Related Experiment Video
Updated: Sep 10, 2026
Reliability of Artificial Intelligence-Based Cone Beam Computed Tomography Integration with Digital Dental Images
Published on: February 23, 2024
Multicentre Evaluation of AI-Assisted Caries Detection on Panoramic Radiographs Among Early-Career Dentists
Yujia Wu1, Xiaowei Hou2, Peng Ding3
1Department of Geriatric Dentistry, Peking University School and Hospital of Stomatology, National Center for Stomatology, National Clinical Research Center for Oral Diseases, National Engineering Research Center of Oral Biomaterials and Digital Medical Devices, Beijing, China.
Introduction And Aims:
Artificial intelligence - assisted diagnostic tools are increasingly introduced into dental practice, yet their impact on clinician performance in routine panoramic radiograph interpretation remains incompletely defined. This multicentre reader study evaluated whether AI assistance improves diagnostic accuracy, efficiency, and inter-reader consistency in caries detection, particularly among early-career dentists.
Methods:
Twelve early-career dentists (≤3 years' experience) and three senior dentists (>10 years' experience) independently interpreted 402 anonymized panoramic radiographs under four conditions: unassisted reading, AI-assisted reading, AI-only analysis, and expert reference. Diagnostic performance was compared with a standardized expert-derived reference standard established by three experienced dentists using pixel-level annotations. Sensitivity, specificity, area under the receiver operating characteristic curve, interpretation time, and inter-reader agreement were analysed.
Results:
AI assistance increased tooth-level sensitivity (82.4% vs 67.4%, P < .001) without compromising specificity (97.4% vs 97.2%), and reduced mean interpretation time (50.37 vs 65.12 seconds, P = .003). Case-level sensitivity improved from 84.7% to 93.3% (P < .001). Inter-reader agreement increased from κ = 0.61 to 0.73. The standalone AI system achieved a sensitivity of 79.2% (95% CI, 76.0-82.4) and specificity of 98.4% (95% CI, 76.0-82.4), and AUC of 0.938 (95% CI, 0.934-0.941).
Conclusion:
AI assistance improved diagnostic sensitivity, efficiency, and consistency among early-career dentists interpreting panoramic radiographs for caries detection, without increasing false-positive rates.
Clinical Relevance:
AI-supported interpretation may help reduce experience-related variability in panoramic radiograph assessment and improve diagnostic efficiency in routine dental practice, particularly in settings with limited access to senior supervision.