Population affinity and variation of sexual dimorphism in three-dimensional facial forms: comparisons between Turkish and Japanese populations

The study was approved by the Research Ethics Committee, Osaka University Dental Hospital (project ID: H25-E37-1) and the Ethics Committee, Ankara University (project ID: 36290600/S5). All experiments were performed in accordance with relevant guidelines and regulations.

Subjects

A total of 272 subjects, which included 72 Turkish (Turkish group; females = 36 [Turkish female subgroup]; male = 36 [Turkish male subgroup]) and 200 Japanese (females = 100 [Japanese female subgroup], males = 100 [Japanese male subgroup]) aged between 18 and 35 years, were recruited from among the students and faculty of Ankara University in Turkey and Osaka University in Japan who met the following selection criteria: no congenital facial deformities including cleft lip or palate, no facial paralysis, no noticeable scars or skin disease in the neck or dentofacial regions (or history thereof), no history of any psychiatric disorder, no subjectively or objectively discernible jaw dysfunction, a body mass index [BMI] that ranged from 18.50 to 24.99, a dental overbite that ranged from 1.0 to 5.0 mm, a dental overjet that ranged from 0.0 to 7.0 mm, and a straight soft-tissue facial profile. For the present Japanese group, we used the samples from our previous study²². Due to recording limitations with 3-D digital cameras, male subjects having thick beards were excluded in advance. A written informed consent form was distributed to and signed by all participants. Informed consent was approved by the Research Ethics Committee, Osaka University Dental Hospital (project ID: H25-E37-1) and the Ethics Committee, Ankara University (project ID: 36290600/S5).

Data acquisition

The participants were asked to sit on a fixed chair with a natural head position without head support. They were then asked to assume a resting posture, which was defined as a relaxed facial posture with the lips in repose and the teeth in light contact in the habitual maximum intercuspation position. Each subject’s face was recorded once with a 3-D digital camera (3dMDcranial System, 3dMD, Atlanta, GA, USA) with a 1.5-ms capture speed and a dimensional accuracy of 0.2 mm²¹.

Each 3D facial image, scaled down to 75% of its actual size, was displayed on a 17-in LCD monitor (1701FP, Dell, Inc., Round Rock, TX, USA). The positions of 10 single and 8 paired landmarks (glabella [Gla], nasion [N], exocanthion [Ex], endocanthion [En], palpebrale superius [Ps], palpebrale inferius [Pi], porion [Po], orbitale [Or], pronasale [Prn], alar curvature point [Ac], subnasale [Sn], labiale superius [Ls], stomion [Sto], cheilion [Ch], labiale inferius [Li], submentale [Sm], pogonion [Pog], gnathion [Gn] (Supplementary Table S4, Supplementary Fig S4) were identified by visual inspection of the image and digitized using a computer mouse cursor and commercial software (Face Rugle, Medic Engineering Co., Kyoto, Japan). The zygomaticus′ [Zy′] and gonion′ [Go′] were mathematically defined as the most lateral point of the mathematically defined facial outline and the most inferior and lateral point of the mandibular facial outline, respectively, where the facial outline was defined as a series of the points with 60° angles between the surface normal vectors and Z-axis²¹. The process was repeated twice for each image, and the landmark coordinates from the two digitizations produced were averaged to yield the final landmark coordinates. A previous study²¹ that investigated the intra-observer reliability confirmed a mean absolute landmark difference of 0.32 mm (range, 0.07 to 0.52 mm) between the repeated measures (Supplementary Text S2). This result falls into the range considered reliable to highly reliable⁵⁸.

A 3D coordinate system identical to that employed in our previous study (Supplementary Fig S4; Informed consent was obtained to publish this image in an online open-access publication.)⁵⁹ was used in the current study. In short, the sagittal plane was defined by the exocanthions and endocanthions, and the axial plane was defined by the exocanthions, porion, and subnasale. The nasion was set as the origin.

Analyses

The 3D soft tissue facial morphology was evaluated by the two kinds of analysis: surface- and sectional-line-and-landmark-based analyses^21,60, which are summarized below.

(1) Surface-based analyses

The morphology of the facial surface was analyzed using the method documented in our previous study^21,60.

Homogeneous modeling

For each facial surface, fitting of high-resolution template meshes or a generic model^60,61 was performed using commercial software (HBM-Rugle, Medic Engineering Co., Kyoto) based on the landmarks assigned to each 3D image. This method automatically generated a homogeneous model that consisted of 6017 points (i.e., fitted mesh or semi-landmark nodes) on the wire mesh for each model with landmark anchors (i.e., Ex, En, Ps, Pi, Prn, Ac, Sn, Ls, Sto, Ch, Li, Sm, and Pog). This technique permits the extraction of relevant surface anatomy from face data while removing and/or smoothing out non-relevant data, yielding high-resolution, 3D surface data that provide enough detail to facilitate a quantitative assessment while maintaining small file sizes that are easily manipulatable and portable to a range of visualization technologies^60,61 (Supplementary Fig S5; Informed consent was obtained to publish this image in an online open-access publication.). The arithmetic means of the coordinate values of each corresponding point on the wire mesh were computed and used to generate the averaged 3D facial images for each male and female subgroup in each population group.

The surface displacement was quantitatively evaluated in each X-, Y-, and Z-axis in two different ways. The actual displacement vectors (male to female) and significance of differences were calculated for the 6017 points on each mesh between the male and female subgroups in each population group. The calculated vectors in millimeters were visualized with color-coding. Thereafter, the arithmetic means of the coordinate values of each corresponding point on the wire mesh were statistically analyzed for significant differences between the male and female subgroups using a two-sample t-test. A significance probability map^60,61 of the X-, Y-, and Z-values was generated to visualize these significant differences (Supplementary Fig S6).

Because previous studies revealed that masticatory muscle function likely influences mandible morphology (mainly in the vertical direction)⁶⁰ and inter-ocular width is less affected by masticatory muscle function⁶², the eyes, which are horizontally separated paired landmarks, were considered candidates for size normalization. A previous study³⁶ also showed that the right and left exocanthions were reliable points for identification. Thus, in the present study, facial size differences between individuals were standardized by normalizing the values of all surface coordinates to the distance between right and left exocanthions.

Sexual dimorphism of accentuated images

To quantitatively infer facial form femininity and masculinity, accentuated averaged faces, (overline{AccA({m}_{w})}) and (overline{AccA({f}_{w})}), were calculated for the male and female subgroups, respectively, to highlight site-specific sexual dimorphism^60,61, where

and (overline{A(m)}), (overline{A(f)}), and (overline{Aleft(allright)}) are the arithmetic means of the coordinate values for the male subgroup, female subgroup, and the sum of both groups, respectively, and w is the arbitrary weight value.

Examination of the variance of each population and each sex, and their interactions

To examine the variance in each population and each sex, and their interactions, we first reduced dimensionality by performing PCA for the 6017 coordinates of the aforementioned surface model⁶³. The significant principal components (PCs) were determined by scree plot analysis. Significant PCs were entered into a MANOVA to test for significance of the factors population affinity and sex. After MANOVA, a dendrogram was computed by applying the single linkage method to the matrix of Mahalanobis distances between subgroup means. Facial morphospace were determined by the significant PCs and used in the following process.

Influence of allometry on SShD

The SShD of the individual face was measured by projection of the individual facial configurations onto an axis connecting the vector between the average facial configurations of males and females in the facial morphospace using the following equation^13,64:

where (overrightarrow{{F}_{i}}) is the vector in the facial morphospace corresponding to an individual face i, and (overrightarrow{{F}_{(m-f)}}) is the vector between male and female facial configurations (male minus female). If SShD < -1, the face is hyperfeminine, and if SShD > 1, the face is hypermasculine.

Furthermore, measures of SShD were mathematically decomposed to allometric and non-allometric components. That is, variations in SShD due to an individual’s size (allometric) and variations that were independent of size (non-allometric) were examined in the overall variation in SShD in each population group using a multivariate regression analysis. CS was used as a measure of an individual’s size. The allometric variation in SShD was calculated by regressing the shapes in the facial morphospace on CS and projecting the estimated values from this regression onto the vector of sex differences. The non-allometric component of SShD was acquired by regressing the shapes in the facial morphospace on CS and then projecting the residualized facial coordinates on the sex difference vector calculated with these residuals¹³.

To assess the differences in CS, SShD, allometric SShD and non-allometric SShD between two populations, a permutation test was conducted as a randomization test, where populations were assigned at random to facial shapes, while the gender assignment of each face and the number of men and women in each sample remained unchanged. A total of 1000 randomized samples were generated within each permutation test.

Furthermore, to understand the effect of each SShD component on the morphospace, measures of each SShD component of individual faces were projected onto the shapes the facial morphospace using multiple regression. Significant coefficient values were evaluated.

(2) Sectional-line-and-landmark-based analyses

To analyze the surface data in detail, five categories of curving lines were extracted from the 3D images (Supplementary Table S5). The curving lines were used to extract 142 measurements that were previously reported (see Supplementary Figs S7, S8, and S9 for definitions of those variables)⁶⁵. In addition, 28 inter-landmark distances and 15 ratios that were previously reported²¹ were determined and employed. Therefore, 185 variables were employed in total. Facial size differences between individuals were also standardized by normalizing the values of all linear variables to the distance between right and left exocanthions.

A t-test was performed to determine whether the mean of each variable significantly differed between the Turkish male and female subgroups. To examine the similarity and dissimilarities of facial forms between the Japanese and Turkish groups, we included the sexual dimorphism results from a Japanese group reported in our previous study²¹. The effect size was calculated for each variable. Values greater than 0.8 for Cohen’s d were considered to have a large effect. Variables that showed significant differences and had large effects were considered biologically significant. Significance level was set to 0.01 due to a power analysis with a power of 0.8.