Machine Learning for Composite Biomaterials in Bone Implants and Scaffolds: A Systematic Review of Manufacturing and the Process–Structure–Property–Performance Relationships

Bone-implant and scaffold research is beginning to use machine learning (ML) alongside biomaterials design and advanced manufacturing, but the literature is scattered across different material classes and stages of development. We therefore reviewed ML studies relevant to composite, multiphase, coated, graded, and multimaterial systems and organized them using the process-structure-property-performance (PSPP) sequence. PubMed/MEDLINE and Scopus were used as the formal bibliographic databases; IEEE Xplore screening, targeted OpenAlex searches, and Google Scholar citation chasing provided supplementary coverage. The frozen dataset contained 46 primary studies. Twenty-five satisfied a strict composite/multimaterial definition, whereas 21 were retained as boundary-supporting evidence for the sensitivity analysis. The studies show that ML is being used for property prediction, surrogate modeling, inverse design, multi-objective optimization, and material screening. Validation, however, did not advance at the same pace as modeling. Physical or experimental confirmation was identified in 32/46 studies (69.6%), in vitro or in vivo experiments, including antibacterial assays, in 11/46 (23.9%), and external validation in only 2/46 (4.3%); none reported prospective clinical validation. Explicit uncertainty analysis and strong reproducibility were also uncommon. Taken together, the literature suggests that the main challenge is no longer demonstrating that an ML model can fit a biomaterials dataset. The more difficult task is showing that a model-guided choice remains useful after fabrication and biological testing. Priorities therefore include better-described multimodal datasets, independent and leakage-resistant validation, characterization of the manufactured rather than only nominal structure, uncertainty-aware design, physics-informed modeling, and prospective biological testing.