the flow rate of blood (L/min), and Hct is the blood hematocrit.
During operation, the rotating assembly reaches a position in which the fluid-generated hydraulic bearing forces are balanced by the EMAG forces exerted by the pump motor. The goal for the coupled EMAG/CFD analyses was to determine this “force-balanced” position of the rotating assembly over a range of pump operating conditions (5).
For the EMAG modeling, a three-dimensional model was developed to predict the effect of the motor current combined with the radial and axial rotor movement on the magnetic radial and axial forces and torques acting on the rotating assembly. These magnetic restoring forces due to rotor movement would then be compared with the CFD-predicted hydraulic forces to determine the “force-balanced” rotating assembly position.
For the coupled full model analyses, the rotating assembly was positioned at the journal bearing predicted radial “force-balanced” location and moved iteratively axially to a location that yielded right pump pressure similar to the in vitro test data. Comparison between the EMAG/CFD predicted rotating assembly position and that measured experimentally was then performed. The full three-dimensional pump model consisted of approximately 15 million elements and was modeled under steady-state flow conditions. Due to symmetry and blade clearance in the volute regions, the “Frozen Rotor” multi-frame of reference model was selected, fixing the rotating assembly in one blade orientation. The other CFD modeling parameters included: (1) a water/glycerin mixture as fluid (density of blood, used to compare with in vitro test data), (2) the Reynolds-averaged Navier-Stokes turbulence model (k/ω SST), and (3) volumetric flow rates of3, 6, and 9 L/min (spanning intended range of use).
The commercial software ANSYS (ANSYS Multiphysics and CFX 12.1, ANSYS Inc., Canonsburg, PA) was used to perform the CFTAH EMAG and flow analyses. The simulations were run using the computing resources at SimuTech Group, Inc.
Figure 2A/B shows the close correlations between MPS and SVR and between MPS and pump flow under non-pulse mode (no modulation mode), 25% sinusoidal modulation mode, and automatic mode, respectively. The MPSs for systemic flow and SVR are expressed as [mean motor power (W)]/[mean pump speed (kRPM)]2 and [mean motor power (W)]/[mean pump speed (kRPM)]3.5, respectively. In automatic mode, an algorithm coded into the controller estimates the SVR based on motor current and speed and uses this calculation to adjust the speed to follow the programmed target flow vs. SVR schedule.
Figure 3A/B shows the close correlations between measured and calculated SVRs and between measured and calculated pump flows, respectively.
Figure 4 shows an example of programmed target systemic flow vs. SVR relationship expressed by the black solid line. In the automatic mode, the controller adjusts pump speed to obtain the target flow in response to the calculated SVR. Black dots are the actual response in automatic mode. Depending on the calculated SVR, the pump speed algorithm follows the red line, changing the mean pump speed between 2,300 and 3,000 rpm to achieve the target flows.
Figure 5 shows the relationship between right and left flow and pressure rise in 2,700 rpm of pump speed and under the same right and left flows. The left pump characteristic remained constant, independent of atrial pressure, whereas the right pump characteristic varied widely, depending on the atrial pressure differences. This result demonstrates the increased right pump performance when right atrial pressure exceeds left atrial pressure, resulting in a self-regulating system.
Figure 6 shows the relationship between normalized flow and pressure rise for a range of various speeds. Because of the self-regulating interaction of the device with the circulatory system, separate right and left flow loops are required to impose a controlled atrial pressure difference on the CFTAH pump. Results show that the left pump’s characteristics are consistent over the full range of pumping conditions, whereas the right pump can vary widely in performance as needed to balance the left and right inlet pressures by passively opening and closing the right pump aperture in response to the hemodynamic environment.
Figure 7 demonstrates the relationship between the SVR/PVR ratio and atrial pressure differences using the full TAH mock circulatory loops. The SVR/PVR ratio was varied between 1 with low SVR and high PVR and 60 with high SVR and low PVR. The pump flow was varied between 3 and 9 L/min, with and without pulsation mode. The atrial pressure difference was maintained within ± 10 mm Hg, which is the one of the pump’s hemodynamic requirements.
Hemolysis was found to be at acceptable values (NIH < 0.01 mg/dL) (Table 1). In the 25% sinusoidally modulated speed mode, hemolysis values increased but remained within an acceptable range.
The key controlling physical parameter in the CFD modeling was the right pump aperture. This aperture is the fluid gap that connects the right pump impeller section with the right volute. For a given flow rate, decreasing the aperture size increased the maximum velocity through the aperture (Fig. 8). This increased aperture velocity led to increased wall shear stress and increased total pressure losses. The aperture functions as a throttling valve, providing the desired flow rates and inlet pressure balancing between the left and right pumps. Over the range of right pump apertures studied, the left pump performance was nearly constant. Reducing the aperture size increased the wall shear stress on the right pump impeller blades and on the pump housing in the aperture region. Based upon the CFD bench validation studies, during normal operation, the rotating assembly freely moves ± .8 mm (1.6 mm total) in the axial direction, but can move more during extreme conditions, such as conduit restriction or suction. Figure 9 indicates the location of the peak residence time within the fluid film of the hydrodynamic bearing. The blood in the hydrodynamic bearing is quickly exchanged and washed out.
The CFD results revealed that, for a fixed rotor position, there was very little difference between the performance of the separated pump sections (left, right, and bearing) and that of the combined pump: the left pump’s performance was nearly constant with the axial position of the rotating assembly, whereas the right pump’s performance varied to balance the left and right inlet pressures. This result was consistent with expectations, as only a small amount of fluid (approximately 40 ml/min) is exchanged between the two pump sections along the connecting journal bearing flow path. The importance of the journal bearing flow path is in coupling the pressure fields between the two pump sections and in providing the majority of the hydraulic radial force.
For the coupled EMAG/CFD full pump solutions, the CFD-predicted hydraulic performance, static pressure taps throughout the pump, and rotor torque matched in vitro data within 5–10%. Radially, the rotating assembly hydraulic forces balanced with the magnetic loads within 5%. An axial force imbalance of 0.3 to 0.4 N towards the left pump was found across the pump operating range. This force difference corresponds to a static pressure difference of 4 to 5 mm Hg. The EMAG/CFD-predicted axial position of the rotating assembly matched the experimental measurements within 0.010”.
The temperature of the blood within the stator’s hydrodynamic bearing cannot be directly measured due to small clearances, but has been modeled using a CFD analysis of the bearing region alone. Due to the thin fluid gap in the journal bearing, the left and right impellers have only a small effect on the flow patterns within the bearing region. Therefore, we focused this heat generation model on the journal bearing flow path by itself. This bearing simulation was performed in the stationary frame of reference at the nominal rotor speed of 2,800 rpm. Opening type boundary conditions, allowing the flow to both enter and exit through the ends of the bearing, were used. A 20 mmHg static pressure difference, based upon previous whole pump results, was applied from the left to right pump sides of the bearing. Blood entering the bearing region was assumed to be at 37 °C. Per this analysis, approximately 2 Watts of heat is generated through viscous dissipation within the fluid film of blood. However, due to the aggressive exchange of fluids through the ends of the bearing, the peak temperature due to viscous dissipation is limited to 0.58 °C above the bulk blood temperature. There is an additional transfer of heat into the blood from conduction and convection of heat generated by electrical losses in the stator winding, approximately 3 Watts, most of which would be conducted away via the large surface areas of the stator housing. Making a conservative assumption that half of the stator losses (1.5 Watts) are conducted through the bearing wall, the analytical estimate of peak fluid film blood temperature rise becomes about 1.0 °C, or 38 °C for a bulk blood temperature of 37 °C.
During the last half of the 20th century, the original pneumatic LVAD, invented in the 1960s, dramatically progressed to implantable centrifugal or axial assist devices with excellent durability, biocompatibility, and smaller size. Worldwide, several LVADs have already been established as the therapeutic option for the adult patient with the end-stage heart failure. As 10–13% of LVAD patients suffer from biventricular failure, requiring additional RVAD support (6–7), and some patients with severe biventricular failure need primary biventricular assist devices (8), the continuous-flow RVAD has also been developed: however, biventricular support by continuous-flow pump devices is still limited because of the difficulty in ensuring that such pumps will reliably respond to left/right flow imbalances caused by mismatching of left/right pump support.
Even though the first successful animal experiment of circulatory mechanical support was the implantation of a TAH (9), the more complex device system and structure failed to develop as quickly as the ventricular assist devices did. Development of the TAH would expand the indication not only for a bridge to heart transplants but also would indicate patients excluded from heart transplants as destination therapy. Moreover, for right ventricular failure incidental to LVAD support, a TAH would be an additional therapeutic option instead of a biventricular assist device. Currently, the AbioCor TAH (Abiomed, Inc., Danvers, MA) as a permanent pulsatile implantable system and the CardioWest TAH (SynCardia Systems, Inc., Tucson, AZ) as a pneumatic system have been approved for clinical use by the U.S. Food & Drug Administration and reported regarding implants for the patients (9–13). However, clinical trials of these large pulsatile devices requiring valves are limited by inconclusive thrombus formation and embolization and the inappropriate fit of the device in children and small women. Based on advanced LVAD technology and our rotary blood pump development programs, we have developed this CFTAH.
Natural cardiac hemodynamics change according to the level of individual activity, and the human heart responds naturally to changes in activity with an increase or decrease in heart rate controlled by the autonomic nervous system.
With a mechanical circulatory device, adjusting the pump speed according to a patient’s activity level is feasible; however, in the current clinical setting, only the Jarvik 2000 LVAD (Jarvik Heart, Inc., New York, NY) can adjust the pump speed and only by a caretaker or the patient manually changing the mean speed. To develop the simpler pump speed control system with high performance, we focused on the close correlation between MPS and flow or SVR; we consider that systemic flow can automatically be adjusted in response to calculated SVR. This study demonstrated that calculated flow and SVR were closely correlated with measured flow and SVR, respectively. Moreover, based on these relationships, this pump and control system succeeded in adjusting the target flow in response to the calculated SVR, changing the pump speed automatically. These close correlations make the sensorless speed control system possible. As shown in Figure 4, we can change the black solid line depending on the patients’ condition. If the patient condition is over this range, vasodilator or vasoconstrictor agents will support CFTAH control.
A sudden change in calculated pump flow may indicate an emergency situation, such as atrial suction caused by cannula obstruction or sudden volume loss, or pump stoppage. If this decreased flow continues for 3 s or more, pump speed immediately decreases at a programmed rate. After the 3 s at this “minimal” flow, pump speed gradually returns to the original or programmed recovery speed level if the situation is corrected. If atrial suction is not corrected, the alarm would continue to appear in the system interface, and the programmed pump speed or target flow would have to be significantly changed to release the atrial suction.
Balancing atrial pressures and pump flow is key in TAH function. This CFTAH design, with specific emphasis on the right side aperture, makes the sensorless passive self-regulation possible. As Figure 6 shows, this analysis of flow vs. pressure coefficients provides a general curve that can be used to determine the pressure-flow relationship of the left pump at any rotational speed (14). In a centrifugal or axial LVAD, pump performance is usually dependent on pump speed. Actually, the left pump’s performance in this CFTAH remains nearly constant, whereas the right pump’s performance varies widely depending on the pressure rise to balance the right and left pump flow by opening and closing the aperture appropriately.
Figure 7 demonstrates the performance of passive self-regulation that results from differential inlet pressure forces on the pump and displays the self-balancing of atrial pressure and right and left flows independently of SVR, PVR, and the SVR/PVR ratio. The atrial pressure differences are maintained within ± 10 mm Hg for the various range of SVR/PVR ratios.
More than 300,000 Americans die every year from heart failure, and approximately 15–20% of them die while waiting for transplantation (13). Therefore, mechanical support devices have been expected to be the substitution of donor heart or bridge to the heart transplants. Frazier et al. have developed a CFTAH system based on the MicroMed DeBakey axial-flow pump (MicroMed Technology, Inc., Houston, TX), performed in vitro evaluations, and showed the feasibility of using two of these continuous-flow devices as a CFTAH (15–16), with emphasis on its smaller size and greater durability. They also have replaced the native heart in calves with two Jarvik 2000 or two HeartMate II axial-flow LVADs (Thoratec Corp., Pleasanton, CA) (17–18). Loebe et al. have performed the initial clinical experience of total cardiac replacement with dual HeartMate-II axial flow LVADs for severe biventricular heart failure (19). Those procedures would develop with TAH development.
Timms et al. have reported on a CFTAH that is similar to our CFTAH in the use of two impellers and a hydrodynamic bearing (20). The movement and position of the rotor in their pump are sensed by using a position flow sensor and actively controlled by motor magnets and bearing electromagnets.
Innovation of our CFTAH would permit its implantation for severe heart failure patients, including women and children. The TAH serves as the bridge to heart transplantations and destination therapy for clinically ill patients with end-stage biventricular heart failure. The TAH will contribute to expand the indication for the patients excluded from the heart transplantation.
In this CFTAH, the atrial pressures are balanced by the passive self-regulation, and pump flow is maintained by the automatic speed control system. The automatic speed control adjusts a target flow in response to calculated SVR. A wide margin of self-regulation in left/right performance was demonstrated in the mock loop, and the modified pump design resulted in acceptable hemolysis values in vitro. The coupled EMAG/CFD analyses were validated with in vitro results.
Supported by National Institutes of Health Grant R01 HL096619-01.
Presented in part at the 19th Congress of the International Society for Rotary Blood Pumps, September 8–10, 2011, Louisville, KY, USA
This study was supported by National Institutes of Health Grant R01 HL096619-01 (Principal Investigator, L.A.R. Golding; Inventor, D.J. Horvath.). Dr. Golding is a staff at the Cleveland Clinic and also is the Chief Medical Officer for Cleveland Heart, Inc..