Machine learning-based prediction and optimisation framework for as-extruded cell viability in extrusion-based 3D bioprinting

Abstract

Extrusion-based 3D bioprinting has revolutionised tissue engineering, enabling complex biostructure manufacturing. However, extrusion imposes substantial shear stress on cells, compromising cell viability. Predicting and optimising cell viability remains challenging due to rheological modelling complexity and cell-type dependency. To address these challenges, this study developed a quantitative framework integrating numerical simulation and machine learning. Support vector regression and simulation were utilised to evaluate alginate ink viscosity and shear stress profiles, while multi-layer perceptron regressors were trained on experimental datasets for diverse cell types to predict as-extruded cell viability based on wall shear stress magnitude and exposure time. Results showed vascular endothelial cells were most susceptible to shear stress, with viability dropping to 80% at 2.05 kPa for 400 ms, while mesenchymal stem, cervical cancer, and embryonic fibroblast cells showed such decrease at 2.65, 2.85, and 3.72 kPa, respectively. This versatile framework enables rapid bioink optimisation across various cell types.

Keywords

• 3D bioprinting
• cell viability
• shear stress
• numerical analysis
• machine learning
• alginate-based bioink