Page 320 - IJB-9-6

International Journal of Bioprinting Rheology-informed machine learning model

Figure 2. (A) Process of data acquisition: investigation of printing resolution and rheological assessment. 3D graphs of the collected data with printing
parameters for the bioink compositions of (B) F127, (C) gelatin/xanthan gum, (D) alginate/CaCl , and (E) alginate/CNC composite.
2
using the algorithm to measure the strand size, as shown material concentration, printing pressure, nozzle diameter,
in Figure S1 (Supplementary File). In particular, the nozzle length, nozzle velocity, and printing resolution
algorithm was composed of two main steps. The first step was stored in the printing dataset. Furthermore, to create
was finding the path line, which can decrease errors by a sub-dataset for the rheological properties, 41 values of
a missed detection of the line. Afterward, the image was measured viscosity and 21 values of storage modulus data
changed to grayscale, and a position where the centers of were acquired at the angular frequency region from 0.1 to
the maximum signals on each line matched the printing 10 rad/s and subsequently, saved with bioink information.
path was found. The second step was to quantify the average
strand size. Specifically, the image was reconstructed to a 2.5. Machine learning model
binary one and cropped with white pixels to leave 2 mm This study utilized two classical machine learning algorithms
from the center lines to each side. Particularly, nine lines (random forest [RF] and support vector machine [SVM]),
in the cropped image were selected using a projection two conventional machine learning models (printing
grade of white pixels. Specifically, when the path and parameter-dependent machine learning model [PDML]
the selected lines were matched up, the average strand and concentration-dependent machine learning model
size of the scaffolds was calculated. The information on [CDML]), and a developed multi-input machine learning

Volume 9 Issue 6 (2023) 312 https://doi.org/10.36922/ijb.1280