Related Experiment Video
Updated: Jun 18, 2026
05:49
Reliability of Artificial Intelligence-Based Cone Beam Computed Tomography Integration with Digital Dental Images
Published on: February 23, 2024
Comparative evaluation of proximal caries detection methods using human, artificial intelligence, and micro-CT
Giovanna Heiden1, Patricia Pereira1, Mirian Lumi Yoshida2
1Department of Restorative Dental Sciences, College of Dentistry, University of Florida, 1395 Center Dr., Gainesville, Fl, 32610-0415, USA.
Journal of Dentistry
|June 16, 2026
Summary
Artificial intelligence (AI)-assisted bitewing analysis shows promise for detecting enamel caries but struggles with dentin lesions. A multimodal approach combining visual and radiographic assessments remains crucial for accurate diagnosis and treatment planning.
Area of Science:
- Dental diagnostics
- Radiology
- Artificial Intelligence in Medicine
Background:
- Accurate diagnosis of proximal caries lesion depth is critical for effective preventive and restorative dental care.
- Traditional methods like visual examination (ICDAS) and bitewing radiography have limitations in precision.
- Artificial intelligence (AI) offers potential for enhancing radiographic interpretation.
Purpose of the Study:
- To evaluate the diagnostic accuracy of ICDAS, conventional bitewing radiography, and AI-assisted bitewing analysis against micro-computed tomography (micro-CT) for proximal caries.
- To assess the reliability, inter-method agreement, and diagnostic performance of these methods at enamel and dentin thresholds.
Main Methods:
- Twenty-seven extracted human teeth were imaged using bitewing radiography and micro-CT.
- Two calibrated raters scored images using an ICDAS-analogous scale; AI-assisted scores were generated from bitewing images.
- Diagnostic accuracy (sensitivity, specificity) was calculated and compared at enamel and dentin thresholds.
Main Results:
- Agreement with micro-CT was substantial for ICDAS (κ=0.67) and bitewing (κ=0.70), and moderate for AI (κ=0.53).
- Bitewing radiography was significantly more accurate than AI (p=0.042) in classifying lesion depth.
- ICDAS excelled at dentin threshold detection (100% sensitivity/specificity), while AI and bitewing showed higher sensitivity for enamel lesions but underestimated depth.
Conclusions:
- No single method proved superior across all diagnostic thresholds.
- AI-assisted and conventional bitewing imaging are effective for enamel lesion detection.
- ICDAS remains essential for identifying dentin involvement, supporting a multimodal diagnostic approach.
