Impeller, pump, and flow dynamics.

Hydrodynamics of the NeoMate pump are achieved by an impeller mounted atop the MagLev rotor body to deliver centrifugal flow (Fig 1B). The impeller (27 mm diameter) contains six backward swept blades covered with a shroud designed to improve hydraulic efficiency (Fig 1C). The rotor body sits below the impeller and houses the rotating magnetic components of the MagLev system. A center opening (8 mm diameter) allows flow recirculation around the rotor body to counteract the lift forces generated by the thrust of the impeller during rotation to keep the impeller/rotor body assembly centered inside the pump housing and prevent axial contact. The impeller/rotor assembly is completely levitated and rotated inside the pump housing via the MagLev system (Fig 1C). The impeller was 3D printed using stereolithography (SLA) with a high-resolution ABS-like clear material (WaterShed XC), and the rotor body was fabricated with a high-density thermoplastic Ultem 1010 PEI (Polyetherimide, Performance Plastics). The pump housing surrounds the impeller/rotor and contains perpendicular barbed inlet and outlet ports (1/4” diameter) along the pump volute for cannula attachment. The stationary components of the MagLev system were mounted to the exterior of the pump housing. The pump housing (14.5 mL priming volume) was split into upper and lower sections to facilitate prototype assembly. The upper pump housing was SLA printed using an ABS-like clear material (WaterShed XC) to facilitate flow visualization, and the lower pump housing was fabricated with Ultem 1010 PEI (Polyetherimide, Performance Plastics). There are three distinct flow paths within the pump (Fig 1D): (1) primary path with flow entering the inlet of the pump housing, then into the impeller eye, through the impeller blades and into the volute, and exiting at the pump outlet; (2) secondary path with flow down the outside of the rotor body and up through the center opening of the rotor body and impeller; and (3) tertiary path at the top of the impeller shroud and the upper pump housing where flow reverses and travels back toward the jet flow at the pump inlet. The secondary and tertiary flow paths are designed to wash the pump surfaces to help reduce stagnate flow and prevent thrombus formation.

Universal controller.

The controller enclosure (Fig 1A) houses the prototype integrated pump, MagLev motor, and all control electronics of the NeoMate System. The electric motor and levitation system were designed to be capable of suspending and rotating the impeller/rotor assembly within the centrifugal blood pump without contact during clinical use. MagLev is accomplished using a voice coil actuator (VCA) with active longitudinal axial positioning and radial stability from the surrounding fluid. Four primary subsystems were developed: (1) 24 VDC isolated power supply with universal alternating current (AC) input, (2) sinusoidal motor controller for rotational control of the impeller/rotor, (3) linear VCA closed-loop position controller for longitudinal position control of the impeller/rotor during levitation, and (4) Hall sensor array for longitudinal position measurement of the impeller/rotor. The prototype systems are also designed with rear panel analog and digital outputs and input for connection to external data acquisition systems, test equipment, and control computers. The rotating components of the motor and levitation system are contained inside the rotor body and the stationary components are contained on the exterior of the pump housing, inside the controller enclosure. The impeller/rotor’s motor is a slot-less brushless DC design and operates over the rotational speed range of 500 RPM to 6000 RPM. After calculating and empirically confirming the torque requirements of the pump to achieve the required flows and pressures [14], a full-custom motor and linear VCA were designed. The system’s power supply is a commercial-off-the-shelf (COTS) design with a universal AC input and 24 VDC output, which provided the current required for longitudinal position control during lift-off and for operating the motor over the required speed range. A sensorless sinusoidal motor controller is used for start-up, open-loop commutation, closed-loop commutation, and speed control. The motor controller and driver are designed to source continuous currents up to 2.0 ADC and short duration peak currents up to 3.0 ADC. Pump speed is controlled via a manual multi-turn speed control knob mounted on the front panel of the controller and is used to limit the slew rate of the impeller/rotor speed. The active longitudinal position control system uses a solenoidal VCA, a permanent sense magnet installed at the base of the rotor assembly, and a Hall effect sensor array located in the base of the pump module receptacle within the controller. The Hall sensor array is aligned to sense the longitudinally oriented magnetic field of the rotor permanent sense magnet and used to monitor instantaneous longitudinal position of the pump rotor. The longitudinal position control system uses the Hall sensor output, which is directly proportional to the instantaneous proximity to the permanent magnet in the rotor to modulate the current in the VCA. This approach functions to maintain the longitudinal position of the impeller/rotor at 0.040” tolerance; however, due to the compact packaging of the magnetic components within the rotor, the VCA was found to contribute peak magnetic fields of ±11mT on top of the sense magnet field yielding poor signal-to-noise ratio. To mitigate this effect, the voltage as a function of VCA instantaneous current was subtracted from the voltage generated by Hall position array to improve the signal-to-noise ratio. A proportional integral derivative (PID) control loop was designed to levitate motor and maintain stable position over the entire operating range.

Static mock flow loop model testing.

Static mock loop experiments were performed to quantify in vitro pump hydraulic performance (head pressure (H) to flow (Q), H-Q curves) of the NeoMate System. Two prototypes (System A and System B) were tested using three different pairs of inflow and outflow cannulae. Six static mock loop experiments were performed in total. Both Systems A and B were each tested in the static mock loop using three different blood-analog volumes and pairs of pediatric cannulae (Medtronic, Minneapolis, MN). The mock loop volumes (with corresponding cannulae pair) were chosen to simulate the blood volume size equivalence to a neonate, infant, and toddler. The volumes (reservoir plus loop prime volume) and cannula pairs tested, from smallest to largest size, were as follows: 1) 300mL volume, inflow cannula 68112, 15”, 12Fr; outflow cannula 77110 9”, 10Fr. 2) 500mL volume, inflow cannula 68116, 15”, 16Fr; outflow cannula 77112, 9”, 12Fr. 3) 2400mL volume, inflow cannula 68124, 15”, 24Fr; outflow cannula 77116, 9”, 16Fr. Pressure gradient and flow data were collected over a range of pump speeds by varying outflow resistance.

The static mock loop was designed as shown in Fig 2A. The NeoMate System was connected to the mock loop circuit using ¼” tubing (ND-100-65 Tygon PVC, ¼”, 8349T42, McMaster-Carr, Elmhurst, IL). A glycerol-water solution of 3.8 cP at 37°C was prepared to mimic the viscosity of blood and validated using a Cannon-Fenske viscometer (7988–37, size 75, ACE Glass, Vineland, NJ). A Hoffman clamp placed on the mock loop tubing was used to adjust the outflow resistance. The mock loop reservoir was placed inside a water bath to maintain the circuit at 37°C. Six fluid-filled transducers (MLT0380/D, ADInstruments, Colorado Springs, CO) were used to acquire pressure data at the inlet and outlet of the pump head, and then both proximal and distal to the inflow and outflow cannulae. A 7mm flow probe (Transonic Systems, Ithaca, NY) was placed on the pump outflow tubing. Hemodynamic data from the pressure transducers and flow module (TS410, Transonic Systems) were collected using a PowerLab 16/35 acquisition system with bridge amps (ADInstruments). Pressure transducers were calibrated using an external manometer (Meriam, Cleveland, OH). Controller output parameters, motor voltage (1 VDC = 1 VDC), motor current (1 VDC = 3 ADC), and motor temperature (4mV/°C), were measured and power was calculated. Data were recorded using LabChart v.8.1.16 (ADInstruments) at a sampling rate of 1000/s.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Mock flow loop models.

Illustration of static (A) and dynamic (B) mock flow loop models for testing hydrodynamic and hemodynamic performance of the prototype Inspired Therapeutics NeoMate System. P, pressure sensor.

https://doi.org/10.1371/journal.pone.0266822.g002

Once the static mock loop was primed with the glycerol-water solution and de-aired, the pump was set to 500 RPM and no resistance. Flow was then decreased in 0.2 L/min increments by increasing outflow resistance (Hoffman clamp) and data were collected at each increment. Pump speed was then increased by 500 RPM increments up to 5000 RPM, and the stepwise data were reacquired at each RPM increment. Data were collected from each of the 6 pressure transducers, the flow probe, and from the controller intrinsic parameters. Delta pressure was calculated across the pump (ΔP pump as measured via the pressure transducers at the pump inlet and outlet, Fig 2A) and across the circuit (pump plus cannulae, ΔP circuit as measured via the pressure transducers post-outflow cannula and pre-inflow cannula, Fig 2A). The above experimental design was repeated for each controller and each cannula pair. Mean values for each parameter at each test condition were calculated and resulting H-Q curves were plotted using Prism v.9.1.2 (226) (GraphPad, San Diego, CA).

Dynamic mock flow loop model testing.

Dynamic mock loop experiments were performed using a pneumatic mock ventricle to quantify the in vitro hemodynamic performance of Systems A and B. Two single left-sided dynamic mock loop experiments were performed. Both Systems A and B were each tested in the dynamic mock loop using a single pair of pediatric cannulae. The cannula pair tested was inflow cannula 68124, 15”, 24Fr; outflow cannula 77116, 9”, 16Fr (both Medtronic, the largest pair tested in the static loop) with a blood-analog volume of 3600mL.

The dynamic mock loop was designed as shown in Fig 2B. The pump and controller were connected to the mock loop circuit using ¼” tubing (ND-100-65 Tygon PVC, ¼”, 8349T42, McMaster-Carr). The outflow cannula was connected to the apex of a silicone ventricle powered by a pneumatic driver (Ventricular Assist Device Pneumatic Drive System, Thoratec, Pleasanton, CA (now Abbott)). Bioprosthetic valves (Mosaic 305 Cinch II, 23mm, Medtronic) were placed in the aortic and mitral positions. The pump outflow cannula fed into the aortic position via a large Tuohy (22828, Qosina, Ronkonkoma, NY). A glycerol-water solution of 3.8 cP at 37°C was prepared to mimic the viscosity of blood and validated using a Cannon-Fenske viscometer (7988–37, size 75, ACE Glass). A Hoffman clamp placed on the mock loop tubing was used to adjust resistance as needed. A 1L compliance chamber was placed proximal to the reservoir, and the reservoir was placed inside a water bath to maintain the circuit at 37°C. Five fluid-filled transducers (MLT0380/D, ADInstruments) were used to acquire pressure data at the inlet and outlet of the pump head, the left ventricle, the aorta, and the venous position of the dynamic mock loop. A 7mm flow probe (Transonic Systems) was placed on the pump outflow tubing. Hemodynamic data from the pressure transducers and flow module (TS410, Transonic Systems) were collected using a PowerLab 16/35 acquisition system with bridge amps (ADInstruments). Pressure transducers were calibrated using an external manometer (Meriam). Controller output parameters, motor voltage (1 VDC = 1 VDC), motor current (1 VDC = 3 ADC), and motor temperature (4mV/°C), were measured and power was calculated. Data were recorded using LabChart v.8.1.16 (ADInstruments) at a sampling rate of 1000/s.

Once the dynamic mock loop was primed with the glycerol-water solution and de-aired, the mock loop was tuned to a heart rate of 150 bpm, mean aortic pressure of 50 ± 5 mmHg, mean aortic flow of 2.0 ± 0.2 L/min, and venous pressure of 15 ± 5 mmHg. Hemodynamic and pump/controller data were acquired with the pump off and clamped, the pump off and unclamped, and then with the pump on at speeds of 500 to 5000 RPM, increased in 500 RPM increments. Data were analyzed using the LabChart Blood Pressure Module (ADInstruments) and plotted using Prism (GraphPad).

Bench testing.

Prototype System C was used in an impromptu test for long-term reliability of the MagLev motor and controller electronics using a static flow loop (different static loop design than described above in part B) with pump inflow and outflow connected to ¾” Tygon tubing via ¼” to ¾” step-up connectors and connected to a static reservoir (20 L volume). The system continues to operate at 500 RPM to provide a flow rate of 0.25 L/min in a temperature-controlled environment (23°C ± 2°C). Low speed, pressure, and flow rate were intentionally used as components of the pump are made from 3D printed material for which the long-term performance and durability are not known.

Acute ovine model testing.

All animals used were housed in a facility in accordance with the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC). The testing protocol (#18393) was approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Louisville. This investigation and all efforts to reduce suffering via sedation, anesthesia, analgesia, and euthanasia are described in detail below conform to the National Research Council. 2011. Guide for the Care and Use of Laboratory Animals: Eighth Edition. Washington, DC: The National Academies Press.

Two acute (non-survival) sheep studies were performed to assess pump performance of Systems A and B to demonstrate early prototype system feasibility in vivo (hemodynamics, surgical feasibility, clinical applicability). Sheep were selected as the animal model, as future chronic survival studies will require externalized cannulae and the use of a jacket to house the Inspired System, and thus an animal model compatible with having externalized components and being tethered and halter-led (unlike pigs) was selected. Two sheep (40.5kg and 43kg Dorset ewes, Noble Life Sciences Inc, Woodbine MD) were sedated with xylazine (0.04 mg/kg IM) and a catheter was placed in an ear vein. Propofol (4 mg/kg IV) was administered, the animals were endotracheally intubated, and an orogastric tube was placed for rumen decompression. Mechanical ventilation was initiated, and tidal volume and respiratory rate were adjusted to maintain a PaCO₂ of 35–45 mmHg under general isoflurane anesthesia (1–3%). The animals were positioned in right lateral recumbency. A transesophageal echocardiography (TEE) probe (X7-2t transducer with Philips iE33, Philips Healthcare, Andover, MA) was placed for intraoperative monitoring and data acquisition/analysis. 0.9% saline was infused for maintenance as needed (5–10 mL/kg/hr IV). Three-lead electrocardiography, body temperature, pulse oximetry, and end-tidal CO₂ were continuously monitored. Rocuronium (1.2 mg/kg IV) was administered just prior to incision. A small incision was created in the left neck for the placement of 7Fr introducers in the left carotid artery and left jugular vein for pressure monitoring, blood draws, and/or drug administration if needed. A thoracotomy was performed at the 5^th intercostal space and the 5^th rib was removed. The pericardium was opened anterior to the phrenic nerve and pericardial stay sutures were placed. A 16mm flow probe (Transonic Systems) was placed around the pulmonary artery (PA) trunk. Pursestring sutures secured fluid-filled pressure lines into the left atrium, right atrium, aorta, and left ventricle for hemodynamic data collection.

Baseline data (pre-LVAD) were acquired with ventilation temporarily suspended. Data included blood chemistry and oximetry (i-STAT, Abbott Point of Care, Chicago, IL), activated clotting time (ACT; ACT II, Medtronic, Minneapolis, MN), plasma free hemoglobin (HemoCue, Ängelholm, Sweden; modified azidemethemoglobin reaction), TEE, hemodynamics from the pressure transducers and PA flow probe, and intrinsic pump parameters. Comprehensive TEE was performed, left and right ventricular systolic and diastolic function was evaluated, all valves were examined for regurgitation or stenosis, and Simpson’s disk method was used to calculate left ventricular volume. Hemodynamic data from the fluid-filled pressure transducers (MLT0380/D, ADInstruments) and flow modules (TS420, Transonic Systems) were collected using a PowerLab 16/35 acquisition system (ADInstruments) with bridge amps (ADInstruments FE224). Pressure transducers were calibrated using an external manometer (Meriam). Controller output parameters, motor voltage (1 VDC = 1 VDC), current (1 VDC = 3 ADC), and temperature (4mV/°C) were measured, and power was calculated. Data were recorded using LabChart v.8.1.16 (ADInstruments) at a sampling rate of 400/s.

The animals were heparinized to achieve an ACT >400s. The pump and tubing were primed with saline and clamped. The pump outflow cannula (77116, 9” length, 16Fr, Medtronic) was placed in the proximal descending aorta. Pledgeted pursestring sutures were placed in the left ventricular apex and the pump inflow cannula (68124, 15” length, 24Fr, Medtronic) was placed into the apex through a small cruciate incision. TEE was used to verify that the inflow cannula was pointed toward the aortic valve and was parallel to the interventricular septum. A 2–3” section of Hemashield graft (Getinge, Göteborg, Sweden) bridged each cannula to the pump tubing (ND-100-65 Tygon PVC, ¼”, 8349T42, McMaster-Carr) to provide a site for atraumatic clamping. A 7mm flow probe (Transonic Systems) was placed on the pump outflow tubing. Fluid-filled pressure transducers were connected to the pump inflow and outflow tubing. The pump and circuit were primed with saline and de-aired, the cannulae were back bled, and the lines were temporarily re-clamped in preparation for the first experimental dataset.

System A was tested on the first sheep and System B was tested on the second sheep. All data collected at baseline (bloodwork, TEE, hemodynamics, pump/controller data) were repeated for each pump setting to be tested. Each dataset consisted of a hemodynamic recording period of ~10–15 seconds with the ventilator suspended and an acclimatization period of 5–10 minutes between each RPM increase prior to each recording. The datasets recorded included: 1) Pump off and clamped; 2) Pump off and unclamped; 3) 3000 RPM; 4) 4000 RPM; 5) 5000 RPM; 6) 6000 RPM. Following completion of the study protocol, the animals were euthanized under general anesthesia with Beauthanasia-D (0.22 mg/kg IV). Necropsies were performed for examination of the pump, cannulae, heart, great vessels, and end-organs.

Data were analyzed using the LabChart Blood Pressure Module (ADInstruments) and plotted using Prism (GraphPad). Hemodynamic data from the animal tests are presented as the mean of 10–20 cardiac cycles worth of data at each test condition.