Related Experiment Video
Updated: Aug 5, 2026
05:49
Reliability of Artificial Intelligence-Based Cone Beam Computed Tomography Integration with Digital Dental Images
Published on: February 23, 2024
External Validation of Two AI Systems for Dentinal Caries Detection: A Retrospective Pilot Clinical Outcomes Study
Owais A Farooqi1,2, Gilbert A Fru3, Yan Ming Gong4
1Chief, Dental Service, VA Greater Los Angeles Healthcare System, Lead Dentist VISN 22, 11301 Wilshire Boulevard, Los Angeles, CA 90073, USA.
Dento Maxillo Facial Radiology
|July 27, 2026
Summary
This study validated two dental artificial intelligence (AI) systems for detecting dentinal caries. The AI systems showed limited sensitivity, highlighting the need for real-world clinical outcome validation in healthcare settings.
Area of Science:
- Dental diagnostics
- Artificial intelligence in healthcare
- Radiographic interpretation
Background:
- Dental caries detection remains a challenge, impacting treatment decisions.
- Artificial intelligence (AI) systems offer potential for improved diagnostic accuracy.
- External validation of AI tools using real-world clinical data is crucial.
Purpose of the Study:
- To evaluate the sensitivity of two commercial dental AI systems for dentinal caries detection.
- To perform longitudinal, outcomes-anchored validation within an integrated healthcare system.
- To establish a framework for AI performance evaluation based on clinical outcomes.
Main Methods:
- Pilot external validation study involving 90 patients in a federal healthcare system.
- Testing of two commercially available AI systems on full-mouth intraoral radiographic series.
- Ground truth established via baseline clinical exams and 6-12 month follow-up documentation.
Main Results:
- Vendor A AI sensitivity for dentinal caries: 48.5% (95% CI: 37.0-60.2).
- Vendor B AI sensitivity for dentinal caries: 57.5% (95% CI: 46.1-68.2).
- Pooled AI sensitivity: 53.0% (95% CI: 44.8-61.1).
Conclusions:
- Outcomes-anchored, longitudinal validation of dental AI for caries detection is feasible in integrated healthcare systems.
- This approach yields clinically relevant sensitivity estimates based on actual treatment outcomes.
- Findings are most applicable to healthcare systems with standardized data capture; private practice variability may differ.