Precision of identifying cephalometric landmarks with cone beam computed tomography in vivo

The study aims were to assess the precision and time required to conduct cephalometric analysis with cone-beam computed tomography (CBCT) in vivo on both three-dimensional (3D) surface models and multi-planar reformations (MPR) images. Datasets from 10 patients scanned with CBCT were used to create two types of images: 1. axial, coronal, and sagittal MPR images and 2. 3D surface models. Eleven observers identified 22 cephalometric landmarks on 3D surface models first and then using 3D in combination with MPR images twice independently. Tracing time was recorded in both methods. Precision was defined as the absolute difference between an observer's repeated measurements and the mean of all measurements per landmark. Inter- and intra-observer agreements were defined as the absolute difference of the observers' measurements from each other and from their repeated measurements averaged over all landmarks, respectively. The precision of measurements ranged between 0.29 ± 0.17 and 2.82 ± 7.53. Adding MPR alongside, 3D surfaces improved the precision of tracing for 15 of 22 of the landmark but it took on average twice as much time. Mean time required to trace one patient was 6:03 ± 2:48 and 10:41 ± 4:01 minutes for 3D and 3D + MPR, respectively (P = 0.0001).