Page 6 - IJB-7-1

Application of Machine Learning in 3D Bioprinting

Figure 1. Example methods in machine learning.

to find the functions between X and Y like in supervised is, therefore, needed because of multiscale complexities of
ML, but through a dynamic interaction with another representing biological tissue models. In addition, ML can
test algorithm called environment. The environment also help predict the compatibility of dissimilar materials
rewards or punishes the agent’s trial and error learning used in bioprinting [5,6] . In processing and post-processing,
so that the ML model becomes more and more accurate it is almost impossible to perform wet experiments when
in predication by maxing the reward. Another similar the number of changing parameters exceeds a certain
method is deep learning, in which the trained learning number, for example, ten parameters. ML is, therefore,
algorithms have multiple hidden layers and are always needed because of multiparameter complexity of finding
applied to new datasets instead of dynamically adjusting optimal protocol of bioprinting. Here, it is envisioned that
agent’s actions from the continuous feedback. Above are ML coupled with Big Data, will solve the multiscale and
example methods in ML. In fact, learning is quite a broad multiparameter complexities and transform present 3D
topic and several techniques have already been used in bioprinting into future 3D bioprinting, which is “heavily
3D printing in general [3,4] . virtual” in nature.

2. Complexities in bioprinting 3. In silico experiment and big data
In general, ML models are preferred when complexities In silico, experiment such as digital fabrication and
arise, because they can account for factors or conditions computational study is believed to be a key innovation driver
not considered in traditional mathematical models, that and play a major role in the new era of tissue engineering [7,8] .
is, they tend to be more robust in the real-world context Although the current challenge is creating more realistic
in terms of predication. Bioprinting coming across virtual experiment with satisfactory accuracy, there are
ML is inevitable for the reason of complexities. The various methods such as statistical tools and techniques
complexities of bioprinting span across the entire process that can be combined with first-principles simulations to
chain, namely, pre-processing, processing, and post- solve it . In the case of bioprinting, ML has already been
[9]
processing (Figure 2). In pre-processing, it is challenging used to optimize the printability of bioinks and drastically
to perform segmentation of tissue images at a single cell reduced the number of experiments from thousands of
level and reconstruct them into a 3D tissue model with possible combinations . Moreover, various mathematical
[10]
cross-scale cellular resolution and tissue properties. ML models on bioink printability have been developed as
2 International Journal of Bioprinting (2021)–Volume 7, Issue 1