Diagnostic accuracy of convolutional neural network for detection of MB2 canal in CBCT: systematic review and meta-analysis

• CNNs showed high accuracy in detecting MB2 canals on CBCT scans • Meta-analysis found 88.9% sensitivity and 97.5% specificity • Risk of bias was low in two studies, with concerns in two others • Certainty of evidence was rated high according to GRADE The presence of a second mesiobuccal canal (MB2) in maxillary first molars is a frequent anatomical finding with significant clinical implications in endodontic success. Failure to detect and treat MB2 canals is one of the primary causes of persistent periapical pathology and retreatment cases. The objective was to evaluate the diagnostic accuracy of deep learning (DL) models based on convolutional neural networks (CNN) in detecting MB2 in maxillary first molars using cone-beam computed tomography (CBCT). A systematic review and meta-analysis was conducted following PRISMA-DTA guidelines. Searches were performed in six databases up to November 2025. Only retrospective observational studies applying CNN-based DL models to CBCT scans for MB2 detection were included. The protocol was registered in PROSPERO (CRD420251113423). Risk of bias was assessed using QUADAS-2, and certainty of evidence was evaluated via GRADE. Four retrospective studies comprising 528 CBCT scans were included. The pooled sensitivity and specificity were 88.9% (95% CI: 83.9–92.4%) and 97.5% (95% CI: 85.2 – 100%), respectively. The area under the summary ROC curve was 0.887. Risk of bias was low in two studies and moderate in the others. Heterogeneity was minimal based on adjusted estimators. CNN-based deep learning models show promising diagnostic accuracy for detecting MB2 canals in maxillary first molars using CBCT. These findings suggest that CNN-assisted analysis may provide useful support for clinicians when evaluating complex root canal anatomy or CBCT scans with challenging visualization.