Imaging Science in Dentistry

Korean Academy of Oral and Maxillofacial Radiology (대한영상치의학회)
권호 표지
DOI QR코드

DOI QR Code

Deep learning-based automatic segmentation of the mandibular canal on panoramic radiographs: A multi-device study

  • Moe Thu Zar Aung (Department of Oral and Maxillofacial Radiology, School of Dentistry and Dental Research Institute, Seoul National University) ;
  • Sang-Heon Lim (Interdisciplinary Program in Bioengineering, Graduate School of Engineering, Seoul National University) ;
  • Jiyong Han (Interdisciplinary Program in Bioengineering, Graduate School of Engineering, Seoul National University) ;
  • Su Yang (Department of Applied Bioengineering, Graduate School of Convergence Science and Technology, Seoul National University) ;
  • Ju-Hee Kang (Department of Oral and Maxillofacial Radiology, Seoul National University Dental Hospital) ;
  • Jo-Eun Kim (Department of Oral and Maxillofacial Radiology, School of Dentistry and Dental Research Institute, Seoul National University) ;
  • Kyung-Hoe Huh (Department of Oral and Maxillofacial Radiology, School of Dentistry and Dental Research Institute, Seoul National University) ;
  • Won-Jin Yi (Department of Oral and Maxillofacial Radiology, School of Dentistry and Dental Research Institute, Seoul National University) ;
  • Min-Suk Heo (Department of Oral and Maxillofacial Radiology, School of Dentistry and Dental Research Institute, Seoul National University) ;
  • Sam-Sun Lee (Department of Oral and Maxillofacial Radiology, School of Dentistry and Dental Research Institute, Seoul National University)
  • Received : 2023.11.13
  • Accepted : 2024.01.10
  • Published : 2024.03.31
Download PDF
⟨ Previous Next ⟩

Abstract

Purpose: The objective of this study was to propose a deep-learning model for the detection of the mandibular canal on dental panoramic radiographs. Materials and Methods: A total of 2,100 panoramic radiographs (PANs) were collected from 3 different machines: RAYSCAN Alpha (n=700, PAN A), OP-100 (n=700, PAN B), and CS8100 (n=700, PAN C). Initially, an oral and maxillofacial radiologist coarsely annotated the mandibular canals. For deep learning analysis, convolutional neural networks (CNNs) utilizing U-Net architecture were employed for automated canal segmentation. Seven independent networks were trained using training sets representing all possible combinations of the 3 groups. These networks were then assessed using a hold-out test dataset. Results: Among the 7 networks evaluated, the network trained with all 3 available groups achieved an average precision of 90.6%, a recall of 87.4%, and a Dice similarity coefficient (DSC) of 88.9%. The 3 networks trained using each of the 3 possible 2-group combinations also demonstrated reliable performance for mandibular canal segmentation, as follows: 1) PAN A and B exhibited a mean DSC of 87.9%, 2) PAN A and C displayed a mean DSC of 87.8%, and 3) PAN B and C demonstrated a mean DSC of 88.4%. Conclusion: This multi-device study indicated that the examined CNN-based deep learning approach can achieve excellent canal segmentation performance, with a DSC exceeding 88%. Furthermore, the study highlighted the importance of considering the characteristics of panoramic radiographs when developing a robust deep-learning network, rather than depending solely on the size of the dataset.

Keywords

Acknowledgement

This work was supported by the Korea Medical Device Development Fund grant, funded by the Korean government (the Ministry of Science and ICT; the Ministry of Trade, Industry and Energy; the Ministry of Health & Welfare; and the Ministry of Food and Drug Safety) (Project Number: 1711194231, RS-2023-KD000011; Project Number: 1711174552, KMDF_PR_20200901_0147; and Project Number: 1711196792, RS-2023-00253380). This work was also supported by the National Research Foundation of Korea (NRF) grant, funded by the Korean government (MSIT) (No. 2023R1A2C200532611).

References

  1. Iwanaga J, Ibaragi S, Takeshita Y, Asaumi J, Horner K, Gest TR, et al. Mandibular canal versus inferior alveolar canal: a Delphi study. Clin Anat 2021; 34: 1095-100. https://doi.org/10.1002/ca.23740
  2. Agbaje JO, Van de casteele E, Hiel M, Verbaanderd C, Lambrichts I, Politis C. Neuropathy of trigeminal nerve branches after oral and maxillofacial treatment. J Maxillofac Oral Surg 2016; 15: 321-7. https://doi.org/10.1007/s12663-015-0843-9
  3. Kushnerev E, Yates JM. Evidence-based outcomes following inferior alveolar and lingual nerve injury and repair: a systematic review. J Oral Rehabil 2015; 42: 768-802. https://doi.org/10.1111/joor.12313
  4. Ai CJ, Jabar NA, Lan TH, Ramli R. Mandibular canal enlargement: clinical and radiological characteristics. J Clin Imaging Sci 2017; 7: 28.
  5. Jung YH, Cho BH. Radiographic evaluation of the course and visibility of the mandibular canal. Imaging Sci Dent 2014; 44: 273-8. https://doi.org/10.5624/isd.2014.44.4.273
  6. Shah N, Bansal N, Logani A. Recent advances in imaging technologies in dentistry. World J Radiol 2014; 6: 794-807. https://doi.org/10.4329/wjr.v6.i10.794
  7. Mallya S, Lam E. White and Pharoah's oral radiology: principles and interpretation. 8th ed. St. Louis: Elsevier, 2018.
  8. Pucilo M, Pucilo A, Safranow K, Nowicka A. The influence of age, sex, and tooth type on the anatomical relationship between tooth roots and the mandibular canal. Imaging Sci Dent 2021; 51: 373-82. https://doi.org/10.5624/isd.20210031
  9. Yang S, Li A, Li P, Yun Z, Lin G, Cheng J, et al. Automatic segmentation of inferior alveolar canal with ambiguity classification in panoramic images using deep learning. Heliyon 2023; 9: e13694.
  10. Ariji Y, Mori M, Fukuda M, Katsumata A, Ariji E. Automatic visualization of the mandibular canal in relation to an impacted mandibular third molar on panoramic radiographs using deep learning segmentation and transfer learning techniques. Oral Surg Oral Med Oral Pathol Oral Radiol 2022; 134: 749-57. https://doi.org/10.1016/j.oooo.2022.05.014
  11. Cha JY, Yoon HI, Yeo IS, Huh KH, Han JS. Panoptic segmentation on panoramic radiographs: deep learning-based segmentation of various structures including maxillary sinus and mandibular canal. J Clin Med 2021; 10: 2577.
  12. Hadilou M, Gholami L, Ghojazadeh M, Emadi N. Prevalence and extension of the anterior loop of the mental nerve in different populations and CBCT imaging settings: a systematic review and meta-analysis. Imaging Sci Dent 2022; 52: 141-53. https://doi.org/10.5624/isd.20220001
  13. Grovik E, Yi D, Iv M, Tong E, Nilsen LB, Latysheva A, et al. Handling missing MRI sequences in deep learning segmentation of brain metastases: a multicenter study. NPJ Digit Med 2021; 4: 33.
  14. Kiljunen T, Akram S, Niemela J, Loyttyniemi E, Seppala J, Heikkila J, et al. A deep learning-based automated ct segmentation of prostate cancer anatomy for radiation therapy planning - a retrospective multicenter study. Diagnostics (Basel) 2020; 10: 959.
  15. Vesal S, Gayo I, Bhattacharya I, Natarajan S, Marks LS, Barratt DC, et al. Domain generalization for prostate segmentation in transrectal ultrasound images: a multi-center study. Med Image Anal 2022; 82: 102620.
  16. Fedorov A, Beichel R, Kalpathy cramer J, Finet J, Fillion robin JC, Pujol S, et al. 3D Slicer as an image computing platform for the Quantitative Imaging Network. Magn Reson Imaging 2012; 30: 1323-41. https://doi.org/10.1016/j.mri.2012.05.001
  17. Soleimanpour N, Mohammadi M. Probabilistic load flow by using nonparametric density estimators. IEEE Trans Power Syst 2013; 28: 3747-55. https://doi.org/10.1109/TPWRS.2013.2258409
  18. Song Z, Zou S, Zhou W, Huang Y, Shao L, Yuan J, et al. Clinically applicable histopathological diagnosis system for gastric cancer detection using deep learning. Nat Commun 2020; 11: 4294.
  19. Targ S, Almeida D, Lyman K. Resnet in resnet: generalizing residual architectures [Internet]. arXiv 2016; 1603.08029 [cited 2023 Nov 10]. Available from https://arxiv.org/abs/1603.08029.
  20. Tan M, Le QV. Efficientnet: rethinking model scaling for convolutional neural networks [Internet]. arXiv 2020; 1905.11946v5 [cited 2023 Nov 10]. Available from https://arxiv.org/abs/1905.11946.
  21. Hu J, Shen L, Albaqnie S, Sun G, Wu E. Squeeze-and-excitation networks [Internet]. arXiv 2019; 1709.01507v4 [cited 2023 Nov 10]. Available from https://arxiv.org/abs/1709.01507.
  22. Lin TY, Goyal P, Girshick R, He K, Dollar P. Focal loss for dense object detection. IEEE Trans Pattern Anal Mach Intell 2020; 42: 318-27. https://doi.org/10.1109/TPAMI.2018.2858826
  23. Jeoun BS, Yang S, Lee SJ, Kim TI, Kim JM, Kim JE, et al. Canal-Net for automatic and robust 3D segmentation of mandibular canals in CBCT images using a continuity-aware contextual network. Sci Rep 2022; 12: 13460.