Identifying the location and value of the smallest airway dimension can be useful in screening and planning treatment for patients with obstructive sleep apnea. Our objectives in this study were to (1) objectively identify the vertical location and value of the minimum sagittal linear dimension (MSLD) on 2-dimensional reconstructed lateral cephalograms (RLCs), (2) compare the location and value of the MSLD on RLCs with the vertical location and sagittal dimension of the minimum cross-sectional area (MCSA), and (3) investigate the association between the MSLD on RLCs and both the MCSA and the airway volume.
Cone-beam computed tomography (CBCT) scans of 91 patients, in 3 age groups (<20, 20-40, and >40 years), were used to perform 3-dimensional assessments of the upper airway and reconstruct lateral cephalograms. Airway volume, MCSA, vertical level, and sagittal dimension of MCSA on the CBCT scans were obtained using Dolphin 3D software (version 11.7; Dolphin Imaging, Chatsworth, Calif). Customized software was used to objectively obtain the location and value of the MSLD of the airway on RLCs.
In all age groups, correlation tests showed significant correlations between the MSLD on RLCs and both the MCSA ( r s ≥0.59; P <0.001) and the airway volume ( r s ≥0.37; P <0.05). Additionally, there were significant correlations between the vertical location of the MSLD and the vertical location of the MCSA ( r s ≥0.41; P <0.05) and between the MSLD and the sagittal dimension of the MCSA ( r ≥0.61; P <0.001). Bland-Altman plots for the MSLD and the sagittal dimension of the MCSA showed much narrower 95% limits of agreement compared with the Bland-Altman plots for the vertical locations of the MSLD and the MCSA.
Two-dimensional images may be used as a screening tool and to identify the sagittal dimension of the smallest airway dimension. However, comprehensive assessment of airway characteristics is better achieved with CBCT-based 3-dimensional evaluation.
We made omputerized measurements on reconstructed lateral cephalograms (RLCs) using MATLAB.
Value and location of the smallest airway dimensions were objectively identified on RLCs.
2D images are more reliable at finding the sagittal dimension of the minimum cross-sectional area compared with its vertical location.
2D images can provide valuable information in treatment planning of OSA patients.
Upper airway assessment is an important component of orthodontic clinical examination because of its influence on growth and craniofacial morphology. This is in addition to its importance during orthodontic and orthognathic surgical treatment planning.
Lateral cephalograms are routinely obtained by orthodontists; for that reason, orthodontists are in a unique position to aid in the early evaluation of airway size and morphology, including the risk for obstructive sleep apnea (OSA). To evaluate the sagittal airway dimension on a lateral cephalogram, various measurements have been proposed: eg, posterior airway space, retropalatal airway space, superior posterior airway space, middle airway space, and inferior airway space. Yet there is no consensus as to which of these measurements is the best approximation of the vertical location of the true minimum sagittal linear dimension (MSLD).
In addition to the conventional method to obtain lateral cephalograms, lateral cephalograms can also be obtained by reconstruction from a cone-beam computed tomography (CBCT) scan, producing reconstructed lateral cephalograms (RLCs). Studies comparing the accuracy of linear and angular measurements using conventional lateral cephalograms and RLCs found no significant differences between them. Nevertheless, both lateral cephalograms remain a 2-dimensional (2D) representation of a 3-dimensional (3D) morphology.
To analyze the upper airway 3 dimensionally, the clinician can use magnetic resonance imaging (MRI), medical computed tomography (CT), or CBCT, which is the current method of choice. An important advantage of 3D imaging is to provide information about the exact location and nature of the airway obstruction in patients with OSA; this is crucial for obtaining an effective treatment plan even in the presence of a sleep study. However, the literature lacks studies aimed to objectively identify the exact location and value of the MSLD in 2D images and to compare them with 3D measurements.
The objectives in this study were to (1) objectively identify the vertical location and the value of the MSLD on 2D RLCs, (2) compare the location and value of the MSLD on RLCs with the vertical location and sagittal dimension of the minimum cross-sectional area (MCSA) on CBCT scans, and (3) investigate the association between the MSLD on RLCs and both the MCSA and the airway volume on CBCT scans.
Material and methods
The sample of this retrospective study consisted of pretreatment CBCT scans of orthodontic patients. The inclusion criteria were (1) age range of 10 to 80 years, (2) Class I skeletal pattern with an ANB angle of 0° to 5°, and (3) CBCT scans showing the entire fourth cervical vertebra. Patients with high mandibular plane angles (FMA >30° or SN-MP >40°), posterior crossbites, previous orthopedic treatment, history of tonsillectomies or adenoidectomies, syndromes, craniofacial anomalies, history of OSA, and mouth breathing were excluded from the study. The University of Illinois at Chicago institutional review board reviewed and approved the study. A total of 1200 medical records were reviewed, and 91 met the eligibility criteria and were analyzed. The sample was divided into 3 groups based on age: under 20, 21 to 40, and over 40 years, with 30, 30, and 31 subjects per group, respectively.
CBCT scans were taken as part of routine initial records in 2 private orthodontic offices of the same orthodontist (E.Y.L.). The CBCT devices came from the same manufacturer (I-CAT; Imaging Sciences International, Hatfield, Pa). The scans were obtained using the following settings: 120 kV, 18.54 mAs, 16 × 13 cm field of view, 0.3 mm 3 voxel, and scanning time of 4.8 to 8.9 seconds. The patients were seated, and the scans were taken in an upright position. Attempts were made to orient in natural head position using a mirror and a laser beam light and by having the Frankfort horizontal plane parallel to the floor. The patients were also instructed to breathe lightly through their noses, avoid deglutition, position the mandible in maximum intercuspation, and rest the tongue in a relaxed position touching the anterior teeth.
The CBCT scans were obtained in DICOM format and uploaded into Dolphin 3D software (version 11.7; Dolphin Imaging, Chatsworth, Calif). Dolphin 3D was used to orient and analyze the 3D images and to obtain RLCs.To orient the scans, guidelines proposed by Guijarro-Martinez and Swennen were adapted. The skull was oriented to the Frankfort horizontal by defining the axial plane by 3 points: right porion, right orbitale, and left orbitale, and checked in 3 reference views. On the right sagittal view, the horizontal reference line was fixed through porion and right orbitale ( Fig 1 , A ). On the frontal view, the horizontal reference line was fixed through the right and left orbitales, and the vertical reference line was set through nasion and the anterior nasal spine ( Fig 1 , B ). On the transverse view with the skull facing down, the vertical reference line was fixed through crista galli and basion ( Fig 1 , C ).
After skull orientation, Dolphin 3D was used to calculate the oropharyngeal airway volume, the MCSA, the sagittal dimension of the MCSA ( Fig 1 , D ), and the vertical location of the MCSA measured from the level of the posterior nasal spine (PNS) in a plane perpendicular to the Frankfort horizontal ( Fig 1 , E ). The line measuring the sagittal dimension of the MCSA was centered in the MCSA transversely with care to ensure that this line was perpendicular to the coronal plane. Dolphin 3D uses a semiautomatic segmentation approach and an interactive threshold technique (1 threshold value for the whole scan that is different from patient to patient). To segment the airway, the limits of the oropharynx were outlined ( Table I ), and a “seed” point was placed in the airway region. Additional “seed” points were placed as needed. The threshold was manually adjusted using a sliding scale function to maximize airway volume and minimize noise. The operator (A.H.A.) evaluated the airway slices in the sagittal, frontal, and transverse views to ensure appropriate threshold sensitivity. To limit extreme threshold variability, the threshold value was predetermined to be between 40 and 80. The threshold value for each CBCT scan was recorded.
|Anterior||Plane, perpendicular to FH, passing through PNS|
|Posterior||Plane, perpendicular to FH passing through the superior posterior boundary of the odontoid process of C2|
|Upper||Plane, parallel to FH passing through PNS|
|Lower||Plane parallel to FH at the level of the tip of epiglottis|
|Lateral||Plane, perpendicular to FH through the lateral walls of medial pterygoid plates|
Obtaining a 2D image was achieved by reconstructing a lateral cephalogram from the 3D scan. To account for the magnification error in conventional lateral cephalograms, the “perspective” projection function was used. RLCs were uploaded to the Dolphin imaging software (version 11.0, Dolphin Imaging, Chartsworth, CA) to trace the oropharyngeal airway (tracing the inner margins of the surrounding borders) and to draw the palatal plane. All RLCs had a bar on the left side indicating 100 mm. The limits of this bar were identified in the software using “Ruler Point 1” and “Ruler Point 2” after setting the ruler option to 100 mm in the setting section. This was done to ensure calibration and standardization ( Fig 2 , A ).
Objective computerized measurements of the location and values of the MSLD were obtained using MATLAB mathematical software (MATLAB 8.4 (2014b); MathWorks, Natick, Mass) written by a software engineer specifically for this study. The written code assumed 100-mm ruler calibration, 1650 × 1275 pixels JPEG images, and a resolution unit of 2. Adobe Photoshop CS5 (version 12.0.1; Adobe Systems, San Jose, Calif) was used to ensure that the required image size and resolution were appropriate before the images were uploaded to MATLAB. Once in MATLAB, the image was calibrated by identifying Ruler Points 1 and 2, PNS was identified, and the posterior and anterior borders of the oropharyngeal airway were retraced. The software calculated the MSLD by identifying the smallest distance between the anterior and posterior borders of the airway. The software also calculated the vertical distance of the MSLD to PNS in a plane perpendicular to the Frankfort horizontal. Both measurements were generated in millimeters up to the fourth decimal place ( Fig 2 , B ).
Measurements for 15 scans, both 3D and 2D, were performed by the principal investigator (A.H.A.) and then repeated 15 days later by the principal investigator and a second assessor (A.I.M.) to determine the intra-rater and inter-rater reliabilities. The intraclass correlation coefficient was used to test reliability. The distribution of the raw data was investigated using the Shapiro-Wilk test of normality. Correlation coefficient tests were used to test the association between the vertical location and the value of the MSLD on the RLCs and the sagittal dimension and the vertical location of the MCSA on the CBCT scans. Correlation coefficient tests were also used to investigate the association between the MSLD on the RLCs and both the MCSA and the volumetric airway measurements on the CBCT scans. Parametric and nonparametric tests, Pearson and Spearman, were used as needed. Statistical significance was noted at α = .05. When evaluating the strength of correlation, the following classification was used: strong if 0.7 < Ι r Ι ≤ 1.0, moderate if 0.4 ≤ Ι r Ι ≤ 0.7, and weak if 0.2 < Ι r Ι < 0.4.
To further investigate the agreement between the vertical location and value of the MSLD on the RLCs and the vertical location and the sagittal dimension of the MCSA on the CBCT scans, Bland-Altman plots were used. Visual inspections of the plots, including the bias and the 95% limits of agreement, were performed to evaluate the reliability of the 2D measurements relative to the 3D measurements. Data analysis was done with SPSS Statistics for Windows (version 22.0; IBM, Armonk, NY).
The ages of the subjects ranged from 11.1 to 75.8 years, with a mean age of 31.48 ± 17.55 years. Of the 91 subjects, 45 (49.45%) were male and 46 (50.55%) were female, with equal sex representations in the 3 age groups. The threshold values for the CBCT scans ranged from 40 to 60 with a mean of 47.22 ± 5.53. Intraclass correlations showed strong correlations (r >0.80) indicating reliability. Table II shows descriptive statistics of the 2D and 3D measurements.
|Age group 1 (<20 y)|
|Oropharyngeal volume (mm³)||30||5762.9||30887.5||15108.9||6089.7|
|MCSA (mm 2 )||30||86.90||586.33||209.66||125.97|
|Sagittal dimension of MCSA (mm)||30||2.20||18.80||9.59||3.99|
|Vertical location of MCSA (mm)||30||3.70||49.00||29.98||11.40|
|Vertical location of MSLD (mm)||30||1.72||53.09||27.92||13.23|
|Sagittal dimension of MSLD (mm)||30||4.25||16.29||8.59||3.11|
|Age group 2 (20-40 y)|
|Oropharyngeal volume (mm³)||30||6713.1||26491.5||15782.0||5631.7|
|MCSA (mm 2 )||30||48.70||364.40||181.49||86.39|
|Sagittal dimension of MCSA (mm)||30||4.20||15.50||8.60||3.26|
|Vertical location of MCSA (mm)||30||23.6||74.40||34.94||11.54|
|Vertical location of MSLD (mm)||30||16.86||56.67||32.83||11.27|
|Sagittal dimension of MSLD (mm)||30||2.58||11.16||6.59||2.33|
|Age group 3 (>40 y)|
|Oropharyngeal volume (mm³)||31||6006.2||101916.0||19430.2||17154.7|
|MCSA (mm 2 )||31||54.60||503.70||175.54||107.81|
|Sagittal dimension of MCSA (mm)||31||3.10||14.20||7.37||3.12|
|Vertical location of MCSA (mm)||31||6.9||62.20||36.37||11.13|
|Vertical location of MSLD (mm)||31||16.17||69.78||36.11||13.59|
|Sagittal dimension of MSLD (mm)||31||3.02||13.05||6.22||2.52|