Assessment of 3-dimensional computer-generated cephalometric measurements

The purpose of this study was to assess the reliability of 3-dimensional computer-generated linear and angular measurements produced by different computer algorithms and various combinations of cephalogram projections compared with direct and CT measurements, A computer program was written to provide 4 computer algorithms and 4 combinations of cephalogram projections generating 22 linear and 10 angular 3-dimensional measurements from 20 landmarks. A new technique to produce biplanar cephalograms from a single x-ray source using a special facebow was developed, and its reliability was assessed. Sets of lateral, frontal, and basilar cephalograms of a human dried skull were taken both with 20 radiopaque landmark markers and without markers. Paired t tests based on marker position demonstrated reliability of the facebow; there were no statistically significant differences in repositioning the skull over time using the facebow at P < .05. In the ideal situation, with minimal head rotation and landmark identification error (with the facebow and radiopaque markers), the average error of linear measurements was 1.5 mm and 3.5 degrees for the angular measurements. Subsequent trials evaluated the errors in head position (within 5 degrees of head rotation) and in landmark identification (by removing all markers); two-way ANOVA with Scheffe groupings concluded that the vector intercept with manual adjustment algorithm using the lateral-frontal biplanar projection provides not only greater accuracy but also clinical practicality for both linear (mean of 2.2 mm error) and angular (mean of 4.0 degrees error) measurements compared with direct or CT measurements (P < .05), The effect of landmark identification error was found to be slightly greater than the head rotation error in the accuracy of 3-dimensional linear and angular measurements (mean, 2.85 mm error for linear and 4.4 degrees error for angular measurements). Lastly, this study concluded that linear measurements in the transverse direction were found to have a slightly larger error than vertical measurements. Anteroposterior measurements have the least error.