Sustainable silk fibroin scaffolds for bone repair: assessing their osteogenic potential via AI-enhanced synchrotron imaging workflow

Effective bone regeneration requires scaffolds capable of guiding and supporting new mineralized matrix formation. In this study, silk fibroin constructs cultured with human mesenchymal stem cells (hBMSCs) in the presence of either Fetal Bovine Serum (FBS) or Human Platelet Lysate (hPL) are evaluated for their osteogenic potential. A distinctive aspect of this work is the combined use of synchrotron X-ray imaging and a convolutional neural network for high-resolution in situ three-dimensional scaffold osteogenic potential assessment. This approach enables precise evaluation of bone matrix arrangement within the scaffold architecture. Two-dimensional analysis reveals increased mineralization in pores with an average radius of ~115 μm, area of ~4.0 × 104 μm2, and eccentricity of ~0.7 in hPL construct. The subsequent three-dimensional analysis extends these findings by quantifying the spatial distribution and connectivity of the mineralized matrix across the scaffold volume. It identifies pores with an equivalent radius between 110 and 120 μm, high surface area, and moderate sphericity (0.65–0.75) as optimal not only for mineral deposition but also for uniform 3D matrix propagation. Moreover, unsupervised clustering analysis also identifies optimal geometric interdependencies between pore size, surface area, and sphericity, offering new insights for rational design of high-performance scaffolds. The study demonstrates both the efficacy of silk fibroin scaffolds cultured with hPL in promoting bone regeneration and the relevance of a combined synchrotron imaging-artificial intelligence approach in quantitatively correlating three-dimensional porous geometry with regenerative outcomes.