Authors
Jay Patel, Mustafa Badi, Ritwik Katiyar, Jie Yang, Marisol Tellez, Lucia Cevidanes
Published in
Oral surgery, oral medicine, oral pathology and oral radiology. Sep 01, 2026. Epub Sep 01, 2026.
Abstract
This study aimed to develop, test, and clinically validate an AI-driven Clinical Decision Support Tool (CDST) to enhance dental radiographic quality, with the goal of reducing retakes and improving diagnosis.
About 188,684 radiographs from 16,703 patients (2020-2024) were manually labeled. Two enhancement strategies were developed: (1) generalized and (2) localized enhancements. For generalized enhancement, two further approaches were developed, including traditional algorithms with the BSRGAN model, and a custom AI pipeline integrating the lighting enhancement-net and enhancement U-net with BSRGAN to remove artifacts, correct exposure, reduce noise, blurriness, and improve sharpness. Localized enhancement involved an ensemble YOLO v8.2 segmentation model followed by LaMa inpainting to remove artifacts. Performance was evaluated on test sets using standard metrics. A panel of dental experts reviewed the original and enhanced radiographs to assess diagnostic quality.
The YOLO model achieved 92% accuracy, while LaMa achieved 96%. Expert evaluation of 1,500 radiographs indicated that 81% were improved, 15% were unchanged, and 3% showed reduced diagnostic quality, with statistically significant differences (P < .05).
This AI-driven, expert-validated CDST enhances global and localized radiographic quality, potentially reducing retakes, lowering radiation exposure, and improving diagnostic reliability.
PMID:
42881064
Bibliographic data and abstract were imported from PubMed on 16 Oct 2026.
Read full publication at:
Please sign in
to see all details.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 4
- Comments 0