Page 10 - IJAMD-1-1

International Journal of AI
for Material and Design ML in 3D bioprinting of cultivated meat

reduce the environmental impact associated with supervised learning – which learns from labeled datasets;
22
traditional livestock farming, which is a major contributor unsupervised learning – which discovers patterns without
to greenhouse gas emissions, deforestation, and water labels; semi-supervised learning – which combines labeled
23
2
24
pollution. Cultivated meat also offers a more humane and and unlabeled data; and reinforcement learning – which
25
ethical alternative for meat consumption, aligning with the trains through trial- and-error processes using reward. Its
growing concern for animal welfare among consumers. unique advantage lies in its ability to analyze large datasets,
Furthermore, cultivated meat has the potential to reduce identify patterns, and make intelligent decisions. It is
the risk of foodborne illnesses associated with traditional important to emphasize the necessity of having extensive
meat production, as it can be produced under controlled data for training both supervised and unsupervised
and sterile conditions and reduce the need for antibiotics learning models. Before initiating the training process for
and other chemicals. The field of cultivated meat is still in an ML model, substantial pre-processing of the acquired
its early stages of development and faces various challenges data is imperative. The utilization of labeled data can
such as cost scalability and public acceptance. Nevertheless, be costly due to the expertise and time required in the
it represents a promising and exciting avenue for the future labeling process. Selecting an appropriate ML model is not
of food production and plays an important role in creating always a straightforward task, as each model comes with
26
a more sustainable and ethical global food system. its own set of risks and benefits, involving trade-offs.
One of the primary challenges hindering the widespread
There is a growing interest in the application of 3D implementation of ML in manufacturing is decision
bioprinting technology across various fields, including validation. The lack of interpretability in ML model outputs
food bioprinting, tissue engineering, and regenerative poses a significant challenge, especially when considering
medicine. The goal of bioprinting is to create 3D the potential financial costs associated with failures in
3-5
artificial tissues by precisely depositing biomaterials and manufacturing operations. Consequently, determining the
living cells at pre-defined positions. 3D bioprinting is trustworthiness of decisions made by ML models remains
6,7
an advanced manufacturing platform that enables the an ongoing research topic, particularly because ML models
precise layer-by-layer deposition of various bioinks, living often operate as black boxes. Historically, human operators
cells, and biomolecules to fabricate 3D structures that gradually develop trust in new software technologies by
closely mimic the spatial arrangement and composition of observing outputs over time, and a similar learning process
native tissues. 8-10 There are three primary categories of 3D is anticipated for ML applications. 26
bioprinting technologies, each with its distinct advantages
In the context of cultivated meat production, ML can be
and limitations: extrusion-based, jetting-based, 12,13 and vat applied to bioprinting techniques – predicting printability,
11
photopolymerization-based bioprinting. 14,15 Jetting-based optimizing printing parameters, and characterizing
bioprinting allows for contactless, on-demand deposition meat flavor – using molecular composition and sensory
of living cells and bioinks to manipulate cell-extracellular array data. In addition, ML ensures quality control by
matrix interactions. Extrusion-based bioprinting can print determining the total viable count (TVC) and detecting
a wide range of bioinks with varying viscosities at high meat adulteration. Real-world analysis of the prints,
fabrication speeds, while vat photopolymerization-based informed by ML techniques, generates substantial data that
bioprinting enables the printing of 3D tissues with high can subsequently be fed back into the models, establishing
resolution and cell density. a synergistic relationship between ML and cultivated meat
Over the years, various sensors have been employed production. This capability positions ML as a valuable
for monitoring and controlling purposes. 16-18 These tool for optimizing material formulation, 3D bioprinting
sensors offer ease of implementation, require minimal of meat scaffolds, regulating meat flavor, and monitoring
computational resources, provide clear and interpretable overall meat quality (Figure 1). An in-depth discussion on
data, and operate independently of extensive datasets. the application of ML in the production of 3D-bioprinted
Nevertheless, conventional sensors may exhibit some cultivated meat is presented in the subsequent sections.
degrees of inflexibility, as they adhere to pre-programmed
rules, necessitating meticulous programming for specific 2. 3D bioprinting of cultivated meat
applications, and they may struggle to adapt without 3D bioprinting has emerged as a revolutionary technology
manual adjustments or identify complex patterns and in the production of cultivated meat, with numerous
anomalies. In contrast, machine learning (ML), which studies exploring 3D food bioprinting. 27-32 The current
is a subset of artificial intelligence, empowers systems to trial-and-error method employed in developing
learn from data, recognize patterns, and make intelligent printable bioink compositions is a time-consuming and
decisions. 19-21 There are four main types of ML algorithms: resource-intensive process. This challenge is exacerbated

Volume 1 Issue 1 (2024) 4 https://doi.org/10.36922/ijamd.2279