An artificial intelligence assisted framework for automated quantified root canal curvature measurement on periapical radiographs.
Authors
Affiliations (9)
Affiliations (9)
- Department of Family Dentistry, Taoyuan Chang Gung Memorial Hospital, Taoyuan, Taiwan.
- Department of Electronic Engineering, Chung Yuan Christian University, Taoyuan, Taiwan.
- Research Center for Semiconductor Materials and Advanced Optics, Chung Yuan Christian University, Taoyuan, Taiwan.
- Program on Semiconductor Manufacturing Technology, Academy of Innovative Semiconductor and Sustainable Manufacturing, National Cheng Kung University, Tainan, Taiwan.
- Department of Electrical Engineering, Ming Chi University of Technology, New Taipei, Taiwan.
- Department of Electronic Engineering, Feng Chia University, Taichung, Taiwan.
- Department of Information Management, Chung Yuan Christian University, Taoyuan, Taiwan.
- Department of Electrical Engineering, National Cheng Kung University, Tainan, Taiwan.
- Ateneo Laboratory for Intelligent Visual Environments, Department of Information Systems and Computer Science, Ateneo de Manila University, Quezon, Philippines.
Abstract
Root canal curvature is a complex anatomical variation that complicates endodontic treatments. Traditionally, dentists rely on two-dimensional periapical radiographs (PA) and manual methods to estimate curvature angles. However, this conventional approach is highly time-consuming and subjective, often leading to inconsistent measurements. This study aims to develop a fully automated, objective framework for precise angular measurement and severity grading of root canal curvature. The YOLO26n-seg model was employed to segment roots and classify single- or multi-rooted teeth from PA. Furthermore, we proposed an image processing algorithm that standardized orientation, extracted centerlines, and applied three-point circle fitting to establish tangential lines for curvature measurement. The framework's performance was validated against a ground truth determined by five experienced dentists. YOLO26n-seg achieved outstanding segmentation precision (98.8%), sensitivity (97.3%), and intersection over union (96.2%). For angular measurement, the framework demonstrated high accuracy with errors of merely 0.7-1.6 degrees compared to manual assessments, yielding a root mean square error of 1.2° and R<sup>2</sup> up to 98.9%. Moreover, the automated system reduced average processing time per PA from 39.9 to 70.8 s to just 0.6-1.7 s, an approximately 44-fold increase in efficiency. The proposed automated framework for root canal curvature measurement provides accurate and consistent geometric data, offering robust support for clinical diagnostic needs. By enhancing evaluation efficiency and reducing the diagnostic burden on clinicians, this system serves as a valuable auxiliary tool that further advances the standardization of dental imaging analysis.