Background
Facial aesthetics are always of general interest. One of the important reasons patients seek orthodontic care is to improve their facial attractiveness. Orthodontic treatment targets the dentition and the maxillomandibular relationships to create a considerable impact on facial esthetics. Orthodontic tooth movement and alveolar bone remodeling can cause facial morphological changes, which are closely related to aesthetic perception, as they interact and collectively determine an individual's facial aesthetic evaluation [
1,
2]. Previous research findings indicate that the changes of lip position after orthodontic treatment (Ls-SnPog', Li-SnPog', and Li-PrnPog') exhibited notable quadratic correlations with the assessment of facial attractiveness in both frontal and lateral profiles [
3,
4]. Furthermore, a study has revealed that the mentolabial sulcus angle increased in the incisor tipping group, whereas it decreased in the incisor translation group, leading to variations in facial aesthetics [
5]. However, previous studies primarily focused on the changes in profile and the lower third of the face, encompassing alterations in teeth and jawbone resulting from orthodontic treatment [
1,
2,
6‐
8], while studies investigating overall facial aesthetic changes have been relatively scarce. In clinical practice, it has been observed that improvements in the profile do not always correlate positively with overall facial aesthetics. Some patients, despite undergoing orthodontic treatment that altered their facial profile through means such as tooth extraction and orthodontic correction, did not experience significant enhancement in overall facial aesthetics, and even encountered aesthetic losses [
7,
9]. Therefore, studies based on overall facial aesthetics are urgently needed.
Despite some studies have been analyzing facial photographs since 1933 [
10], cephalometric analysis is still an important basis for developing orthodontic treatment [
11]. Comprehensive assessment of facial aesthetics requires consideration not only of changes in dental and skeletal parameters but also the integration of the patient's overall facial morphology and baseline characteristics [
7]. From an aesthetic point of view, facial soft tissues are more judgmental. However, there is a lack of both methods and metrics for soft tissue measurement and analysis relative to cephalometric measurements. The more commonly used clinically are still E-line [
12], angle of facial convexity [
13], and nasolabial angle [
14]. With the rocketing progress of smart devices in recent years, facial morphological changes before and after orthodontic treatment can be easily evaluated even by smartphone-based facial scanning that could be a viable tool for facially driven orthodontics [
15]. Hence, the study of measurement and analysis of facial photographs can help in designing valuable analytical tools.
The aim of this study was to investigate the impact of orthodontic treatment induced changes in facial morphology on facial aesthetics enhancement and analyze the key factors involved, which is crucial for optimizing treatment plans. In this study, we utilized pre- and post-orthodontic treatment photographs of patients, combined with expert evaluations, to assess facial aesthetics from the frontal and profile views, as well as overall assessment. We identified two important clinical facial features in orthodontic treatment that are associated with facial aesthetics. Our comprehensive understanding of the aesthetic outcomes related to orthodontic treatment can assist in optimizing treatment planning, enhancing patient satisfaction, and advancing the field of orthodontics and providing more accurate and comprehensive guidance for clinical practice and aesthetic evaluation.
Methods
Study design and sample recruitment
This study is retrospective research. A total of 73 eligible patients were recruited from the Department of Orthodontics, Ninth People's Hospital, Shanghai Jiao Tong University School of Medicine, who underwent orthodontic treatment between January 1st, 2019, and December 31st, 2020. For each participant, we collected frontal, 45-degree left and right, and 90-degree profile facial photographs before and after orthodontic treatment.
Inclusion criteria encompassed:
-
(1) Participants of any gender who had undergone at least 2 years of orthodontic treatment and possessed complete data information.
-
(2) Participants underwent orthodontic treatment throughout the entire process under the supervision of the Chief Physician, which represents a relatively higher level of orthodontic treatment expertise.
Exclusion criteria comprised:
(1) Participants who had undergone plastic surgery during the interval between the two orthodontic photographs or (2) had a history of maxillofacial trauma during this period.
This study received ethical approval from the Ethics Committee of Shanghai Ninth People's Hospital, Shanghai Jiao Tong University School of Medicine (Approval No.: SH9H-2021-TK461-1) and was registered with the Chinese Clinical Trial Registry (Registration No.: CTR2100050216).
Facial photographs processing
All facial photographs were taken under standard conditions for treatment comparison. When taking facial photos, the patient should sit upright with both eyes looking straight ahead, the head parallel to the Frankfort horizontal plane (natural head position), habitual occlusion, and relaxed lips and facial muscles. For capturing frontal images: the camera should be strictly positioned at a fixed distance and location, aligned with the horizontal plane passing through the orbits and ears, and facing the midline of the face. For capturing profile images: the camera should be aligned with the ear canal, ensuring that the ears are not covered by hair. To facilitate a comprehensive evaluation of facial morphology, in addition to the aforementioned frontal and profile images, it is common to include photographs of the patient smiling and semi-profile (45°) images.
Due to the potential influence of the patient's clothing on subsequent evaluations, for photograph at each angle of the sample, we only cropped and retained the facial region to minimize the cofounding factors of clothing. We used the trichion point (tri) as the highest point, the gnathion point (gn) as the lowest point, and the contours of the face on both sides as the cropping boundaries for the left and right sides.
For each sample, we retained the frontal view photographs (Supplementary Fig.
1a), 90° right profile view photographs (Supplementary Fig.
1b), and overall view photographs (the combination of 90° right, 45° right, frontal, 45° left, and 90° left) (Supplementary Fig.
1c) for subsequent aesthetic ratings.
Facial aesthetic ratings (FAR)
We recruited 10 experienced orthodontists as evaluators to give the facial aesthetics ratings (FAR). The FAR were on a scale of 0 to 10, where 0 represented the least attractive and 10 represented the most attractive. To avoid instrument and time-related systematic errors, the experts independently rated the photographs on the same screen simultaneously. For each participant, three FAR were obtained for frontal, profile, and overall views, respectively, at a specific time point (before and after orthodontic treatment). During the rating process, experts were not allowed to discuss with each other.
To maintain a natural appearance of facial aesthetic, experts were not given standardized training before the evaluation. They were only asked to assess the photos based on their own experience and subjective judgment. For each view of the face, we randomly included 5% (7 photographs) as duplicates for quality control purposes, to assess the consistency and reliability of the experts' ratings.
We defined the global facial phenotypes and clinical features based on the facial landmarks, specifically:
A total of 48 facial landmarks on frontal photographs were manually placed by experienced orthodontists according to the definition of traditional anthropometric measurements, including 35 facial soft tissue landmarks and 13 skeletal landmarks (Supplementary Fig.
2a, Supplementary Table
1). A total of 31 facial landmarks (24 soft tissue landmarks and 7 skeletal landmarks) were placed on right profile photographs (Supplementary Fig.
2b, Supplementary Table
1). For each photograph, coordinates for each landmark were acquired. Generalized Procrustes analysis (GPA) was then performed on the group of facial landmarks to eliminate any differences in position, orientation, and size of shapes, resulting in normalized facial landmarks.
Subsequently, we used two approaches to extract the facial phenotypes.
(1) Regarding the global facial phenotypes, we applied a dimensionality reduction approach by conducting Principal Component Analysis (PCA) on the coordinates of the facial landmarks for frontal and profile views separately. PCA was performed using:
$$X\approx {U}_{k}{\Sigma }_{k}{V}_{k}^{T}$$
where
\({X}_{n\times p}\) is a matrix of normalized facial landmarks with
n samples and
p coordinates of landmarks,
k is the number of retained principal components (PCs);
\({\Sigma }_{k}\) is a diagonal matrix of the largest
k singular values; and the column vectors of
\({U}_{k}\) and
\({V}_{k}^{T}\) are the corresponding
k left and right singular vectors, where
\({U}_{k}\) stand for principal components and
\({V}_{k}^{T}\) stand for loadings. Here, we retained the first
k = 15 principal components (
\({U}_{k}\)) for further analysis to describe global facial shape variations associated with changes in facial aesthetics following orthodontic treatment.
(2) Regarding the clinical features, we extracted a total of 20 profile features and 17 frontal features, which were specifically selected based on their relevance to aesthetics or clinical orthodontic indicators. These features were derived from analyzing the proportions and angles of the facial landmarks (Supplementary Table 2). The purpose of extracting these features was to conduct a thorough analysis to determine which specific facial characteristics were correlated with aesthetic changes. By examining these selected features, we aimed to identify the key factors that contribute to changes in facial aesthetics following orthodontic treatment.
Average face generation
The facial average face was constructed using the triangulation method. Initially, we applied an affine transformation to align each photograph, eliminating differences in position, size, and angle. Then, we calculated the average coordinates of the marked landmarks and performed Delaunay triangulation. For each triangle obtained by segmenting each photograph, we computed the affine transformation to map it to the corresponding triangle in the average shape. Finally, we calculated the pixel average of all photographs for each triangle, resulting in the generation of the average face.
Data analysis
Association between FAR and global facial phenotypes
For both frontal and profile views, we performed multi-variate regression analysis to examine the association between the FAR and the facial principal components (
\({U}_{k}\)) obtained through dimensional reduction of the facial landmarks. The multi-variate regression model can be expressed as follows:
$$FAR \sim {\beta }_{0}^{g}+\sum_{i=1}^{k}{\beta }_{i}^{g}{U}_{i}+{\varepsilon }_{g}$$
where
\({\beta }_{i}^{g}\) denote the regression coefficients, which represent the influence of each facial principal component on the FAR;
\(\varepsilon\) stand for the error terms, capturing the unexplained variability in the FAR. The
P-values of multivariate linear regression were calculated using the F-statistic (two-tailed), which quantifying the significance of the relationship between FAR and facial principal component features. Additionally, we defined weighted sums of PC loadings multiplied by regression coefficients as facial aesthetic vector (FAV) to quantify the changes in facial morphology corresponding to varying FAR, the formula can be expressed as follows:
FAV= {V}_{k}{\Sigma }_{k}{\beta }_{k}
Association between FAR and clinical features
For each facial feature, we employed three methods to assess its influence on facial aesthetics.
-
(1) Assuming that the average value had higher FAR [
3,
4,
16], we performed regression analysis of the FAR with both the features and their squares to determine their correlation with FAR:
$$FAR \sim {\beta }_{0}^{c}+{\beta }_{1}^{c}c+{\beta }_{2}^{c}{c}^{2}+{\varepsilon }_{c}$$
where
c denotes the particular clinical features, the
P-values of were calculated using the F-statistic (two-tailed).
All the three tests above were conducted with a significance level of P-value < 0.01, and phenotypes showing significance in all three tests were considered more crucial in orthodontic treatment.
Multifactor analysis of clinical features' impact on FAR
We employed multivariate regression methods to analyze the correlation between overall FAR and age, gender, and the three important clinical features identified using the test above. This step aimed to identify the most significant facial features related to aesthetic, the formula can be expressed as follows, the
P-values for each clinical feature were calculated using T-statistic (two-tailed):
$$FAR={\beta }_{0}^{c}+\sum_{i=1}^{n=3}{\beta }_{i}^{c}{c}_{i}+ {\beta }_{age}{c}_{age}+{\beta }_{gender}{c}_{gender}+{\varepsilon }_{c}$$
Discussion
Orthodontics is the conventional treatment for dental and facial malocclusions, usually focusing more on changes in the profile and lower third of the face. In this study we focused on the quantified changes in facial aesthetic induced by orthodontic treatment. By collecting pre- and post-orthodontic facial photographs from a clinical cohort, facial features were systematically extracted using based on facial landmarks, and in combination with FAR given by experts, the impact of orthodontics on facial aesthetics was analyzed. The findings not only underscore the substantial impact of orthodontic treatments, particularly from profile perspectives, but also emphasize the pivotal role of specific clinical features like pg.sm.hori and pg.n.ls in facial aesthetics. These findings could help orthodontists to better understand and optimize the aesthetic outcomes of orthodontic treatment.
Our study utilized two types of phenotyping approaches based on facial landmarks. The global facial phenotypes involved all landmarks, which give a whole facial change with FAR increase. This allows us to receive intuitive feeling of the most important facial area with facial aesthetic. Here is lip area in both profile and frontal view. However, because lack of specific indicator, this approach is difficult to applicate to clinical practice. Thus, we collected 37 clinical features, and used three test to identify the key features which are both associated with facial aesthetic and orthodontic treatment. In results, pg.sm.hori and pg.n.ls were found to be the key factors. Although these two phenotyping approaches were designed for different purposes, their results were consistent and mutually supported each other, reinforcing their significance in understanding the relationship between orthodontics and facial aesthetics.
Our results showed that patients' FAR improved after orthodontic treatment, whether in profile, frontal, or overall view. We observed that orthodontic treatment led to more pronounced changes in the profile aspects, primarily in the lip and mandibular regions. After orthodontics, facial morphology changes included lip retraction and increased mandibular fullness. Conversely, orthodontic changes in the frontal aspect were relatively minor, but the main alterations also focused on the lip region, mainly manifesting as a slight reduction in lip height, consistent with previous studies and with the current clinical investigation [
17]. One possible explanation is that in our sample, the majority of patients have relatively symmetrical facial features. Further study is needed to explore the improvement of frontal facial aesthetics in other specific patients (such as those with laterognathia) undergoing orthodontic treatment. By considering tooth extraction as an evaluation factor, we discovered that the improvement in FAR was statistically significantly higher in the extraction group compared to the non-extraction group. Recent reflections on extraction orthodontics and concerns about "tooth extraction face" have led to increased patient anxiety [
18‐
20]. Our results show that tooth extraction is beneficial for the overall aesthetic evaluation of patients, consistent with previous research findings [
21]. Regarding of orthodontic treatment methods, although we observed greater aesthetic improvement in patients with lingual appliances, this may be attributed to that the majority of these patients underwent premolar extraction treatment (22/27, 81.5%), while fewer patients underwent premolar extraction with labial appliances (19/31) and invisible aligners (0/15). Further study with a larger sample size is needed to determine if there are any differences in aesthetic improvement among different types of orthodontic methods under similar conditions.
We summarized and analyzed previously reported orthodontic clinical features that are associated with facial aesthetics. Overall, similarly to the above conclusions, profile clinical features showed more statistically significant correlations with facial aesthetics in orthodontic patients. Furthermore, for profile aspects, although numerous metrics, including the upper and lower lip to E-line [
22], facial convexity [
23], have been correlated with aesthetic scores, only the phenotypes with strong correlations with both orthodontics and aesthetics were pg.sm.hori and pg.n.ls. These two phenotypes underwent statistically significant changes before and after orthodontics and were correlated with FAR. We found that the highest FAR for pg.n.ls was around 8°, which differs from previous research suggesting the optimal angle is 5.9° [
24]. We attribute this difference to variations in the patient population and suggest that the reference standard for this angle can be set within a range, rather than being a specific value. For frontal aspects, most phenotypes related to aesthetic scores underwent minimal changes before and after orthodontics. Specifically, sto.sm/sn.gn experienced changes and was correlated with FAR, indicating that the proportion of the lower lip and lower third of the face is a crucial factor in frontal aesthetics, consistent with previous literature [
25,
26]. Besides, clinical features like sto-sm/sn-gn, puR-puL/ex-en, alR-alL/ex-en et al., which are related to eye and nose, have a larger correlation with frontal FAR [
27,
28]. These features did not have a statistically significant change after orthodontic treatment, suggesting that clinical features more related to frontal facial aesthetics may not be improved through orthodontics. Other approaches such as plastic surgery may be needed to improve frontal facial aesthetics.
Moreover, our study found that there was a strong consensus among experts in assessing facial aesthetics, with the highest consistency observed in profile views, suggesting that orthodontists may share more consistent opinions on facial attractiveness in profile aspects due to systematic and traditional training as well as aesthetic perception education [
29]. However, the consistency in frontal views was lower, indicating that there may be more individual factors affecting aesthetic evaluations in the frontal view, such as the evaluator's gender, age, occupation, and preferences [
30,
31].
In addition, this study has some limitations. Firstly, being a retrospective study, the sample size is limited, and lack statistical power to detect some previously reported features related to orthodontic aesthetics. We have initiated a prospective study that included a larger sample size and grouped patients according to type of malocclusion, gender and age for more comprehensive research that provides a comprehensive picture of the relationship between orthodontics and facial aesthetics. Additionally, this study only recruited orthodontists for FAR, therefore, the facial aesthetic assessments in this study only could represent orthodontists. Some studies have indicated differences in aesthetic judgment criteria for facial aesthetics among orthodontists, patients, and the general population [
32,
33]. It would be informative to include ratings from plastic surgeons, patients, and the general individuals to investigate whether there are similar or different aesthetic perception among different groups. Furthermore, with the advancements and increasing availability of 3D imaging technologies, future studies could utilize 3D facial scans to obtain more precise data about the true facial morphology of subjects, and to fabricate customized appliances for patients who have undergone orthodontic treatments, which could be used to improve both the functional and aesthetic outcomes of treatment [
34].
Conclusion
In conclusion, this study utilized facial photographs of orthodontic patients and expert ratings to systematically investigate the impact of orthodontics on facial aesthetics. The results demonstrated that orthodontic treatment generally improves facial aesthetic, especially for profile aspect showing a larger improvement. Moreover, we explored the relationship between clinical features and facial aesthetics, discovering that profile features, especially pg.sm.hori and pg.n.ls, had the most significant correlations with facial aesthetics. These two phenotypes underwent substantial changes after orthodontics and were correlated with FAR. These findings provide essential references and guidance for the clinical application of orthodontic treatment and facial aesthetics research. While there are limitations to the study, by expanding the sample size, considering ratings from patients and the general population, and utilizing 3D facial scanning, we can further investigate the relationship between orthodontics and facial aesthetics, contributing to the development of facial aesthetics and enhancing the quality of life for individuals.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.