Enhanc­ing per­for­mance of the Pedi­aFlow heart pump

by Mansur Zhussupbekov

Pedi­aFlow is an implantable heart pump, also known as ven­tric­u­lar assist device (VAD), intended to provide hemo­dy­nam­ics support for infants and young children with heart failure. The fifth-gen­er­a­tion Pedi­aFlow (PF5) is a minia­ture, mag­net­i­cally lev­i­tated pump approx­i­mately the size of an AA battery, capable of deliv­er­ing flows ranging from 0.5 to 4.0 L/​min. Devel­op­ing minia­tur­ized blood pumps like the Pedi­aFlow requires nav­i­gat­ing a delicate balance of hydraulic per­for­mance, bio­com­pat­i­bil­ity, and man­u­fac­tura­bil­ity. This problem lends itself well to auto­mated shape opti­miza­tion, hence why in our latest work, we turned to CFD-driven auto­mated shape opti­miza­tion using CAESES to tackle a key chal­lenge — improv­ing pressure recovery in the pump’s flow path while main­tain­ing excel­lent biocompatibility.

A key feature of the Pedi­aFlow is its fully mag­net­i­cally lev­i­tated rotor, which elim­i­nates mechan­i­cal bearings and sig­nif­i­cantly enhances bio­com­pat­i­bil­ity. However, achiev­ing suf­fi­cient magnetic stiff­ness requires an extended rotor length, result­ing in an uncon­ven­tional design with a long axial gap between the impeller and diffuser stages. This extended path causes energy losses as blood tra­verses from impeller to diffuser, further exac­er­bated by Taylor vortices that develop within the annular gap. To address these losses, we proposed adding a set of sta­tion­ary diffuser blades (front diffuser) imme­di­ately down­stream of the impeller to convert dynamic head into static pressure more effi­ciently. Our hypoth­e­sis was that this would allow the pump to achieve its desired oper­at­ing point at a lower rotor speed.

When opti­miz­ing blood pumps, any gains in hydraulic per­for­mance must be care­fully weighed against poten­tial impacts on hemo-com­pat­i­bil­ity (the pump’s com­pat­i­bil­ity with blood). Of par­tic­u­lar concern is hemol­y­sis – the damage to red blood cells caused by exposure to supra-phys­i­o­logic levels of shear stress. This inherent trade-off between per­for­mance and blood damage makes this task par­tic­u­larly well-suited for multi-objec­tive optimization.

Diffuser Para­me­ter­i­za­tion and Opti­miza­tion Worflow

Using CAESES, we were able to fully para­me­ter­ize the geometry of these new diffuser blades, see the effects of the selected para­me­ters and their ranges in the picture above. The CAESES module was coupled with OpenFOAM for meshing and CFD and con­trolled the opti­miza­tion process. This allowed us to sys­tem­at­i­cally explore the design space and identify the optimal diffuser con­fig­u­ra­tion that max­i­mized pressure recovery while min­i­miz­ing hemolysis.

Results and Findings

Our key findings from this effort include:

• Diffuser designs with fewer blades (2 – 3) out­per­formed those with more blades (4 – 5) in terms of pressure recovery.

• Increas­ing the blade axial length gen­er­ally improved pressure rise but also increased hemol­y­sis (expressed as RHI – relative hemol­y­sis index).

• There was an optimal range for the wrap angle to axial length ratio that balanced pressure recovery and blood damage.
• Plotting pressure recovery vs hemol­y­sis revealed a set of Pareto-optimal solu­tions that balanced hydraulic per­for­mance and hemo-com­pat­i­bil­ity. Using the utility function that combined the two objec­tives into the ratio of -∆P to RHI allowed select­ing the best design candidate.

• The best can­di­date from the Explo­ration stage was further opti­mized using the T‑search algo­rithm. The T‑search opti­miza­tion improved the ratio -∆P/RHI from 23.76 to 24.0, pri­mar­ily attrib­uted to the reduc­tion in hemol­y­sis from RHI 1.65 to 1.63.

By com­bin­ing these insights, we con­verged on an opti­mized 2‑blade diffuser design that produced a 39 mmHg pressure rise in an isolated stage – a 26% increase from the baseline pressure head at 14,000 RPM. Impor­tantly, we were also able to assess the hemolytic per­for­mance of this design and identify the optimal com­pro­mise between pressure recovery and blood damage.

To verify the per­for­mance of the opti­mized front diffuser in the full pump, we incor­po­rated it into a CFD sim­u­la­tion of the complete PF5 flow path. The results were remark­able: the pump with the added front diffuser stage achieved the target OP of 1.5 L/​min at 160 mmHg while running at just 14,000 RPM, compared to the 16,000 RPM required by the baseline design. This reduc­tion in oper­at­ing speed, combined with improved flow guidance, yielded two sig­nif­i­cant benefits: the hydraulic effi­ciency increased from 26.3% to 32.5%, and pre­dicted hemol­y­sis decreased by 31%. These results suggest that the mid-stator not only enhances the pump’s per­for­mance but also improves its hemo-compatibility.

Con­clu­sion and Outlook

CAESES’s robust opti­miza­tion frame­work, coupled with OpenFOAM, proved instru­men­tal in this work. The ability to rapidly iterate on the blade geometry and seam­lessly connect to high-fidelity CFD sim­u­la­tions allowed us to navigate the complex design space effi­ciently. Looking ahead, we plan to expand our opti­miza­tion to consider off-design oper­at­ing con­di­tions and assess the impact on overall pump effi­ciency and hemo-com­pat­i­bil­ity. As we continue refining and val­i­dat­ing this solution, we hope it will con­tribute to our broader efforts to deliver safer and more effec­tive cir­cu­la­tory support devices for pedi­atric patients.

About the Author

Notable con­trib­u­tors to this project are: Jingchun Wu, PhD (Advanced Design Opti­miza­tion, LLC, Irvine, CA, USA), Greg W Burgreen, PhD (Center for Advanced Vehic­u­lar Systems, Mis­sis­sippi State Uni­ver­sity, Starkville, MS), Jeongho Kim, PhD (Depart­ment of Bio­med­ical Engi­neer­ing, Daejeon Insti­tute of Science and Tech­nol­ogy, Daejeon, Korea), led by James F Antaki, PhD (Meinig School of Bio­med­ical Engi­neer­ing, Cornell Uni­ver­sity, Ithaca, NY, USA).

Learn More

See this overview for the pos­si­bil­i­ties and capa­bil­i­ties that CAESES offers for medical applications.