Abstract
The accuracy of 3D reconstructions of the craniomaxillofacial region using cone beam computed tomography (CBCT) is important for the morphological evaluation of specific anatomical structures. Moreover, an accurate segmentation process is fundamental for the physical reconstruction of the anatomy (3D printing) when a preliminary simulation of the therapy is required. In this regard, the objective of this study is to evaluate the accuracy of four different types of software for the semiautomatic segmentation of the mandibular jaw compared to manual segmentation, used as a gold standard. Twenty cone beam computed tomography (CBCT) with a manual approach (Mimics) and a semi-automatic approach (Invesalius, ITK-Snap, Dolphin 3D, Slicer 3D) were selected for the segmentation of the mandible in the present study. The accuracy of semi-automatic segmentation was evaluated: (1) by comparing the mandibular volumes obtained with semi-automatic 3D rendering and manual segmentation and (2) by deviation analysis between the two mandibular models. An analysis of variance (ANOVA) was used to evaluate differences in mandibular volumetric recordings and for a deviation analysis among the different software types used. Linear regression was also performed between manual and semi-automatic methods. No significant differences were found in the total volumes among the obtained 3D mandibular models (Mimics = 40.85 cm3, ITK-Snap = 40.81 cm3, Invesalius = 40.04 cm3, Dolphin 3D = 42.03 cm3, Slicer 3D = 40.58 cm3). High correlations were found between the semi-automatic segmentation and manual segmentation approach, with R coefficients ranging from 0,960 to 0,992. According to the deviation analysis, the mandibular models obtained with ITK-Snap showed the highest matching percentage (Tolerance A = 88.44%, Tolerance B = 97.30%), while those obtained with Dolphin 3D showed the lowest matching percentage (Tolerance A = 60.01%, Tolerance B = 87.76%) (p < 0.05). Colour-coded maps showed that the area of greatest mismatch between semi-automatic and manual segmentation was the condylar region and the region proximate to the dental roots. Despite the fact that the semi-automatic segmentation of the mandible showed, in general, high reliability and high correlation with the manual segmentation, caution should be taken when evaluating the morphological and dimensional characteristics of the condyles either on CBCT-derived digital models or physical models (3D printing).
Highlights
Cone beam computed tomography (CBCT) represents the three-dimensional (3D) imaging method of choice in oral and maxillofacial fields [1,2]
The null hypothesis was the absence of significant differences in both the accuracy and usability of semi-automatic segmentation software compared to manual segmentation
Small differences were found in the volumetric measurements of mandibular 3D models obtained with different software, ranging from 40.85 cm3 (Mimics) to 42.03 cm3 (Dolphin 3D); such differences were not statistically significant, as reported by the one-way analysis of variance (ANOVA)
Summary
Cone beam computed tomography (CBCT) represents the three-dimensional (3D) imaging method of choice in oral and maxillofacial fields [1,2]. Presents a higher resolution and a higher contrast-to-noise-ratio [3,4], CBCT scans offer adequate images for diagnostic purposes, at a lower risk of radiation exposure for patients [5,6,7]. In this respect, a 3D reconstruction of the craniomaxillofacial region using CBCT is employed in the diagnosis and planning of orthodontic treatment [8], maxillofacial surgery [9] and oral implantology [10], often involving the transformation of a physical object (3D printing). The software automatically selects the voxel of the interested region and excludes the surrounding structures
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
