A deep learning approach for error detection and quantification in extrusion-based bioprinting

Quality control in extrusion-based bioprinting (EBB) represents a crucial step to: i) reduce the trial-and-error process and associated material consumption, ii) achieve standard results across different set-ups and laboratories to comply with relevant health standards, and iii) so accelerate the translation of Tissue Engineered products to more impactful clinical applications. In this context, machine learning algorithms represent a key enabling technology that is currently being explored in literature for quality control in EBB, thanks to their ability to learn relevant features from a training dataset and generalize to new, unseen data. In this work, we present a novel application of a deep learning model to EBB, namely a convolutional Long Short-Term Memory (LSTM) autoencoder, to extract a relevant quality measure from videos taken from a frontal view during the printing process. In particular, a comprehensive dataset was built by varying multiple printing parameters and using different EBB set-ups. The data was then used to train the model and validate it using videos containing different types of errors (i.e., under- or over-extrusion). Results highlight that the approach can effectively detect relevant extrusion-related problems in a proportional way to the error magnitude, and so can be applied as a quality control solution for the EBB process.