The accuracy of cephalometric tracing superimposition.

PURPOSE: The purpose of this study was to compare the accuracy of 4 methods for cephalometric tracing superimposition. They are the FH@Porion method, S-N@Sella method, least-squared averaged 5 landmarks (LS-5) method, and manual geometric method.
MATERIALS AND METHODS: Eight lateral cephalometric radiographs were used. Cephalometric tracing was performed by 2 examiners. One had extensive experience in landmark digitization while the other had minimal experience. The radiographs were scanned and the reference landmarks ANS, Point A, Point B, and Pogonion were digitized, creating 8 master tracings. Then 6 digital copies of each master tracing were made, 3 for each examiner. Subsequently, the examiners were asked to digitize and trace predetermined cranial base landmarks and structures. Tracings occurred at 1-month intervals. As a result, 3 separate tracings of each set were obtained from each examiner. The tracings of each set were superimposed using 4 different methods in the CASSOS software (SoftEnable Technology Ltd, Hong Kong SAR, China). For each method of superimposition, the coordinates of ANS, Point A, Point B, and Pogonion were recorded. Their means and variances were calculated. The variance represents the variability of the superimposition method. A general linear model for repeated measures was computed to test whether there were statistically significant differences among the 4 superimposition methods, 2 examiners, 4 reference landmarks, and 2 directions. Because the distribution of the variances was skewed, they were transformed to log variances. Finally, the errors of the superimposition in millimeters for each given examiner, superimposition method, reference landmark, and direction (X, Y) were calculated.
RESULTS: There was a statistically significant difference in measurement variability among the 4 superimposition methods (P < .001). For both examiners, the variability of the different superimposition methods from the highest to the lowest was: Frankfort Plane registered at Porion method, Sella-Nasion registered at Sella method, least-square averaged 5 landmarks method, and the manual geometric method. In addition, there was a statistically significant difference in the magnitude of superimposition errors between the 2 examiners (P < .001). The experienced examiner was consistently more precise than the inexperienced examiner across all methods. Moreover, there was a statistically significant difference among 4 reference landmarks (P < .001). For both examiners, the recorded variability of each given reference landmark from the lowest to the highest was: ANS, Point A, Point B, and Pogonion. Furthermore, the variability differences between horizontal and vertical directions did not reach a conventional level of significance (P = .123). Finally, the recorded errors in millimeters for each superimposition method were summarized. A smaller error in millimeters represented a higher accuracy in superimposition. The error of using manual geometric or LS-5 methods for both examiners was less than 0.50 mm, while the error of using the other 2 methods was up to 0.99 mm for the experienced examiner and 2.88 mm for the inexperienced examiner.
CONCLUSION: The error of both manual and LS-5 methods was within 0.5 mm. The LS-5 method had its advantage because it could be automated by the computer.