Bayesian optimization of 3D bioprinted polycaprolactone/magnesium oxide nanocomposite scaffold using a machine learning technique

Abstract

3D bioprinting of polycaprolactone (PCL) is an additive manufacturing technique, fabricating 3D scaffolds with widespread applications in biomedical bone regeneration. PCL has favorable properties such as tunable mechanical, biological, cytocompatibility, and good printability. In addition, adding magnesium oxide (MgO) nanoparticles effectively enhance bioactivities and bone formation. However, several researchers are reported that PCL-MgO nanocomposite may face challenges in printability. Therefore, this study has focused on optimizing printing parameters to achieve enhanced mechanical and osteoconductivity properties, printability, and print resolution. A newly developed and cost-effective method, Bayesian optimization (BO), has been applied to achieve this objective. The developed model investigates and accelerates the optimization of printing parameters, including air pressure, printing speed, and nozzle temperature on printability and print resolution. Despite the wide search spaces of printing parameters, the BO model drastically reduces the number of experiments to 11 iterations in each target width. There is a good agreement between the model-predicted and actual values (91% in width). Besides, this model can be used to find optimum process parameters in printing gradient width filament to fabricate 3D gradient scaffolds.