Machine Learning-Guided Three-Dimensional Printing of Tissue Engineering Scaffolds.

Area of Science:

• Biomaterials Engineering
• Additive Manufacturing
• Computational Biology

Background:

• 3D printing enables scaffold fabrication for tissue engineering using diverse materials.
• Optimizing printing conditions for new materials is experimentally intensive and costly.

Purpose of the Study:

• To develop a Machine Learning (ML) framework to predict 3D print quality based on material composition and printing parameters.
• To differentiate between printing configurations yielding low-quality versus promising prints for tissue engineering scaffolds.
• To lay the groundwork for a recommendation system for optimal 3D printing conditions.

Main Methods:

• Investigated two ML approaches: direct classification and indirect regression, both using Random Forests.
• Trained and evaluated models on a dataset from a previous extrusion-based 3D printing study of porous polymer scaffolds.
• Input features included material composition and printing parameters; output was predicted print quality (low/high).

Main Results:

• Both Random Forest-based ML models accurately classified the majority of tested printing configurations.
• A simpler linear ML model proved ineffective for predicting print quality.
• Analysis indicated data redundancies in full factorial designs for ML applications.

Conclusions:

• ML, specifically Random Forests, is effective in predicting 3D printing quality for tissue engineering scaffolds.
• The study highlights the potential for ML to guide the selection of printing parameters, reducing experimental effort.
• A more efficient data collection strategy is proposed to optimize ML model training for 3D printing applications.