Several imaging software programs with different tools are available for upper airway (UAW) analysis by means of cone beam computed tomography. Because of this wide variability, this study aimed to compare the reproducibility of two of the most used software programs on UAW segmentation, that is, Dolphin Imaging (Dolphin Imaging and Management Solutions, Chatsworth, Calif) and Mimics Research (Materialise, Leuven, Belgium).
The sample consisted of 50 scans of adult subjects with Class III malocclusion; pharyngeal volume and minimal cross-sectional area (mCSA) were assessed twice by 2 experienced evaluators using both software programs. Intra- and intersoftware and/or evaluator agreement were calculated using the intraclass correlation coefficient (ICC). Bland-Altman analysis was used to visualize the intersoftware and interevaluator agreement.
Using Mimics Research and Dolphin Imaging, the respective mean values of volume (cm 3 ) were 27.1 ± 8.4 and 24.7 ± 8.1 ( P < 0.05) and of mCSA (mm 2 ) were 167.8 ± 95.5 and 176.5 ± 102.3 ( P < 0.05). Although differences between software packages were observed on both the analyzed variables, the results showed a high intersoftware agreement, with ICC of 0.87 (volume) and 0.97 (mCSA) and mean bias of 24.6 (volume) and −8.8 (mCSA). A high intrasoftware agreement was also observed, with values varying from 0.83 to 0.99. Excellent intra- and interevaluator agreement was also obtained, with ICC values from 0.93 to 0.99.
Dolphin Imaging and Mimics Research individually provide highly reproducible results, with clinically acceptable agreement between them for UAW segmentation, providing consistent values for volume and mCSA.
Comparison of 2 of the most used software programs for upper airway assessment.
Significant sample size (n = 50) composed using strict inclusion criteria.
Results show that both software programs are reliable for airway assessment.
Three-dimensional (3D) evaluation of the upper airway (UAW) by means of cone beam computed tomography (CBCT) has been the subject of several studies in the recent past, being largely used in surgical , and orthodontic outcomes assessment, airway analysis related to sleep apnea, , as well as for the characterization of the UAW in different craniofacial morphologies. , In the past, 2-dimensional (2D) cephalograms were the most used method for UAW volumetric evaluation. However, the overlap of 3D structures resulting in 2D images has been considered as one of the main limitations of the method.
Since its development, the CBCT has become widely used for the assessment of volumetric data, providing high quality images with low-dose radiation. , , , Many imaging software packages have been developed since that time, presenting different graphical user interfaces, tools, and commands that make imaging analysis relatively simple. Nevertheless, the accuracy and reproducibility of airway segmentation in the literature is controversial. Some studies consider the method to be reliable, , , whereas others show evidence of relatively unsatisfactory results. Accordingly, a recent study by Tsolakis et al has shown a high correlation between 3D tomographic assessment of the UAW and acoustic rhinometry and pharyngometry.
Considering commercial imaging software packages, Mimics Research (Materialise, Leuven, Belgium) and Dolphin Imaging (Dolphin Imaging and Management Solutions, Chatsworth, Calif) are among the most popular in clinical practice as well as in research. Recently, Ghoneima and Kula demonstrated a great accuracy and reliability of UAW analysis on CBCTs using Dolphin. In addition, Weissheimer et al showed several advantages for both imaging programs, highlighting the wide use of Mimics Research in biomedical engineering and reporting that Dolphin Imaging is considered very popular among clinicians.
Although a previous study by our group has demonstrated similar pharyngeal dimensions comparing a commercial software with an open source software, it is still relevant to question the reliability among imaging software programs. Thus, considering the variety of studies characterizing the UAW and the different imaging softwares available, the present study aimed to evaluate the inter- and intrasoftware agreement as well as the reproducibility of Mimics Research and Dolphin Imaging softwares assessing the volume and the minimal cross-sectional area (mCSA) of the UAW in adult subjects by means of CBCT.
Material and methods
This study was approved by the Institutional Review Board of the University of São Paulo, Bauru, São Paulo, Brazil. In an initial search, approximately 2000 CBCT scans were assessed. From that, 421 scans presented proper field of view (>13 cm height) necessary to completely assess the length of the UAW. The inclusion criteria were the presence of anatomic structures such as the pituitary saddle, hyoid bone, and C4 vertebra covering the UAW extension.
Considering that the complexity of the airway morphology can affect the final segmentation volume, only subjects with the same skeletal pattern (Class III malocclusion) and ethnic group were included in the sample. The following additional inclusion criteria were considered: CBCT scans of adult subjects (mean age, 22.8 years; age range, 18-36 years) and images obtained using i-CAT scanner (KaVo Dental, Berlin, Germany). Scans of subjects with a body mass index of ≥30 kg/m 2 and with hypertrophic tonsils or adenoids were excluded from the sample. This search resulted in 50 scans, obtained from a maxillofacial surgery practice.
All images were obtained for surgical planning purposes through an i-CAT Next Generation scanner (ISI-i-CAT Imaging System, beam cone, Next Generation i-CAT; Imaging Sciences International, Hatfield, Pa) with the following specifications: field of view of at least 16 × 13 cm, exposure time of 26.9 seconds, 120 kV, 37 mA, and a minimal resolution of 0.25 voxels. These images were imported as a Digital Imaging and Communications in Medicine–formatted file and displayed using Mimics Research (version 17.0) and Dolphin Imaging (version 11.7).
To assess pharyngeal dimension in Mimics Research, a mask was created, with a threshold range of −1024 to −400, consistent with the density of air. This tool allows the pharynges to be filled to distinguish them from other structures such as soft and hard tissues. The semiautomatic segmentation was performed by removing structures of noninterest for UAW analysis and adding areas that could not be selected by the threshold, in the coronal, sagittal, and axial axes. The superior pharyngeal limit was defined as the most inferoposterior point of the inferior turbinate, and the inferior pharyngeal limit was the anterior limit of the fourth cervical vertebra ( Fig 1 , A ). After selecting the area of interest, the software creates the 3D reconstruction, which is smoothed and compatible with a colored pharyngeal airway ( Fig 1 , B ). To determine mCSA, Mimics Research calculates the area of each axial slice, and the operator selects the smallest one ( Fig 1 , C ).
To assess the pharyngeal dimensions using Dolphin Imaging software, an area of interest was created. From the midsagittal plane, 4 anatomic landmarks were located and marked: Ba (basion), S’ (anterosuperior edge of the pituitary saddle), C4 (anterior limit of the fourth cervical vertebra), and H (anterior limit of the hyoid bone), creating a polygon that encompassed the pharyngeal area ( Fig 2 , A ). Subsequently, the airway contained in this polygon was marked with a seed point tool, which automatically colored the area of interest. Subsequently, the software generated a 3D image of the UAW in which the numerical values for volume, expressed in cm 3 ( Fig 2 , B ), and mCSA, expressed in mm 2 ( Fig 2 , C ), were obtained.
Airway assessment was performed twice by 2 experienced investigators. Evaluator no. 2 assessed the entire sample to verify the interevaluator reproducibility. The second analysis was performed 30 days later. Intra- and intersoftware and intra- and interevaluator agreement were calculated using the intraclass correlation coefficient (ICC), using the following score: ICC <0.50, poor agreement; ICC = 0.50-0.75, moderate agreement; and ICC >0.75, high agreement. Furthermore, Bland-Altman analysis was performed to determine and visualize the intersoftware and interevaluator agreement ; the mean bias, 95% confidence intervals, and 95% limits of agreement were calculated. Student t test was used to assess differences between the groups. P values <0.05 were considered statistically significant.
Considering that both variables assessed were normally distributed, the results were expressed as means ± standard deviations. In addition, considering the excellent reproducibility found between the evaluators (ICC = 0.93-0.99), the mean values of the most experienced evaluator were used for major intra- and intersoftware analyses.
Mean values of volume (cm 3 ) on Mimics Research and Dolphin Imaging were 27.1 ± 8.4 and 24.7 ± 8.1, respectively. This difference was statistically significant ( P < 0.05) because volumes obtained from Mimics Research were as high as 9% when compared with Dolphin Imaging. Mean values of mCSA (mm 2 ) were 167.8 ± 95.5 and 176.5 ± 102.3 for Mimics Research and Dolphin Imaging, respectively. The difference was statistically significant ( P < 0.05). However, at this time, Dolphin Imaging provided greater values of area (5%) when compared with Mimics Research. The means and standard deviations are provided in Table I .