INTRODUCTION

Assessing symmetry is important in any esthetic evaluation of the craniofacial region. A previous study by Katsumata et al.1 described a system for investigating skeletal asymmetry on three-dimensional head radiographs using a skeletal landmark-based system. Such a system could allow accurate comparison of a given orthodontic patient to normative values, and make asymmetry diagnosis more objective.

There are many articles that posit a relationship between malocclusion and cranial skeletal asymmetry. One relationship frequently cited in the literature is the Class II subdivision patient. Janson et al.2 found that a Class II subdivision patient has the maxillary midline on with the face and the mandibular midline off toward the Class II side 61% of the time, the mandibular midline on with the face and the maxillary midline off 18% of the time, and a combination is present 20% of the time. Alavi et al.3 examined Class II subdivision patients and primarily found asymmetry in the mandible of these patients.

Another relationship that has been previously investigated is the relationship between ANB angle and facial or skeletal asymmetry. A study by Good et al.4 divided patients by ANB angle with groups of less than 3 degrees, 3–4 degrees, and greater than 4 degrees, and examined facial photographs for mandibular asymmetry. They found significantly more asymmetry in the less than 3 degrees ANB angle group. A retrospective study by Severt et al.5 looked at 1460 orthognathic cases treated at the University of North Carolina and found an increased percentage of chin deviation of at least 2 mm from the midline in Class III patients. The purpose of this study is to compare the skeletal symmetry of orthodontic patients who have a high ANB angle to those with average ANB angles using CBCT images. The hypothesis is that patients with a skeletal Class II growth pattern, which is a readily identifiable skeletal variation, may also have increased levels of skeletal asymmetry.

Three-dimensional cone beam computed tomography (CBCT) images are uniquely suited to assess asymmetry. These images use a built-in reconstruction algorithm to correct known distortions due to projection geometry,6 and allow clinicians to assess skull anatomy either through three-dimensional surface renderings or through accurate two-dimensional slices through the skull. Previous methods of asymmetry assessment relied on two-dimensional radiographs with inherent problems not present in CBCT images. Two-dimensional images have variations in magnification with areas closer to the film magnified less than those nearer to the radiation source.7 Cook8 found that a posteroanterior cephalometric radiograph with patient head rotation of 5 degrees can cause the apparent side of an asymmetry to switch. Lateral cephalometric radiographic asymmetry studies can be affected by asymmetric external auditory meati locations, which would lead to head rotation and erroneous results. A potential drawback to asymmetry assessment using CBCT images is a lack of a database of symmetric patients as a basis for comparison and asymmetry diagnosis. Such a database would be difficult to create because of concerns about unnecessary radiation exposure of people who may not require a CBCT scan.

There are unique difficulties and advantages associated with identifying craniometric landmarks on three-dimensional CBCT images. Some landmarks can be easily identified in one or two planes of space, but identification in the third plane of space can be difficult. In these cases observers tend to locate the landmarks in the planes of easiest visualization.9,10 Difficulty is created by inclusion of the medio-lateral dimension to landmark location, a dimension which was not considered when lateral cephalometric landmarks were created.10 A benefit of CBCT landmark identification is a lack of superimposed structures, which creates easier visualization of certain skull regions. Also, CBCT images are not plagued by geometric distortion.11–13 Another advantage to CBCT landmark identification is visualization using the stacks of CBCT slices, known as the multiplanar reconstruction (MPR) images. This allows inspection from any of the three planes of space to help identify landmarks from multiple views. Grauer et al.14 recommends identifying landmarks on the MPR images rather than on the three-dimensional surface rendering as the rendered image is dependent on factors such as image contrast, movement during acquisition, signal-to-noise ratio of the image, and threshold filters applied by the operator.

There are several variables that could influence accuracy of linear measurements reported in the literature. Voxel size, type of image detector, scanning time, reconstruction time, radiation dose, and head position have been mentioned as variables that could affect CBCT linear measurements.9,15 Two previous studies have investigated the effect of head position on linear measurements and found it insignificant.16,17 Another study compared measurement accuracy using two different voxel settings with no difference in accuracy between the two groups.17 Overall, various studies have investigated the accuracy of landmark localization on CBCT images. Many have found the process both accurate and reproducible10,17–19 Other studies have investigated linear measurements using CBCT images and support the accuracy of such measures.15,20–25

MATERIALS AND METHODS

This study was reviewed and approved by the Ethics Committee at the University of Minnesota. This study examined CBCT images of 70 patients presenting for routine orthodontic treatment at the University of Minnesota obtained as part of diagnostic record gathering. The use of CBCT was not based on any clinical assessment of asymmetry. IRB approval was obtained to review these CBCT images. ANB angles were measured on simulated lateral cephalometric images extracted from the CBCT radiographs and digitized using Dolphin 11.5 betabuild 15 (Dolphin Imaging and Management Solutions, Chatsworth, Calif). The Class I group consisted of 30 patients with ANB angles ranging from −0.5 degrees to 3.4 degrees. The Class II group consisted of 30 patients with ANB angles of 4.5 and higher. A third group of 10 patients was included that had ANB angles between 3.4 and 4.4 degrees to enable the examination of ANB as a continuous variable. The inclusion criteria were patients age 11 years 0 months to age 15 years 11 months with CBCT films who fit into a required ANB group angle. Patients were excluded if they had been identified during clinical examination, or through examination of the photographs, to have a CR/CO shift, a posterior crossbite, a first molar in dental crossbite, a lack of a first molar, or a craniofacial anomaly or syndrome including cleft lip or cleft palate. The patients were consecutively selected starting from July 2008 until the required patient groups were filled.

CBCT images were acquired with a Next Generation iCAT CBCT unit (Imaging Sciences International, Hatfield, Penn). The device was set for 18.54 mAs and 120 kV. The scan was made with a single 360-degree rotation, 8.9-second scan with a voxel size of 0.3 mm. The head was oriented with the midsagittal plane perpendicular to the floor. Patients were biting in centric occlusion during exposure. DICOM files obtained from the CBCT scan were reconstructed by Dolphin 3D, which is part of Dolphin 11.5 (betabuild 15); all landmark identifications and measurements were made using this software. Images were visualized on a Dell 1908FP monitor (Dell Inc, Round Rock, Tex) with a resolution of 1280 X 1024 at 60 Hz.

Asymmetry assessment for this study was conducted in a manner similar to a study by Katsumata et al.1 Craniometric landmarks were identified as defined in Table 1. Figure 1 demonstrates the MPR slices and three-dimensional reconstructions used to locate each landmark. All landmarks in this study were identified by the primary author. This study strove to use whichever views were most advantageous for localizing each landmark. The views varied from landmark to landmark, but were constant between subjects. For most landmarks, identification was limited to MPR images only. In identifying orbitale, the three-dimensional surface rendering was used as it allowed better medio-lateral localization. In identifying gonion, the three-dimensional surface rendering was used as it gave a broad picture of the ramus and body, making anterior-posterior and superior-inferior identification more accurate. This study did not include the landmark pterygomaxillary fissure as included by Katsumata et al.1 as preliminary investigation demonstrated difficulty in accurate identification.

View Figure

• Download as Powerpoint
• Download Figure

Table 1 Landmark Definitions for Asymmetry Assessment

View Fullscreen

The landmarks Sella, Nasion, and Dent were used to create the three reference planes used to measure linear landmark distances (Figure 2). The mid-saggital plane was defined as passing through the landmarks Sella, Nasion, and Dent. The axial plane was defined as passing through Sella and Nasion and being perpendicular to the mid-saggital plane. The coronal plane was defined as being perpendicular to the other two planes and passing through Dent. For the remaining landmarks, the distance to each plane was calculated by Dolphin 3D in millimeters by orienting the coordinate plane system to the reference planes. This gave a dx, dy, and dz value for that landmark, with dx the distance to the saggital plane, dy the distance to the axial plane, and dz the distance to the coronal plane.

View Figure

• Download as Powerpoint
• Download Figure

For solitary landmarks located near the midline, the asymmetry indices (AI) were defined as the distance to the sagittal plane (AI  =  dx). For bilateral landmarks the asymmetry indices were calculated using the following formula, where R  =  right, L  =  left:

Thirty patients were randomly selected using a random number generator to have landmarks and reference planes re-identified and asymmetry indexes recalculated two weeks later to test the intra-examiner reliability of the presented method.

RESULTS

The Class I group had a mean ANB angle of 1.7 degrees (SD 1.2, range −0.5 to 3.4). The mean age was 13 years 3 months (SD 1y 1mo, range 11y 2mo to 15y 1mo). The Class II group had a mean ANB angle of 6.3 degrees (SD 1.4, range 4.6 to 10.2). The mean age for this group was 13 years 3 months (SD 1y 3mo, range 11y 5mo to 15y 11mo). The group of 10 patients had a mean ANB angle of 4.0 degrees (SD 0.3, range 3.5 to 4.4). The mean age for this group was 13 years 4 months (SD 1y 1mo, range 11y 9mo to 15y 1mo).

Pearson correlation coefficients measuring the linear relationship of ANB to asymmetry index score were calculated for each landmark across all patients and the results are presented in Table 2. None of the coefficients are statistically significant for a linear relationship between ANB and asymmetry index score for any landmark. Visual inspection of scatter plots gives no apparent support for any non-linear relationships between ANB and asymmetry index score.

Table 2 Pearson Correlation Coefficients of ANB and Asymmetry Index Score Across All Patients

View Fullscreen

ANOVA results compared mean asymmetry index scores between Class I and Class II ANB angle groups of patients. Table 3 gives F and P values for each ANOVA test for each landmark. No significant differences in the mean asymmetry index values were found for any landmark.

Table 3 ANOVA Analysis of the Relationship Between Angle Classification and Asymmetry Index

View Fullscreen

Because there were no significant differences in asymmetry index scores relative to ANB angle, the scores from all groups were combined to find an asymmetry index mean score and standard deviation for each landmark (Table 4). Reproducibility of asymmetry index scores was assessed through repeat measurement of 30 randomly selected patients. Altman-Bland plots26 compared original and repeated measurement data for each landmark as well as for dx, dy, and dz. Table 5 presents a summary of the Altman-Bland plots. The data does not demonstrate average bias for any landmark.

Table 4 Asymmetry Index Means ± SD with All Patients Combined

View Fullscreen

Table 5 Altman-Bland Analysis with Bias and 95% CI

View Fullscreen

DISCUSSION

This study used a previously developed three-dimensional asymmetry analysis system to investigate the relationship between skeletal Class II patients and asymmetry. The results of this study do not support a relationship between increasingly positive ANB angles and skeletal asymmetry. The data does not approach significance for any craniometric landmark. A future research study could investigate three-dimensional asymmetry among groups of patients known to be asymmetric, such as Class II subdivision patients or unilateral posterior crossbite patients, to test the sensitivity of this method. It is noted that the range of agreement for each landmark limits this index's ability to diagnose asymmetry on an individual patient. It may be best utilized as it was in this study, for comparing large groups and testing asymmetry differences for a specific trait.

Katsumata et al.1 used the mean asymmetry index scores and standard deviations to establish ranges for symmetry, asymmetry, and marked asymmetry for a particular landmark. Their baseline for calling a landmark asymmetric was the mean for the landmark plus one standard deviation. They defined a landmark as markedly asymmetric if its asymmetry index score was twice the baseline level for asymmetry. Figure 3 demonstrates graphically the boundaries for asymmetry and marked asymmetry diagnosis using the method of Katsumata et al.,1 as well as the new boundaries for diagnosis if means and standard deviations from the present study are used. The boundaries are higher in the present study as Katsumata's study utilized patients who were selected for lack of asymmetry. Using either set of norm values it may be more advantageous to report the number of standard deviations from the mean rather than specific cutoffs for asymmetry and marked asymmetry which can oversimplify the information provided from the index. Future studies of known asymmetric populations using CBCT may clarify recommended values for asymmetry diagnosis using this index.

View Figure

• Download as Powerpoint
• Download Figure

Landmark identification errors are likely the main source of error in this study. Altman Bland analysis indicated that for this study reproducibility was poorest in the mediolateral dimension (dx). There are extra difficulties in identifying landmarks on living human scans which can lead to errors, such as reductions in image quality due to soft-tissue attenuation, metallic artifacts, and patient motion.15 Creating the three reference planes requires identifying the landmarks sella, nasion, and dent. Each of these landmarks requires identifying a centroid of a specific three-dimensional structure and thus qualifies as a “fuzzy landmarks.”17,18 Any identification error of these landmarks leads to a built-in error for all asymmetry index measures. Even the relatively easy to identify landmark ANS has a range of 95% agreement for the asymmetry index of nearly two units in either direction. It is possible that this range of agreement for ANS asymmetry index scores is mostly due to variability in landmark identification of the reference landmarks and consequently variability in the reference planes. Ideally, a system of reference planes would be based upon easy to identify anatomic landmarks which should increase the accuracy and reliability of this system.

Orienting the reference planes using Dolphin 3D created another source of error. The reference planes had to be manually oriented to pass through the center of each of the required reference landmarks. Dolphin 3D is able to identify how far the reference planes are located from the center of each of the landmarks, and this error was minimized by not allowing the reference planes to pass greater than 0.2 mm away from the center of the landmarks before accepting the plane locations.

Dolphin 3D automatically calculated the linear distances from each landmark to each of the reference planes in millimeters. Because the linear measurement portion of the analysis was automated, it is not a likely source of further error in this study.

CONCLUSION

• Based on this study, the discrepant jaw growth resulting in a Class II skeletal pattern results in no more skeletal asymmetry than Class I skeletal patterns.