Dong-Hyun Kim, YuWen Li, Kyungmin Clara Lee
Three-dimensional cephalometric evaluation of the craniofacial morphology in Korean population utilizing cone-beam computed tomography.
- Korean J Orthod 2025;55:254-265
With the increasing use of cone-beam computed tomography (CBCT) in orthodontics, its three-dimensional (3D) nature provides higher-quality information than conventional two-dimensional (2D) lateral cephalograms. However, directly applying the normative values of traditional cephalograms—established over decades as standards for orthodontic diagnosis and treatment evaluation—to 3D CBCT analyses remains challenging. In this regard, the authors’ effort to establish new 3D reference standards is highly meaningful. I sincerely appreciate their valuable contribution and would like to raise the following questions.
Q1. When defining 3D landmarks, were there any items for which the definitions had to be modified because they could not be directly matched with conventional 2D landmarks? Ultimately, to develop an artificial intelligence (AI)-based automated system, are there particular considerations that should receive greater attention during 3D landmark identification compared with 2D analysis?
Q2. Some measurements appear to differ from previously reported normative values for Korean cephalograms. In particular, linear measurements relative to the N–P line seem to be larger than those in 2D norms. Could this be related to differences in head orientation between CBCT and conventional cephalograms, and did you observe any consistent trend in this regard?
Q3. The transverse measurements presented in this study are newly proposed parameters. As their linear values may be influenced by skeletal size or sex, how might these factors be accounted for or standardized in future research?
Questioned by
Hyo-Won Ahn
Department of Orthodontics, College of Dentistry, Kyung Hee University, Seoul, Korea
A1. We sincerely appreciate your interest in our study and your thoughtful question. As you correctly pointed out, the definition of landmarks must differ between 2D and 3D analyses, which inherently makes 3D landmark identification more challenging.In 2022, our team developed an AI-based automatic landmark identification program using 3D CBCT volume-rendered images and encountered similar difficulties during its development. While most 3D landmarks were derived from conventional 2D anatomical definitions extended into 3D space, certain points required modified criteria because of differences in reference planes or structural configurations. Landmarks located along structural borders such as gonion and menton demanded careful inspection from multiple viewing angles for accurate localization, and soft-tissue landmarks required even greater attention due to lower contrast and surface variability compared with skeletal structures.For future AI-based automated landmark detection systems, it will be essential to consider the continuity and precision of landmark identification on 3D structures, as well as potential errors from image reconstruction. Establishing clear and precise definitions, setting reliable reference planes, and carefully localizing each landmark to account for anatomical variability are crucial steps to ensure reproducibility and achieve accurate 3D analysis.
A2. The measurements “A point to N–P (mm)” and “Pog to N–P (mm)” appeared slightly larger than those reported in 2D norms in the McNamara analysis. In our study, however, the “N–P plane (N-perpendicular plane)” was defined somewhat differently from the conventional 2D N–P line. Specifically, it refers to a plane passing through Nasion and perpendicular to both the horizontal plane (the plane passing through N and parallel to the Frankfort horizontal plane) and the midsagittal plane (the plane passing through N and Ba, and perpendicular to the horizontal plane). Thus, while the differences you mentioned could partly be influenced by variations in head orientation, they are more likely attributable to the difference in the definition of the reference plane itself. Accordingly, it may not be necessary to directly compare these measurements with 2D normative values. In addition, during CBCT acquisition, the patient’s head was stabilized in a manner similar to that used with a 2D cephalostat.
A3. One of the greatest advantages of 3D cephalometric analysis using 3D landmark identification is the ability to perform a more detailed and accurate evaluation of transverse dimensions, including facial asymmetry. In this study, we focused on establishing normative values and therefore limited our analysis to symmetric participants. In future research, we plan to extend this work by incorporating transverse dental axes to assess transverse dental compensation in patients with facial asymmetry. Although not included in the manuscript, our data indicated significant sex differences in transverse skeletal dimensions. Specifically, gonial width was larger in males (100.75 ± 6.09 mm) than in females (92.53 ± 5.26 mm), and zygomatic width was also greater in males (140.63 ± 4.96 mm) compared with females (130.77 ± 4.87 mm). Since only symmetric participants were analyzed, no significant sex difference was found in Me–MSP distance. As you mentioned, transverse measurements can only be accurately obtained using 3D imaging, and additional development of related parameters will be essential as the sample size expands. For future studies, it would be beneficial to include only strictly symmetric participants and to normalize measurements according to craniofacial morphology, such as the cephalic index—the ratio of cranial width to depth—by classifying the participants into brachycephalic, mesocephalic, and dolichocephalic groups. This approach would be analogous to differentiating high-angle and low-angle patterns in conventional 2D analysis.
Replied by
Kyungmin Clara Lee
Department of Orthodontics, School of Dentistry, Chonnam National University, Gwangju, Korea