File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
Supplementary

Article: Artificial intelligence-assisted identification and assessment of mandibular asymmetry on panoramic radiography

TitleArtificial intelligence-assisted identification and assessment of mandibular asymmetry on panoramic radiography
Authors
Issue Date5-Mar-2025
PublisherElsevier
Citation
American Journal of Orthodontics and Dentofacial Orthopedics, 2025 How to Cite?
Abstract

Introduction

Mandibular symmetry is crucial in orthodontic diagnosis and treatment planning. This study aimed to establish an artificial intelligence (AI) method to automatically and accurately identify mandibular landmarks and assess asymmetry via orthopantomography (OPG) radiographs.

Methods

A total of 1038 OPG radiographs (451 mixed and 587 permanent dentitions) were collected and annotated to develop the AI model for identifying mandibular landmarks. First, the mesiodistal widths of the bilateral mandibular first molars were compared to categorize images as horizontally distorted or nondistorted. Next, the efficacy and robustness of the model were assessed through landmark identification, measurement, and asymmetry diagnostics accuracy using successful detection rates and interclass correlation coefficient.

Results

The AI model achieved an average landmark detection error of 0.86 ± 0.95 mm, with 0.97 ± 0.99 mm for bony landmarks and 0.54 ± 0.84 mm for dental landmarks. The successful detection rates at 1, 2, and 3 mm were 75.33%, 93.11%, and 96.72%, respectively. The accuracy exhibits region-specific variations: vertical errors were larger in condylar landmarks, whereas horizontal errors were more pronounced in the mandibular gonial angle (P <0.05). The AI and manual methods show high consistency in all measurements (interclass correlation coefficient >0.983). Condyle landmarks were more accurate in permanent dentition, whereas mandibular angle landmarks were more precise in mixed dentition (P <0.05). Furthermore, the model achieved 82.52% and 75.24% diagnostic accuracy when using gonial angle and total ramal height.

Conclusions

The AI model accurately identifies anatomic landmarks and assesses mandibular asymmetry in OPG radiographs, demonstrating generalizability and robustness across different dentitions and showcasing potential as a promising diagnostic tool in clinical practice.

Facial symmetry plays a vital role in esthetics, attractiveness,1 and orofacial functionality.2 Mild facial asymmetry is common and considered a natural outcome of normal growth and development; severe facial asymmetry can negatively impact oral functions, facial attractiveness, and psychosocial well-being.3-5 Mandibular asymmetry is the primary contributor to facial asymmetry,6 with variations found in different studies ranging 17.43%-72.95%.7-11 Notably, a study reported that more than half of children present with different degrees of asymmetry in the height and width of the mandibular ramus.12 Early identification of mandibular asymmetry is crucial for pediatric patients to prevent the progression of skeletal abnormalities into adulthood.13,14 Similarly, detecting mandibular asymmetry in adults during orthodontic planning can enhance treatment outcomes, minimize adverse effects such as temporomandibular joint (TMJ) disorders and occlusal canting, and facilitate effective communication between patients and clinicians.15,16 Therefore, timely detection and diagnosis of mandibular asymmetry are important to designing individualized orthodontic treatment plans.

Currently, mandibular asymmetry analysis primarily relies on posteroanterior (PA) radiographs17 or cone-beam computed tomography (CBCT).18-21 However, these approaches are not routinely used because of their limitations: (1) PA has multiple overlapping structures, making analysis difficult, and (2) CBCT has high radiation exposure, which prevents its adoption as a standard screening method, and is usually performed after an initial diagnosis.22 In contrast, panoramic radiograph, also known as orthopantomography (OPG), is the most commonly used radiograph in dentistry and serves as an essential tool for initial diagnostic screening. It offers a broad field of vision of the entire dentition, jaws, and surrounding structures, making it potentially suitable for early screening of mandibular asymmetry.23

Several studies have been conducted to assess mandibular asymmetry using OPG,24-26 after Hebets’ establishment of mandibular asymmetry indexes based on OPG.27 The reliability and validity of OPG in evaluating mandibular asymmetry have been verified through comparisons with PA cephalograms28 and CBCT.29 Although the diagnostic role of OPG in mandibular asymmetry is confirmed, its accurate and efficient analysis remains a challenge, especially for young dentists with less experience.30 Traditional method requires dentists to manually landmark and calculate asymmetry indexes in OPG, which is time-consuming and subject to individual variations. In addition, the measurements may be affected by the quality of images, including horizontal distortion and artifacts. Therefore, an efficient and objective tool is needed to assist dentists in effectively analyzing mandibular asymmetry using OPG images.

Deep learning is an artificial intelligence (AI) method which can serve as a powerful tool in analyzing medical images. It can automatically learn generalized representations from provided datasets and make accurate predictions on unseen data.31 In dentistry, deep learning models have been used to perform tasks including tooth segmentation, anatomic landmark identification, and disease diagnosis.32 Automated asymmetry analysis using deep learning can provide efficient, objective, and standardized measurements that enhance the reliability of diagnostic and treatment planning processes. To the best of our knowledge, no studies have yet explored deep learning models for the assessment of mandibular asymmetry using OPG radiographs. Therefore, this study aimed to establish a high-precision deep learning model that can identify tooth and bony markers on OPG radiographs, detect horizontal distortions, and calculate the asymmetry index. Ultimately, this model is expected to facilitate the early screening and diagnosis of mandibular asymmetry.


Persistent Identifierhttp://hdl.handle.net/10722/355354
ISSN
2023 Impact Factor: 2.7
2023 SCImago Journal Rankings: 1.283

 

DC FieldValueLanguage
dc.contributor.authorQu, Wanting-
dc.contributor.authorQiu, Zelin-
dc.contributor.authorLam, Kwong Chuen-
dc.contributor.authorSakaran, Koshla Guna-
dc.contributor.authorChen, Hao-
dc.contributor.authorLin, Yifan-
dc.date.accessioned2025-04-04T00:35:19Z-
dc.date.available2025-04-04T00:35:19Z-
dc.date.issued2025-03-05-
dc.identifier.citationAmerican Journal of Orthodontics and Dentofacial Orthopedics, 2025-
dc.identifier.issn0889-5406-
dc.identifier.urihttp://hdl.handle.net/10722/355354-
dc.description.abstract<h3>Introduction</h3><p>Mandibular symmetry is crucial in orthodontic diagnosis and treatment planning. This study aimed to establish an artificial intelligence (AI) method to automatically and accurately identify mandibular landmarks and assess asymmetry via orthopantomography (OPG) radiographs.</p><h3>Methods</h3><p>A total of 1038 OPG radiographs (451 mixed and 587 permanent dentitions) were collected and annotated to develop the AI model for identifying mandibular landmarks. First, the mesiodistal widths of the bilateral mandibular first molars were compared to categorize images as horizontally distorted or nondistorted. Next, the efficacy and robustness of the model were assessed through landmark identification, measurement, and asymmetry diagnostics accuracy using successful detection rates and interclass correlation coefficient.</p><h3>Results</h3><p>The AI model achieved an average landmark detection error of 0.86 ± 0.95 mm, with 0.97 ± 0.99 mm for bony landmarks and 0.54 ± 0.84 mm for dental landmarks. The successful detection rates at 1, 2, and 3 mm were 75.33%, 93.11%, and 96.72%, respectively. The accuracy exhibits region-specific variations: vertical errors were larger in condylar landmarks, whereas horizontal errors were more pronounced in the mandibular gonial angle (<em>P</em> <0.05). The AI and manual methods show high consistency in all measurements (interclass correlation coefficient >0.983). Condyle landmarks were more accurate in permanent dentition, whereas mandibular angle landmarks were more precise in mixed dentition (<em>P</em> <0.05). Furthermore, the model achieved 82.52% and 75.24% diagnostic accuracy when using gonial angle and total ramal height.</p><h3>Conclusions</h3><p>The AI model accurately identifies anatomic landmarks and assesses mandibular asymmetry in OPG radiographs, demonstrating generalizability and robustness across different dentitions and showcasing potential as a promising diagnostic tool in clinical practice.</p><p>Facial symmetry plays a vital role in esthetics, attractiveness,<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>1</sup></a> and orofacial functionality.<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>2</sup></a> Mild facial asymmetry is common and considered a natural outcome of normal growth and development; severe facial asymmetry can negatively impact oral functions, facial attractiveness, and psychosocial well-being.<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>3-5</sup></a> Mandibular asymmetry is the primary contributor to facial asymmetry,<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>6</sup></a> with variations found in different studies ranging 17.43%-72.95%.<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>7-11</sup></a> Notably, a study reported that more than half of children present with different degrees of asymmetry in the height and width of the mandibular ramus.<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>12</sup></a> Early identification of mandibular asymmetry is crucial for pediatric patients to prevent the progression of skeletal abnormalities into adulthood.<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>13</sup></a><sup>,</sup><a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>14</sup></a> Similarly, detecting mandibular asymmetry in adults during orthodontic planning can enhance treatment outcomes, minimize adverse effects such as temporomandibular joint (TMJ) disorders and occlusal canting, and facilitate effective communication between patients and clinicians.<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>15</sup></a><sup>,</sup><a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>16</sup></a> Therefore, timely detection and diagnosis of mandibular asymmetry are important to designing individualized orthodontic treatment plans.</p><p>Currently, mandibular asymmetry analysis primarily relies on posteroanterior (PA) radiographs<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>17</sup></a> or cone-beam computed tomography (CBCT).<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>18-21</sup></a> However, these approaches are not routinely used because of their limitations: (1) PA has multiple overlapping structures, making analysis difficult, and (2) CBCT has high radiation exposure, which prevents its adoption as a standard screening method, and is usually performed after an initial diagnosis.<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>22</sup></a> In contrast, panoramic radiograph, also known as orthopantomography (OPG), is the most commonly used radiograph in dentistry and serves as an essential tool for initial diagnostic screening. It offers a broad field of vision of the entire dentition, jaws, and surrounding structures, making it potentially suitable for early screening of mandibular asymmetry.<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>23</sup></a></p><p>Several studies have been conducted to assess mandibular asymmetry using OPG,<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>24-26</sup></a> after Hebets’ establishment of mandibular asymmetry indexes based on OPG.<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>27</sup></a> The reliability and validity of OPG in evaluating mandibular asymmetry have been verified through comparisons with PA cephalograms<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>28</sup></a> and CBCT.<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>29</sup></a> Although the diagnostic role of OPG in mandibular asymmetry is confirmed, its accurate and efficient analysis remains a challenge, especially for young dentists with less experience.<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>30</sup></a> Traditional method requires dentists to manually landmark and calculate asymmetry indexes in OPG, which is time-consuming and subject to individual variations. In addition, the measurements may be affected by the quality of images, including horizontal distortion and artifacts. Therefore, an efficient and objective tool is needed to assist dentists in effectively analyzing mandibular asymmetry using OPG images.</p><p>Deep learning is an artificial intelligence (AI) method which can serve as a powerful tool in analyzing medical images. It can automatically learn generalized representations from provided datasets and make accurate predictions on unseen data.<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>31</sup></a> In dentistry, deep learning models have been used to perform tasks including tooth segmentation, anatomic landmark identification, and disease diagnosis.<a href="https://www.ajodo.org/article/S0889-5406(25)00056-3/fulltext#"><sup>32</sup></a> Automated asymmetry analysis using deep learning can provide efficient, objective, and standardized measurements that enhance the reliability of diagnostic and treatment planning processes. To the best of our knowledge, no studies have yet explored deep learning models for the assessment of mandibular asymmetry using OPG radiographs. Therefore, this study aimed to establish a high-precision deep learning model that can identify tooth and bony markers on OPG radiographs, detect horizontal distortions, and calculate the asymmetry index. Ultimately, this model is expected to facilitate the early screening and diagnosis of mandibular asymmetry.</p>-
dc.languageeng-
dc.publisherElsevier-
dc.relation.ispartofAmerican Journal of Orthodontics and Dentofacial Orthopedics-
dc.rightsThis work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.-
dc.titleArtificial intelligence-assisted identification and assessment of mandibular asymmetry on panoramic radiography-
dc.typeArticle-
dc.identifier.doi10.1016/j.ajodo.2025.01.018-
dc.identifier.scopuseid_2-s2.0-86000180481-
dc.identifier.eissn1097-6752-
dc.identifier.issnl0889-5406-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats