Related Experiment Video
Updated: Jun 25, 2025
Author Spotlight: 3D Movement Assessment of Maxillary Posterior Teeth in Clear Aligner Treatment
Published on: February 23, 2024
AI model to detect contact relationship between maxillary sinus and posterior teeth
Wanghui Ding1, Yindi Jiang2, Gaozhi Pang3
1Stomatology Hospital, School of Stomatology, Zhejiang University School of Medicine, Zhejiang Provincial Clinical Research Center for Oral Diseases, Key Laboratory of Oral Biomedical Research of Zhejiang Province, Cancer Center of Zhejiang University, Hangzhou, China.
A novel deep learning network, MSF-MPTnet, accurately assesses the relationship between the maxillary sinus floor (MSF) and posterior teeth (MPT) using panoramic radiographs. This AI model outperforms dentists and radiologists in detecting these anatomical relationships.
Area of Science:
- Dentistry
- Radiology
- Artificial Intelligence
- Medical Imaging
Background:
- The relationship between the maxillary sinus floor (MSF) and maxillary posterior teeth (MPT) is crucial for dental implant planning.
- Accurate assessment traditionally relies on complex imaging like Cone-beam computed tomography (CBCT), necessitating advanced analysis.
Purpose of the Study:
- To develop and validate a deep learning network, MSF-MPTnet, for automated assessment of the MSF-MPT relationship using panoramic radiographs (PRs).
- To compare the diagnostic accuracy of MSF-MPTnet against human experts (dentists and radiologists).
Main Methods:
- A dataset of 1035 PRs and 1035 CBCT images was collected.
- Relationships were classified into non-contact (Class I) and contact (Class II) groups based on CBCT.
- A subset of 350 PRs was used to test the MSF-MPTnet model, dentists, and radiologists.
Main Results:
- MSF-MPTnet demonstrated reliable performance with sensitivity (0.682-0.852) and accuracy (0.890-0.951).
- The AI model achieved higher accuracy in detecting Class I relationships compared to dentists and radiologists across various tooth positions (p < 0.05).
- Overall accuracy for MSF-MPTnet ranged from 79.7% to 90.3%, surpassing human experts.
Conclusions:
- MSF-MPTnet significantly enhances the accuracy of detecting the MSF-MPT relationship from PRs.
- The AI model minimizes false positive contact identifications and can potentially reduce the need for CBCT scans.
- This deep learning approach offers a reliable and efficient tool for pre-surgical assessment in dentistry.
