A Deep Learning Quality Control Loop of the Extrusion-based Bioprinting Process.

Area of Science:

• Biomaterials Engineering
• Additive Manufacturing
• Artificial Intelligence in Medicine

Background:

• Extrusion-based bioprinting (EBB) is a widely used deposition technology with versatile applications.
• Optimizing EBB parameters often involves extensive trial-and-error, hindering standardization and clinical translation.
• Machine learning (ML) shows promise for enhancing quality control in bioprinting processes.

Purpose of the Study:

• To develop and validate a deep learning-based control loop for automated optimization and online monitoring of EBB.
• To reduce material waste and processing time by enabling early detection of printing errors.
• To accelerate the translation of bioprinted products to clinical applications through improved process control.

Main Methods:

• Collected a comprehensive dataset of EBB prints using high-resolution webcam recordings.
• Trained a custom convolutional neural network (CNN) on diverse printing scenarios, controlling for overfitting and optimizing prediction time.
• Integrated the trained ML model into a control loop for real-time process monitoring and automated parameter adjustment.

Main Results:

• The deep learning model effectively monitored the EBB process, identifying potential errors for early termination.
• The integrated control loop successfully optimized printing parameters by combining ML predictions with a mathematical EBB model.
• Demonstrated the feasibility of ML for automating extrusion-based bioprinting, enabling a complete quality control loop.

Conclusions:

• Machine learning techniques can automate extrusion-based bioprinting processes.
• A deep learning-based control loop offers a robust solution for quality control in EBB.
• This approach paves the way for standardized, efficient, and accelerated clinical translation of bioprinted constructs.