1. Introduction
Since its introduction to dentistry in 1998, cone beam computed tomographic (CBCT) imaging has become an essential radiographic technique for orthodontic diagnosis and treatment planning, particularly in situations requiring the evaluation of complex anatomical relationships within the maxillofacial skeleton [
1,
2,
3]. The adoption of CBCT has been driven by its advantages in providing three-dimensional (3D) imaging over traditional two-dimensional radiography [
4,
5,
6,
7,
8].
It is recognized that CBCT imaging in orthodontics and pediatric dentistry exposes children to three to five times more radiation than adults during the same scanning session. However, patients with severe malocclusions and dentofacial deformities often require combined orthodontic treatment and maxillofacial surgery [
9,
10]. In these cases, CBCT integrated with virtual surgical planning allows surgeons to simulate different treatment options and predict the postoperative facial profile [
11,
12].
Previous research has used comparisons of 3D visualizations of cranial models and configurations of landmarks on the skull and face for surgical planning [
13,
14]. Basic forms of comparison, such as the evaluation of metric features and the calculation of distances between landmarks on 3D models, help determine the relative agreement or similarity between external structures or identify significant differences [
15]. More complex data processing, such as landmark-based analysis through geometric morphometrics or surface-based processing, facilitates the exploration and representation of spatial relationships between craniofacial components [
16,
17].
Geometric morphometrics processes discrete points (expressed in
X,
Y,
Z Cartesian coordinates) using General Procrustes Analysis (GPA) for alignment. Once aligned, these coordinates are used to compute the Procrustes distance (PD), a metric that indicates the dissimilarity or similarity between two or more shapes [
18,
19]. Conversely, comparing two or more 3D surfaces [
20] often requires more computationally intensive registration algorithms such as Iterative Closest Point (ICP), with differences expressed as nearest point-to-point distances or cumulative measures such as the Hausdorff distance.
The size and shape of the facial skeleton are fundamental determinants of facial appearance. Small asymmetries in skeletal morphology are noticeable, but the changes resulting from orthodontic therapy or maxillofacial surgery are usually much more significant. Soft tissues, in particular, can either mask or accentuate deformities, but they do not always correspond directly to the underlying skeleton.
Given these circumstances, orthodontics and maxillofacial surgery rely on long-term imaging techniques to monitor both soft and hard head tissues, primarily the profile, skeleton, and dentition, for therapy planning and orthodontic and orthognathic treatment. These techniques can also be applied to anthropology and forensic dentistry [
21,
22]. The 3D treatment plan for orthodontic patients, which includes both soft and hard tissue considerations along with orthognathic surgery, allows for an accurate prognosis of the outcome when the skeletal basis of the face is available in two or more three-dimensional image sets.
The aim of our study was to compare 3D visualizations of skull and facial models before and after surgery, to assess the degree of agreement or similarity, and to identify any morphological differences [
23,
24]. All of these methods use advanced sensor systems, traditional signal and image processing tools, and computational intelligence methods to support the needs of 3D CBCT models in dentistry [
25,
26]. The new 3D technology proposed in our research is based on CBCT recordings taken before and after orthodontic therapy, offering advantages such as (i) the accurate long-term planning of future maxillofacial surgery over several years and (ii) the ability to visualize future positions of the soft and hard tissues of the face in 3D space.
2. Materials and Methods
The study consisted of three separate parts, each involving a group of orthodontic patients from whom CBCT data and 3D facial scans were obtained before and after treatment. All participants provided informed consent before the study, were in good general health, and the research was conducted according to the guidelines of the American Dental Association (ADA) and the tenets of the Declaration of Helsinki. Data confidentiality and anonymity were strictly maintained. The study was approved by the Ethics Committee (EK-973IGA 1.12/11). Our methodology focused on comparing three-dimensional images of the skull or its components, supplemented by 3D facial scans to enhance the robustness of our findings.
The proposed methodology includes:
taking cranial measurements on 3D cranial models,
collecting discrete feature points on 3D cranial models,
using surface face models and discrete feature points collected on 3D face models.
Our study focused on adult orthodontic patients, both male and female. A 3D treatment plan was prepared for each participant, with surgical intervention as a core component of the therapy. Inclusion criteria included planned therapy, surgical intervention, comparison with a diagnostic standard, and desired outcomes (PICO). The study aimed to contribute to a methodology that can enhance quality of life and facial aesthetic appearance.
The initial set of patients included 45 individuals, selected based on age, general health condition, orthodontic defects, and surgery types. This set was later reduced to 30 individuals from whom 3D scans of both soft and hard facial tissues were obtained. Analyses of these scans included: (i) surgically altered cranial metrics, (ii) hard tissue landmark assessment, and (iii) soft tissue surface and landmark matching. No facial soft tissue scans were available for 15 patients in the first group, which was closely related to upper and lower jaw therapy, leading to their exclusion from this part of the study.
The surgically altered cranial metrics in the first part of the study consisted of 45 individuals who had undergone corrective surgery of the upper and lower jaw. Two medical imaging data sets were available for each patient, obtained by cone beam computed tomography (CBCT). The sample consisted of 19 women and 26 men. The mean age at the start of treatment was 22.21 years (females: 22.31, males: 22.13). The mean age after surgery was 23.01 years. The mean time between data sets was 0.85 years (approximately 10 months), ranging from 4.5 months to 2 years and 11 months. According to a Mann–Whitney test, no statistically significant differences in age or time interval were found between males and females at the 5% significance level.
For each image dataset, a set of 14 linear measurements were taken using the functionalities available in the Anatomage Invivo software (
Figure 1). Eight of these measurements were bilateral (comprising four pairs), while the remaining six were unilateral. The differences between the preoperative (
) and postoperative (
) data were analyzed using univariate statistical tests, both unpaired and paired. In addition, identical tests were performed separately for males and females. Furthermore, the signed relative preoperative to postoperative differences
were calculated using the formula:
These differences were evaluated using t-tests and Mann–Whitney tests concerning biological sex, the time interval between image datasets, and patient age at baseline. Finally, the overall differences were analyzed by multivariate analysis of variance with biological sex and time interval as factors. For bilateral linear distances, side differences were assessed using both unpaired and paired tests. Within-group variations according to biological sex and preoperative and postoperative status were also examined.
The landmark-based comparison of hard tissues in the second part of the study consisted of 30 individuals who had undergone corrective surgery of the upper and lower jaw as in the previous case. Two medical imaging data sets were available for each patient, obtained by cone beam computed tomography (CBCT). The sample consisted of 11 females and 19 males. The mean age at the start of treatment was 23.30 years—25.30 years for women and 22.14 years for men. A Mann–Whitney test showed no statistically significant differences in age or time interval between men and women at the 5% significance level.
For all participants, facial geometry was recorded using CBCT protocol. For each CBCT dataset, a set of 12 landmarks (or dx/sin, zy dx/sin, go dx/sin, al dx/sin, na, inc, pg, gn) was manually collected using the functionalities available in the Anatomage Invivo software (
Figure 2). The entire dataset was first registered using partial Procrustes analysis, an alternative to generalized Procrustes analysis that standardizes the variation in position and rotation while preserving the variation in size. Once registered, the dataset was analyzed for differences between the preoperative and postoperative groups using one-way PERMANOVA. In addition, the centroid size was extracted from the coordinates and compared between the preoperative and postoperative sets using both a paired Wilcox test and an unpaired Mann–Whitney test.
In the second step, the postoperative point configuration was fitted to the preoperative configuration using Ordinary Procrustes Analysis (OPA), which preserves the size variation. OPA involves the superimposition of two sets of points (as opposed to generalized Procrustes analysis, which typically involves three or more sets of points). The Procrustes distance was then calculated between the pairs of registered coordinates and was analyzed for differences based on the biological sex and age of the patient. In addition, Procrustes residuals were computed pairwise, tested for sex and age differences, and plotted on a mean point configuration to illustrate the magnitude and variance of changes associated with maxillofacial surgery.
Surface and landmark-based comparison of soft tissues in the third part of the study was used on 30 patients (the same as in the second part of this study) from the Department of Stomatology, Second Medical Faculty, with a mean age of 23.57 years, resulting in a total of 60 facial scans. Each patient was scanned using the Vectra M3 3D non-invasive, ambient light-based scanner (Canfield Scientific Inc., Fairfield, NJ, USA). All individuals were scanned from the front and were instructed to maintain a neutral facial expression throughout the process. Raw data were processed using Mirror PhotoTools software, version 7.4 (Canfield Scientific Inc.), which generated a surface model represented as a mesh of several hundred thousand triangles. All images were taken with the patients in their natural upright head position to ensure that the Frankfort plane was parallel to the floor.
A set of 26 features was manually collected for each facial model (
Figure 3). For each surgical patient, a pair of preoperative and postoperative 3D facial models were directly compared. Initially, the postoperative 3D model was designated as the primary (reference) model, while the preoperative model was designated as the secondary (comparison) model. Iterative closest point (ICP) registration was then performed to align the secondary model to the primary model, preserving the size differences. The closest point distances were then calculated for each vertex of the primary model. Statistical descriptors, including root mean square (RMS) and 75th percentile (PERC75), were extracted and used as measures of deviation. These distances were visualized using color-coded deviation maps.
In the next stage, the postoperative point configuration was aligned with the preoperative configuration using ordinary Procrustes analysis, while maintaining the size variation. The Procrustes distance was then calculated between the pairs of registered coordinates as a measure of morphological dissimilarity.
4. Discussion
CBCT is a precise imaging technique that provides a three-dimensional analysis of head and neck anatomy to assist clinicians in the planning of maxillofacial and orthognathic surgery, as well as postoperative assessment [
7,
8]. CBCT provides clinicians with multi-planar views of the skull, facilitating more accurate treatment planning, diagnosis and therapy while minimising X-ray radiation exposure compared to traditional computed tomography [
3,
8].
In this study, we used CBCT to identify changes associated with maxillofacial surgery. Our research highlighted that data derived from 3D CBCT scans, including measurements and spatial points, can overcome the limitations of data obtained from 2D orthopantomogram (OPG) images, such as lack of depth, real-world scale, and perspective distortion [
3,
8,
27]. This is consistent with a retrospective pilot study that supported the replacement of two-dimensional analysis with three-dimensional methods in orthognathic surgery planning [
11,
18]. We have shown that CBCT scans are compatible with both basic measurements and more sophisticated exploration and analysis methods, such as surface-based and landmark-based data processing.
Our manuscript confirms the fact that oral and maxillofacial surgery (OMFS) can transform the current concepts of reconstruction and rehabilitation of maxillofacial defects using Artificial Intelligence (AI) by providing various 3D techniques that can improve the accuracy, efficiency, and outcomes of surgical procedures [
28]. Computer-aided design and computer-aided manufacturing (CAD/CAM) uses three-dimensional (3D) CBCT and facial scans not only for treatment planning, but also for precise reconstruction of maxillofacial defects. All 3D systems were responsible for more precise function and aesthetics of the therapy [
29].
The results of our manuscript also demonstrated quantitative and qualitative differences in the power of each methodological approach. The initial study using cranial linear distances revealed significant postoperative changes, particularly an increase in mandibular width (go dx-go sin measurement). The subsequent landmark-based approach, while confirming this increase, effectively depicted both numerical and visual changes in the configuration of the maxilla and mandible.
The third study design focused on exploring 3D facial scans using landmark-based and surface-based approaches. Unlike the complex 3D cranial data, 3D facial scans, which take the form of an open-ended digital shell, are easier to process as polygonal models or point clouds. The surface-based approach allows postoperative changes to be investigated independently of preoperatively selected feature points or linear distances. This allows practitioners the freedom to examine changes without bias, and to view them both numerically and visually. For example, this approach has shown how regions such as the chin, cheeks, and nasal area have been affected by surgery and postoperative treatment.
The results in
Table 1,
Table 2,
Table 3,
Table 4,
Table 5 and
Table 6 confirm our new clinical strategy based on 3D planning and treatment for the precision, optimal control, and prediction of treatment outcomes. For all interventions, we propose 3D treatment planning, which is much more precise than 2D imaging, following the guidelines of the American Dental Association (ADA). Techniques and technologies are repeatable in orthodontic and maxillofacial treatment, but the individual approach for each patient is necessary.