Optimization of hemocompatibility metrics in ventricular assist device design using machine learning and CFD-based...

Area of Science:

• Biomedical Engineering
• Cardiovascular Devices
• Computational Fluid Dynamics

Background:

• Ventricular assist devices (VADs) are critical for end-stage heart failure.
• Optimizing VAD design requires balancing hydraulic efficiency and hemocompatibility to minimize blood damage.
• Current VAD designs face challenges in preventing adverse blood-material interactions.

Purpose of the Study:

• To develop and validate a multi-objective optimization framework for improving VAD rotor hemocompatibility.
• To identify optimal design parameters for VAD rotors that minimize blood damage.
• To enhance the safety and efficacy of VADs for patients with heart failure.

Main Methods:

• Integrated computational fluid dynamics (CFD) with Random Forest Regression (RFR) and Bayesian optimization.
• Utilized a D-optimal design of experiments to optimize seven key VAD rotor parameters.
• Employed a Carreau-Yasuda blood model and mesh independence analysis in CFD simulations.

Main Results:

• The RFR surrogate model achieved high predictive accuracy (R² > 0.84) for hemocompatibility metrics.
• Optimized design reduced shear stress exposure, achieving 97.24% of blood flow below 50 Pa.
• Demonstrated significant reductions in hemolysis index (0.01%) and platelet activation (1x10⁻⁶%).

Conclusions:

• The proposed optimization framework systematically enhances VAD hemocompatibility.
• The optimized VAD rotor design offers substantial improvements in blood-damage reduction.
• Further experimental validation is recommended to confirm computational findings under physiological conditions.