Machine learning for early detection and risk prediction in peri-implantitis: A review

Abstract

Peri-implantitis causes progressive bone loss around dental implants and arises from interacting microbial, host, behavioral and implant factors. Conventional clinical and radiographic criteria capture this complexity only partially, whereas machine learning can integrate heterogeneous data across early detection, risk stratification and outcome prediction. Convolutional neural networks detect peri-implant bone loss on radiographs with high reported accuracy. Supervised models, such as random forests, combine clinical and patient-level data to estimate individual risk, often outperforming conventional statistics. Thus, data shows that evidence rests largely on small, single-center, retrospective datasets, so machine learning should augment rather than replace clinical judgment until standardized, externally validated models are available.