Rheology-informed machine learning for filament-width prediction and process-space screening in handheld coaxial bioprinting

Flow-rate-controlled handheld coaxial bioprinting offers flexible filament deposition, but reproducible control of deposited geometry remains challenging because material properties and process conditions interact with device motion. In this study, we developed a rheology-informed, machine-learning-assisted workflow for predicting and screening filament deposition under controlled translational motion. A handheld coaxial bioprinter was mounted on a motorised X-axis stage to minimise motion-related variability, while the inner-channel conditions were fixed and the outer GelMA concentration and flow rate were systematically varied. Rheological measurements were converted into physically interpretable descriptors using generalised Maxwell and Carreau-Yasuda models and used to predict deposited filament width. The regression model reliably predicted filament width across multiple validation schemes. A complementary categorical model was also used to screen width-defined deposition regimes based on the predicted filament width. Prospective testing at previously untested intermediate GelMA concentrations showed good agreement between predicted and experimentally measured filament widths and deposition regimes. Overall, the proposed workflow provides a rheology-informed, prediction-guided framework for screening printing conditions and evaluating filament-width reproducibility in controlled-motion coaxial bioprinting, while defining its applicability to conditions beyond the experimentally sampled process space.