Page 12 - IJB-6-1
P. 12
Machine learning in bioprinting
positioned cells, curved layers, and microstructure accuracy can be analyzed by machine learning in
errors in a fabrication process which can monitor advance, the final fabricated bio-parts can then be
the whole bioprinting process. Figure 4 shows an guaranteed in good quality. The process is similar,
example of using CNN to detect flaws, errors, or as shown in Figure 4, while the input data are
defects in 3D bioprinting. The input data can be different.
the images from high-quality cameras during the
bioprinting process. Then, the data are analyzed 3.4 Material property design or prediction
by the CNN model to predict a defect or other In traditional 3D printing, Gu et al. proposed
[29]
objectives. a machine learning-enabled method to design
3.3 Dimensional accuracy analysis hierarchical composites for fabrication, trained
with a database of enough structures from finite
In the traditional 3D printing process, Francis and element analysis. Hamel et al. presented
[30]
Bian developed a deep learning method that can a machine learning method to design active
[22]
accurately predict the distortion of parts in laser- composite structures where target shape shifting
based AM. Similarly, Khanzadeh et al. proposed responses can be achieved in a 4D printing process.
[23]
an unsupervised machine learning approach Li et al. proposed a predictive modeling method
[31]
(self-organizing map) to quantify the geometric with machine learning that can predict the surface
deviations of additively manufactured parts in roughness of FDM printed parts with high accuracy.
fused filament fabrication processes. In addition, Currently, Jiang et al. used backpropagation
[32]
Zhu et al. also developed a strategy coupled neural network to analyze and predict printable
[24]
with machine learning to address the modeling bridge length in FDM processes.
[38]
of shape deviations in AM. Tootooni et al. Similarly, machine learning can also be
[25]
compared six machine learning techniques (sparse used to design or analyze material properties
representation, k-nearest neighbors, neural in 3D bioprinting process. For example, tissue-
network, naïve Bayes, support vector machine, engineered scaffolds are very important in 3D
and decision tree) with regard to the accuracy bioprinting, whose structures should be carefully
of predicting dimensional variation in fused designed for successful cell growth and function
deposition modeling (FDM) printed parts. Based achievement. For example, if we want to bioprint
on their study, the sparse representation approach an organ in the future, the scaffolds should
has the best classification performance. have the specific structure for the successful
In 3D bioprinting, similarly, machine learning growth of cells to form a functional organ. In
can be used for analyzing the accuracy of the case here, what kind of complex scaffold
fabricated bio-parts. For example, the tissue- structure is the most suitable? How the properties
engineered scaffolds are generally very complex (e.g., density and strength) of the scaffolds will
because they supporting cell growth in an expected influence the printed organ function? If machine
way to achieve corresponding functions. If the learning can be used to generate these qualified
Figure 4. An example of using machine learning (convolutional neural network) in three-dimensional
bioprinting.
8 International Journal of Bioprinting (2020)–Volume 6, Issue 1
