Original Article

Concordance and reproducibility in the location of reference points for a volumetric craniofacial analysis: Cross-sectional study

Introduction

Cephalometric analysis is the preferred diagnostic study in the orthodontic area.^1-3 Over the years, it has been implemented to identify the cause of malocclusion. Due to its application in orthodontics, radiographic evaluation should create a representation that is as close to the reality of the patient as possible, allowing the clinician to understand the causes of the existing craniofacial alterations and the alternatives for their correction.^1,4 This diagnostic practice has been based mainly on the study of lateral skull radiographs, but other cranial radiographs are added to the diagnostic process.⁵ On the other hand, taking into account the limitations of visualization that occur even with the use of radiographs, the cone beam computed tomography (CBCT) becomes more attractive to make the diagnosis, and consequently an assertive treatment plan.³ Cephalometric analysis is designed to be performed on a 2D image,⁶ so conventional cephalometric analysis tends to be performed on images that have three-dimensional (3D) information, and the additional information contained in a CBCT is not used.

CBCT represents a radiation exposure that could be of higher dose,⁶ however, this depends on the quality of the equipment and the settings by the manufacturers or by adjusting the image parameters that the CBCT machine allows manually. The acquisition of three-dimensional images with very low radiation doses and sufficient quality for diagnosis has already been reported.^7–9

In this area, some studies describe the marking of reference points in the CBCT and make the respective comparison of reliability and reproducibility between the two-dimensional and three-dimensional methods,¹⁰ which yield favorable data for the three-dimensional analysis. However, there is a need to adjust the description of volumetric points for a three-dimensional image and make a diagnostic tool.¹¹ The description of the points in the planes, x, y is already done and now z is necessary to be describe. It has been described in the literature that the points that present great difficulty and variability in their identification in the three planes of space are some such as the porion, condilion, orbital, basion, gonion, anterior nasal spine, posterior nasal spine, and the apex of the lower incisor.¹²

Justification

Visualization tools (zoom and contrast) are useful for marking reference points with high reliability.¹³ A step forward is taken, because the technology seeks to facilitate and simplify these processes.⁶ Due to the information obtained from the CBCT, it is necessary to develop methods that allow its use through the standardization of cephalometric measurements in a 3D reconstruction. The objective of this study was to indicate 31 reference points describing them in the three planes of space in a bimaxillary CBCT and to evaluate the intra and interobserver reproducibility and concordance.

Methods

CBCT of patients who attended the [Universidad del Valle, Cali, Colombia] dental school seeking orthodontic treatment were selected. The tomography of patients with complete permanent dentition and in which it was visualized from the nasion point to the menton point was included. CBCT of patients with skeletal abnormalities, syndromes, with a history of craniofacial trauma, and a history of orthognathic surgery were excluded.

Using the following formula:

We determined the size of the adequate sample to detect a difference > 0.5 mm between the reference points.³ From a database of 108 bimaxillary CT scans taken between June 2019 and March 2020, 20 met the selection criteria, using a simple random sampling the six tomographies were selected.

The images were generated by an iCAT 1719 equipment, FOV 12.6 cm, voxel size 0.25. resolution 0.25 voxel per 26.9 sec exposure + 37.07 kVp 120 and size of the reconstructed volume diameter of 16 cm. The images were taken with the patients adopting a corrected posture with a straight back and head at rest, with the Frankfort plane parallel to the floor.

Tomographies were exported in DICOM files for viewing in Horos software (version 4.0.0; The Horos Project & OsiriX Team) in multiplanar 3D mode (MPR). The marking of each point was verified three dimensionally (Tables 1-3). The point was represented in the three windows of the spatial planes x, y, and z, each observer collected the information on the coordinates of the reference points employing image captures to record the information in tables of Microsoft Excel for Mac version 16.52 ().

Close

Figure 1. Reference points marked in the multiplanar view taking into consideration the three planes of space (sagittal, axial, and coronal). A. Sella point, B. Basion point, C. Mandibular base point

Three observers were involved, two of them were orthodontists, and one was a dental radiologist. 31 reference points were chosen, and these were reviewed in two meetings in which pilot tests were carried out to describe them three-dimensionally, considering anatomical and technological criteria. A consensus was sought among the three observers that the landmarks were clearly described. These 31 reference points were marked in the multiplanar view in triplicate by each of the observers individually.

Statistical analysis

The Friedman test was used to make a comparison of paired or related intraobserver samples by marking the 31 reference points in their three coordinates x, y, and z in triplicate. Interobserver concordance was determined by the intraclass correlation coefficient (ICC) with the ranges > 0.75 (excellent), between 0.60 and 0.74 (good), between 0.40 and 0.59 (sufficient), and < 0.40 (poor). The P value to determine statistical significance for differences between coordinates was set at < 0.05 and the difference in millimeters of the reference points found with poor concordance was obtained.

Results

The evaluation of the points in the 3 planes of space x, y, and z was carried out using the Friedman test to identify if these differences were statistically significant, with which no significant differences were found (P > 0.05) for 19 landmarks (Table 4).

A high ICC ( > 0.75%) was obtained by comparing the x, y, and z values at the location of landmark points. Excellent ICC > 0.75 was for 81.7% and poor < 0.40 was 7.5% in the interobserver evaluation. Data showed that 25 points had excellent concordance on the x-plane, 25 on the y-plane, and 26 on the z-plane (0.75%) (Table 5). One point had poor concordance on the x-plane condylium R, 4 on the y-plane; nasion, the frontomalar suture R, gonion L and menton, and 2 on the z-plane, menton and pogonion. Differences in millimeters for points of poor concordance ranged from 0.3–5 mm (Table 6).

Discussion

Previous studies that dealt with the evaluation of the reliability and reproducibility of 3D reference points, used their anatomical definitions to locate them in the CBC.^5,14-16 In this study, a detailed description was made of how to locate the reference points in each of the planes of space, as suggested by de Oliveira et al.¹⁷ We show the sequence in which the cutting planes must be modified, and we mention the structures to take into account the location of the landmark point, thus creating a universal concept to achieve the marking of the point consistently.

In this study, 31 reference points were marked, most of them described in the literature,¹⁸ except for the mandibular right and left base points, Mx3 right and left and Mx6, right and left. These reference points were described for the first time in this study. Since variations have been found due to the poor description of the reference points.^17,19,20 Here, the reference points were described three dimensionally on rigid structures to make their marking easier, modifying the way of locating the sella turcica.

The results indicate that repeatability and concordance are achieved by marking the reference points proposed in this research. In a systematic review on the reliability of marking points in CBCT, the differences in a cephalometric analysis became clinically significant when the difference was greater than 0.5 mm.³

In the intraobserver analysis of this study, the points did not present significant variations in more than one plane, except the right and left mandibular base point, which varied in the three planes of space. This point was described for the first time in this study and the description and understanding may be complex to achieve a good degree of precision. In the evaluation of interobserver concordance, a high correlation was found in 81.7% of points, a good correlation in 3%, a sufficient 7%, and a poor concordance in 7%. In this study, as well as in other studies where the repeatability and reliability of new reference points have been evaluated, it has been found that more than 80% of the points show high concordance.^13,21,22

In the points that showed poor ICC, there was variation in one of the three planes, except for the menton point, with poor concordance in the y and z planes. On the y and z planes, there points were cataloged with poor or high correlation, showing the thin line between concordance and not. On the z-plane, the points with high concordance were marked more frequently. The highest frequency of poor reliability was on the y plane as indicated by de Oliveira et al.¹⁷

Nasion has been reported in the literature as one of the least variable points, since being on the midsagittal plane, it has been routinely visualized in the 2D view, which makes it easier to understand and locate,²³ this is consistent with other studies where it turns out to have high reliability ICC (x: 0.87; y: 0.98; z: 0.97)¹⁷ (x: 0.98; y: 1; z: 0.98).²⁰ In contrast, in our study, nasion was found to be located with more difficulty in the y-plane (ICC = 0.07), This may be due to the difficulty in finding the deepest point in the concavity of the floor of the sella turcica.

On the x-plane for condylium R, while in other studies it was located with high concordance (ICC > 0.75) (x: 0.97), In the study by Chien et al¹² bilateral evaluation is not reported they speak only of a condylium point while in this study we evaluated right and left condylium; they also omitted the z-plane to make the comparison with the 2D image. In Oliveira and colleagues’ study, left and right condylium are evaluated and the ICC for condylium R on the x-plane(0.46) shows sufficient concordance but is closer to our finding; on the y-plane, the data of high concordance coincide, and differs again on the z-plane(ICC = 0.28), which contrasts with the z-plane of our study (ICC = 0.99),¹⁷ the poor concordance of condylium R on the x-plane was also found in the study by Neiva et al¹³ (ICC = 0.21).

The point located on the frontomalar suture R is evaluated in a few studies; in Neiva et al¹³ poor concordances were found on the x-plane and high on y,z; while in this study poor concordance was found on the y-plane and high in x,z and de Oliveira et al¹⁷ found that the point is concordant in x,y,z.Other points with poor concordance in this study were menton and pogonion, results that radically contrast with the results of high concordance found in the literature.^12,13,16,17

Gonion is one of the points reported with high marking difficulty, but in this study, it turned out to be a reliable point, perhaps due to the description that was made in the three planes of space using reference structures such as the mandibular ramus and base.²⁴ However, the error associated with the gonion point is in the vertical plane as seen with the ICC of Gonion L on the y-plane (0.32) which may favor the presence of errors in measurements on the vertical plane having gonion as reference. The fact that there are three coordinates for the identification of a reference point adds to the risk of variability in the repeatability of the data. Therefore, the selection of the segment for marking the reference point must be carefully guided by the structures to be evaluated and properly managed by the visualization tools.

Points with poor concordance in this study do not follow a pattern when compared to other studies. Generally, one of the planes shows poor concordance, while the remaining two show sufficient, good, or excellent. In this evaluation, discrepancies are observed in different planes, and two points were observed, nasion and menton with poor concordance in the two planes. It has already been mentioned that it is impossible to locate a point of reference according to its definition without a margin of error, however, all efforts are aimed at minimizing non-concordance.²⁵

In the points where poor concordance was found, the average difference in millimeters varies from 0.3 mm to 5 mm. The greatest discrepancy in this study is reported for condilium R on the x-plane with an average difference of 5 mm, other authors such as Lagravère et al²⁶ also report variations greater than 1 mm.

With the frequent variability that can be found on the y-plane and for points that are located on convex surfaces such as the Gonion point, an average difference is reported for Gonion L in y of 2.6 mm, while previously it had been reported greater difference (5.5 mm).²⁶ Nasion point, in this study, had poor concordance in the y-plane and with an average difference of 0.3 mm, being found in the points with smaller difference averages.

When analyzing the repeatability of the intraobserver marking pattern, suggests that with a complete description of the reference points, a clear training, a person with basic knowledge of image analysis and software management can locate the reference points properly.

It has been shown that the location of the reference points in a CBCT is carried out in a reliable and repeatable way,³ however, due to the variations that occur in the manual location of the reference points, studies have been carried out that claim that one of the multiple functions of 3D craniofacial analysis is the automatic identification of points in the CBCT image.²⁷ No consensus has been reached regarding the description of reference points for three-dimensional cephalometric analysis, therefore, the next step in this line of research will be to apply a three-dimensional analysis as a diagnostic tool. With the incursion of artificial intelligence tools and deep learning,²⁸ an additional panorama opens for segmentation techniques and automatic identification of craniofacial structures, which will further facilitate this field of intuitive diagnosis.

Conclusion

Intraobserver concordance analysis indicated that the location of anatomical reference points on bimaxillary CBCT is performed with great reproducibility by interpreting their location with a clear description in the three planes of space. There are still important reference points for morphological analysis, that because they are located in convex areas with variable bone densities, are difficult to standardize. Variability in the manual marking of reference points may be determined by different visualization factors and clinical experience. The systematized location of the reference points would contribute to the reduction of such variability.

Competing Interests

The authors declare that there is no conflict of interest regarding the publication of this article.

Ethical Approval

This study was approved by the ethics committee of the Universidad del Valle under the act number 234-019.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

References

1. Principales análisis cefalométricos utilizados para el diagnóstico ortodóntico. Rev Cient Odontol 2006; 2(1):11-27.
2. A comparison between 2D and 3D cephalometry on CBCT scans of human skulls. Int J Oral Maxillofac Surg 2010; 39(2):156-60. doi: 10.1016/j.ijom.2009.11.017 [Crossref]
3. Reliability of different three-dimensional cephalometric landmarks in cone-beam computed tomography: a systematic review. Angle Orthod 2019; 89(2):317-32. doi: 10.2319/042018-302.1 [Crossref]
4. A new methodology for automatic detection of reference points in 3D cephalometry: a pilot study. Int Orthod 2018; 16(2):328-37. doi: 10.1016/j.ortho.2018.03.013 [Crossref]
5. Ludlow JB, Gubler M, Cevidanes L, Mol A. Precision of cephalometric landmark identification: cone-beam computed tomography vs conventional cephalometric views. Am J Orthod Dentofacial Orthop 2009;136(3):312.e1-312.e10. 10.1016/j.ajodo.2008.12.018.
6. The reliability and reproducibility of cephalometric measurements: a comparison of conventional and digital methods. Dentomaxillofac Radiol 2012; 41(1):11-7. doi: 10.1259/dmfr/37010910 [Crossref]
7. Cone beam computed tomographic imaging in orthodontics. Aust Dent J 2017; 62 Suppl 1:33-50. doi: 10.1111/adj.12479 [Crossref]
8. Cone beam computed tomography for dental and maxillofacial imaging: technique improvement and low-dose protocols. Radiol Med 2017; 122(8):581-8. doi: 10.1007/s11547-017-0758-2 [Crossref]
9. Novel low-dose protocols using cone beam computed tomography in dental medicine: a review focusing on indications, limitations, and future possibilities. Clin Oral Investig 2019; 23(6):2573-81. doi: 10.1007/s00784-019-02907-y [Crossref]
10. A comparative study of the reproducibility of landmark identification on posteroanterior and anteroposterior cephalograms generated from cone-beam computed tomography scans. Korean J Orthod 2019; 49(1):41-8. doi: 10.4041/kjod.2019.49.1.41 [Crossref]
11. Comparison of linear and angular measurements using two-dimensional conventional methods and three-dimensional cone beam CT images reconstructed from a volumetric rendering program in vivo. Dentomaxillofac Radiol 2011; 40(8):492-500. doi: 10.1259/dmfr/15644321 [Crossref]
12. Comparison of reliability in anatomical landmark identification using two-dimensional digital cephalometrics and three-dimensional cone beam computed tomography in vivo. Dentomaxillofac Radiol 2009; 38(5):262-73. doi: 10.1259/dmfr/81889955 [Crossref]
13. Evaluation of cephalometric landmark identification on CBCT multiplanar and 3D reconstructions. Angle Orthod 2015; 85(1):11-7. doi: 10.2319/120413-891.1 [Crossref]
14. Reliability of traditional cephalometric landmarks as seen in three-dimensional analysis in maxillary expansion treatments. Angle Orthod 2009; 79(6):1047-56. doi: 10.2319/010509-10r.1 [Crossref]
15. Precision of identifying cephalometric landmarks with cone beam computed tomography in vivo. Eur J Orthod 2013; 35(1):38-44. doi: 10.1093/ejo/cjr050 [Crossref]
16. Accuracy and reliability of landmark-based, surface-based and voxel-based 3D cone-beam computed tomography superimposition methods. Orthod Craniofac Res 2017; 20(4):227-36. doi: 10.1111/ocr.12205 [Crossref]
17. Observer reliability of three-dimensional cephalometric landmark identification on cone-beam computerized tomography. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2009; 107(2):256-65. doi: 10.1016/j.tripleo.2008.05.039 [Crossref]
18. A new mandible-specific landmark reference system for three-dimensional cephalometry using cone-beam computed tomography. Eur J Orthod 2016; 38(6):563-8. doi: 10.1093/ejo/cjv088 [Crossref]
19. Precision of manual landmark identification between as-received and oriented volume-rendered cone-beam computed tomography images. Am J Orthod Dentofacial Orthop 2017; 151(1):118-31. doi: 10.1016/j.ajodo.2016.06.027 [Crossref]
20. 3D landmarks of Craniofacial Imaging and subsequent considerations on superimpositions in orthodontics-the Aarhus perspective. Orthod Craniofac Res 2019; 22 Suppl 1:21-9. doi: 10.1111/ocr.12299 [Crossref]
21. Development and validation of novel three-dimensional craniofacial landmarks on cone-beam computed tomography scans. J Craniofac Surg 2019; 30(7):e611-e5. doi: 10.1097/scs.0000000000005627 [Crossref]
22. Cephalometric landmark identification consistency between undergraduate dental students and orthodontic residents in 3-dimensional rendered cone-beam computed tomography images: a preliminary study. Am J Orthod Dentofacial Orthop 2017; 151(1):157-66. doi: 10.1016/j.ajodo.2016.06.034 [Crossref]
23. Reliability of 3D dental and skeletal landmarks on CBCT images. Angle Orthod 2019; 89(5):758-67. doi: 10.2319/082018-612.1 [Crossref]
24. Cephalometric landmark variability among orthodontists and dentomaxillofacial radiologists: a comparative study. Imaging Sci Dent 2015; 45(4):213-20. doi: 10.5624/isd.2015.45.4.213 [Crossref]
25. 3D cephalometric landmark detection by multiple stage deep reinforcement learning. Sci Rep 2021; 11(1):17509. doi: 10.1038/s41598-021-97116-7 [Crossref]
26. Intraexaminer and interexaminer reliabilities of landmark identification on digitized lateral cephalograms and formatted 3-dimensional cone-beam computerized tomography images. Am J Orthod Dentofacial Orthop 2010; 137(5):598-604. doi: 10.1016/j.ajodo.2008.07.018 [Crossref]
27. Cephalogram synthesis and landmark detection in dental cone-beam CT systems. Med Image Anal 2021; 70:102028. doi: 10.1016/j.media.2021.102028 [Crossref]
28. Artificial Intelligence for radiographic image analysis. Semin Orthod 2021; 27(2):109-20. doi: 10.1053/j.sodo.2021.05.007 [Crossref]