Abstract
Organoids produce through traditional manual pipetting methods face challenges such as labor-intensive procedures and batch-to-batch variability in quality. To ensure consistent organoid production, 3D bioprinting platforms offer a more efficient alternative. However, optimizing multiple printing parameters to achieve the desired organoid size remains a time-consuming and costly endeavor. To address these obstacles, machine learning is employed to optimize five critical printing parameters (i.e., bioink viscosity, nozzle size, printing time, printing pressure, and cell concentration), and develop algorithms capable of immediate cellular droplet size prediction. In this study, a high-throughput cellular droplet bioprinter is designed, capable of printing over 50 cellular droplets simultaneously, producing the large dataset required for effective machine learning training. Among the five algorithms evaluated, the multilayer perceptron model demonstrates the highest prediction accuracy, while the decision tree model offers the fastest computation time. Finally, these top-performing machine learning models are integrated into a user-friendly interface to streamline usability. The bioprinting parameter optimization platform develops in this study is expected to create significant synergy when combined with various bioprinting technologies, advancing the scalable production of organoids for a range of applications.