This study was approved by the institutional (Mohammed Bin Rashid University IRB-2023-45) and regional health authority (Dubai Scientific Research Ethics Committee-GL22-2023) review boards and was performed in accordance with institutional guidelines and regulations.
Dataset
Craniofacial CBCT datasets (n = 11) of subjects (4 male,7 female; 18.7 ± 5.9 years) with unilateral (6 left-sided, 5 right-sided) intraosseous palatally impacted canines acquired with the i-CAT 17–19 (Imaging Sciences International, Hatfield, PA, USA) (n = 10) and Veraviewepocs 3D R100 (J. Morita Corporation, Osaka, Japan) (n = 1) scanners. Scan settings for the i-CAT and the Veraviewepocs 3D units were 120 kVp, 5 mA & 8.9 seconds and 90 kVp, 2 mA, & 9.3 seconds respectively. Informed consent was obtained from all subjects and/or their legal guardians prior to scanning.
The scans with suboptimal quality, partial inclusion of the region of interest, bilateral canine impactions, buccally impacted canines, history of maxillofacial trauma, post-orthognathic surgery, presence of mini-plates, presence of dental implants, craniofacial syndromes, and CLP were excluded. The scans were anonymized and new identification numbers were assigned prior to uploading them on a cloud-based segmentation and image analysis software (Relu BV, Leuven, Belgium).
Automatic Segmentation
Once uploaded (Fig. 1A), sequential steps were executed to facilitate the segmentation of specific regions of interest and volumetric measurements. Initially, the automatic segmentation function of the software was used to segment the edentulous maxillary complex (Mx), impacted canine (IMxC), and non-impacted canine (NIMxC)(Fig. 1B).Then, simultaneous visualization of the volume-rendered (Fig. 1C) and multiplanar reformatted (MPR) views (Figs. 1D,E,F) were enabled in the editor interface of the software.
Investigator-guided segmentation: With automatic segmentation of the edentulous maxillary complex the anterior, posterior, and inferior boundaries were pre-set at the anterior nasal spine (ANS), posterior nasal spine (PNS) and alveolar crest region, respectively. Further segmentation of the edentulous maxillary complex was investigator-guided, wherein boundary conditions were established by identifying specific anatomic landmarks simultaneously in the MPR views by manually toggling between slices. The re-sliceable axis feature of the software was initially utilised to reorient the scans in the MPR views (Figs. 2A1, 2A2, & 2A3) by aligning the ANS, nasopalatine canal (NC), and PNS. Then, in the midline region of the edentulous maxillary complex, the most superior surface, distal to the nasopalatine canal as observed in the sagittal (Fig. 2B1) and coronal (Fig. 2B2) views was chosen as the boundary to guide superior segmentation (Fig. 2B3). The most distal point of the lateral wall of the maxillary sinus as observed in the axial (Fig. 2C1) and coronal (Fig. 2C2) views was chosen as the lateral limit to guide lateral segmentation (Fig. 2C3). Once scan reorientations and landmarks were finalized, smart features of the software interface were activated to segment and remove extraneous regions from the maxillary complex (Fig. 3A) while preserving the areas of interest (Fig. 3B). For segmentation (Fig. 3C1 - C3), the dimensions of the cylindrical eraser tool were set to their maximum value (size:200, depth:100).Finally, ANS, PNS, and NC were again aligned in the MRP views to guide segmentation of the two halves of the maxilla, namely the maxillary half with no impaction (NIMx) and the maxillary half with the impacted canine (IMx) (Fig. 3D).
Volumetric measurements: The volume calculation tool was utilized to measure the voxel-based volumes of the segmented:1)maxillary complex (Mx), 2)maxillary half with the impacted canine (IMx), 3)maxillary half with no impaction (NIMx), 4)canine on the impacted side (IMxC), and 5)non-impacted canine (NIMxC).All measurements were performed independently by three investigators (AB-orthodontic resident; KG-recently graduated orthodontist; and SP-experienced orthodontist),repeated after a two-week memory washout period, and recorded on a spreadsheet.
Statistical Analysis
Data from the spreadsheet were imported, and statistical analyses performed using SPSS for Mac (version 29.0; SPSS Inc., Chicago, IL, USA). The intraclass correlation coefficient (ICC) was calculated to determine the magnitude of the inter- and intra-investigator measurement errors. The median of all measurements was used for further statistical analyses. Paired t-tests were used to compare the mean volume of the structures on the impacted (IMxC and IMx) and non-impacted halves (NIMxC and NIMx) of the maxilla.