Abstract
Bioprinting of cell-laden hydrogels is a rapidly growing field in tissue engineering. The advent of digital light processing (DLP) three-dimensional (3D) bioprinting technique has revolutionized the fabrication of complex 3D structures. By adjusting light exposure, it becomes possible to control the mechanical properties of the structure, a critical factor in modulating cell activities. To better mimic cell densities in real tissues, recent progress has been made in achieving high-cell-density (HCD) printing with high resolution. However, regulating the stiffness in HCD constructs remains challenging. The large volume of cells greatly affects the light-based DLP bioprinting by causing light absorption, reflection, and scattering. Here, we introduce a neural network-based machine learning technique to predict the stiffness of cell-laden hydrogel scaffolds. Using comprehensive mechanical testing data from 3D bioprinted samples, the model was trained to deliver accurate predictions. To address the demand of working with precious and costly cell types, we employed various methods to ensure the generalizability of the model, even with limited datasets. We demonstrated a transfer learning method to achieve good performance for a precious cell type with a reduced amount of data. The chosen method outperformed many other machine learning techniques, offering a reliable and efficient solution for stiffness prediction in cell-laden scaffolds. This breakthrough paves the way for the next generation of precision bioprinting and more customized tissue engineering.