Physics-based computational fluid dynamics-machine learning framework for evaluating cell-material interactions in blood-contacting devices
Blood-contacting materials and device surfaces experience complex interactions among geometry, fluid dynamics, blood rheology, and shear-induced blood damage. This study presents a physics-based computational fluid dynamics (CFD)-machine learning (ML) framework for exploratory assessment of shear stress and model-derived hemolysis in a customized blood-shearing device. CFD simulations were performed across variations in channel geometry, rotational speed, and prescribed blood viscosity to evaluate their effects on shear-stress response. Statistical analysis showed that channel width had a notable marginal association with average shear stress, while rotational speed demonstrated a positive association within the investigated design space. Hemolysis was estimated using an empirical computational formulation to provide an initial assessment of potential blood-damaging conditions. A hybrid convolutional neural network (CNN) was also explored as a surrogate for CFD-based design analysis. Overall, integrating CFD and ML shows promise for preliminary evaluation and optimization of blood-contacting devices.
Authors
- R.D.K. Misra (ORCID: https://orcid.org/0000-0002-0918-8180)
- K. Veldhi
Institutions
- Lawrence Technological University (US)
Publication Details
- Journal
- Materials Technology
- Published
- 2026-11-21
- DOI
- https://doi.org/10.1080/10667857.2026.2767988
- Primary Topic
- Mechanical Circulatory Support Devices
- Type
- article
- Field-Weighted Citation Impact
- 0.00