The material for this study consisted of pairs of CBCT scans of 30 children patients (8 ~ 11y) retrieved from the Department of Orthodontics, The stomatological hospital of the Hebei Medical University.CBCT database of patients who underwent orthodontic treatment. The first CBCT scan was performed before treatment while the second was taken after treatment, on average 24 ~ 26 months later. The inclusion criteria were: 1) cervical vertebral maturation (CS1-CS3); 2) Craniofacial basic symmetry;3) mandibular basic symmetry; 4) no temporomandibular joint disorders; 5) no history of maxillofacial trauma; 6) no history of maxillofacial surgery; 7) no systemic disease. 8) clear 3D image. The study protocol was approved by the Medical Ethical Commission of the Hebei Medical University. All patients signed the informed consent.
The CBCT scans were obtained using the Dental Volumetric Tomograph KaVo 3D eXam (Imaging Sciences International LLC, Hatfield, PA, USA) set at 120 kVp, 18.54mA, a field of view of 23 x 17 cm and 0.3 mm voxel size, and scan time of 8.9 s. The scans before treatment were performed under the condition that the patient’s mandibular plane were parallel to floor(Fig. 1). Data from the CBCT were exported in Digital Imaging and Communications in Medicine (DICOM) format to InvivoDental software 5.1.3 ( Anatomage, Inc, San Jose, USA). For each subject, a 3D virtual model was created in a three-dimensional coordinate system. The position and orientation of the 3D model in the coordinate system depended on the head position of the patient when the CBCT scanned. In the "section" module, using axial, coronal, and sagittal views on the left side, the 3D virtual model on the right side was oriented: in the lateral view, bilateral structures, such as the orbits, external auditory canals and other structures as much as possible overlap, and the Frank-fort horizontal plane was oriented horizontally(Fig. 2).
For the fully automatic voxel-based rigid registration, the "superimposition" module in InvivoDental software was used.Data from the CBCT scans after treatment were exported in DICOMformat tothe module and initially superimposed with the image before treatment. After that, clicking on the "voxel registration" button in the top left corner, the 3D surface models before treatment returned to the original posture(Fig. 3), while three white boxes will pop upon three different views of the scan (sagital, coronal, axial). The boxes were used to select the superimposition area5 encompassing from the internal part of thesymphysis to the distal aspect of the first molars at thelevel of basal bone (Fig. 3). Then, clicking on the "start" button and the automatic andvoxel-based superimposition of 3D images before and after treatment begins.The superimposition process took approximately 3 minutes.
A total of 10 landmarks were selected (Table I and Figur4) and defined criteria were established for each landmark.Closing the"automatic registration" box, the models returned to the adjusted posture again (Fig. 3). Moving the mouse pointer to any one of the views, and then turning the mouse wheel to select the most appropriate slice in the view. The three-dimensional coordinates of each landmark were defined by completing the most appropriate selection in the axial, coronal, and sagittal views. Only two of the three views were required, the X, Y, and Z coordinate values of each landmark can be read. For those landmarks that were difficult to defined in a special view, a 3D virtual rendering on the right side can be used as a reference to help the observer define the landmarks(Fig. 5).
Three observers (an orthodontist, anorthodontics master, and a dental radiologist) were trained and calibratedto complete the entire superimposing measurement process using a set of 30 CBCT scans not included in this study. The 3 calibratedobservers worked independently and refined the 9 anatomic landmarks in the CBCT volume before and after treatment. Using the sagittal, coronal, and axial views, the X, Y, and Z coordinate values of each landmark were read and recorded,respectively.The digitized data were then exported to a Microsoft Excel spreadsheet (Microsoft Corporation, Redmond, WA, USA). Each observer repeated the superimposing measurement 3 times at intervals of 5 days, yielding 90 sets for each observer.
Statistical analyses were done by using the Statistical Package for the Social Sciences (version 21.0; SPSS, Chicago. IL, USA). The first step was to calculate the coordinate difference: subtract the X, Y, and Z coordinate values of the landmarks before the treatment from the coordinate value of the corresponding landmarks after the treatment, and the change of the landmarks was represented. The intraobserver agreement was estimated using Intraclass correlation coefficients (ICCs). The difference in mean values were used to assess the precision of the method and the interobserver reproducibility errors.