Related Experiment Video
Updated: Jul 22, 2026
09:10
Precision of In Vivo Quantitative Tooth Wear Measurement Using Intra-Oral Scans
Published on: July 12, 2022
3.4K
Longitudinal assessment of an AI-based software for interproximal caries detection in bitewing radiographs
Nicole Rodrigues1, Monica Bonfanti-Gris1, Shizhu Bai2
1Department of Conservative Dentistry and Bucofacial Prostheses, Faculty of Dentistry, Complutense University of Madrid. Plaza Ramón y Cajal, S/N. 28040, Madrid, Spain.
Journal of Dentistry
|November 21, 2025
Summary
This study tracked AI dental diagnostic software over time. While newer versions improved accuracy in identifying decay, sensitivity decreased, highlighting the need for ongoing monitoring of AI tool performance in clinical practice.
Area of Science:
- Artificial Intelligence in Dentistry
- Radiographic Interpretation
- Diagnostic Accuracy
Background:
- AI tools are increasingly used for dental diagnostics, particularly in interpreting radiographs and detecting pathologies.
- Longitudinal studies are essential to understand the performance evolution of these AI technologies over time.
- Ensuring the clinical reliability and effective use of AI in dentistry requires understanding its developmental trajectory.
Purpose of the Study:
- To longitudinally assess the performance changes of an AI-based dental diagnostic software.
- To evaluate the impact of software updates on the accuracy of detecting carious lesions from digital bitewing radiographs.
- To understand the implications of AI tool evolution for clinical reliability and integration.
Main Methods:
- A longitudinal observational study reanalyzed a 2021 dataset using a 2025 version of AI software.
- 300 digital bitewing radiographs were evaluated, with clinical validation as the reference standard.
- Diagnostic accuracy metrics including Sensitivity, Specificity, F1-score, and Area Under the ROC Curve were calculated using modified ICDAS criteria and McNemar's test.
Main Results:
- The latest AI software version demonstrated improved specificity (93%) and AUC (0.794) but lower sensitivity (80%) compared to the previous version.
- There was a decrease in error-free surfaces (65.7%) and an increase in bounding box and omission errors.
- McNemar's test indicated a statistically significant difference (p < 0.001) in paired classifications between the AI software versions.
Conclusions:
- Longitudinal assessment shows AI software shifting towards higher specificity and reduced overdiagnosis, but with decreased sensitivity and more false negatives.
- Clinicians must be aware that AI diagnostic tools evolve with data exposure, impacting their performance over time.
- Continuous longitudinal performance monitoring is crucial for the safe and effective integration of AI tools in real-world dental settings.
