Answers are grounded in this paper and cite their sources.
Volumetric additive manufacturing (VAM) has emerged as a powerful biofabrication platform capable of rapidly generating complex, freeform, cell-laden constructs. These advantages have positioned VAM as a promising technology for tissue engineering. However, widespread adoption for bottom-up tissue engineering and bioassembly applications remains limited by optical dose heterogeneities, material-dependent polymerization behavior, and post-fabrication swelling or shrinkage, all of which reduce dimensional accuracy and reproducibility. Here, we present two complementary, open-source correction strategies to improve print fidelity in VAM: (i) an attention-based machine learning (ML) framework and (ii) a real-time video-guided correction workflow. The ML framework employs a dimensionally reduced representation of volumetric print data encoded as stacked 2D arrays with print metadata, enabling efficient model training without specialized computing hardware. Using random shape arrays printed in gelatin methacryloyl (GelMA) and poly(ethylene glycol) diacrylate (PEGDA), we trained a metadata-informed attention U-Net to generate corrected print arrays that compensate for material-specific printing errors. ML correction improved fine-feature fidelity in GelMA constructs, enhancing corner sharpness and volume recapitulation. However, the ML approach did not provide benefits for PEGDA formulations. In parallel, we developed a video-based correction workflow that directly compensates for z-axis polymerization heterogeneities without requiring material-specific model training. A universal video correction model generated from a PEGDA phantom was successfully applied to optically transparent GelMA bioinks, enabling accurate fabrication without recalibration. Using this approach, hundreds of highly uniform hydrogel spheroids were fabricated in less than 30 s and integrated into melt electrowritten (MEW) scaffolds for bioassembly applications. Cell-laden U87 GelMA spheroids maintained high viability, exhibited extensive F-actin organization and collagen VI deposition, and initiated early spheroid fusion following both video corrected VAM and bioassembly processes. Collectively, these results establish both ML- and video-corrected VAM as facile workflows for reproducible spheroid fabrication for bottom-up biofabrication of complex tissues across a broad range of biomaterials.
No takes yet. Share an insight, caveat, or question.
Wheeler et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: