A Classification-Based Machine Learning Approach to Understand and Infer the Ultimate Successful Bioprinting Process

Abstract 3D bioprinting emerges as a prominent tool for regenerative medicine and biomedical applications. Among extrusion-based, laser-assisted, and drop-on-demand techniques, extrusion-based bioprinting receives greater attention due to its capability to handle a variety of material types while supporting higher cell densities. Three major tasks in a successful bioprinting process are bioink formulation, optimizing print process parameters (e.g., extrusion pressure, nozzle diameter, print speed and distance, and environment) to obtain the desired construct, and finally targeted cell survivability and growth in the construct. Most existing works focus on each task category and consider each process optimization individually. However, the goal is to make bioprinting successful with the sequential success of all these three task categories. In this work, we have considered all three task categories (e.g., bioink formulation, optimizing printing parameters, and fostering targeted cell growth and survivability) to understand the success of the bioprinting process. In this article, we developed a classification-based machine learning model to predict the success of extrusion-based bioprinting using 72 experimental datasets comprising process parameters, rheological properties, filament fidelity, and cell viability. The model achieved an 83% accuracy rate, demonstrating its potential to assist researchers in identifying optimal printing conditions and reducing trial-and-error experimentation. By providing a data-driven assessment of bioprinting outcomes, this approach can enhance efficiency and resource utilization in bioprinting research and applications.