Diagnostic Accuracy of Artificial Intelligence and Machine Learning in Dental Age Estimation: A Systematic Review and Meta-Analysis.

Abstract

Background: Traditional dental age estimation methods rely heavily on subjective human evaluation, which is prone to intra- and inter-observer variability. The application of artificial intelligence (AI), specifically convolutional neural networks (CNNs) and vision transformers, to radiographic imaging offers a promising avenue for automated, objective, and accurate forensic age estimation. Objectives: To systematically review and meta-analyze the diagnostic accuracy and prediction performance of AI-based models in estimating dental age from radiographic images, with a focus on the 18-year legal threshold. Methods: Following PRISMA guidelines (PROSPERO CRD420261471359), a comprehensive search of PubMed, IEEE Xplore, Cochrane Library, Science Direct, and grey literature was conducted for studies published up to 2026. Studies evaluating AI for dental age estimation using panoramic or tomographic radiographs were included. Risk of bias was assessed using the QUADAS-2 tool. Bivariate random-effects models were utilized to calculate pooled sensitivity, specificity, Area Under the Curve (AUC), and Diagnostic Odds Ratio (DOR). Results: A total of 55 studies were included in the meta-analysis. The pooled sensitivity and specificity of AI models for age estimation were 91.2% (95% CI: 88.5–93.4) and 90.5% (95% CI: 87.9–92.8), respectively. The summary ROC curve demonstrated excellent discriminatory ability with a pooled AUC of 0.92 and a DOR of 105 (95% CI: 85–130). Risk of bias was predominantly high in the domains of patient selection (49%) and flow and timing (58%), largely due to reliance on retrospective, single-ethnicity datasets. Conclusions: Artificial intelligence provides highly accurate and reliable diagnostic performance for dental age estimation. However, the generalizability of these models is limited by ethnically homogenous training datasets. Future research must prioritize multicenter, diverse datasets and explainable AI frameworks.