Abstract: Objective: To evaluate the reliability of identification of CBCT-based skull landmarks and to provide reference for the three-dimensional cephalometric analysis in the future. Methods: 30 pretreatment CBCT data were randomly selected from the CBCT database. Dolphin 11.5 was used to identify 14 skull landmarks in the sagittal, coronal,and axial slices on 3D images by five observers. A principal observer repeated five times and the other four observers three times within a 1-week interval. Intra-class correlation coefficient (ICC) for the values of the coordinate x, y, z of each landmark was calculated to determine Intra-observer and inter-observer reliability. Results: The ICC was ＞0.90 for 64.3%,＞0.75 for 90.5% ,and≤0.45 for only 2.4% of inter-observer assessments. The ICC of the intra-observer assessments was ＞0.90 for 88.1%,＞0.75 for 95.2% ,and none of intra-observer ICC was ≤0.45. Conclusion: Using Dolphin software to identify landmarks on CBCT images can offer excellent reliability with appropriate landmark definition and well-trained observers.

Key words: Cone-beam computed tomography, Landmarks, Reliability, Cephalometry