基于全景X光片的深度学习预测下颌第三磨牙与下牙槽管邻近关系
Deep learning-based prediction of close proximity between the mandibular third molar and the inferior alveolar canal using panoramic radiographs
ResNet18模型以最高AUROC(0.829±0.022)、准确率(0.800±0.027)、敏感度(0.817±0.029)成功预测下颌第三磨牙与下牙槽管邻近关系(CBCT定义根-管距离≤1 mm)。
- Article
- Open access
- Published: 18 November 2026
- MyeongGyun Jang1 na1,
- Hakjin Lee1 na1,
- Jinmo Byeon1,
- Dong Ok Kim1,
- Yeongmin Kim1,
- Inyong Jeong1,
- Ho-Kyung Lim2,
- Bo-Yeon Hwang3,
- In-Seok Song3 &
- …
- Hwamin Lee1
Scientific Reports (2026) Cite this article
We’re sharing this article early to provide faster access to peer-reviewed, accepted research. It is citable and carries a permanent DOI. This version is subject to further edits and will be replaced automatically by the final Version of Record. All legal disclaimers apply.
Abstract
Assessment of the relationship between the mandibular third molar and the inferior alveolar canal is essential for preoperative treatment planning. This retrospective study evaluated whether panoramic radiographic regions of interest (ROIs) contain sufficient information to predict close proximity, defined as a minimum root–canal distance of 1 mm or less on cone-beam computed tomography (CBCT). We analyzed 110 mandibular third molars from 63 patients. CBCT images were used only to establish ground truth and define ROIs, whereas model training and inference used panoramic ROI images exclusively. EfficientNet-B0, DenseNet121, ConvNeXt-Tiny, and ResNet18 were compared using patient-level five-fold stratified group cross-validation, and Gradient-weighted Class Activation Mapping was used for model interpretation. ResNet18 achieved the numerically highest mean area under the receiver operating characteristic curve (AUROC) (0.829 ± 0.022), with an accuracy of 0.800 ± 0.027, sensitivity of 0.817 ± 0.029, and specificity of 0.780 ± 0.045. The model primarily focused on regions corresponding to the third molar root and inferior alveolar canal. These preliminary findings, derived from folds also used for model selection, suggest that panoramic ROIs contain information associated with CBCT-defined close proximity but cannot determine buccolingual canal position due to the two-dimensional nature of panoramic radiography; independent validation is needed.
Acknowledgements
This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (RS-2026-25477630) and the Korea Institute for Advancement of Technology (KIAT) grant funded by the Korean government (MOTIE) (RS-2025-02214034, HRD Program for Industrial Innovation).
Funding
This research was supported by the National Research Foundation of Korea (NRF), funded by the Korean government (MSIT) (RS-2026-25477630 and RS-2026-25477844).
Author information
Author notes
MyeongGyun Jang and Hakjin Lee have contributed equally to this work.
Authors and Affiliations
Department of Biomedical Informatics, Korea University College of Medicine, 161 Jeongneung-ro, Seongbuk-gu, Seoul, 02708, Republic of Korea
MyeongGyun Jang, Hakjin Lee, Jinmo Byeon, Dong Ok Kim, Yeongmin Kim, Inyong Jeong & Hwamin Lee
Department of Oral & Maxillofacial Surgery, Korea University Guro Hospital, Seoul, Republic of Korea
Ho-Kyung Lim
Department of Oral & Maxillofacial Surgery, Korea University Anam Hospital, 73 Goryeodae-ro, Seongbuk-gu, Seoul, 02841, Republic of Korea
Bo-Yeon Hwang & In-Seok Song
Authors
- MyeongGyun Jang
- Hakjin Lee
- Jinmo Byeon
- Dong Ok Kim
- Yeongmin Kim
- Inyong Jeong
- Ho-Kyung Lim
- Bo-Yeon Hwang
- In-Seok Song
- Hwamin Lee
Corresponding authors
Correspondence to In-Seok Song or Hwamin Lee.
Ethics declarations
Competing interests
The authors declare no competing interests.
Ethical approval
This retrospective observational study was approved by the Institutional Review Board of Korea University Anam Hospital (IRB No. 2026AN0357) and was conducted in accordance with the ethical standards of the Declaration of Helsinki and its later amendments.
Consent to participate
The requirement for informed consent was waived by the Institutional Review Board because of the retrospective nature of this study.
Consent for publication
Not applicable.
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Jang, M., Lee, H., Byeon, J. et al. Deep learning-based prediction of close proximity between the mandibular third molar and the inferior alveolar canal using panoramic radiographs. Sci Rep (2026). https://doi.org/10.1038/s41598-026-81280-9
Received: 29 July 2026
Accepted: 13 November 2026
Published: 18 November 2026
DOI: https://doi.org/10.1038/s41598-026-81280-9
Keywords
来源:Nature 健康护理主题 · nature.com