The midsagittal plane (MSP) is the foundation for 3-dimensional (3D) cephalometric analyses. This article aimed to provide a protocol to construct a reliable and accurate MSP for 3D cephalometric craniofacial analysis.
Cone-beam computed tomography data of 16 adult patients without obvious bilateral asymmetry were collected. The model of the anterior cranial base and sphenoid bone was constructed, and the candidate MSP was determined on the basis of the symmetry of this model. Intraclass correlation coefficients were used to assess intra- and interexaminer reliability of the candidate MSP. To investigate the accuracy of this candidate MSP, we constructed a true plane of symmetry of craniomaxillofacial structure and a control plane on the basis of 3 cranial midline points. We then compared these with the candidate MSP.
This candidate MSP resembled the true plane of symmetry with all the mean absolute errors <1 mm, and all the absolute errors for the candidate MSP were significantly smaller than the control plane ( P ≤0.002). The relative intra- and interexaminer reliability for this candidate MSP was almost perfect (intraclass correlation coefficients >0.9).
The candidate MSP constructed using this method was thought to be reliable and accurate for 3D cephalometric analysis in patients without obvious cranial asymmetry.
We constructed a reliable and efficient model for the anterior cranial base and sphenoid bone.
We constructed a reliable and accurate midsagittal plane based on the model.
The midsagittal plane resembles the true symmetrical plane of the maxillofacial region.
The midsagittal plane is more accurate than a plane through 3 cranial midline points.
Three-dimensional (3D) cephalometric analysis provides the ability to overcome the limitations of 2-dimensional cephalometrics, such as image distortion and magnification. It also provides a comprehensive analysis of asymmetric structures and enables segmental movements using 3D digital operations and comparisons of preoperative and postoperative facial profiles. , As a baseline plane, the midsagittal plane (MSP) is the foundation for 3D cephalometric analyses. It affects the validity of skull orientation, landmark location, quantification measurement of asymmetry, and the superimposition of radiographs.
There have been several approaches to construct or extract the MSP in the literature. Some defined the MSP by appointing 3 points of central landmarks (sella turcica, nasion, basion, crista gali, incisive foramen, anterior and posterior nasal spine, etc.), , or midpoints of bilateral landmarks (anterior clinoid process, orbitale, porion, etc.), , , or a combination of those mentioned above. Some suggested determining the MSP as a plane passing through 2 midline points and perpendicular to a horizontal plane (eg, Frankfort horizontal plane) , , or passing through a midline point, bisecting and perpendicular to a horizontal line connecting the bilateral zygomaticofrontal sutures. In addition, certain studies proposed morphometric methods by Procrustes analysis. Damstra et al determined the MSP by partial ordinary Procrustes analysis with 4 pairs of bilateral landmarks around the supraorbital region as references. Shin et al constructed a symmetrical midsagittal reference plane by ordinary and generalized Procrustes analyses with various craniofacial region landmarks. All these approaches were based on a set of points extracted from the craniofacial region. Although simplifying the structure, they may deviate from the true plane of symmetry because of the deviation of points plotting and selection. It stands to reason that an ideal MSP would consider as many useful structure characteristics as possible. Recently, Zhang et al suggested determining the MSP by iterative closest point registration of the model of the anterior cranial base and its mirrored model. As a 3D model incorporates sufficient symmetrical characteristics, this method can provide more accurate and reliable cephalometric analysis.
The purposes of this study were to (1) provide a protocol to construct a 3D model of the relevant cranial base (Model CB ), (2) provide a method to determine an MSP based on the symmetry of this model (MSP CB ), and (3) evaluate and compare the accuracy and validity of MSP CB .
Material and methods
Cone-beam computed tomography (CBCT) images of 16 patients (6 men and 10 women) who visited orthodontic and orthognathic departments from August 2019 to September 2019 were collected retrospectively. The average age was 21.9 years (18-27 years). The inclusion criteria were patients without clinically detectable asymmetry, defined as <1 mm of menton and subnasale deviation from the facial midline. The exclusion criteria were cleft lip or palate, craniofacial syndrome, and facial trauma. Approval was obtained from the local institutional review board.
CBCT images were taken from the NewTom VGiMK4 scanner (Aperio, Milan, Italy). Image acquisition conditions were as follows: 110 kV tube voltage, 2 mA tube current, voxel size of 0.3 mm, field of view of 15 cm × 15 cm ×15 cm. The images were obtained in Digital Imaging and Communications in Medicine format and imported into a third-party software Mimics 19 (Mimics innovation suite; Materialise, Leuven, Belgium).
The radiographs were visually oriented to make bilateral orbital cavities at roughly the same horizontal position. The landmarks, 3D models, and MSPs used for analysis in this study are presented in Table I . All the landmarks were plotted on multiplanar views carefully, and N, ANS, A, B, Pog, Gn, and Me were further refined on the volume rendering view.
|S||Sella. The center of sella turcica from a sagittal view centered mediolaterally on sella turcica from the axial and coronal views|
|N||Nasion. The most anterior midpoint of the frontonasal suture|
|Ba||Basion. Middorsal point of the anterior margin of the foramen magnum|
|ANS||Anterior nasal spine. The most anterior point of the anterior nasal spine. If bilateral processes are visible, their midpoint is selected|
|A||The point at the deepest midline concavity on the maxilla between the anterior nasal spine and prosthion|
|B||The midpoint of the greatest concavity on the anterior border of the symphysis|
|Pog||Pogonion. The most anterior midpoints on the symphysis of the mandible|
|Gn||Gnathion. The lowest, most anterior midline point on the symphysis of the mandible|
|Me||Menton. The most inferior midpoint on the symphysis of the mandible|
|Model CB||The 3D model constructed on the basis of the anterior cranial base and sphenoid bone except for pterygoid processes|
|Model CMF||The 3D model constructed on the basis of the whole craniomaxillofacial skeleton|
|MSP CB||The MSP based on the symmetry of Model CB|
|MSP CMF||The MSP based on the symmetry of Model CMF|
|MSP S-N-Ba||The plane S-N-Ba, which is constructed through 3 landmarks S, N, and Ba|
Construction of Model CB
Three points were selected as the indicators of the bilateral and posterior boundary of the model. These points were the left and right intersection points of the greater wing of the sphenoid bone, the frontal bone, and the zygomatic bone (SFZ-L, SFZ-R), and the midpoint of the left and right apices of petrous pyramids (mid-APP) ( Fig 1 , A and B ). SFZ-L (or SFZ-R) was plotted in the fronto-sphenoid suture on the axial slice where the shape of the zygomatico-sphenoid process of the frontal bone presented a rounded triangle, and it was fine-tuned into the fronto-zygomatic suture on the coronal slice where the zygomatico-sphenoid process of the frontal bone was tapered ( Fig 1 , A ). A sphere was defined by clicking SFZ-L, SFZ-R, and mid-APP to determine a circle and then clicking any of these points to set this circle as the maximum cross-section of the sphere. The mask, which was the color-labeled region of interest, was calculated from the sphere ( Fig 1 , B ). On the middle sagittal slice, the part below the line connecting the frontonasal suture and the floor of the sphenoid sinus was erased with a lasso-type brush using the multiple slice editing tool, and then the setting was copied to all sagittal slices.
Consequently, the mask was restricted to a hemisphere ( Fig 1 , C ). A bone mask was created by selecting optimal thresholding gray values to segment the bone structure. The optimal thresholding gray values were determined individually for each patient and for different bony regions; the thin bone in the middle region usually corresponds to a lower thresholding value, whereas thick bone in the outer region corresponds to a higher value. If the artifacts of the bone mask were obvious, they were erased on the certain slices. Subsequently, the intersecting part between the bone mask and the hemisphere mask was calculated to acquire the final mask through a Boolean operation algorithm ( Fig 1 , D ). Finally, the 3D model of the relevant cranial base (Model CB ) was calculated from the final mask ( Fig 1 , E ). The detailed procedure is shown in Video 1 (available at www.ajodo.org ).
Model CB appropriately incorporates the stable and symmetrical structures of the anterior cranial base and most of the sphenoid bone (cribriform plate, crista gali, frontal crest, orbital plates of the frontal bone, anterior and posterior clinoid processes, lesser wings of the sphenoid, hypophysial fossa, and most of the greater wings of the sphenoid) ( Fig 1 , E ).
Construction of MSP CB
Model CB was copied and pasted into 3-matic software version 11 (Mimics innovation suite; Materialise, Leuven, Belgium). An S-N-Ba plane was created through the 3 points S, N, and Ba to serve as a preliminary symmetrical plane ( Fig 2 , A ). Subsequently, a mirrored model (Model CB ′) of Model CB was generated, and S-N-Ba was copied to acquire a new identical plane named S-N-Ba′ ( Fig 2 , B ). In this study, S-N-Ba was used to mark the spatial location of Model CB , whereas S-N-Ba′ marks the location of Model CB ′. The 2 models were aligned by global registration with enough iterations; wherein the settings were Model CB fixed, Model CB ′ movable, and S-N-Ba′ moving along with Model CB ′. Consequently, Model CB ′ translated and rotated to match the surface morphology of Model CB until no change was observed, during which the transform matrix of Model CB ′ was recorded, and S-N-Ba′ made the same transformation as Model CB ′ ( Fig 2 , C ). Because S-N-Ba and S-N-Ba′ represented the locations of the Model CB and Model CB ′, the symmetrical plane of the 2 planes was equal to the symmetrical plane of the composite of the 2 models. Because the composite of the 2 models was now fully matched and completely symmetrical, the symmetrical plane of the composite of the 2 models could reveal the symmetry to the greatest extent for both Model CB and Model CB ′. Therefore, MSP CB , the true symmetrical plane of Model CB , was generated as an average of planes S-N-Ba and S-N-Ba′ using the creating datum plane function ( Fig 2 , D ). The detailed procedure is shown in Video 2 (available at www.ajodo.org ).