Accuracy of Artificial Intelligence-Assisted CBCT Interpretation for the Detection of Endodontic Pathologies and Findings: A Systematic Review and Meta-analysis
Artificial intelligence (AI) has increasingly been applied to cone-beam computed tomography (CBCT) interpretation in endodontics. However, reported diagnostic performance varies across endodontic findings and AI-based approaches. This systematic review and meta-analysis evaluated the diagnostic accuracy of AI-assisted CBCT interpretation for detecting endodontic pathologies and findings. We searched PubMed/MEDLINE, Embase, Scopus, Web of Science, and the Cochrane Library for studies published between January 2016 and May 2026. We included studies evaluating AI-assisted detection, classification, or segmentation of endodontic pathologies on dental CBCT in human subjects. We extracted data on sensitivity, specificity, and accuracy, and assessed methodological quality using QUADAS-2. We used random-effects models to calculate pooled estimates, and assessed heterogeneity using Cochran’s Q and I² statistics. We performed subgroup analyses by diagnostic finding. We included 12 studies, contributing 31 diagnostic evaluations for sensitivity and 25 for specificity and accuracy. The pooled sensitivity, specificity, and accuracy were 79.82% (95% CI: 73.86–85.77), 86.80% (95% CI: 82.69–90.91), and 84.74% (95% CI: 80.53–88.96), respectively. Sensitivity was highest for periapical osteolytic lesions (97.10%) and apical periodontitis (88.27%), while specificity was highest for apical periodontitis (96.27%). For periapical radiolucency, pooled sensitivity, specificity, and accuracy were 87.89%, 71.46%, and 78.89%, respectively. Sensitivity, specificity, and accuracy differed significantly across diagnostic subgroups (all P < 0.001). Substantial heterogeneity was observed across the analyses. AI-assisted CBCT interpretation demonstrated clinically relevant diagnostic accuracy for endodontic findings, with performance varying by pathology.