Hierarchical Machine Learning for High-Fidelity 3D Printed Biopolymers

Area of Science:

• Biomaterials Engineering
• Machine Learning Applications
• 3D Bioprinting

Background:

• 3D bioprinting of alginate hydrogels requires precise control over numerous parameters to achieve high fidelity.
• Current methods often rely on extensive trial-and-error, limiting scalability and efficiency.

Purpose of the Study:

• To develop a hierarchical machine learning (HML) framework for predicting optimal 3D bioprinting parameters.
• To enhance the fidelity of printed alginate hydrogel features by minimizing dimensional errors.

Main Methods:

• Systematic variation of print speed, flow rate, ink concentration, and nozzle diameter to create a dataset of 48 prints.
• Image analysis to score print fidelity against CAD models, defining high fidelity as <10% error.
• Application of the least absolute shrinkage and selection operator (LASSO) for variable selection within the HML framework.

Main Results:

• The HML framework achieved a model fit of R² = 0.643, demonstrating good performance between predicted and observed print fidelity.
• Identified dominant build parameters influencing print error and predicted optimal settings for high-fidelity features.
• Developed process maps to guide designers in minimizing errors based on input variables.

Conclusions:

• The HML framework effectively predicts and optimizes build parameters for high-fidelity 3D bioprinting of alginate hydrogels.
• A trade-off exists when optimizing fidelity for multiple features within a single construct, necessitating advanced predictive tools.
• This approach offers a scalable solution for 3D bioprinting by reducing experimental iterations and improving print quality.