Prediction of cell viability in dynamic optical projection stereolithography-based bioprinting using machine learning

Stereolithography (SLA)-based bioprinting can fabricate three-dimensional complex objects accurately and efficiently. However, the ultraviolet (UV) irradiation in the SLA-based bioprinting process is a significant challenge, which may damage the cells. Physics-based models are not able to predict cell viability with high accuracy because of the complexity of cell biological structures and cell recovery. To overcome this challenge, we developed a predictive model using machine learning to predict cell viability. We designed a set of experiments considering the effects of four critical process parameters, including UV intensity, UV exposure time, gelatin methacrylate concentration, and layer thickness. These experiments were conducted under varying bioprinting conditions to collect experimental data. An ensemble learning algorithm combining neural networks, ridge regression, K-nearest neighbors, and random forest (RF) was developed aiming at predicting cell viability under various bioprinting conditions. The performance of the predictive model was evaluated based on three error metrics. Finally, the importance of each process parameter on cell viability was determined using RF. The predictive model has been demonstrated to be able to predict cell viability with high accuracy as well as determine the significance of each process parameter on cell viability in SLA-based 3D bioprinting.