Page 524 - IJB-9-6

International Journal of Bioprinting Advances for 3D-printed oral drug delivery systems

3.4. Digital technologies geometry. The results demonstrated that it was not possible
This classification includes a total of 14 publications to predict a geometry with the required length, width,
and is characterized by the use of digital technologies to height, and underlying geometry, with ANN .
[99]
improve and enhance processes and dosage formulations
with machine learning, decreasing the risk of frauds with 3.4.2. Cybersecurity
cybersecurity and novel ways of adding information to the Cybersecurity involves protecting systems, networks, and
printed forms by quick response (QR). programs to reduce or avoid the risk of cyberattacks. The
most common target of cyberattacks is accessing, changing,
or destroying sensitive information [100] . The supporting
3.4.1 Machine learning software and systems for the formulation or production of
Machine learning (ML) can be used as a predictive method 3D-printed dosage could be a target to cyberattacks as they
for formulation and release profiles in 3D-printed dosage might contain sensitive information related to the patient
forms, making it a possible quality control tool. ML can or the dosage form.
automate and reduce development times while maintaining
a good accuracy of the design parameters. Due to the possible cyber risks of remote digital transfer
of an electronic prescription to the 3D printer while
ML was employed alongside 3D printing to provide on- printing dosage forms, Kok et al. explored the application
demand manufacturing and quality control of orodispersible of DEFEND3D, a technology to enhance cybersecurity
films. O’Reilley et al. developed orodispersible films and intellectual property protection [101] . DEFEND3D is a
with direct ink writing (DIW) and complex geometries. patented secure streaming transfer protocol (SSTP) and a
These films were classified by active ingredient using ML virtual inventory communication interface with controlled
algorithms and NIR spectrums. Based on the results of the reproduction. Different shapes were created using remote
subsequent partial least square algorithm, it was stated that fused deposition modeling. The authors concluded that
ML, 3D printing, and NIR have the potential to automate DEFEND3D can remotely 3D-print various designs at
orodispersible film workflows and enable rapid drug and various infill densities [101] .
dose verification .
[97]
3.4.3. Quick response and binary digits (bits)
In 2021, Obeid et al. also tried to predict the diazepam
release of tablets with artificial neutral networks . Tablets Quick response (QR) codes allow the storage of information
[98]
in small surface areas; the information is easy to access as
of different shapes were printed using fused deposition the QR code just needs to be scanned with a smartphone
modeling. Self-organizing maps and multi-layer perceptron or QR scanner.
were applied to model the influence of tablet surface area
to volume (SA/V) ratio and printing parameters (infill In 2019, Trenfield et al. printed QR codes and data
density and infill pattern) on the release of diazepam. The matrices on the surface of paracetamol printlets to
results showed the ability of the multi-layer perceptron generate a unique track-and-trace measure for product
network to predict drug release behavior as a function of authenticity [102] . The QR code can be scanned with
infill density and SA/V ratio . a smartphone, and the encoded information can be
[98]
Ong et al. created a balanced database of 1594 personalized to illustrate data related to the drug product
formulations with in-house and literature data of hot (batch, expiration date, active ingredient, etc.), patient
melt extrusion and fused deposition modeling to predict (age, birth, and gender), and prescriber (name). The results
[102]
formulation outcomes using ML. The models were able to demonstrated a novel anti-counterfeit mechanism .
predict hot melt extrusion and fused deposition modeling Oh et al. developed a 3D-printed QR-coded
processing temperatures with a mean absolute error of orodispersible film (QRODF) in a one-step process using
5.5°C and 8.4°C, and the printability and mechanical a hot melt pneumatic process. The QRODF was loaded
characteristics of the filaments with an accuracy of 84%. with aripiprazole and can be read with a smartphone in a
The optimized models were added to the FabRx web- QR scanning application to obtain additional information
application software M3DISEEN . about the film. QRODF may be a promising approach for
[66]
tailored drug formulations as they are easy to scan and are
Mazur et al. obtained specific dosage and release profile [103]
using fused deposition modeling dosage forms employing not easily broken due to their flexibility .
artificial neural networks (ANN) to predict appropriate Two years later, Windolf et al. used a different approach
geometries. With the in vitro dissolution results and the using QR codes to store information on 3D-printed
mathematical description of the API release profiles, ANN geometries and ensure batch traceability [104] . The dosage
architectures were created to predict the most suitable forms were fabricated with fused deposition modeling, a

Volume 9 Issue 6 (2023) 516 https://doi.org/10.36922/ijb.1119