Related Experiment Video
Updated: Nov 22, 2025
09:10
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
2.0K
Deep-learning for predicting C-shaped canals in mandibular second molars on panoramic radiographs
Su-Jin Jeon1, Jong-Pil Yun2, Han-Gyeol Yeom3
1Department of Conservative Dentistry, Wonkwang University Daejeon Dental Hospital, Daejeon, South Korea.
Dento Maxillo Facial Radiology
|January 6, 2021
Summary
A deep-learning system accurately predicts C-shaped canals in mandibular second molars using panoramic radiographs. This convolutional neural network (CNN) analysis shows high diagnostic performance, aiding dental diagnostics.
Area of Science:
- Dentistry
- Radiology
- Artificial Intelligence in Medicine
Background:
- C-shaped canals in mandibular second molars present diagnostic challenges.
- Accurate identification is crucial for successful endodontic treatment.
Purpose of the Study:
- To evaluate a convolutional neural network (CNN) system for predicting C-shaped canals.
- To assess the diagnostic performance of deep learning on panoramic radiographs.
Main Methods:
- A CNN model (Xception) was trained on 2040 mandibular second molars (887 C-shaped).
- Cone beam CT (CBCT) served as the gold standard for diagnosis.
- Diagnostic metrics (accuracy, sensitivity, specificity, precision) and ROC curves were analyzed. Grad-CAM visualized model attention.
Main Results:
- The CNN model achieved 95.1% accuracy, 92.7% sensitivity, 97.0% specificity, and 95.9% precision.
- Grad-CAM indicated the model focused on root canal convergence and furcation anatomy.
Conclusions:
- Deep learning demonstrates significant accuracy in predicting C-shaped canals from panoramic radiographs.
- This AI approach offers a promising tool for dental diagnostics.
Keywords:
C-shaped canalConvolutional neural networkDeep learningDiagnostic imagingPanoramic radiograph
