Machine learning-based microstructure identification in lithium silicate dental glass‒ceramics

This study explores the application of machine learning (ML) to identify and quantify the microstructure of lithium silicate based dental glass‒ceramics (DGCs), a material at the forefront of prosthetic dental applications due to its good mechanical properties and aesthetics. Despite the increasing demand for LS-based DGCs, the design process has seen little innovation, largely due to the complexity of the potential glass compositions and the associated microstructures. This research builds on previous efforts to model microstructure of glass‒ceramics by utilizing scanning electron microscopy (SEM) and X-ray diffraction data from 10 commercial LS-based DGCs. The microstructures were analyzed after various heat treatments, and ML techniques, including Weka segmentation, were employed to classify and label the phases within the SEM images. These labeled images were then used as inputs for an ML algorithm aimed at identifying each microstructure, paving the way for further optimization of DGCs through advanced computational methods.