Clinical Readiness of Artificial Intelligence Models for MB2 Canal Detection in Maxillary Molars: A Scoping Review

The second mesiobuccal (MB2) canal detection in maxillary molars represents a notable challenge in endodontics. This scoping review (SR) attempts to map the existing research on artificial intelligence (AI)-based models in MB2 detection, delineating the current stage of clinical readiness and gaps associated with routine clinical implementation.

A comprehensive, independent search was conducted in three databases (PubMed, Embase, and Scopus) to synthesize evidence from inception to 20 February 2026, adhering to PRISMA 2020 guidelines and PRISMA-Scr checklist. Study designs, including observational, diagnostic accuracy, as well as experimental designs investigating human permanent maxillary molars employing AI approaches, including machine learning or deep learning models, were included.

From 1218 identified records, 5 studies met inclusion criteria. All 5 studies reported strong diagnostic performance of AI models in detecting MB2 canals with favourable diagnostic performance in both deep learning and machine learning models. The clinical readiness assessment using the Technology Readiness Level - Implementation Science framework reported progression of all 5 studies to level 4. Overall risk of bias (RoB) assessment using the QUADAS-2 tool reflected some concerns, highlighting unclear or insufficient methodological components and low overall applicability concerns. RoB assessment was visualized via the ROBVIS tool.

Overall, AI models demonstrate favourable diagnostic performance for MB2 canal identification; however, current evidence highlights important limitations in their translational readiness.

AI-based tools show near-expert accuracy for MB2 detection on cone beam computed tomography, but all current models remain at the prototype stage without real-world validation. They should be used as decision-support adjuncts, not standalone diagnostic systems.