Related Experiment Video
Updated: Aug 6, 2026
05:49
Reliability of Artificial Intelligence-Based Cone Beam Computed Tomography Integration with Digital Dental Images
Published on: February 23, 2024
PanMB2-Net: a deep learning framework for preoperative MB2 canal risk assessment on panoramic radiographs
Xuzhu Wang1, Jiajun Xiao2, Zhan Shi1
1DUT School of Software Technology, the DUT-RU International School of Information Science Engineering, Dalian University of Technology, Dalian, Liaoning, China.
BMC Oral Health
|July 17, 2026
Summary
This study developed an AI tool, PanMB2-Net, to assess the risk of the second mesiobuccal (MB2) canal in maxillary molars using panoramic radiographs. The AI tool aids in preoperative decision-making, highlighting potential MB2 canals without extra radiation.
Area of Science:
- Dentistry
- Artificial Intelligence
- Radiology
Background:
- The second mesiobuccal (MB2) canal in maxillary first molars is often missed on standard radiographs.
- Accurate preoperative assessment of MB2 canals is crucial for successful endodontic treatment.
- Existing methods may involve additional radiation exposure or costs.
Purpose of the Study:
- To develop and validate an artificial intelligence (AI) framework for preoperative risk assessment of the MB2 canal.
- To utilize panoramic radiographs for MB2 canal identification, reducing radiation exposure.
- To create a clinically applicable tool to aid endodontic treatment planning.
Main Methods:
- Retrospective collection of 388 panoramic radiographs.
- Utilized YOLOv5 for maxillary molar localization and region of interest cropping.
- Developed the Panoramic-based MB2 Prediction Network (PanMB2-Net) using ResNet50 architecture with attention mechanisms and specialized loss functions.
- Validated the model using accuracy, precision, recall, and F1-score, alongside ablation studies.
Main Results:
- PanMB2-Net achieved 72.9% validation accuracy, significantly outperforming the statistical baseline.
- MB2 prediction yielded precision, recall, and F1-scores of 0.687, 0.673, and 0.679, respectively.
- Ablation studies confirmed the effectiveness of individual components like attention mechanisms and regularization losses.
Conclusions:
- PanMB2-Net is a feasible AI tool for MB2 canal risk assessment on panoramic radiographs.
- This AI tool can support preoperative decision-making by identifying teeth needing further attention, without increasing radiation or cost.
- PanMB2-Net complements clinician judgment and Cone Beam Computed Tomography (CBCT), rather than replacing them.
