Annals of Biomedical Engineering

, Volume 39, Issue 1, pp 324–336 | Cite as

Numerical, Hydraulic, and Hemolytic Evaluation of an Intravascular Axial Flow Blood Pump to Mechanically Support Fontan Patients

  • Amy L. Throckmorton
  • Jugal Y. Kapadia
  • Steven G. Chopski
  • Sonya S. Bhavsar
  • William B. Moskowitz
  • Scott D. Gullquist
  • James J. Gangemi
  • Christopher M. Haggerty
  • Ajit P. Yoganathan


Currently available mechanical circulatory support systems are limited for adolescent and adult patients with a Fontan physiology. To address this growing need, we are developing a collapsible, percutaneously-inserted, axial flow blood pump to support the cavopulmonary circulation in Fontan patients. During the first phase of development, the design and experimental evaluation of an axial flow blood pump was performed. We completed numerical modeling of the pump using computational fluid dynamics analysis, hydraulic testing of a plastic pump prototype, and blood bag experiments (n = 7) to measure the levels of hemolysis produced by the pump. Statistical analyses using regression were performed. The prototype with a 4-bladed impeller generated a pressure rise of 2–30 mmHg with a flow rate of 0.5–4 L/min for 3000–6000 RPM. A comparison of the experimental performance data to the numerical predictions demonstrated an excellent agreement with a maximum deviation being less than 6%. A linear increase in the plasma-free hemoglobin (pfHb) levels during the 6-h experiments was found, as desired. The maximum pfHb level was measured to be 21 mg/dL, and the average normalized index of hemolysis was determined to be 0.0097 g/100 L for all experiments. The hydraulic performance of the prototype and level of hemolysis are indicative of significant progress in the design of this blood pump. These results support the continued development of this intravascular pump as a bridge‐to‐transplant, bridge‐to‐recovery, bridge-to-hemodynamic stability, or bridge-to-surgical reconstruction for Fontan patients.


Artificial right ventricle Blood pump Cavopulmonary assist device Heart pump Intravascular blood pump Mechanical cavopulmonary assist Pediatric circulatory support Single ventricle physiology 


Worldwide over 1 million children are born each year with congenital heart anomalies.2,5 A special heterogeneous subset of this group encompasses babies born with malformations of their heart chambers and connecting vessels who have only one functional ventricle. The incidence of patients born with functional univentricular physiology is approximately 2 per every 1000 births. Without surgical intervention, this combination of cardiac anomalies is fatal within the first 2 weeks of life.11,27 For these patients, the current treatment paradigm consists of three staged, open heart surgeries, which culminate in the Fontan physiology, as originally described in 1971 for patients with tricuspid atresia.7,8

The Fontan procedure has remained the prominent means of surgical palliation for these patients with modifications and staging evolving over the past three decades. Each patient presents unique challenges to surgical teams, and thus numerous variations of these palliative procedures have emerged, such as the creation of a total cavopulmonary connection (TCPC) as an example of the extracardiac Fontan. This procedure connects the inferior vena cava (IVC) and superior vena cava (SVC) directly to the main right pulmonary artery.22

In a univentricular Fontan circulation, there is no subpulmonary power source to drive systemic venous blood through the lungs. As a result, systemic venous pressure is progressively elevated, and, under failing conditions, ventricular end-diastolic volume is reduced and suboptimal.12 To address these issues, surgical optimization of the TCPC has been the focus of clinicians who seek to streamline cavopulmonary designs and to minimize energy losses through the vessel configuration.6,18 In recent years, progress in surgical advances of the TCPC has reached a plateau. Therefore, a rising interest in the implementation of mechanical circulatory assistance in the cavopulmonary connection has emerged for the growing population of ailing Fontan patients.27

As a new therapeutic alternative for patients with a failing Fontan physiology, we are developing a collapsible, percutaneously-inserted, magnetically levitated axial flow blood pump to support the cavopulmonary circulation.9 This pump is intended to serve as a bridge-to-transplant, bridge-to-hemodynamic stability, or bridge-to-surgical reconstruction. It is designed to provide 4 weeks of temporary mechanical circulatory support. Implantation of the intravascular axial flow pump involves insertion into a femoral vein and advancement toward the IVC or extracardiac conduit at the entrance of the TCPC. This study presents the numerical design and development of the impeller component of the cavopulmonary assist device. The impeller is a critical region of a rotary blood pump as its unique geometry and blade configuration define the operating capabilities. Common design practices were employed to develop the geometry and blade configuration of the impeller and to evaluate its performance using a three-pronged approach. We performed numerical modeling of the impeller using computational fluid dynamics (CFD) analysis, manufactured a pump prototype, completed the hydraulic performance testing of the pump, and executed blood bag experiments (n = 7) to measure the levels of hemolysis produced by the pump.

Materials and Methods

Conceptual Pump Design—Long-Term Goal

Figure 1 illustrates the conceptual design of the intravascular blood pump. The axial flow pump is designed for percutaneous positioning in the IVC. The outer protective cage has radially arranged filaments that serve as touchdown surfaces to protect the vessel wall from the rotating components. Each filament is hydrodynamically designed to reduce drag and to maximize energy production from the rotating, engineered impeller blades. Currently, the rotating pump consists of an impeller with four uniquely designed and helically wrapped blades to maximize energy transfer. Pump rotation will be induced through a motor-magnetic bearing suspension, which levitates and rotates the impeller within the protective cage of filaments. An outlet nose is also located at the outflow of the pump to physically limit the axial movement of the impeller, to connect the cage filaments, and to house bearings which support the impeller during operation. The blade tip-to-tip diameter of the first generation design is approximately 14 mm in the fully open configuration. In this study, for the purposes of measuring hydraulic performance, the pump prototype was mounted to a drive-shaft that was supported by mechanical bearings. The target design for the intravascular pump is to generate flow rates of 0.5–4 L/min with pressure rises of 2–25 mmHg for operating rotational speeds of 3000–8000 RPM.

Conceptual design of the axial flow blood pump: (a) the device consists of a protective sheath with cage filaments, a rotating shaft and catheter, an impeller blades, diffuser region, and inlet and outlet sections. (b) Position of the cavopulmonary assist device in the IVC of the TCPC for Fontan patients. It is designed to augment pressure and thus flow in IVC and subsequently drive blood into the left and right pulmonary arteries (LPA and RPA) while supporting the incoming flow from the SVC

Impeller Geometry and Configuration

The focus of this study is on the design and evaluation of the impeller blades to enable the appropriate generation and transfer of rotational energy to the fluid. This group has previous experience in the design of impellers for ventricular assist devices28 and has applied that knowledge-base to develop a new category of intravascular blood pumps for mechanical cavopulmonary assist. The basis of the pump design in this study originated, in part, from this prior work. Nevertheless, the design of an impeller region for any rotary blood pump requires the execution of an iterative process based on target operating conditions, size constraints, and the tools available for numerical and experimental evaluation.

Traditional impeller blade geometry consists of the hub diameter, blade root, blade tip, leading and trailing edge angles, blade thickness, chord, pitch, skew, number of blades, and blade area (mean-width ratio).10,25 The leading and trailing edge angles of the blades are carefully selected to achieve a particular hydraulic performance and range of operating conditions. These angles are considered to be the most critical parameters to ensure hydraulic performance. Blade thickness and tip clearance between the blade and the housing for blood pumps, given the millimeter scale, are dictated in part by manufacturing tolerances and constraints as well as by the motor-drive configuration. Energy transfer across an axial blood pump occurs as the blade imparts rotational energy to fluid. Newton’s law of motion, as applied to an impeller in the form of fluid traversing the rotor, states that the torque on the impeller is equal to the changing rate of an angular momentum of fluid.10,25

Standard pump design equations10 were employed to derive the dimensions and characteristics of the impeller blade geometry based on operating specifications (flow, pressure rise, RPM). For the same head and capacity requirements, pump operational speed is inversely proportional to impeller diameter; thus, a smaller diameter requires a higher rotational speed to generate the same head and flow performance. For a blood pump, higher rotor speed translates into higher fluid shear stress, which could have a traumatic effect on blood, especially in the narrow clearance gaps between rotating and stationary surfaces. The larger the clearance gaps, the smaller the fluid stresses for a particular rotational speed. However, wider clearance gaps lead to energy loss, potential flow separation, and recirculation. Therefore, during this initial design process, we determined a balance among fluid stress levels, hydraulic performance, and size constraints.

In further consideration of the impeller design and in contrast to ventricular assist of the systemic circulation, mechanical cavopulmonary assist requires much lower pressure generations (almost 20 times less depending upon a patient’s body surface area). A pressure augmentation of 2–5 mmHg may be sufficient to improve the hemodynamics of a Fontan patient. For our application, the demands for hydraulic efficiency are reasonably low and a moderate degree of recirculation is desirable to facilitate surface washing of impeller blades and cage filaments. Risk of thrombosis in this low pressure environment is a critical consideration. The design of a non-obstructive flow path using the vessel wall as the pump housing is expected to assist in the maintenance of a continuous wash over all surfaces and minimize regions of recirculation or stagnant flow that would encourage protein or platelet deposition.

All of these design principles have been carefully considered during this preliminary development phase. In this study, the initial starting point for the basic geometry was based on previous impeller designs.28 Optimization of the blade geometry was performed through sequential parametric analysis in which CFD simulations provided valuable information about the impact of geometric adjustment on hydraulic performance and fluid stress levels. The findings for a given geometric adjustment served as the foundation, from which, to optimize the impeller design. This analytical process culminated in the manufacture of a pump prototype that was subsequently evaluated in a hydraulic test-rig for performance and hemolytic levels.

Numerical Analysis

During optimization of the impeller geometry, we used ANSYS CFX 11.0 (ANSYS Incorporated, Canonsburg, PA, USA) to simulate the flow through the pump models. The CFD analysis involves dividing or discretizing the complex 3D fluid domain (model) into smaller volumetric regions or mesh elements with common nodes of linkage (Fig. 2). We are able to characterize the behavior of the fluid at each nodal location and to solve the fluid dynamics of the entire domain of interest. A grid density and a convergence study were completed for grid quality assurance. For the prototype design, the mesh consisted of 898,500 hexahedral elements.

Computational model of axial flow impeller: (a) meshed model having four regions—inlet pipe, spindle, impeller with four blades, and outlet. (b) Numerical analysis involves dividing the complex 3D fluid model into smaller volumetric regions or mesh elements with nodes to mathematically characterize the fluid dynamics of the pump

Following generation of the mesh, boundary conditions for the computational flow model were set to initiate the simulations. Flow through the pump was defined to be steady with constant boundary conditions and velocities. The no-slip boundary condition was assigned to the stationary vessel walls such that the fluid velocity values along the boundary would equal zero. The inlet vessel walls, vessel shroud of the pump, and outlet vessel wall surfaces were defined as stationary. The pump domain was defined to be in the rotating reference frame, whereas the inlet vessel and outlet section were specified to be in the stationary reference frame. The spindle at the inlet, rotor surface at the outlet, and shaft were defined to be spinning surfaces. The frozen rotor interface linked the regions of differing reference frames. A uniform mass inflow rate and rotational speed were specified for each simulation. The outflow pressure was set to be a constant static value to establish the outlet boundary condition. A constant physiologic fluid viscosity and density value of 3.5 cP and 1050 kg/m3, respectively, were applied for each simulation.

Turbulent flow conditions are expected to dominate in the axial flow blood pump with Reynolds (Re) numbers above 6000. The global Re number was approximated by using the impeller diameter (D), as a characteristic length as per Re = ρωD 2/μ for pump design, where ρ signifies the fluid density, μ the fluid viscosity, and ω the angular speed of the rotor. We also determined the Taylor number (Ta) for the impeller to be above 397. The Ta number is an indicator of flow instability, which is in turn a precursor to turbulent flow conditions. According to Schlichting,23 experiments have demonstrated that flow conditions become turbulent for Taylor numbers greater than 400. These findings support the modeling of turbulent flow conditions in the axial flow pump.

The selection of a turbulence model is important to the physical accuracy of the CFD simulations, yet there is no definitive conclusion as to which turbulence model is appropriate for miniature blood pumps. The k–ε turbulence model has been used for several years in the evaluation of our prior pump designs with experimental validation.28 Thus, this turbulence model was employed for this study as well.

The simulations were carried out to predict the pump performance for flow rates of 0.5–6 L/min for rotational speeds of 3000–6000 RPM. It is never expected that the pump would need to operate at 6 L/min; this range was examined to capture the full operability of the blood pump. In addition, we used CFD to estimate a scalar fluid stress, which identifies regions of high stress (possible hemolysis) and low stress levels (risking thrombosis).1,24 We have adopted the approach, as developed by Bludszuweit,3 to account for the 3D flow field and calculated the scalar fluid stress (σ) levels. The exposure time to such levels of stress is also important; therefore, we analyzed fluid streamlines as indicative of numerically predicted fluid residence times within the cavopulmonary assist device.17

Prototype Manufacturing

A 3D computer-aided design model of the blood pump was generated from the numerical model using SolidWorks (SolidWorks Corporation, Concord, MA, USA). The prototype was manufactured using a rapid prototyping technique called stereolithography that was equipped with solid-state (Nd:YVO) lasers. Figure 3a shows the prototype with four helically wrapped blades around the hub. The prototype measured 30 mm in length with an impeller blade height of 2.5 mm and maximum hub diameter of 9 mm. The clearance between the blade tip and vessel wall was 2.0 mm. The IVC or extracardiac conduit vessel was constructed from an Acrylic resin with a diameter of 18 mm. The inlet vessel before the pump prototype had a length of 120 mm to obtain a fully developed flow conditions. Figure 3b displays the components of the pump housing, and Table 1 lists the design characteristics for this impeller.

Physical prototypes manufactured for performance testing: (a) axial impeller pump prototype with helically-wrapped blades, (b) Acrylic housing with an entrance length of pipe and ports to remove air bubbles from the system through filling and draining

Table 1

Design specifications for the impeller pump

Design characteristic


Number of blades


Hub diameter (leading edge) (mm)


Hub diameter (trailing edge) (mm)


Blade height (mm)


Tip clearance (mm)


Length (mm)


Vessel diameter (mm)


Length of inlet pipe (mm)


Hydraulic Test-Rig

To evaluate the performance of the pump prototype, a hydraulic test-rig (Fig. 4) was designed and constructed. The flow loop consists of inlet and outlet reservoirs, inlet vessel with an entrance length of pipe, a motor, its controller, differential pressure transducer, flow probe, clamp, and pump prototype. The 4-bladed prototype was placed into the vessel near the outlet reservoir. The blood analog fluid traveled from the inlet reservoir through a nozzle into the inlet pipe until reaching the pump. Flow then continued across the rotating impeller blades into the outlet reservoir tank. The fluid returned from the outlet reservoir to the inlet via connecting Tygon tubing. A clamp was located on the flexible tubing to adjust the resistance of flow in the loop and subsequently flow rate. An ultrasonic flowmeter (Transonic Systems Inc., Ithaca, NY, USA) was employed to measure flow rates. The 4-bladed impeller prototype was supported on a rear-mounted, drive shaft that passed through the outlet tank and a hydraulic fluid seal into housing of the prototype. The shaft was itself supported and aligned by a mechanical ball bearing and needle bearing (McMaster-Carr, Atlanta, GA, USA) and then connected to a high-speed, DC brushless motor (MicroMo Electronics Inc., Clearwater, FL, USA) with a controller. Pressure taps were located in the inlet and outlet tanks such that a differential pressure transducer (Validyne Engineering, Northridge, CA, USA) was able to capture the pressure rise across the pump. Data collection software (LabJack Corporation, Lakewood, CO, USA) was used to simultaneously measure the pressure rise and flow rate at a sampling rate of 50 Hz. The data acquisition software also enabled us to measure rotational speed (RPM) from the motor controller. A water/glycerin mixture (60/40 wt%) was employed as the fluid medium and blood analog solution. A Cannon–Fenske viscometer and hydrometer were used to verify the fluid properties of the water/glycerin mixture: a viscosity of 3.542 ± 0.176 cP and a specific gravity of 1.1 ± 0.002.

Experimental test loop to evaluate the hydraulic performance of the axial impeller pump: Components consist of inlet and outlet reservoirs, inlet vessel, a motor, its controller, differential pressure transducer, flow probe, clamp, and impeller prototype

To complete a quantitative comparison of the numerical predictions and experimental data, a nondimensional analysis was performed. This analysis involved collapsing the experimental and computational data into pressure coefficients and flow coefficients28:
$$ {{\psi}} = C_{{{\psi}}} {\frac{\Updelta P}{{\rho N^{2} R^{2} }}} $$
$$ \phi = C_{\phi } {\frac{Q}{{NR^{3} }}} $$
where ψ represents the pressure coefficient, ϕ the flow coefficient, ρ the density of the fluid (kg/m3), R the radius of the impeller (mm), N the rotational speed (RPM), Q the flow rate (L/min), ΔP the pressure rise of the pump (mmHg), C ψ a pressure factor equal to 1.2157 × 1010, and C ϕ a flow factor equal to 1.5195 × 105. The purpose of the C ψ and C ϕ coefficients is to facilitate the nondimensionality of Eqs. (1) and (2). A regression analysis was then performed using a polynomial trendline for each coefficient data set. The statistical regression model generated characteristic constants (β0, β1, β2, and β3) as described in the following equation, where the subscript ‘type’ indicates either the CFD or experimental data:
$$ \psi_{\text{TYPE}} = \beta_{3} \left( {\phi_{\text{TYPE}}^{3} } \right) + \beta_{2} \left( {\phi_{\text{TYPE}}^{2} } \right) + \beta_{1} \left( {\phi_{\text{TYPE}} } \right) + \beta_{0} $$
Data normality was assessed, and an F-test was performed to determine the suitability of the regression analysis. A student t-test was completed on each characteristic constant to ensure significance.

Hemolysis Experiments

After rigorous cleaning, the hydraulic test-rig (Fig. 4) was used for hemolysis testing. The accumulated levels of plasma-free hemoglobin (pfHb) were measured over time to characterize the degree of hemolysis for the impeller prototype. Experiments (n = 7) were performed according to industrial and recommended test standards based on the American Society for Testing and Materials (ASTM) standards F1841-97, F1830-97 and F756-00.15 The internal aluminum surfaces of the hydraulic test-rig were anodized to provide a resilient coating on the surfaces of the rig and reduce sharp metallic edges.

The rig was partially submerged into a warm water bath throughout the experiment to maintain the temperature of the blood around 37 °C. A thermocouple was placed into the inlet reservoir to measure the temperature of the blood during the experiments. An inlet and an outlet NPT fitting on each reservoir were used as access ports to sample blood. The flow loop was filled with phosphate-buffered saline (PBS) solution to wet the internal surfaces of the rig.

In accordance with the Virginia Commonwealth University’s Institutional Animal Care and Use Committee (IACUC) approved study (AM#10184), fresh bovine blood was collected from a healthy donor by venipuncture using a large bore needle. The bovine blood was collected in a standard 500 mL bags containing citrate phosphate dextrose adenine anticoagulant, and the blood bags were gently placed into a water bath for slow warming to 37 °C after collection and arrival to the lab.

After draining the PBS solution, blood was introduced into the loop through a 150-µm filter. Once full, a blood sample was collected for the baseline measurement (initial condition, time = 0). The pump was started; the operating rotational speed was achieved by gradually increasing the RPM. Within 15 min the pump reached 3000 RPM, and the flow rate was set to 3.7 L/min. At each hourly interval, approximately 4 mL of blood was collected as a sample from the inlet tank. From each collected sample, 2 mL of blood was used to measure the sample hematocrit (Hettich Haematokrit 210 centrifuge, Oxford, CT, USA). Two measurements of hematocrit (4000 RPM for 8 min) were performed and averaged.

In addition, two separate 2 mL samples were transferred to tubes containing ethylenediamine tetra acetic acid. These samples were centrifuged for 8 min at 4000 RPM to separate the plasma layer from the packed red cells. Using a pipette, the plasma layer was transferred into a 1.5-mL polystyrene disposable cuvette for spectrophotometric analysis. The optical density or absorbance was determined for three wavelengths: 576.5, 560, and 593 nm. The spectrophotometer (Genesys 10Vis Spectrophotometer, ThermoFisher Scientific, Waltham, MA, USA) was calibrated using a blank cuvette prior to each analysis of the sample for three wavelengths. We calculated the pfHb based on the weighted difference in absorbance measurements, according to the Cripps method14:
$$ {\text{pfHb}}\;({\text{mg}}/{\text{dL}}) = 177.6 \times \left[ {A_{1} - (A_{2} + A_{3} )/2} \right] $$
where A 1 corresponds to the absorbance at wavelength (λ1) of 576.5 nm, A 2 the absorbance at wavelength (λ2) of 560.0 nm, and A 3 the absorbance wavelength (λ3) of 593.0 nm. The normalized index of hemolysis (N.I.H) was determined using the following equation:
$$ {\text{N}} . {\text{I}} . {\text{H}}\;({\text{g}}/100\;{\text{L}}) = V * \Updelta {\text{pfHb}} * {\frac{{(100 - {\text{Hct}})}}{100}}*{\frac{100}{\Updelta t * Q}} $$
where ΔpfHb corresponds to the increase in pfHb (g/L) over the sampling time interval, Hct represents hemotocrit (%), Δt the sampling time (min), Q the flow rate (L/m), and V the circulating volume of the hydraulic loop (L). For the statistical analysis, a 1-way ANOVA using Minitab 16.0 (Minitab Incorporated, State College, PA, USA) and pairwise comparison at each hourly interval using a Tukey simultaneous analysis were carried out. All statistical analyses were based on a preset α value of 0.05.


Computational Predictions

Figure 5 illustrates the numerical predictions for the hydraulic performance of the blood pump as determined for flow rates of 0.5–6 L/min over rotational speeds of 3000–6000 RPM. The pressure rise across the pump for a given rotational speed decreased with increasing flow rate, as expected due to flow losses. A pressure rise range of 2–30 mmHg was achieved for the pump design over these operating conditions.

Hydraulic performance results for axial flow impeller pump, illustrating the numerical predictions in comparison to the experimental findings from testing the prototype

Figure 6 displays the scalar stress distribution along the impeller surfaces and the fluid streamlines through the pump model at 3 L/min and 4000 RPM. The highest fluid stresses were found at the leading edge of the impeller blades and along the blade tip surfaces. These fluid stress levels did not exceed 61 Pa. At the outlet of the pump, a strong rotational component was found in the flow leading to higher shear stresses. All streamlines exited the model within 0.19 s, indicating a short residence time within the pump.

Scalar stress distribution along the rotating impeller surfaces of the pump and flow streamlines: (a) scalar fluid stresses on the impeller with the highest levels along the blade tip surfaces (61 Pa), as would be expected. (b) Fluid or particle streamlines through the model predicting a strong rotational component to the flow at the outlet of the pump. All streamlines exited the model within 0.19 s

Measured Hydraulic Performance

In these experiments, we measured a pressure rise across the pump prototype for each flow rate and rotational speed. The prototype delivered a pressure rise range of 2–26 mmHg over flow rates from 0.4–7 L/min for 3000–6000 RPM. Figure 5 illustrates the measured performance curves for the blood pump prototype in comparison to the numerical predictions. Trends in the experimental performance data met expectations with a decrease in pressure rise for increasing flow rate at a given rotational speed.

Quantitative Comparison of Numerical Predictions and Experimental Data

To carry out the nondimensional analysis and compare the numerical findings to the prototype performance, Fig. 7 shows the calculated pressure coefficients as function of the flow coefficients for the experimental and CFD data. The regression trendlines for each data set were determined to be:

Nondimensional analysis for comparison of numerical predictions to experimental data of the hydraulic performance of the axial flow pump prototype

$$ \Uppsi_{{{\rm EXPERIMENTAL}}} = - 1.3749(\phi)^{3} + 1.4255(\phi)^{2} - 0.5657(\phi) + 0.1885 $$
$$ \Uppsi_{{{\rm CFD}}} = - 0.6591\left(\phi\right)^{3}\,+\, 0.9391\left(\phi \right)^{2}\,-\,0.4953\left(\phi\right) + 0.1805 $$
Table 2 lists the results of the regression analysis and the statistical evaluation of the coefficients. Normality for both data sets was found. The F-test indicated a strong significance (p < 0.001) in the correlation of the polynomial trendlines to capture the trends of the experimental and CFD data. The correlation coefficients (R 2) and adjusted R 2 values were above 0.94 for the trendline of the experimental data and above 0.98 for the CFD predictions. A Student’s t-test indicated the significance for each of the coefficients in the regression models. Using these models, the average and maximum deviation between the experimental and numerical data sets were determined to be 4 and 6%, respectively.
Table 2

Regression analysis and the statistical evaluation of the coefficients











R 2

R adj 2

Exp. data












CFD data












Hemolysis Results

During the blood bag experiments (n = 7), duplicate data measurements were performed for each sample interval. Variability in the rotational speed of the prototype, flow rate, and temperature of the loop was minimal during all experiments. Values of sample hematocrit remained between 26.5 and 30% with a slight decline over the 6-h period for all experiments. Figure 8 demonstrates the pfHb levels as a function of time (hour) for the six experiments. A consistent, linear increase in the pfHb was found (p < 0.001). The maximum pfHb level was determined to remain below 21 mg/dL. Based on confirmed data normality, the 1-way ANOVA indicated that all of the pfHb data had a strong dependence (p < 0.001) on time (hour). The Tukey simultaneous analysis with pairwise comparison further indicated that the mean pfHB at each hourly interval was significantly different (p < 0.001) from the other time points, except for hour three and four relative to each other. Table 3 lists the N.I.H levels for each of the 6-h of blood bag experiments. The average and maximum N.I.H levels were determined to be 0.0097 and 0.0107 g/100 L, respectively.

pfHb concentration during hemolysis testing over a 6-h period

Table 3

N.I.H for 6-h blood bag experiments (n = 7)


N.I.H value (g/100 L)




















The treatment of single ventricle anomalies presents a significant challenge for clinicians caring for patients with congenital heart disease. Clinicians have theorized that a mechanical pump specifically designed to augment pressure from the great veins through the lungs would ameliorate the poor physiology of the failing univentricular circulation.27 Current mechanical blood pumps were designed and developed for adult patients with congestive heart failure and to support the systemic circulation, not the unique anatomic physiology of the cavopulmonary connection. These devices produce pressures far exceeding the desired range to be used for cavopulmonary support. A pressure augmentation of 2–5 mmHg may be sufficient to augment the energy required to drive blood flow in the cavopulmonary circulation, reduce systemic venous pressure, and improve ventricular preload.

Several institutions are pursuing research and development of cavopulmonary assist devices. Rodefeld et al. 20 have successfully demonstrated the use of the axial flow Hemopump to assist cavopulmonary flow in animals. This research group is also developing an innovative percutaneously-implantable, expandable propeller blood pump as a cavopulmonary assist device.26 Limitations of this design include a wide distance between the rotor and blade-tip, increasing shear stresses, and an extremely short contact time with the thin propeller blades, preventing flow control downstream. A new design, recently presented21 by this team, indicates their exploration of a unique bi-conical impeller pump or disk to stabilize flow directionality in the TCPC junction. Riemer et al. 19 at the Stanford University have used a sheep model of the TCPC to test the response to the Thoratec HeartMate II axial flow blood pump (Thoratec Corporation, Pleasanton, CA, USA). These studies demonstrated a baseline return of cardiac output, IVC flow, and arterial pressures. Similarly, the research team at the University of Colorado has made steady progress through numerical and in vitro studies on the development of an axial flow pump for proposed use in the IVC only.13 The axial flow pump being developed by Lacour-Gayet et al. 13 requires invasive implantation similar to a ventricular assist device with connection cannulae, titanium pump housing, and blood-contacting components. As proven by these research teams, the standard axial flow pump design has suitable characteristics for cavopulmonary assist, but the use of a conventional ventricular assist device creates risks of flow obstruction in the event of device failure and implantation requires invasive surgery.

In addition, existing intravascular blood pumps, such as the Impella Technology, have limitations for use as cavopulmonary assist devices. The Impella technology represents a specific group of intravascular blood pumps, which are designed to augment the pressure and flow in the systemic circulation. For example, the Abiomed Impella Recover 2.5 (4 mm in fixed-diameter pump) is designed to support the systemic circulation by generating 2.5 L/min at 33,000 RPM for a pressure rise of 100 mmHg. It has been developed to operate for 5 days and to sit across the aortic valve into the native reservoir of the left ventricle; this configuration is substantially different from cavopulmonary assist. In contrast to fixed-diameter pumps, our collapsible axial flow pump will have a protective cage and rotating blades that extend into 90% of the vessel diameter. This characteristic will allow for direct control of blood flow and its forward thrust. The vessel wall itself serves as the housing for our blood pump, contrary to cannulae connections in most fixed-diameter intravascular pumps. In further contrast, standard fixed-diameter, intravascular axial flow pumps that are placed in the IVC would require an occlusive mechanism to prevent recirculation around the pump body and retrograde flow in the fluid layers away from the pump and closer to the vessel wall. In these fluid layers away from the fixed-diameter impeller blades, blood flow will respond to the pressure gradient; the higher static pressure at the pump outlet will result in retrograde flow if no occlusive mechanism is present, as observed by Rodefeld et al. 20 Such an occlusive mechanism is obstructive to flow and creates regions of irregular flow patterns and potential flow stagnation, risking hemolysis and thrombosis. Furthermore, the use of pumps that were originally designed for the systemic circulation in the IVC for cavopulmonary assist would require a reduction in rotational speed in an attempt to reach a reasonable pressure for cavopulmonary assist, which may limit the flow range and result in consistent operation off-design. Operation of any blood pump consistently off-design increases the risk of angular mismatch of the fluid velocity and blades, which produces irregular flow patterns and high fluid shear stresses.

In support of this effort to develop alternative therapeutic options for Fontan patients, we are developing an intravascular, axial flow blood pump with a magnetically-levitated rotor and a uniquely-shaped protective cage. In this study, we present the results of an impeller design, including a numerical analysis of the pump, hydraulic evaluation of a prototype, and hemolysis testing. The numerical model and pump prototype both achieved the targeted hydraulic performance by generating flow rates of 0.5–4 L/min with pressure rises of 2–25 mmHg for rotational speeds of 3000–6000 RPM. An excellent agreement between the numerical predictions and the experimental data from prototype testing was obtained. The maximum deviation between CFD and experimental results was 6%, which is better than the standard 10–20%.9 The scalar stress levels within the blood pump were also well below the design criterion of 425 Pa for the development of axial flow assist devices.27 The linear and increasing trend of the pfHb concentration during the hemolysis experiments met expectations.3 The design objective for the maximum N.I.H level for an adult left ventricular assist device (LVAD) during a 1-month support duration is 0.01 g/100 L.16 The axial flow pump prototype in this study demonstrated an average N.I.H of 0.0097 g/100 L for all experiments and a maximum N.I.H level of 0.0107 g/100 L. A direct comparison of this design objective to our results for the repeated 6-hr study is not encouraging. However, since our flow rate is much lower (3.7 L/min versus 6 L/min for an LVAD), our N.I.H. measurements will always be 1.5–2 times higher (see Eq. 5). Correcting for this difference in flow rate yields hemolysis results that meet expectations and are comparable to clinically used blood pumps. This study represents a strong initial effort to develop an intravascular cavopulmonary assist device.


This study, however, has several limitations to be addressed. For instance, the size of the axial flow impeller is currently too large for intravascular assist; however, the current design exceeds expectations and, thus, provides inherent design flexibility to allow for a reduction in size without compromising target performance requirements. In collapsible form, the pump diameter would need to be reduced to a maximum of 7–9 mm for use in adolescent or adult Fontan patients. Of course, an outer diameter of 7 mm would be preferable since the 9 mm design would likely require stay sutures. The construction of the hydraulic components for a pump of this size presents few obstacles to manufacturing and assembly. In contrast, the inclusion of a motor-magnetic suspension is anticipated to be challenging, but possible. The magnetic bearing designs can be made axially longer to balance radial space when considering a cantilever configuration. Work on the magnetic suspension and motor design are ongoing, and novel manufacturing and precision machining techniques continue to improve everyday.

In addition, biocompatible and flexible materials must be explored and identified to construct the impeller blades and for design of the protective cage of filaments. Progress has been made to incorporate the use of flexible polyurethanes and nitinol-based cage geometries. Numerical analyses of the support system, consisting of the impeller pump, protective cage of filaments, and catheter, would provide insight into the internal fluid dynamics, shear stresses, flow vorticity, and overall hydraulic performance; progress in this area is also ongoing.1,2,9 Moreover, we are conducting additional prototype testing through particle image velocimetry (PIV) to identify the regions of irregular flow patterns in need of removal.4 Acute animal studies to be conducted in the near future will provide insight into the interactive dynamics with the cardiovascular system and pump implantability.

The use of the indwelling catheter for placement and support of the pump introduces limitations related primarily to patient mobility, especially if the full duration of support must be employed (4 weeks). Patient mobility after open heart surgery or a heart transplant has been demonstrated to improve patient recovery. Given the intravascular target design of the pump, however, the device and catheter could be removed for hours to days until a new pump is inserted. This off-pump time period could then facilitate patient mobility. While this indwelling catheter restricts mobility, placement of pump will not require invasive open heart surgery, as in the use of conventional ventricular assist devices.

The findings of this study demonstrated that this impeller geometry produces the hydraulic energy necessary to support Fontan patients. The hemolysis measurements indicate a reasonable level. The use of the test-rig for the hydraulic performance and hemolysis testing, while feasible and cost-effective, may have contributed to rig-induced hemolysis as measured in these studies. The metallic surfaces of the aluminum tanks were anodized and visually inspected for defects, but surface scanning at the micron level, where roughness could trigger red cell lysis, was not performed. In addition, the entire volume of the test loop is 1 L, and data samples consisted of 10 mL with 4 mL of this amount being used for measurements. The sample size for measurements was less than 10% of the total circulating volume in the test-loop, which may adversely impact measurement sensitivity. Nevertheless, the repetition of the blood bag experiment (n = 7) supports reasonable data consistency. Moreover, the location of the sampling port for the measurement data, however, was not in a region where stagnant flow conditions occur in the rig; therefore, the last 4 mL of the 10 mL drawn off of the port for measurements represents the newly captured volume of blood flow around the port. According to Mueller et al.,15 the volume of the test loop must be small, but not undersized such that it gives rise to secondary effects, including the accumulation of cell fragments and metabolites. A new test loop, specifically used for the hemolysis, is currently being built of disposal polyurethane plastics with a filling volume expected to be less than 500 mL to improve measurement sensitivity and better balance the circulating volume to sample size ratio for the future hemolytic evaluation of prototypes with a cage of filaments, outlet nose, and other pump design features.


The performance evaluation using CFD, the experimental testing of a pump prototype, and blood bag testing in this study indicated an acceptable design from which to build upon and optimize. The next step will involve parallel efforts, PIV measurements of the flow field, evaluation of the motor-magnetic suspension system design, and development of other system components, such as the collapsible protective cage and filaments, catheter and sheath, and electronic platform from which to monitor and adjust operation while in use. This blood pump will improve the clinical treatment of patients with failing Fontan physiology and provide a unique catheter-based therapeutic approach as a bridge-to-recovery or transplantation.



The authors wish to acknowledge the financial support for this work provided by the Thomas F. and Katie Jeffress Memorial Trust, Phase I and Phase II Award (Grant Number: J-874), American Heart Association Beginning Grant-in-Aid (Grant Number: 0865320E), National Science Foundation (Grant Number: EEC-0823383), 2009 Oak Ridge Associated Universities (ORAU) Ralph E. Powe Junior Faculty Enhancement Award, and the U.S. Department of Education GAANN Interdisciplinary Graduate Engineering Education and Research (I-GEEAR) fellowship awards (S. S. Bhavsar and S. G. Chopski), and the Mendel Family Diary Farm in Amelia County, VA.


  1. 1.
    Bhavsar, S. S., J. Y. Kapadia, S. G. Chopski, and A. L. Throckmorton. Intravascular mechanical cavopulmonary assistance for patients with failing Fontan physiology. Artif. Organs 33(11):977–987, 2009.CrossRefPubMedGoogle Scholar
  2. 2.
    Bhavsar, S. S., W. B. Moskowitz, and A. L. Throckmorton. Interaction of an idealized cavopulmonary circulation with mechanical circulatory assist using an intravascular rotary blood pump. Artif. Organs (in press).Google Scholar
  3. 3.
    Bludszuweit, C. Three-dimensional numerical prediction of stress loading of blood particles in a centrifugal pump. Artif. Organs 19:590–596, 1995.CrossRefPubMedGoogle Scholar
  4. 4.
    Chopski, S. G., E. A. Downs, S. S. Bhavsar, J. Y. Kapadia, C. M. Haggerty, A. P. Yoganathan, and A. L. Throckmorton. Particle image velocimetry measurements of an idealized total cavopulmonary connection with mechanical circulatory assistance in the inferior vena cava. In: 6th International Conference on Pediatric Mechanical Circulatory Support Systems and Pediatric Cardiopulmonary Perfusion, Boston, MA, USA, May 6–8, 2010.Google Scholar
  5. 5.
    Christianson, A., C. P. Howson, and B. Modell. March of Dimes Global Report on Birth Defects: The Hidden Toll of Dying and Disabled Children. White Plains, NY: March of Dimes Birth Defects Foundation, 2006.Google Scholar
  6. 6.
    de Leval, M. R., G. Dubini, F. Migiliavacca, et al. Use of computational fluid dynamics in the design of surgical procedures: application to the study of competitive flows in the cavopulmonary connections. J. Thorac. Cardiovasc. Surg. 111:502–511, 1996.CrossRefPubMedGoogle Scholar
  7. 7.
    Ensley, A. E., A. Ramuzat, T. M. Healy, C. Lucas, S. Sharma, R. Pettigrew, and A. P. Yoganathan. Fluid mechanic assessment of the total cavopulmoary connection using magnetic resonance phase velocity mapping and digital particle image velocimetry. Ann. Biomed. Eng. 28:1172–1183, 2000.CrossRefPubMedGoogle Scholar
  8. 8.
    Hosein, R., A. J. Clarke, S. P. McGuirk, M. Griselli, O. Stumper, J. V. De Giovanni, D. J. Barron, and W. J. Brawn. Factors influencing early and late outcome following the Fontan procedure in the current era. The ‘Two Commandment’s? Eur. J. Cardiothorac. Surg. 31:344–353, 2007.CrossRefPubMedGoogle Scholar
  9. 9.
    Kapadia, J. Y., K. C. Pierce, A. K. Poupore, and A. L. Throckmorton. Hydraulic testing of intravascular axial flow blood pump designs with a protective cage of filaments for mechanical cavopulmonary assist. ASAIO J. 56:17–23, 2010.CrossRefPubMedGoogle Scholar
  10. 10.
    Karassik, I., W. C. Krutzsch, W. H. Fraser, and J. P. Messina. Pump Handbook (2nd ed.). New York: McGraw-Hill Book Company, 1986.Google Scholar
  11. 11.
    Kayatas, M., N. Ozdemir, H. Muderrisoglu, M. Ulucam, M. Turan, and N. Hizel. Comparison of the non-invasive methods estimating dry weight in hemodialysis patients. Renal Failure 28:217–222, 2006.CrossRefPubMedGoogle Scholar
  12. 12.
    Khairy, P., S. M. Fernandez, J. E. Mayer, J. K. Triedman, E. P. Walsh, et al. Long-term survival, modes of death, and predictors of mortality in patients with Fontan surgery. Circulation 117:85–92, 2008.CrossRefPubMedGoogle Scholar
  13. 13.
    Lacour-Gayet, F. G., C. J. Lanning, S. Stoica, et al. An artificial right ventricle for failing Fontan: in vitro and computational study. Ann. Thorac. Surg. 88:170–176, 2009.CrossRefPubMedGoogle Scholar
  14. 14.
    Malinauskas, R. A. Plasma hemoglobin measurement techniques for the in vitro evaluation of blood damage caused by medical devices. Artif. Organs 21(12):1255–1267, 1997.CrossRefPubMedGoogle Scholar
  15. 15.
    Mueller, M. R., H. Schima, H. Engelhardt, A. Salat, D. B. Olsen, U. Losert, and E. Wolner. In vitro hematological testing of rotary blood pumps: remarks on standardization and data interpretation. Artif. Organs 17(2):103–110, 1993.CrossRefPubMedGoogle Scholar
  16. 16.
    Nosé, Y. Design and development strategy for the rotary blood pump. Artif. Organs 22(6):438–446, 1998.CrossRefPubMedGoogle Scholar
  17. 17.
    Paul, R., J. Apel, S. Klaus, F. Schugner, P. Schwindke, and H. Reul. Shear stress related blood damage in laminar Couette flow. Artif. Organs 27(6):517–529, 2003.CrossRefPubMedGoogle Scholar
  18. 18.
    Pekkan, K., D. de Zélicourt, L. Ge, F. Sotiropoulos, D. Frakes, M. A. Fogel, and A. P. Yoganathan. Physics-driven CFD modeling of complex anatomical cardiovascular flows—a TCPC case study. Ann. Biomed. Eng. 33(3):284–300, 2005.CrossRefPubMedGoogle Scholar
  19. 19.
    Riemer, K., G. Amir, S. H. Reichenbach, and O. Reinhartz. Mechanical support of total cavopulmonary connection with an axial flow pump. J. Thorac. Cardiovasc. Surg. 130:351–354, 2005.CrossRefPubMedGoogle Scholar
  20. 20.
    Rodefeld, M., J. H. Boyd, B. J. LaLone, et al. Cavopulmonary assist: circulatory support for the univentricular Fontan circulation. Ann. Thorac. Surg. 76:1911–1916, 2003.CrossRefPubMedGoogle Scholar
  21. 21.
    Rodefeld, M. D., B. Coats, T. Fisher, J. Brown, and S. H. Frankel. Cavopulmonary assist using a percutaneous, bi-conical, single impeller pump: a new spin for Fontan circulatory support. In: 89th Annual Meeting of the American Association for Thoracic Surgery, Boston, MA, USA, May 9–13, 2009.Google Scholar
  22. 22.
    Ryu, K., T. M. Healy, A. E. Ensley, S. Sharma, C. Lucas, and A. P. Yoganathan. Importance of accurate geometry in the study of the total cavopulmonary connection: computational simulations and in vitro experiments. Ann. Biomed. Eng. 29:844–853, 2001.CrossRefPubMedGoogle Scholar
  23. 23.
    Schlichting, H. Boundary-Layer Theory (7th ed.). New York: McGraw-Hill, Inc., 1979.Google Scholar
  24. 24.
    Song, X., A. L. Throckmorton, H. G. Wood, J. F. Antaki, and D. B. Olsen. Quantitative evaluation of blood damage in a centrifugal VAD by computational fluid dynamics. J. Fluid Eng. 126:410–418, 2004.CrossRefGoogle Scholar
  25. 25.
    Stepanoff, A. Centrifugal and Axial Flow Pumps (2nd ed.). New York: Krieger Publishing Company, 1957.Google Scholar
  26. 26.
    Throckmorton, A. L., K. K. Ballman, C. D. Myers, S. H. Frankel, J. W. Brown, and M. D. Rodefeld. Performance of an expandable propeller pump for cavopulmonary assist in a univentricular Fontan circulation. Ann. Thorac. Surg. 86:1343–1347, 2008.CrossRefPubMedGoogle Scholar
  27. 27.
    Throckmorton, A. L., and S. G. Chopski. Pediatric circulatory support systems: current strategies and future directions. Biventricular and univentricular mechanical assistance. ASAIO J. 54:491–497, 2008.CrossRefPubMedGoogle Scholar
  28. 28.
    Throckmorton, A. L., A. Untaroiu, P. E. Allaire, H. G. Wood, D. S. Lim, M. A. McCulloch, and D. B. Olsen. Numerical design and experimental hydraulic testing of an axial flow VAD for infants and children. ASAIO J. 53:754–761, 2007.CrossRefPubMedGoogle Scholar

Copyright information

© Biomedical Engineering Society 2010

Authors and Affiliations

  • Amy L. Throckmorton
    • 1
  • Jugal Y. Kapadia
    • 1
  • Steven G. Chopski
    • 1
  • Sonya S. Bhavsar
    • 1
  • William B. Moskowitz
    • 2
  • Scott D. Gullquist
    • 2
  • James J. Gangemi
    • 3
  • Christopher M. Haggerty
    • 4
  • Ajit P. Yoganathan
    • 4
  1. 1.Department of Mechanical Engineering, School of EngineeringVirginia Commonwealth UniversityRichmondUSA
  2. 2.The Division of Pediatric Cardiology, Medical College of VirginiaVirginia Commonwealth UniversityRichmondUSA
  3. 3.The Division of Thoracic and Cardiovascular Surgery, School of MedicineUniversity of VirginiaCharlottesvilleUSA
  4. 4.The Wallace H. Coulter School of Biomedical EngineeringGeorgia Institute of Technology and Emory UniversityAtlantaUSA

Personalised recommendations