Page 191 - IJB-7-4

Fu, et al.
A B

Figure 5. Validation on probability space. Three probability hyperplanes, 25%, 50%, and 75%, were illustrated based on Figure 3B. (A)
Validation point selected at 20 w/v% concentration, 30°C nozzle temperature, and 0.4 mm path height. (A-1) Microscopic image evaluating
the filament width. (A-2) Printing cube structure (5 × 5 × 5 mm ). (A-3) Top view of the printed grid structure (5 × 5 × 5 mm ). (B) Validation
3
3
point at 25 w/v% concentration, 29°C nozzle temperature, and 0.32 mm path height. (B-1) Microscopic image evaluating the filament width
index. (B-2) Printing cube structure (5 × 5 × 5 mm ) with top view and side view. (B-3) Printing pyramid structure (10 mm × 10 mm × 8 mm)
3
with top and side view. Scale bar for (A-1) and (B-1) is 1.87 mm and 2 mm for the rest.
of a scaffold is a critical information in supervised ML. creates the continuity between bioprinters and can
This problem raises the importance of a standardized be used to eliminate the need for mass testing when
metric for printability within the bioprinting community. optimizing the bioprinting of new bioinks.
With a standard evaluation method, ML models could be
more generalized and applied across different materials, Acknowledgments
printers, and applications. This expansion would greatly We acknowledge the financial support from SunP Biotech
increase the usefulness of ML in bioprinting and allow for company research grant (Drexel University–260676).
high fidelity prints using new materials without the labor-
intensive testing required to continuously build new ML Conflicts of interest
models.
We declaration no conflict of interest.
5. Conclusions

In this paper, the effects of path height, nozzle temperature, Author contributions
nozzle gauge, and composition on printability were Z.F. conceptualized the study, wrote the ML section,
determined for PL 127 inks. Path height was shown to performed SVM and visualization, reviewed and edited
have an significant impact on printability, while nozzle the manuscript. V.A. conducted experiments, wrote
temperature and composition affect the rheological the sections related to evaluation printing parameters
properties of PL 127, and thus, affect the printability. on print outcome, reviewed and edited the manuscript.
Nozzle gauge alone was shown to have no effect. W.S. conceptualized the study, reviewed, and edited the
Rheological data and an investigation into how these manuscript
parameters affect printability revealed the importance of
viscosity in optimizing parameters and their interactions. References
Based on 12 UD training data, a ML model was
then built and validated to create a recommendation for 1. Sun W, Starly B, Daly AC, et al., 2020, The Bioprinting
optimal combinations of specific printing parameters Roadmap. Biofabrication, 12:022002.
for extrusion printing of acellular PL 127 bioink. This http://doi.org/10.1088/1758-5090/ab5158
is our first step and a preliminary study to explore 2. Suntornnond R, Tan EY, An J, et al., 2017, A highly Printable
the application of ML toolkit to extrusion-based 3D and Biocompatible Hydrogel Composite for Direct Printing
bioprinting. Future work includes standardizing the of Soft and Perfusable Vasculature-Like Structures. Sci Rep,
quantification of “printability,” incorporation of cell
viability into the metric (ML labeling), and selection 7:16902.
of higher dimensional feature space with more relevant https://doi.org/10.1038/s41598-017-17198-0
parameters. Building a more generalized ML model 3. Kolesky DB, Truby RL, Gladman AS, et al., 2014, 3D

International Journal of Bioprinting (2021)–Volume 7, Issue 4 187