AI-Assisted Detection of Dental Caries Using Radiographs for Operative Dentistry Applications
Dental caries; Radiographs; Operative dentistry; Artificial intelligence; YOLOv8; Object detection
AuthorsAbstractDental caries is still a leading cause of restorative treatment need and radiographs play a pivotal role in the decision-making process for operative dentistry especially when the caries are subtle, anatomically overlapped or proximal. In this study, a preprocessing-integrated detection framework (Caries-YOLOv8-CE) was developed and evaluated, which combined the contrast-limited adaptive histogram equalization with the YOLOv8s-based object detection for radiographic localization of caries. A radiographic dataset of 4,658 images and 6,566 caries annotations was divided into three subsets: training, validation and independent test. The proposed framework is compared with the raw baseline (YOLOv8s) and Faster R-CNN ResNet50-FPN using precision, recall, F1-score, [email protected], [email protected]:0.95, qualitative prediction review, and error analysis. Caries-YOLOv8-CE achieved the highest default precision (0.879) and F1-score (0.841), with competitive [email protected] (0.881) and [email protected]:0.95 (0.526). The box level error analysis yielded 549 true positive, 171 false positive and 97 false negative results. These results suggest that CLAHE enhanced YOLO based detection could be a valuable assistive method for localizing caries in radiographs, but external validation and clinical reader studies are needed before it can be practically implemented.
Received-13-05-2026 Revised-20-06-2026 Accepted-23-06-2026
1. Introduction A caries is a progressive oral health burden that is not detected and managed at an appropriate stage can lead to cavitation, pulpal involvement, pain and restorative intervention, and is caused by a mineral imbalance that occurs as a result of the presence of a biofilm [1]. Caries is a public health and service delivery challenge, as well as a biological problem, and is a high prevalence disease worldwide with populations of people at risk of caries being untreated across all age groups and health systems [2]. A diagnostic problem is particularly significant in operative dentistry, as the choice of treatment (monitor, remineralize, seal, restore or replace) relies on accurate lesion detection, establishment and monitoring of the size and activity of the lesion. Radiographs offer a lot of information which can't be gained from visual examination alone, especially for lesions in the proximal areas, for recurrent caries next to restoration margins, and for lesions that overlap with other structures and may not be accessible clinically. Operative intervention should also be guided by a conservative understanding of caries process rather than by radiographic appearance alone, which can result in unnecessary removal of tooth structure and/or delayed diagnosis of caries allowing lesion progression [3]. The tension translates to a demand for reproducible tools to support radiography that will enhance localization, but not overestimate diagnostic certainty. The increasing relevance of AI to dental image analysis stems from the ability of deep neural networks to learn patterns in images that are correlated with anatomy and disease, from annotated examples [4]. AI systems are •••••••••••••••••••••••••••••••• ejprd.org- Published by Riset Publishing Services LLC.
EJPRD
Copyright © 2026 by Riset Publishing Services LLC