Artificial intelligence in the detection of dental pulp calcifications: a scoping review

Dental pulp calcifications, including pulp stones and diffuse calcific changes, can complicate endodontic access, canal negotiation, and treatment planning. Artificial intelligence may support radiographic detection of these findings, but the available evidence remains limited and methodologically heterogeneous. This scoping review mapped peer-reviewed studies that applied artificial intelligence to detect, classify, or segment dental pulp calcifications on two-dimensional radiographs and cone-beam computed tomography.

A scoping review was conducted in accordance with the Joanna Briggs Institute methodology and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews. PubMed, Scopus, Web of Science, IEEE Xplore, and Google Scholar were searched for English-language studies published from 1 January 2015 to 29 September 2025. After duplicate and ineligible records were removed, 450 records underwent title and abstract screening, 32 full-text reports were assessed, and seven peer-reviewed studies were included. Data were charted descriptively according to imaging modality, artificial intelligence architecture, task type, validation design, annotation approach, and reported performance. Owing to heterogeneity in tasks, datasets, and outcome metrics, no meta-analysis or inferential statistical testing was performed.

The included studies were retrospective, single-centre investigations published between 2023 and 2025. Two-dimensional radiographic studies using detection or classification pipelines reported high internal performance, with accuracy ranging from 95.4% to 96.5% and F1 scores ranging from 78.9% to 96.6%. One panoramic segmentation study reported Dice scores of 0.84 for pulp and 0.759 for pulp stones. In cone-beam computed tomography, one three-dimensional U-Net study reported accuracy of 72.8% and an area under the curve of 0.74, with reduced sensitivity for micro or diffuse calcifications. No study used external validation, and public code or datasets were generally unavailable.

Artificial intelligence methods show technical potential for detecting dental pulp calcifications, particularly in internally validated two-dimensional radiographic studies. However, the evidence is still preliminary. Clinical translation is limited by single-centre designs, small and heterogeneous datasets, lack of external validation, inconsistent reporting, and limited reproducibility. Future research should prioritize multicentre datasets, lesion-size stratification, standardized reporting, open benchmarking, calibration assessment, and prospective workflow-based evaluation.

Not applicable. This study was a scoping review and did not involve a clinical trial.