Related Experiment Video
Updated: Jul 11, 2026
10:50
A Finite Element Approach for Locating the Center of Resistance of Maxillary Teeth
Published on: April 8, 2020
9.7K
Development and verification of a convolutional neural network-based model for automatic mandibular canal
Xiao Pan1,2, Chengtao Wang1, Xuhui Luo1
1Department of Dentomaxillofacial Radiology, Nanjing Stomatological Hospital, Affiliated Hospital of Medical School, Institute of Stomatology, Nanjing University, Zhong Yang Road 30, Nanjing, Jiangsu Province, 210008, China.
BMC Oral Health
|August 21, 2025
Summary
A new deep learning model accurately locates the mandibular canal on cone beam CT scans. This convolutional neural network shows strong potential for clinical use in dental imaging.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Radiology
Background:
- Accurate localization of the mandibular canal (MC) is crucial for dental implant surgery.
- Cone beam computed tomography (CBCT) is widely used for dental imaging, but MC localization can be challenging.
- Automating MC localization can improve efficiency and accuracy in clinical practice.
Purpose of the Study:
- To develop and verify a deep learning (DL) model using convolutional neural networks (CNNs) for automated mandibular canal (MC) localization.
- To evaluate the model's performance on multicenter CBCT images from various manufacturers.
Main Methods:
- A CNN model was developed using 836 CBCT scans from one manufacturer (training, validation, internal testing).
- The model was externally validated on 220 CBCT scans from four different manufacturers.
- Performance was quantitatively assessed using Average Symmetric Surface Distance (ASSD) and Symmetric Mean Curve Distance (SMCD), and qualitatively via visual scoring.
Main Results:
- Internal testing showed average ASSD of 0.486 mm and SMCD of 0.298 mm.
- External testing achieved visual scores ≥ 4 points in 86.8% of scans, with average ASSD of 0.438 mm and SMCD of 0.185 mm for high-scoring scans.
- Automatic MC localization took an average of 8.52 seconds, significantly faster than manual methods.
Conclusions:
- The developed CNN model demonstrates high accuracy and generalizability for mandibular canal localization.
- The model shows excellent clinical application potential for automated MC localization on multicenter CBCT data.
- Further research can explore integration into clinical workflows for enhanced dental imaging analysis.
