Related Experiment Video
Updated: Apr 11, 2026
09:10
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
2.4K
Deep learning-based segmentation of caries, implants, fixed prosthesis, and restorations on bitewing radiographs: A
Amisha Parekh1, Rohan Jagtap2, Yalamanchili Samata3
1Department of Biomedical Materials Science, School of Dentistry, University of Mississippi Medical Center, Jackson, MS, USA.
Science Progress
|April 10, 2026
Summary
An AI system shows diagnostic accuracy comparable to radiologists for detecting dental restorations, prostheses, and implants on bitewing radiographs. While effective for most findings, its performance for caries detection was slightly lower but still promising.
Area of Science:
- Dentistry
- Radiology
- Artificial Intelligence
Background:
- Bitewing radiographs are crucial for detecting dental pathologies.
- Accurate interpretation of dental radiographs is essential for diagnosis and treatment planning.
- AI offers potential for automating and improving the efficiency of radiographic analysis.
Purpose of the Study:
- To evaluate the diagnostic accuracy of an AI system for detecting teeth, caries, implants, restorations, and fixed prostheses on bitewing radiographs.
- To compare the AI system's performance against human expert interpretation.
Main Methods:
- A retrospective study analyzed 407 bitewing radiographs from 315 adult patients.
- An AI system (VELMENI Inc.) was used, with results compared to annotations by two oral and maxillofacial radiologists.
- Inter-rater reliability (Cohen's kappa) and diagnostic accuracy (sensitivity, specificity) were assessed.
Main Results:
- The AI system demonstrated substantial to near-perfect agreement with human experts for restorations (κ=0.812-0.871) and prostheses (κ=0.882-0.940) detection.
- Moderate agreement was observed for caries (κ=0.454-0.508), while high agreement was found for implants (κ=0.763-0.974).
- High sensitivity and specificity were achieved for implants, prostheses, and restorations; slightly lower specificity was noted for caries detection.
Conclusions:
- The AI system exhibits diagnostic performance comparable to radiologists for detecting multiple dental findings on bitewing radiographs.
- The AI system shows potential as a clinical tool to enhance efficiency and consistency in dental imaging interpretation.
- Further validation may be needed, particularly for caries detection.
