Classification of dental implants using supervised deep learning from dental radiographs: A scoping review.

Purpose: This scoping review examined the evidence on the use of supervised deep learning models for the classification of dental implants using radiographic images. Materials and Methods: A preliminary search was conducted in PubMed, Google Scholar, PROSPERO, JBI Evidence Synthesis, and the Open Science Framework, identifying a small number of relevant records. A comprehensive search was subsequently performed across 7 databases using adapted strategies without filters. Studies were included if they evaluated implant classification using supervised deep learning models applied to panoramic or periapical radiographs. Studies were excluded if they did not involve implant classification, were review articles, involved children, were unavailable in full text, or did not apply artificial intelligence methods. Data extraction was conducted by 2 independent reviewers, with disagreements resolved by a third reviewer. Descriptive statistics were used for data analysis. Results: Of 274 records, 9 studies met the inclusion criteria. Studies published between 2020 and 2024 evaluated deep learning and machine learning approaches for dental implant identification and classification from radiographic images. A range of models was applied, predominantly convolutional neural networks. Dataset sizes ranged from 355 to 156,965 radiographs and included multiple implant brands. Few studies addressed ethical considerations related to recent data protection regulations. Conclusion: This scoping review indicates that most deep learning-based approaches to implant classification using dental radiographs primarily rely on implant brand or manufacturer as the labeling strategy. This reliance may limit model generalizability and long-term applicability due to the frequent discontinuation of implant systems. Future studies should focus on intrinsic radiographic characteristics, including macrogeometry and prosthetic connection types.