Flow simulation-based particle swarm optimization for developing improved hemolysis models

• Published in: Biomechanics and modeling in mechanobiology 2023-04, Vol.22 (2), p.401-416
• Main Authors: Torner, B., Frank, D., Grundmann, S., Wurm, F.-H.
• Format: Article
• Language: English
• Subjects: Biological and Medical Physics; Biomedical Engineering and Bioengineering; Biophysics; Blood flow; Capillary flow; Capillary tubes; Computer Simulation; Cost analysis; Engineering; Flow simulation; Heart-Assist Devices; Hemolysis; Hemolysis - physiology; Humans; Mathematical models; Medical devices; Medical equipment; Model accuracy; Multiple objective analysis; Numerical prediction; Optimization; Original Paper; Particle swarm optimization; Predictions; Reliability; Reproducibility of Results; Stress (physiology); Stress, Mechanical; Support systems; Theoretical and Applied Mechanics
• ISSN: 1617-7959; 1617-7940
• DOI: 10.1007/s10237-022-01653-7

The improvement and development of blood-contacting devices, such as mechanical circulatory support systems, is a life saving endeavor. These devices must be designed in such a way that they ensure the highest hemocompatibility. Therefore, in-silico trials (flow simulations) offer a quick and cost-effective way to analyze and optimize the hemocompatibility and performance of medical devices. In that regard, the prediction of blood trauma, such as hemolysis, is the key element to ensure the hemocompatibility of a device. But, despite decades of research related to numerical hemolysis models, their accuracy and reliability leaves much to be desired. This study proposes a novel optimization path, which is capable of improving existing models and aid in the development of future hemolysis models. First, flow simulations of three, turbulent blood flow test cases (capillary tube, FDA nozzle, FDA pump) were performed and hemolysis was numerically predicted by the widely-applied stress-based hemolysis models. Afterward, a multiple-objective particles swarm optimization (MOPSO) was performed to tie the physiological stresses of the simulated flow field to the measured hemolysis using an equivalent of over one million numerically determined hemolysis predictions. The results show that our optimization is capable of improving upon existing hemolysis models. However, it also unveils some deficiencies and limits of hemolysis prediction with stress-based models, which will need to be addressed in order to improve its reliability.