Diagnostic Accuracy of Deep Learning Models for the Detection of Pulpal Calcifications in Dental Radiographs: A Systematic Review and Meta-analysis

Introduction Pulpal calcifications are a relatively common finding in dental radiographic examinations and are often associated with factors such as dental trauma, aging, chronic inflammation, and systemic conditions. The study objective is to evaluate the diagnostic accuracy of convolutional neural network (CNN) based deep learning (DL) models for detecting pulpal calcifications on dental radiographs. Materials and Methods A systematic review and meta-analysis was conducted under PRISMA-DTA guidelines (PROSPERO: CRD420251102558). Searches were performed in PubMed, Embase, Scopus, Web of Science, Google Scholar, Clinicaltrials.gov, and Cochrane Library (November 2025). Eligible studies assessed CNN-based DL models. Data extraction and risk of bias (QUADAS-2) were performed independently. Pooled diagnostic values were estimated with a bivariate random-effects model, and certainty was graded with Grading of Recommendations Assessment, Development and Evaluation (GRADE). Results Six retrospective studies with 6,916 radiographs were included, applying CNN architecture. Pooled sensitivity was 87% and specificity 92%, with an AUC of 0.95. Risk of bias was low, with some heterogeneity. Conclusion CNN-based DL models show high accuracy for pulpal calcification detection, supporting their potential use in endodontics and systemic health screening. However, all included studies were retrospective and based on internal validation, which may limit the generalizability of these findings to real-world clinical settings.