Related Experiment Video
Updated: Jun 11, 2026
07:05
Viability of Bioprinted Cellular Constructs Using a Three Dispenser Cartesian Printer
Published on: September 22, 2015
Artificial intelligence in 3D bioprinting and biofabrication: a validation-stringency assessment.
Jae-Seong Lee1, Ju-Hee Cha1, Byoung Soo Kim1,2,3,4
1Department of Information Convergence Engineering (Biomedical Convergence Engineering Major), Pusan National University, Yangsan 50612, Republic of Korea.
Biofabrication
|June 9, 2026
Summary
Artificial intelligence (AI) and machine learning (ML) are advancing 3D bioprinting, but most models lack broad validation. Developing standardized datasets is crucial for reproducible AI in bioprinting across diverse labs and conditions.
Area of Science:
- Bioprinting and Biofabrication
- Artificial Intelligence in Medicine
- Tissue Engineering
Background:
- Artificial intelligence (AI) and machine learning (ML) are increasingly integrated into 3D bioprinting and biofabrication processes.
- Current AI/ML models in bioprinting are often trained and tested within limited experimental conditions (single batch, printer, or lab), hindering generalizability.
- The lack of standardized validation frameworks limits the understanding of AI model performance under varying conditions.
Purpose of the Study:
- To introduce a comprehensive framework for evaluating AI/ML methods in 3D bioprinting and biofabrication.
- To assess the current validation rigor of AI/ML applications across the bioprinting pipeline and various tissue systems.
- To identify key challenges and propose solutions for enhancing the reproducibility and reliability of AI in bioprinting.
Main Methods:
- Development of a three-tier taxonomy for classifying AI methods based on learning approach, model type, and task.
- Application of a five-level Validation Ladder, adapted from engineering and clinical AI, to grade evaluation rigor.
- Systematic review and analysis of 40 primary studies in 3D bioprinting, categorized by AI tier and validation level across seven tissue systems.
Main Results:
- The majority of reviewed studies (83%) are at Validation Level 0 or 1, indicating limited testing beyond internal datasets.
- Only 15% of studies reached Level 2 (cross-condition testing), and just one study (2.5%) reached Level 3 (multi-laboratory testing); no studies reached Level 4 (real-time deployment).
- Higher validation levels require data from multiple independent batches and printer configurations, highlighting experimental scope limitations rather than method immaturity.
Conclusions:
- Current AI/ML applications in 3D bioprinting exhibit low validation levels, restricting their real-world applicability.
- The development and adoption of community benchmark datasets, annotated with AI tier and validation level, are essential.
- Standardized datasets will facilitate the creation of reproducible and generalizable AI models for bioprinting, advancing tissue engineering and regenerative medicine.
