Landmarks in three dimensions: reconstruction from cephalograms versus direct observation.

A method for generating three-dimensional landmark locations from routine biplane pairs of cephalograms was previously introduced. In this article we compare the locations thus reconstructed to the same configurations as directly recorded through a redundant network of interpoint distances measured with calipers. Six mandibular landmarks were located by both methods on each of 10 dry skulls. With the caliper distances is associated a standard error that can be estimated without explicit remeasurement by the method of "adjustment of networks" familiar from surveying. These standard errors are consistent with the remeasurement error both of the caliper measurements and of the stereolocation from cephalograms; the methods appear to have the same precision, about 0.4 mm per distance. The bias (systematic shift) of the biplane reconstructions with respect to the points used for laying down the calipers may be estimated by regressions of distance discrepancies on the direction cosines of the separations between pairs of landmarks. The caliper tips placed condylion approximately 10 mm medially and a bit forward of where we chose to reconstruct it from biplane cephalograms. The caliper locations of gonion average about 1.6 mm back of their cephalometric position, while those at menton and lower incisal edge are forward by some 1.4 mm. We conclude that the biplane reconstruction (the "three-dimensional cephalogram") is sufficiently accurate for routine clinical and surgical application.