Abstract
Background and Objectives: AI is considered one of the most innovative technologies of this century. Its introduction into healthcare has transformed the industry, significantly impacting various aspects such as education, teaching, diagnosis, treatment planning, and patient care. Researchers have tested the accuracy of various generations of AI models for detecting peri-implant bone loss using radiographic images. While studies have reported promising outcomes, there remains significant potential for improving these models. This systematic review aims to critically analyze the existing published literature on the accuracy of AI models in detecting peri-implant bone loss and to evaluate the current state of knowledge in this area. Methods: The guidelines established by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) were pivotal and provided a framework for preparing, implementing, and recording this systematic review. The protocol for this review was registered in PROSPERO. Four electronic databases (PubMed, Scopus, Web of Science, and Cochrane) were diligently searched on 5-6 January 2025, targeting articles published between January 2000 and December 2024. The PIRD elements (population, index test, reference test, diagnosis of interest) that helped in structuring the protocol of the present review were: P: X-ray images of humans demonstrating the bone loss around the dental implant; I: Artificial intelligence models used for detecting radiographic peri-implant bone loss; R: Expert opinions and reference standards; D: Radiographic peri-implant bone loss. The Quality Assessment and Diagnostic Accuracy Tool (QUADAS-2) was used to assess the quality of each included study. Results: Seven studies met the selection criteria and were included in the qualitative analysis. A self-designed table was used to tabulate all the relevant study characteristics. The included studies were reported to have a moderate level of certainty of evidence as assessed by the GRADE assessment. In general, all studies included in this review demonstrated a low risk of bias. Overall accuracy of the AI models varied and ranged between 61% and 94.74%. The precision values ranged from 0.63% to 100%. Whereas sensitivity and specificity values range between 67% and 94.44%, and 87% and 100%, respectively. Conclusions: The present systematic review highlights that AI models demonstrate high accuracy in detecting peri-implant bone loss using dento-maxillofacial radiographic images. Thus, AI models can serve as effective tools for the practicing dentist in confirming the diagnosis of peri-implant bone loss, ultimately aiding in accurate treatment planning and improving treatment outcomes.