모바일 메뉴

KJO Korean Journal of Orthodontics

Open Access

pISSN 2234-7518
eISSN 2005-372X
QR Code QR Code

퀵메뉴 버튼

Article

home All Articles View
Split Viewer

Article In Press

Korean J Orthod

Published online July 5, 2021

Copyright © The Korean Association of Orthodontists.

One-step Automated Orthodontic Diagnosis Model using a Convolutional Neural Network and Lateral Cephalograms from 10 Nationwide Multi-Centers

Sunjin Yim a, Sungchul Kim b, Inhwan Kim c, Jae-Woo Park d, Jin-Hyoung Cho e, Mihee Hong f, Kyung-Hwa Kang g, Minji Kim h, Su-Jung Kim i, Yoon-Ji Kim j, Young Ho Kim k, Sung-Hoon Lim l, Sang Jin Sung m, Namkug Kim n, Seung-Hak Baek o

a Graduate student (PhD), Department of Orthodontics, School of Dentistry, Seoul National University, Seoul, Republic of Korea
b Graduate student (MSc), Department of Convergence Medicine, Asan Medical Institute of Convergence Science and Technology, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea
c Graduate student (PhDc), Department of Convergence Medicine, Asan Medical Institute of Convergence Science and Technology, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea
d Private practice, Department of Orthodontics, Kooalldam Dental Hospital, Incheon, Republic of Korea
e Professor and Chair, Department of Orthodontics, Chonnam National University School of Dentistry, Gwangju, Republic of Korea
f Assistant Professor, Department of Orthodontics, School of Dentistry, Kyungpook National University, Daegu, Republic of Korea
g Professor, Department of Orthodontics, School of Dentistry, Wonkwang University, Iksan-si, Republic of Korea
h Associate Professor, Department of Orthodontics, College of Medicine, Ewha Womans University, Seoul, Republic of Korea
i Professor, Department of Orthodontics, Kyung Hee University School of Dentistry, Seoul, Republic of Korea.
j Assistant Professor, Department of Orthodontics, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea
k Professor, Department of Orthodontics, Institute of Oral Health Science, Ajou University School of Medicine, Suwon-si, Gyeonggi-do, Republic of Korea
l Professor, Department of Orthodontics, College of Dentistry, Chosun University, Gwangju, Republic of Korea,
m Professor, Division of Orthodontics, University of Ulsan Asan Medical Center, Seoul, Republic of Korea
n Professor, Department of Convergence Medicine, University of Ulsan, College of Medicine, Asan Medical Center, Seoul, Republic of Korea
o Professor, Department of Orthodontics, School of Dentistry, Dental Research Institute, Seoul National University, Seoul, Republic of Korea

Correspondence to:Seung-Hak Baek, Department of Orthodontics, School of Dentistry, Dental Research Institute, Seoul National University, Daehak-no #101, Jongno-gu, Seoul, 03080, South Korea, Tel:+82-2- 2072-3952, Fax:+82-2-3672-2678, e-mail: drwhite@unitel.co.kr; Orcid number: 0000-0002-6586- 9503
Namkug Kim, Department of Convergence Medicine, Department of Convergence Medicine, University of Ulsan, College of Medicine, Asan Medical Center, 88, Olympic-ro 43-gil, Songpa-gu, Seoul, 05505, South Korea, Tel:+82-2-3010-6573; Fax:+82-2-3010-6196; e-mail: namkugkim@gmail.com; Orcid number: 0000-0002-3438-2217

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Objective: The purpose of this study was to investigate the accuracy of one-step automated orthodontic diagnosis of the skeletal and dental discrepancies using a convolutional neural network (CNN) and lateral cephalogram images from 10 multi-centers.
Methods: Among retrospectively collected 2,174 lateral cephalograms, 1,993 cephalograms from two hospitals were used for training set and internal test set and 181 cephalograms from eight other hospitals were used for external test set. They were divided into three classification groups according to the anteroposterior skeletal discrepancy (Class I, II, and III), vertical skeletal discrepancy (normo-, hypo-, and hyper-divergent) and vertical dental discrepancy (normal overbite, deep bite, and open bite) as a gold standard. Pre-trained DenseNet-169 were used as CNN Classifier model. Diagnosis performance was evaluated by receiver operating characteristic (ROC) analysis, t-stochastic neighbor embedding (t-SNE), and gradient-weighted class activation mapping (Grad-CAM).
Results: In the ROC analysis, the mean area under curve and the mean accuracy of all classification were high with both internal and external test sets (all higher than 0.89; higher than 0.80, respectively). In the t-SNE analysis, our model succeeded in creating a well-separation between the three classification groups in each diagnosis. Grad-CAM figures showed different patterns and sizes of the focus areas according to the three classification groups in each diagnosis.
Conclusion: Since the accuracy of our model was validated with both internal and external test sets, it shows a possibility of one-step automated orthodontic diagnostic tool. However, it still needs technical improvement in classification of vertical dental discrepancy.

Keywords: one-step automated orthodontic diagnosis, skeletal and dental discrepancy, convolutional neural networks, lateral cephalogram, multi-center study.

Article

ahead

Korean J Orthod

Published online July 5, 2021

Copyright © The Korean Association of Orthodontists.

One-step Automated Orthodontic Diagnosis Model using a Convolutional Neural Network and Lateral Cephalograms from 10 Nationwide Multi-Centers

Sunjin Yim a, Sungchul Kim b, Inhwan Kim c, Jae-Woo Park d, Jin-Hyoung Cho e, Mihee Hong f, Kyung-Hwa Kang g, Minji Kim h, Su-Jung Kim i, Yoon-Ji Kim j, Young Ho Kim k, Sung-Hoon Lim l, Sang Jin Sung m, Namkug Kim n, Seung-Hak Baek o

a Graduate student (PhD), Department of Orthodontics, School of Dentistry, Seoul National University, Seoul, Republic of Korea
b Graduate student (MSc), Department of Convergence Medicine, Asan Medical Institute of Convergence Science and Technology, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea
c Graduate student (PhDc), Department of Convergence Medicine, Asan Medical Institute of Convergence Science and Technology, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea
d Private practice, Department of Orthodontics, Kooalldam Dental Hospital, Incheon, Republic of Korea
e Professor and Chair, Department of Orthodontics, Chonnam National University School of Dentistry, Gwangju, Republic of Korea
f Assistant Professor, Department of Orthodontics, School of Dentistry, Kyungpook National University, Daegu, Republic of Korea
g Professor, Department of Orthodontics, School of Dentistry, Wonkwang University, Iksan-si, Republic of Korea
h Associate Professor, Department of Orthodontics, College of Medicine, Ewha Womans University, Seoul, Republic of Korea
i Professor, Department of Orthodontics, Kyung Hee University School of Dentistry, Seoul, Republic of Korea.
j Assistant Professor, Department of Orthodontics, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea
k Professor, Department of Orthodontics, Institute of Oral Health Science, Ajou University School of Medicine, Suwon-si, Gyeonggi-do, Republic of Korea
l Professor, Department of Orthodontics, College of Dentistry, Chosun University, Gwangju, Republic of Korea,
m Professor, Division of Orthodontics, University of Ulsan Asan Medical Center, Seoul, Republic of Korea
n Professor, Department of Convergence Medicine, University of Ulsan, College of Medicine, Asan Medical Center, Seoul, Republic of Korea
o Professor, Department of Orthodontics, School of Dentistry, Dental Research Institute, Seoul National University, Seoul, Republic of Korea

Correspondence to:Seung-Hak Baek, Department of Orthodontics, School of Dentistry, Dental Research Institute, Seoul National University, Daehak-no #101, Jongno-gu, Seoul, 03080, South Korea, Tel:+82-2- 2072-3952, Fax:+82-2-3672-2678, e-mail: drwhite@unitel.co.kr; Orcid number: 0000-0002-6586- 9503
Namkug Kim, Department of Convergence Medicine, Department of Convergence Medicine, University of Ulsan, College of Medicine, Asan Medical Center, 88, Olympic-ro 43-gil, Songpa-gu, Seoul, 05505, South Korea, Tel:+82-2-3010-6573; Fax:+82-2-3010-6196; e-mail: namkugkim@gmail.com; Orcid number: 0000-0002-3438-2217

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Objective: The purpose of this study was to investigate the accuracy of one-step automated orthodontic diagnosis of the skeletal and dental discrepancies using a convolutional neural network (CNN) and lateral cephalogram images from 10 multi-centers.
Methods: Among retrospectively collected 2,174 lateral cephalograms, 1,993 cephalograms from two hospitals were used for training set and internal test set and 181 cephalograms from eight other hospitals were used for external test set. They were divided into three classification groups according to the anteroposterior skeletal discrepancy (Class I, II, and III), vertical skeletal discrepancy (normo-, hypo-, and hyper-divergent) and vertical dental discrepancy (normal overbite, deep bite, and open bite) as a gold standard. Pre-trained DenseNet-169 were used as CNN Classifier model. Diagnosis performance was evaluated by receiver operating characteristic (ROC) analysis, t-stochastic neighbor embedding (t-SNE), and gradient-weighted class activation mapping (Grad-CAM).
Results: In the ROC analysis, the mean area under curve and the mean accuracy of all classification were high with both internal and external test sets (all higher than 0.89; higher than 0.80, respectively). In the t-SNE analysis, our model succeeded in creating a well-separation between the three classification groups in each diagnosis. Grad-CAM figures showed different patterns and sizes of the focus areas according to the three classification groups in each diagnosis.
Conclusion: Since the accuracy of our model was validated with both internal and external test sets, it shows a possibility of one-step automated orthodontic diagnostic tool. However, it still needs technical improvement in classification of vertical dental discrepancy.

Keywords: one-step automated orthodontic diagnosis, skeletal and dental discrepancy, convolutional neural networks, lateral cephalogram, multi-center study.