Fast computational screening of extrusion parameters for bioprinting of skeletal myoblasts in alginate hydrogel

Abstract

Background: Real-time selection of extrusion bioprinting parameters requires rapid prediction of an operating window that balances material throughput against mechanically induced cell damage. Full numerical simulations can resolve the relevant flow mechanics, but their computational cost limits repeated evaluation during process adjustment. In this work, a finite element-Gaussian process (GP) framework is developed for rapid operating-window prediction in skeletal myoblast bioprinting for tissue engineering purposes. Methods: An axisymmetric model of inertia-retained, shear-thinning flow of a 1.5% (w/v) alginate suspension of L8 rat myoblasts is coupled to a previously documented damage law that separates nozzle shear from entrance extensional injury of the same cell model. The dataset comprises 32 combinations of four nozzle diameters (100-330 {micro}m) and eight applied pressures (25-400 kPa). GP surrogates are developed to predict flow rate, wall shear stress and total mechanical damage. Results: GP surrogates respond in under one millisecond per parameter combination, compared with approximately 1.5 hours per finite element case. In leave-one-out validation, the surrogates reproduce the simulated damage with a mean absolute error of 0.34 percentage points and the flow rate with a mean absolute percentage error of 6.8%. With a whole nozzle diameter withheld, the damage error rises to 0.88 percentage points. With screening limits of predicted damage below 20% and a flow rate of at least 1 {micro}L/s, 17 of the 32 simulated conditions lie within the operating window. The damage decomposition also shows the dependency of nozzle selection on the imposed constraint: at 200 kPa, the predicted damage increases from 9.0% to 18.0% with nozzle diameter increasing from 200 {micro}m to 330 {micro}m. The 330 {micro}m nozzle carries a 15.8-fold higher flow rate than the 200 {micro}m nozzle, which raises the effective extensional stress at the nozzle entrance. At a matched flow rate of 1 {micro}L/s, the predicted damage instead decreases from 5.5% to 4.6%. Conclusions: The developed framework supports real-time computational screening of printing parameters within the sampled domain and suggests the need for careful selection of nozzle diameter and applied pressure together, since the ordering of nozzles by predicted damage reverses between fixed pressure and matched flow rate.