Reproducibility and Validation Challenges in AI-Driven Scaffold Design for Bone Regeneration

Abstract

Artificial intelligence (AI) is increasingly applied to bioink formulation and bioprinting process control in tissue engineering (TE). Yet, translational progress remains constrained by small datasets, limited cross-platform validation, and weak links between computational predictions and biological outcomes. This review critically evaluates experimentally validated AI applications across scaffold-based bone regeneration, spanning materials design, fabrication control, and biological assessment. Physics-informed models are often more robust and generalizable than purely data-driven approaches, whereas transfer learning remains challenging due to variability in cellular responses and fabrication conditions. To address gaps in existing reporting standards, to address gaps in current reporting recommendations, we propose the AI–TE reporting framework, incorporating TE–specific requirements for biological validation, fabrication documentation, and translational readiness.

Impact Statement

Despite the rapid growth of AI-driven scaffold design for bone regeneration, reproducibility, and validation remain insufficiently addressed across current studies. This review identifies key methodological gaps and synthesizes practical validation and reporting priorities to improve reliability and translational relevance.