Artificial intelligence assisted multitask framework for dental detection and root canal filling assessment on periapical radiographs.
6 affiliations
- 1Department of Pediatric Dentistry, Chang Gung Memorial Hospital, Linkou Branch, Taoyuan, Taiwan.
- 2Graduate Institute of Dental and Craniofacial Science, Chang Gung University College of Medicine, Taoyuan, Taiwan.
- 3Department of Computer Science and Information Engineering, National Central University, Taoyuan, Taiwan.
- 4Program on Semiconductor Manufacturing Technology Academy of Innovative Semiconductor and Sustainable Manufacturing, National Cheng Kung University, Tainan, Taiwan.
- 5Department of Electronic Engineering, Feng Chia University, Taichung, Taiwan.
- 6Ateneo Laboratory for Intelligent Visual Environments, Department of Information Systems and Computer Science, Ateneo de Manila University, Quezon, Philippines.
Abstract
Background/purpose
Accurate evaluation of root canal filling (RCF) quality is paramount for successful endodontic outcomes. However, conventional radiographic interpretation is often subject to subjective variability. This study developed an automated, multitask artificial intelligence (AI) framework using YOLO11 and advanced image pre-processing to identify multiple dental structures and provide quantitative assessments of RCF quality on periapical radiographs.Materials and methods
We proposed an image enhancement pipeline, which integrate intensity normalization, white top-hat transform, and linear contrast stretching to refine dental contours for training the YOLO11 multitask model. The framework simultaneously detected primary, permanent, dental crowns, and RCFs. Quantitative metrics, including root canal filling length (RCFL) and filling rate (RCFR), were calculated using pixel-accumulation and landmark localization algorithms.Results
The proposed image enhancement significantly improved model performance, with the total mAP 50 increasing from 92.3% to 99.6%. The model achieved excellent diagnostic reliability, with intraclass correlation coefficients exceeding 0.96 across all tasks, notably reaching 0.997 for primary teeth. In clinical validation, the AI-assisted framework demonstrated high diagnostic agreement with the consensus of four dentists in classifying RCFL as adequate or underfilled. Furthermore, the AI-calculated RCFR for permanent (73.96%) and primary teeth (77.44%) showed high consistency with expert observations, while AI confidence levels consistently exceeded 95%.Conclusion
The integrated YOLO11 framework provides a highly accurate, objective, and automated system for the quantitative evaluation of root canal obturation. By delivering reliable second opinions and precise apical limit localization, this system has strong potential to enhance diagnostic consistency and efficiency in clinical endodontic practice.
Related
- An artificial intelligence assisted framework for automated quantified root canal curvature measurement on periapical radiographs.
- Diagnostic accuracy of artificial intelligence in detecting periapical lesions: A comparative study with cone-beam computed tomography and conventional periapical radiography.
- The accuracy of large language models in assessing tooth detection and identification, dental caries, periapical status, and tooth development from mixed dentition radiographs
- Development and clinical validation of an artificial intelligence-assisted radiographic system for predicting periodontal pocket severity using full-mouth radiographs.
- ROI-based binary caries segmentation in intraoral periapical radiographs using a ResNet-conditioned U-Net with FiLM-modulated decoder
BibTeX
@article{hsu2026artificial,
title = {{Artificial intelligence assisted multitask framework for dental detection and root canal filling assessment on periapical radiographs.}},
author = {Chia-Lin Hsu and Li Wei Ou-Yang and Chih-Hao Chang and Yuan-Jin Lin and Po-Lin Huang and Kai-Syun Chiu and Heng-Chia Chen and Tsung-Yi Chen and Jia-Ching Wang and Wei-Chen Tu and Patricia Angela R Abu},
journal = {Journal of dental sciences},
year = {2026},
month = oct,
doi = {10.1016/j.jds.2025.07.3528},
url = {https://europepmc.org/article/MED/42924174},
note = {CC BY-NC-ND},
}
RIS
TY - JOUR
TI - Artificial intelligence assisted multitask framework for dental detection and root canal filling assessment on periapical radiographs.
AU - Chia-Lin Hsu
AU - Li Wei Ou-Yang
AU - Chih-Hao Chang
AU - Yuan-Jin Lin
AU - Po-Lin Huang
AU - Kai-Syun Chiu
AU - Heng-Chia Chen
AU - Tsung-Yi Chen
AU - Jia-Ching Wang
AU - Wei-Chen Tu
AU - Patricia Angela R Abu
PY - 2026
DA - 2026/10/01
JO - Journal of dental sciences
DO - 10.1016/j.jds.2025.07.3528
UR - https://europepmc.org/article/MED/42924174
AB - Background/purpose Accurate evaluation of root canal filling (RCF) quality is paramount for successful endodontic outcomes. However, conventional radiographic interpretation is often subject to subjective variability. This study developed an automated, multitask artificial intelligence (AI) framework using YOLO11 and advanced image pre-processing to identify multiple dental structures and provide quantitative assessments of RCF quality on periapical radiographs.Materials and methods We proposed an image enhancement pipeline, which integrate intensity normalization, white top-hat transform, and linear contrast stretching to refine dental contours for training the YOLO11 multitask model. The framework simultaneously detected primary, permanent, dental crowns, and RCFs. Quantitative metrics, including root canal filling length (RCFL) and filling rate (RCFR), were calculated using pixel-accumulation and landmark localization algorithms.Results The proposed image enhancement significantly improved model performance, with the total mAP 50 increasing from 92.3% to 99.6%. The model achieved excellent diagnostic reliability, with intraclass correlation coefficients exceeding 0.96 across all tasks, notably reaching 0.997 for primary teeth. In clinical validation, the AI-assisted framework demonstrated high diagnostic agreement with the consensus of four dentists in classifying RCFL as adequate or underfilled. Furthermore, the AI-calculated RCFR for permanent (73.96%) and primary teeth (77.44%) showed high consistency with expert observations, while AI confidence levels consistently exceeded 95%.Conclusion The integrated YOLO11 framework provides a highly accurate, objective, and automated system for the quantitative evaluation of root canal obturation. By delivering reliable second opinions and precise apical limit localization, this system has strong potential to enhance diagnostic consistency and efficiency in clinical endodontic practice.
SN - 1991-7902
SN - 2213-8862
ER -