Coupling machine learning with 3D bioprinting to fast track optimisation of extrusion printing *

3D bioprinting, a paradigm shift in tissue engineering holds a promising perspective for regenerative medicine and disease modelling. 3D scaffolds are fabricated for subsequent cell seeding or incorporated directly to the bioink to create cell-laden 3D constructs. A plethora of factors relating to bioink properties, printing parameters and post print curing play a significant role in the optimisation of the printing process. Although qualitative evaluation of printability has been investigated largely, there is a paucity of studies on quantitative approaches to assess printability. Hence, this study explores machine learning as a novel tool to evaluate printability quantitatively and to fast track optimisation of extrusion printing in achieving a reproducible 3D scaffold. Bayesian Optimisation, a machine learning method has been employed for optimising 3D bioplotting with a scoring system established to assess the printability of gelatin methacryloyl (GelMA) and hyaluronic acid methacrylate (HAMA) bioinks. The performance of two fundamental criteria encountered in the printing process: the filament formation of the bioink and the layer stacking of the 3D scaffold have been incorporated in the scoring metric. The optimal print parameters for GelMA containing inks with ranging concentrations (10%, 7.5% & 5% (w/v)) were obtained in 19, 4 & 47 experiments whereas for GelMA:HAMA (10:2%, 7.5:2% & 5:2% (w/v)) 32, 25 & 32 experiments were required respectively. This number of experiments is drastically reduced compared to the 60 0 0 to 10 0 0 0 possible combinations in the Bayesian algorithm. Hence, this study will be a stepping-stone into unravelling the benefits of machine learning in this rapidly developing area of 3D bioprinting. (c) 2020 Published by Elsevier Ltd.

Automated summary

This study explores the use of machine learning to quantitatively evaluate printability in 3D bioprinting, focusing on GelMA and HAMA bioinks. Bayesian Optimisation was employed to optimize the printing process, achieving optimal print parameters for different concentrations of GelMA and GelMA:HAMA. The study demonstrates the potential of machine learning in improving the efficiency and reproducibility of 3D bioprinting.