European Journal of Orthodontics 1 of 10 © The Author 2011. Published by Oxford University Press on behalf of the European Orthodontic Society.doi:10.1093/ejo/cjr033 All rights reserved. For permissions, please email: journals.permissions@oup.com

Introduction

There exists a significant body of orthodontic literature exploring gender differences in facial form between different populations, but the vast majority has been undertaken using two-dimensional (2D) assessments (Miyajima et al., 1996; Hwang et al., 2002, Hamdan, 2010). Three-dimensional (3D) facial shape/form research in a normal population is sparse.

Farkas and Posnick (1992), Farkas et al. (1992a,b), and Farkas (1994) used direct anthropometry to establish 3D soft tissue normative values for different populations to study growth and development of the face. Farkas and Cheung (1981) investigated soft tissue facial asymmetry in a normal population using the midfacial landmarks as reference points. A major concern with this technique is that it is time consuming and errors could be introduced by the measuring tools. Ferrario et al. (1997, 1999, 2000) employed 3D computerized mesh diagram analysis to study normal growth and development of facial soft tissues from 6 years to adulthood. They published linear, angular, and facial ratios relating to growth and development from a mixed longitudinal and cross-sectional study on healthy Italian subjects of different age groups. They concluded that facial maturity is reached at 13–14 years of age in females and 15–

Three-dimensional gender differences in facial form of children

in the North East of England

I. Bugaighis*, C. Mattick**, B. Tiddeman*** and R. Hobson*****Arab Medical University, Benghazi, Libya, **Royal Victoria Infirmary, Newcastle upon Tyne, ***Department of Computer Science, Aberystwyth University and ****Windmill Dental Suite, Newcastle upon Tyne, UK

Correspondence to: Dr I. Bugaighis, PO Box 595, Benghazi, Libya. E-mail: isbugaighis@yahoo.com

SUMMARY The aim of the prospective cross-sectional morphometric study was to explore three dimensional (3D) facial shape and form (shape plus size) variation within and between 8- and 12-year-old Caucasian children; 39 males age-matched with 41 females. The 3D images were captured using a stereophotogrammeteric system, and facial form was recorded by digitizing 39 anthropometric landmarks for each scan. The x, y, z coordinates of each landmark were extracted and used to calculate linear and angular measurements. 3D landmark asymmetry was quantified using Generalized Procrustes Analysis (GPA) and an average face was constructed for each gender. The average faces were superimposed and differences were visualized and quantified. Shape variations were explored using GPA and PrincipalComponent Analysis. Analysis of covariance and Pearson correlation coefficients were used to explore gender differences and to determine any correlation between facial measurements and height or weight. Multivariate analysis was used to ascertain differences in facial measurements or 3D landmark asymmetry.

There were no differences in height or weight between genders. There was a significant positive correlation between facial measurements and height and weight and statistically significant differences in linear facial width measurements between genders. These differences were related to the larger size of males rather than differences in shape. There were no age- or gender-linked significant differences in 3D landmark asymmetry. Shape analysis confirmed similarities between both males and females for facial shape and form in 8- to 12-year-old children. Any differences found were related to differences in facial size rather than shape.

18 in males. They found no significant correlation between 3D linear distances, ratios, and volumes of the mouth and nose. However, their method is time consuming and has a large inter-operator error of approximately 2 mm.

Several other cross-sectional studies have published 3D normative control data. Garrahy (2002) and White (2005) examined Scottish children and infants without oral clefts. Their samples were the control groups matched with subjects with oral clefts. A 3D stereophotogrammetric assessment tool was employed for image acquisition. The assessment included angular and linear measurements, ratios, and 3D asymmetry using Generalized Procrustes Analysis (GPA). Landmark asymmetry was undertaken by applying partial Procrustes analysis and the outcome was presented as scores after scaling all the landmarks to a unit size. White (2005) reported the correlation between height, weight, head circumference, and body measurement. A limitation of their method was that it was not possible to express facial asymmetry quantitatively (Garrahy, 2002; White, 2005).

Recently, there has been a growing interest in the subjective and objective assessment of facial shape and form in three dimensions, by constructing average faces for normal populations using optical-laser scanner systems.

The European Journal of Orthodontics Advance Access published April 29, 2011 by Im

an Bugaighis on D

ecember 28, 2011

http://ejo.oxfordjournals.org/D

ownloaded from

I. BUGAIGHIS ET AL.2 of 10

Toma et al. (2008) constructed an average face for 15-year-old females and males and Kau et al. (2006) used average faces as a template for comparison of facial skeletal disproportions. Moreover, it has been used for discriminating between male and female average faces of two different Caucasian populations (Božic et al., 2009; Gor et al., 2010). Their method described by Kau and Richmond (2008) had an image acquisition time of 7.5 seconds which could introduce errors due to movement of the subject. Each scan was generated by matching separate right and left facial scans which may affect the accuracy of the final image and their 3D model did not include a texture map.

Emerging developments in 3D imaging modalities and evolving technology in computer graphics have led to innovative statistical approaches that have the potential to enhance qualitative and quantitative evaluation of facial shape and form, in combination with conventional assessment tools.

This study aimed to evaluate the 3D facial shape and form of 8- to 12-year-old Caucasian children employing 3D facial measurements and to determine possible correlations with height and weight. In addition, assessment of 3D landmark asymmetry and construction of 3D photorealistic average facial models for each of the male and female subjects was carried out using a 3D non-invasive acquisition system and advanced statistical tools.

Subjects and methods

Ethical approval was granted by the Northumberland Local Research Ethics Committee (06/Q0902/36) and informed written consent was obtained from both parents/guardians and children.

The sample comprised 80 Caucasian children aged 8–12 years from the North East of England; 39 males and 41 females. The number of males in each age group was 3, 5, 9, 9, and 13, respectively, and the number of females in each age group was 5, 6, 11, 10, and 9. The subjects were recruited from Newcastle Dental Hospital. Care was taken to ensure that the children had harmonious balanced faces with a Class I dental occlusion, competent lips, no craniofacial abnormalities, including hypodontia, and had not undergone orthodontic treatment. This sample was a control group for a larger study to evaluate the effect of oral cleft repair on facial form (Bugaighis et al., 2010).

Recording of height and weight

The standing height (without shoes) of each child was measured in centimetres to the nearest millimetre using a portable digital height measuring scale (Soehnle Professional GmbH & Co. KG, Backnang, Germany). The weight of children was recorded in kilograms (kg) to the nearest gram using a digital weighing scale re-calibrated to zero between each patient (Soehnle).

Image acquisition and processing

3D facial scans were captured using 3dMD stereopho-togrammetry photographic capture equipment (3dMD, Atlanta, Georgia, USA). The subjects were seated 95 cm in front of the unit with the Frankfort plane raised upwards by 10 degrees to the horizontal to ensure a clear image of the area under the nose. The subjects were instructed to relax with their lips together if possible without straining and with their eyes open without stretching the forehead. The capture time was 2 ms, which minimized changes in position or facial expression. The scan resolution was 2 megapixels. All cameras and flashes were synchronized to capture the entire surface of the face from ear to ear. The 3dMD software generates one continuous point cloud from both sets of cameras. This reduces inaccuracy by merging the right and left camera images and point clouds.

The image was captured and transported directly to a connected computer. The software algorithm then generates the data geometry constructed from the captured point cloud. This takes approximately about 1 minute. The captured image was composed of approximately 40 000 vertices to create a polygonal mesh model of the captured face.

Landmark identification

Thirty-nine anthropometric homologous landmarks were used to characterize facial and nasiolabial form (Figure 1). First the landmarks (Table 1) were defined mainly based on work of Farkas (1994) and Hajeer (2003). The landmarks were recorded using MorphAnalyser software (V 2.07), a 3D software package developed by Tiddeman et al. (2000). This software allows 3D visualization of faces, interactive placement of landmarks, and calculation of the x, y, and z coordinates of each landmark. From these, linear and angular measurements were calculated, 3D landmark symmetry assessed, and average faces constructed by matching corresponding surfaces between the landmarks.

A validation study found that the overall error from image acquisition, the MorphAnalyser software, and operator error was on average 0.5 mm. This value is comparable with that of other 3D morphometric validation studies (Garrahy, 2002; Hajeer, 2003).

3D linear and angular measurements

MorphAnalyser software was employed to extract 3D linear and angular measurements. The linear measurements were calculated using the x, y, and z coordinates of two selected landmarks,

2 2 22 1 2 1 2 1Inter-landmark distance= ( ) ( ) ( )x x y y z z− + − + −

where x1, y1, z1 are the values of the coordinates of the first landmark and x2, y2, z2 are the values of the coordinates of the second landmark.

by Iman B

ugaighis on Decem

ber 28, 2011http://ejo.oxfordjournals.org/

Dow

nloaded from

3 of 10 3D STUDY OF FACIAL FORM IN ENGLISH CHILDREN

3D asymmetry

3D asymmetry of facial landmarks was assessed by reflecting each facial scan on itself and comparing the original and reflected versions. Procrustes analysis was used to translate, rotate, and scale them to best fit while

Table 1 Definition of the facial soft tissue landmarks used in the study.

Landmarks (R = Right, L = Left) Definitions

Upper face landmarks Endocanthion (enR, enL) Right, and left, located at the inner commissure of the eye fissure Exocanthion (exR, exL) Located at the outer commissure of the eye fissure Nasion (n) Point in the midline of the nasal root identical to hard tissue nasionNasolabial landmarks Alare curvature (acR, acL) Most lateral point in the curved baseline of each ala Alare (alR, alL) Most lateral point on each alar contour where the nostril starts to curve laterally Columellar high point (cR, cL) Highest point on the columella crest Inner alare (ali’R, ali’L) Inner marking level at the midportion of the alae where the thickness of each ala is measured Outer alare (alo’R, alo’L) Outer marking level at the midportion of the alae where the thickness of each ala is measured Pronasale (prn) Most protruded point of the apex of the nose Subalare (sbalR, sbalL) Point at the lower limit of each alar base where it joins the skin of the upper lip Subalare′ (sbali’R, sbali’L) Point at the inner lower limit of each alar base Subnasale (sn) Midpoint of maximum concavity where upper lip skin meets columella base Subnasale inner (sniR, sniL) Midpoint of columella on each side at the bottom line where the thickness of the columella is measured Cheilion (chR-chL) Point located at each labial commissure Christa philtri (cphR, cphL) Point on each elevated margin of upper lip at the junction of the vermillion line of upper lip and white roll line Laberale superiorus (ls) Midpoint of upper vermillion line Laberale inferiorus (li) Midpoint of the lower vermillion line Superior labial sulcus (sls) Deepest midline point between the mouth and the nose Stomion Superiorus (stos) Most inferior midpoint of the vermillion border of the upper lip Stomion Superiorus (stoi) Most superior midpoint of the vermillion border of the lower lipLower face landmarks Sublabialis (sl) Determines lower border of the lower lip and the upper border of the chin Pogonion (pog) The most anterior midpoint of the chin Menton (me) The lowest median point on the lower border of the mandibleAural landmarks Subtragion (strR, strL) Most inferior point on the anterior inferior margin of the helix attachment of the face Otobasion inferiorus (otbiR, otbiL) Most inferior point on the ear lobe at the attachment to the cheek

Figure 1 The anthropometric landmarks used in the study digitized on an average face of 30 subjects.

retaining their shape. The distance between each landmark and the corresponding reflected landmark was measured in millimetres and the asymmetry was calculated for each landmark.

Construction of average faces

An average face for males and females was constructed. This was created for each group by calculating the mean position of each vertex of each face using MorphAnalyser software. This method removes variations in translation, rotation, and scale. The software also calculates the 3D positional distribution around each surface point in the mean. Subsequently, male and female average faces were compared visually (Figure 2). The female face was then superimposed on the male face by surface-based registration using GPA with 3dMD software. A colour-coded image displays the subjective differences between the two surfaces. The descriptive statistical differences in millimetres are shown in Figure 3.

An average face of each group was constructed. The same procedure was repeated again after 1 week. Objective differences between each pair of superimposed average faces were displayed in a histogram; the average mean value, standard deviation (SD), and the root mean square (RMS) distances were equal in both the trials. Subjectively, MorphAnalyser software displayed

by Iman B

ugaighis on Decem

ber 28, 2011http://ejo.oxfordjournals.org/

Dow

nloaded from

I. BUGAIGHIS ET AL.4 of 10

Figure 3 A colour-coded image of superimposed female control average face (difference) on the male controls average face (reference) (A) and histogram (B). The histogram shows the differences in the colour-coded images. Red represents the area were the female average face is protrusive, green retrusive, and blue no difference.

identical homogenous colours for the entire superimposed facial surfaces indicating superimposition of identical surfaces.

Shape analysis

Morphologika v2.5 software was used to carry out shape analysis. Morphologika is a validated integrated software package developed by O’Higgins and Jones (1998). It allows geometric morphometric analysis of landmark configurations using landmark coordinates. Differences in shape between and within groups can be explored using GPA and Principal Component Analyses (PCA).

To assess shape differences between the groups, the landmark coordinates of subjects were submitted to GPA to

Figure 2 (A) Rendered, right, front and left average control male face visual comparison to female (B) rendered, right, front and left average face.

register the best fit by standardizing the translational and rotational differences and scaling them to a unit centroid size. The resulting shape coordinates were tangent projected to minimize the deformation of the distributions following GPA (O’Higgins and Jones, 1998) and used to explore within and between gender variability. Permutation analysis was undertaken to assess statistically significant differences between the groups. PCA, which summarizes variation in the correlated variables in the uncorrelated axis of shape, was undertaken (Mitteroecker et al., 2004). Shape discrimination was achieved by comparison of the full set of recorded landmarks of each subject. The 3D geometric information of the subjects in each group was retained independent of size.

by Iman B

ugaighis on Decem

ber 28, 2011http://ejo.oxfordjournals.org/

Dow

nloaded from

5 of 10 3D STUDY OF FACIAL FORM IN ENGLISH CHILDREN

Statistical analysis

The Statistical Package for Social Sciences version 15.0 (SPSS Inc. Chicago, Illinois, USA) was used for statistical analyses. Analysis of covariance (ANCOVA) was used to determine any significant differences in the height or weight between genders.

Linear and angular measurements

Shapiro–Wilk test was used to investigate the distribution of the data and Levene’s test to explore the homogeneity of the variables. Multivariate analysis of covariance (MANCOVA) was conducted to discriminate facial linear and angular measurements and ratios between males and females for each measurement using age as the covariate. The mean, SD, and mean difference of each measurement for each gender were calculated. Pearson’s correlation coefficient was used to ascertain correlations between linear and angular facial measurements and height and weight. Paired Student’s t-tests were used to confirm any significant differences between paired facial measurements within each gender. Significance was set at 0.05.

MANCOVA was undertaken to examine 3D landmark asymmetry between males and females. The mean, SD, and mean difference of each landmark asymmetry of both genders were calculated.

Results

The results showed the similarity of both genders in height, weight, and 3D landmark symmetry. Differences in linear measurements were mainly related to differences in size rather than shape and were mainly observed in width measurements. There were no significant differences between the mean age of males at 10.6 years (SD = 1.3) and females at 10.4 years (SD = 1.3) at P = 0.436.

ANCOVA revealed that there was no significant difference in height between males at 147.77 cm (SD = 10.32) and females at 147.76 cm (SD = 10.17;P = 0.997). Also the difference in weight between males at 42.99 kg (SD = 11.47) and females at 41.27 kg (SD = 8.93) was not statistically significant (P = 0.682).

Shapiro–Wilk test revealed that the data were normally distributed. Levene’s test established homogeneity of the variables. MANCOVA showed that while there were significant differences in all variables (P = 0.006), there was no significant difference between males and females (P = 0.120) and no significant association between age and variables (P = 0.660).

Fourteen transverse facial linear measurements were found to be statistically significantly different between the genders. Overall, differences were mainly in facial width, with wider faces occurring in males. On the other hand, there were no significant differences in facial ratios or angular measurements between the groups (P > 0.051).

Paired t-tests showed that most of the paired measurements were similar on both sides of the face in both genders except for one in males and three in females (P ≤ 0.011). Left lower lip length at 23.45 mm was significantly greater in males than right lower lip length at 23.05 mm (P = 0.003). Significant differences were observed between the following paired measurements; in females right and left nasion-endocanthion widths (20.07 and 19.81 mm, respectively), right and left hemiphiltral widths (5.68 and 5.76 mm, respectively), and right and left lower lip lengths at 23.14 and 23.51 mm, respectively (P ≤ 0.018).

Comparison between the male and female facial measurements was as follows: significantly greater binocular width was found in males (87.93 mm) compared with females at 85.70 mm (P = 0.007; Table 2). The binocular width measurement was moderately but significantly correlated with height and weight (0.424 and 0.490) as shown by Pearson correlation coefficient (P ≤ 0.001). Also the right and left endocanthion-nasion widths in males, at 20.95 and 20.80 mm, were found to be greater than in females at 20.07 and 19.81 mm (P ≤ 0.008). The correlation between these width measurements and height and weight was weak, and ranged from 0.227 to 0.296, but the correlation remained significant (P ≤ 0.026).

Table 2 shows similar nasolabial measurements for both genders except for the philtrum width measurements and upper cutaneous lip height. The philtrum was significantly wider in males at 11.01 mm compared with females at 10.52 mm (P = 0.034). Philtrum width was weakly but significantly correlated to height and weight (0.287 and 0.296; P = 0.008). Moreover, upper cutaneous lip height was significantly longer in males (14.60 mm) compared with females (13.29 mm; P = 0.001). This measurement had a non-significant (P = 0.054) and weak correlation with height and weight.

Male faces were wider than females (Table 3). Significant differences were found between male total facial width measured from the paired subtragion and otobasion inferiorus landmarks at 134.75 and 128.73 mm, respectively, compared with females at 130.34 and 124.95 mm, respectively (P ≤ 0.025). There were significant correlations between total facial width measured between the bilateral subtragion and otobasion inferiorus (0.488 and 0.539, respectively) and weight and height (0.488 and 0.374, respectively) at P ≤ 0.001. Furthermore, the left naso-aural width and the left subnasosubtragion width at 116.49 and 112.73 mm, respectively, in males were significantly greater than in females at 114.12 and 110.10 mm, respectively (P ≤ 0.043). On the other hand, only the male’s right subnasosubtragion width at 112.71 mm was significantly greater than females at 110.18 mm (P = 0.035). All these midfacial width measurements were moderately to strongly correlated with height and weight (range from 0.436 to 0.568) and P < 0.001.

by Iman B

ugaighis on Decem

ber 28, 2011http://ejo.oxfordjournals.org/

Dow

nloaded from

I. BUGAIGHIS ET AL.6 of 10

3D landmark asymmetry

There were no statistically significant differences in 3D facial landmark asymmetry (P = 0.058), 3D landmark asymmetry between males and females (P = 0.120) or between age and landmark asymmetry (P = 0.568).

Table 4 summarizes the asymmetry of each landmark for both genders. 3D asymmetry of facial landmarks in both groups was overall less than 2 mm. The midsagittal landmarks were less asymmetric (≤1 mm). Of 25 landmarks, the only one that was statistically significantly different in

Table 2 Comparison of three-dimensional (3D) face height, nasal and oral measurements (SD, mm), for males (n = 39) and females (n = 41).

Male mean Female mean Mean difference P

Upper facial and face height measurements Binocular width (exR-exL) 87.93 (3.71) 85.70 (2.83) 2.23 0.007 Intercanthal width (enR-enL) 28.94 (2.53) 27.97 (2.10) 0.97 0.086 Right endocanthion–nasion (n-enR) 20.95 (1.67) 20.07 (1.01) 0.88 0.008 Left endocanthion-nasion (n-en L) 20.80 (1.65) 19.81 (1.07) 0.99 0.004 Upper face height (n-stos) 67.89 (3.52) 67.02 (4.34) 0.87 0.519 Total face height (n-me) 107.68 (5.06) 105.88 (6.50) 1.80 0.306 Lower face height (sn-me) 61.99 (3.19) 60.24 (4.34) 1.75 0.077Nasal measurements Soft tissue nasal width (alR-alL) 30.11 (2.46) 29.53 (2.19) 0.58 0.468 Right alar thickness (alRo’alRi’) 4.32 (0.4.9) 4.35 (0.44) −0.03 0.661 Left alar thickness (alLo-′alLi′) 4.37 (0.44) 4.39 (0.41) −0.02 0.727 Nasal base width (sbalR-sbalL) 17.61 (1.24) 17.05 (0.90) 0.56 0.052 Length of columella (right) (cR-sn) 9.09 (0.99) 9.06 (0.91) 0.03 0.875 Length of columella (left) (cL-sn) 9.05 (0.99) 9.06 (0.89) −0.01 0.740 Width of columella (sn’R-sn’L) 4.98 (0.36) 4.90 (0.50) 0.08 0.455Oral measurements Upper lip height (sn-stos) 20.59 (1.48) 19.81 (2.02) 0.78 0.070 Upper cutaneous lip height (sn-ls) 14.60 (1.46) 13.29 (1.86) 1.31 0.001 Upper vermillion lip height (ls-stos) 7.09 (0.93) 7.40 (1.12) −0.31 0.108 Lower vermillion lip height (stoi-li) 7.58 (0.89) 7.92 (1.13) −0.34 0.061 Mouth width (chR-chL) 40.78 (3.78) 40.41 (2.66) 0.37 0.897 Philtrum width (cphR-cphL) 11.01 (1.00) 10.52 (0.78) 0.49 0.034 Right hemiphiltrum width (cphR-ls) 6.05 (0.66) 5.68 (0.48) 0.63 0.009 Left hemiphiltrum width (cphL-ls) 6.10 (0.66) 5.76 (0.50) 0.58 0.021 Right lower lip length (chR-li) 23.05 (2.62) 23.14 (2.03) −0.09 0.529 Left lower lip length (chL-li) 23.45 (2.64) 23.51 (2.06) −0.06 0.623

Table 3 Comparison of three-dimensional (3D) facial width measurements, angular measurements (°), and facial ratios (%, SD) and mean difference (mm) for males (n = 39) and females (n = 41).

Males mean (SD) Females mean (SD) Mean difference P

Facial width measurements Right nasosubtragion width (n-strR) 116.20 (4.90) 114.06 (3.80) 2.14 0.059 Left nasosubtragion width (n-strL) 115.94 (4.66) 113.59 (4.01) 2.35 0.034 Right naso-aural width (n-otbiR) 116.79 (4.75) 114.70 (4.18) 2.09 0.078 Left naso-aural width (n-otbiL) 116.49 (4.79) 114.12 (4.26) 2.37 0.043 Right subnasosubtragion width (sn-strR) 112.71 (5.60) 110.18 (3.70) 2.53 0.035 Left subnasosubtragion width (sn-strL) 112.73 (4.92) 110.10 (4.05) 2.62 0.022 Right subnaso-aural width (sn-otbiR) 109.84 (5.58) 107.52 (3.89) 2.32 0.061 Left subnaso-aural width (sn-otbiL) 109.95 (5.16) 107.36 (4.34) 2.59 0.033 Right mandibular length (otbiR-pog) 112.02 (6.45) 109.71 (4.86) 2.31 0.138 Left mandibular length (otbiL-pog) 112.10 (6.31) 109.39 (5.26) 2.71 0.077 Intersubtragion width (strR-strL) 134.75 (7.16) 130.34 (4.50) 4.41 0.003 Inter otobasion inferiorus width (otbiR-otbiL) 128.73 (7.87) 124.95 (5.39) 3.78 0.025Facial angles and facial ratios Maxillary prominence angle (sl-n-sls) 7.44 (1.74) 7.18 (1.67) 0.26 0.346 Nasolabial angle (prn-sn-ls) 136.29 (7.10) 135.51 (8.02) 0.78 0.508 Facial convexity angle including the nose (n-prn-pog) 133.12 (3.99) 132.83 (4.17) 0.29 0.658 Facial convexity angle excluding the nose (n-sn-pog) 161.87 (5.10) 161.74 (4.80) 0.13 0.693 Facial proportion (LFH/TFH) 57.58 (1.84) 56.88 (1.60) 0.70 0.053 Philtrum width mouth width ratio (cphcph/chch) 27.11 (2.35) 26.10 (2.00) 1.01 0.051

by Iman B

ugaighis on Decem

ber 28, 2011http://ejo.oxfordjournals.org/

Dow

nloaded from

7 of 10 3D STUDY OF FACIAL FORM IN ENGLISH CHILDREN

the level of asymmetry between genders was superior labial sulcus (males 0.59 mm, females 0.41 mm; P = 0.034). However, the difference was only 0.18 mm.

The landmarks in both genders with asymmetry greater than 1 mm were aural (subtragion and otobasion inferiorus), cheilion, and exocanthion. Subtragion and otobasion inferiorus were asymmetric in males at 1.58 and 1.63 mm and in females at 1.67 and 1.84 mm, respectively. There were no differences in these landmark asymmetries between genders (P ≥ 0.363). Exocanthion was more asymmetric in females (1.44 mm) compared with males (1.22 mm) but the difference was not statistically significant (P = 0.121). Cheilion was the most asymmetric landmark in the nasolabial region in both groups at a similar level of 1.30 mm (P = 0.651).

3D average faces of males and females

Visual comparison (Figure 2) between male and female average faces showed that the average female face had a more prominent chin, a fuller upper lip, prominent zygomatic region, and a more convex profile.

Superimposition of both faces by surface registration (Figure 3) showed a more protrusive zygomatic region and cheek surface, fuller lips, protruded nasal tip, and protruded mentolabial surface in the average female face. On the other hand, the nasal bridge surface and the right and left sides of the face were very similar. The histogram displays the objective differences (Figure 3). The output illustrates that for this colour-coded image, 8536 valid

distance measurements were undertaken between the superimposed average faces. The RMS of the distance measurements between both average faces was 2.350 mm with an SD of 2.283 mm. The histogram curve was mainly blue indicating similarity between the two superimposed surfaces.

3D shape analysis

Permutation tests showed that the Procrustes distance between the mean male and female subjects was not significant (P = 0.388).

The x, y, z coordinates of the landmarks of the male and female subjects included were submitted to Procrustes analysis to be translated, rotated, and scaled to best fit. Subsequently, shape differences were examined. The first three PCs accounted for 46 per cent of shape variation between both groups (PC1 21 per cent, PC2 14 per cent, and PC3 11 per cent). Figure 4 shows variability within and between both groups. Plotting PC1 shape from GPA/PCA with PC2 showed that there was no clear Procrustes distance between the means of both groups and that individuals in both groups overlapped.

Discussion

Several 3D assessment tools have been employed to record 3D facial surface morphometry. Precision and accuracy are the main features required when selecting an approach for spatial data acquisition and analysis (Farkas, 1994). In this

Table 4 Comparison of the three-dimensional landmark asymmetry and mean difference (mm) for males (n = 39) and females (n = 41).

Landmarks Males mean (SD) Females mean (SD) Mean difference P

Excanthion (ex) 1.22 (0.53) 1.44 (0.69) −0.22 0.121Endcanthion (en) 1.07 (0.37) 0.96 (0.34) −0.10 0.205Nasion (n) 0.61 (0.50) 0.70 (0.62) −0.09 0.425Sublabialis (sl) 0.46 (0.28) 0.47 (0.45) −0.01 0.938Pogonion (pog) 0.36 (0.31) 0.51 (0.49) −0.14 0.151Menton (me) 0.50 (0.29) 0.57 (0.46) −0.07 0.657Subtragion (sbtr) 1.58 (0.71) 1.67 (1.08) −0.09 0.763Otobasion inferioris (obi) 1.63 (1.63) 1.84 (1.07) −0.21 0.363Pronasale (prn) 0.48 (0.43) 0.70 (0.63) −0.23 0.068Subnasale (sn) 0.45 (0.37) 0.58 (0.46) −0.15 0.105Alare (al) 0.89 (0.38) 0.93 (0.39) −0.04 0.626Alare curvature (ac) 0.93 (0.74) 0.84 (0.60) 0.08 0.380Subalare (sbal) 0.70 (0.92) 0.77 (0.29) −0.06 0.360Subalare′ (sbal′) 0.70 (0.92) 0.77 (0.29) −0.07 0.267Columella′ (c′) 0.73 (0.40) 0.71 (0.77) 0.03. 0.584Alare outer (alo′) 0.87 (0.38) 0.85 (0.32) 0.02 0.792ALare inner (ali′) 0.79 (0.43) 0.82 (0.32) −0.03 0.657Subnasale inner (sn′) 0.71 (0.40) 0.71 (0.37) −0.02 0.815Christa philtri (cph) 0.68 (0.37) 0.59 (0.26) 0.10 0.176Cheilion (ch) 1.30 (0.39) 1.25 (0.67) 0.05 0.651Superior labial sulcus (sls) 0.59 (0.37) 0.41 (0.30) 0.18 0.034Laberale superioris (ls) 0.49 (0.28) 0.48 (0.46) 0.03 0.727Stomion superioriss (stos) 0.35 (0.25) 0.43 (0.39) −0.08 0.323Stomion inferioris (stoi) 0.32 (0.29) 0.43 (0.40) −0.11 0.178Labialis inferioris (li) 0.41 (0.28) 0.43 (0.44) −0.02 0.990

by Iman B

ugaighis on Decem

ber 28, 2011http://ejo.oxfordjournals.org/

Dow

nloaded from

I. BUGAIGHIS ET AL.8 of 10

study, a 3D non-invasive imaging system acquiring an image in 2 ms was employed and validated and a combination of traditional 3D measurements and more advanced 3D shape analysis tools were used.

The normative data presented is for the control group of a larger study into the affects of oral clefts. The oral cleft age group (8–12 years old) was selected to examine the effect of primary and secondary lip and palate surgical repairs undertaken prior to the implementation of the Clinical Standards Advisory Group (1998) recommend-ations. The children’s age range was selected so that subjects had not started significant orthodontic treatment yet allowed adequate growth and development to have taken place. There were difficulties in recruitment in that study due to the size and social demography of the cleft population. As the control group was matched to the cleft group, this lead to skewed numbers in the control group, which may have influenced the outcome. However, this research contains comprehensive and innovative analytical methods that can be applied in further studies with larger sample sizes to avoid potential skew or bias of the analysis.

Both males and females had mostly similar facial form and symmetry. There were no significant differences in facial ratio or in any of the angular measurements, indicating similarity in shape between the genders. Significant differences in linear measurements were related to the larger size of males rather than differences in shape. This was confirmed by shape analysis. These results might have been influenced by the sample age range, where each group is at a different stage of maturity. The mean age in females was on the verge of the stage of the pubertal growth spurt, whereas this was yet to occur in males. The findings of the study are in agreement with Ferrario et al. (1997) where similar nasal measurements were found in both males and females at 11–12 years of age as a result of the earlier growth spurt in females. On the other hand, Ras et al. (1994) reported that 9-year-old males had a more asymmetric nose

Figure 4 Variability within and between male (diamond) and female (circle) control group. The combination of the first Principal Component (PC1) versus the second Principal Component (PC2) shows overlapping distribution with no clear difference between group means.

compared with females. This difference might have occurred as a result of the method used in assessing the asymmetry in their study. Ras et al. (1994) assessed the asymmetry in their sample in relation to an assumed facial plane of symmetry constructed midway between the exocanthion landmarks.

Both males and females showed symmetric paired linear facial width measurements. The significant differences between the paired lip length measurements and the paired endocanthion nasion width were less than 0.5 mm, which is considered within the range of landmark reproducibility error, although there were individual cases of fluctuating asymmetry between the two facial halves. Garrahy (2002) found no significant differences between paired facial width measurements in 3-year-old Scottish children using a similar 3D technique. Her examined group was much younger than that in the current study but the findings seem to be similar. On the other hand, Farkas and Cheung (1981) and Koff et al. (1981) reported right side facial dominance in their study groups. Also Farkas and Cheung (1981) found that the upper facial third demonstrated greater asymmetry. Furthermore, they reported that differences of up to 3 per cent in paired facial width measurements can be found in aesthetically pleasing faces. The examined groups in both former studies (Farkas and Cheung, 1981; Koff et al., 1981) were older and from different Caucasian populations which might have affected the outcome.

In the present study, 3D facial asymmetry was assessed by two different methods. Firstly, by comparing the 3D paired facial linear measurements. This method provided information about the difference in size between the selected landmarks but not their configuration. Therefore, comparing paired linear measurements is less informative than applying geometric morphometry. Similar paired linear measurements do not necessarily mean that the landmarks employed are symmetrical (Shaner et al., 2000; Garrahy, 2002). Secondly, the degree of asymmetry of each 3D landmark was assessed by quantifying (in millimetres) the distance between each landmark and the corresponding reflected point, without relying on a symmetrical facial plane. This allowed assessment of the contribution of each facial region to the overall facial asymmetry. The asymmetry of the paired exocanthion landmarks found in this study raises questions about the validity of the method of Ras et al. (1995) of constructing a symmetric plane based on these landmarks. Garrahy (2002) and White (2005) applied partial Procrustes analysis for the evaluation of asymmetry. However, the limitation of their method was that it was not possible to express the ‘asymmetry score’ quantitatively. Stauber et al. (2008) undertook a similar Procrustes analysis approach in assessing 3D landmark asymmetry in German children with anilateral cleft lip and palate, but the distance between the landmarks and the constructed facial plane was quantified in millimetres and not between each landmark and its reflection. This appears to be the first time quantification of 3D

by Iman B

ugaighis on Decem

ber 28, 2011http://ejo.oxfordjournals.org/

Dow

nloaded from

9 of 10 3D STUDY OF FACIAL FORM IN ENGLISH CHILDREN

landmark asymmetry in conventional metric values has been reported. This allowed the assessment of the contribution of each facial region to the overall facial asymmetry.

The findings of the current study indicate that asymmetry is present in aesthetically pleasing faces. The midline landmarks did not coincide when the original and reflected images were superimposed. These results echo the findings of a number of anthropometric studies on healthy individuals (Farkas and Cheung, 1981; Ferrario et al., 1994, 2001; Garrahy, 2002); therefore, the use of the ‘midfacial plane’ as a plane of asymmetry is not possible as it is not a true plane. This is significant, especially in craniofacial deformities where the assessment of facial asymmetry is one of the challenges faced by surgeons and orthodontists.

The sample was selected to have a harmonious, balanced, and aesthetically pleasing face. Analysis revealed that the mean asymmetry for all the recorded landmarks was less than 2 mm, indicating that this level of asymmetry does not influence visual perception of facial harmony and balance. It would be interesting to investigate the relationship between the visual perception of facial harmony and the level of asymmetry in a future study.

In the current study, the 39 recorded landmarks on each facial scan were used to guide the formation of the dense correspondence between the scans in each gender, but the dense correspondence occurred between all the vertices of the image so they serve as landmarks. This aids in the precision and accuracy of the construction of the average face and keeps distortion to a minimum. In addition, it explains the display of parts of the image in which landmarks were not digitized, such as the cheeks.

The method of constructing 3D photorealistic average faces used in the present study compares favourably with other techniques reported in the literature because of the short capture time (2 ms) and the creation of each scan from one point cloud, which minimizes the introduction of errors by merging right and left shells. The construction of 3D photorealistic average faces also includes a fully texture mapped surface.

Average faces provide an interesting perspective for diagnosis, treatment planning, and comparison of facial shapes between controls and groups with craniofacial deformities. This can provide surgeons with valuable 3D templates in planning the reconstruction of craniofacial anomalies. Orthodontists could use a soft tissue reference for understanding facial soft tissue movement in relation to dentoalveolar and skeletal changes occurring with treatment and growth, in addition to predicting facial growth for every gender and age group instead of relying on 2D radiographic templates.

Superimposition of average faces offers a method of comparison of facial surface areas that cannot be achieved by other assessment techniques, such as the fuller upper lip and the prominent zygomatic region in the female average

face. Also it provides a relevant tool in evaluating orthodontic and surgical procedure outcomes by superimposing pre- and post-treatment facial images. Furthermore, this method can be employed in studying longitudinal soft tissue growth changes in cohort studies.

Shape analysis was found to be highly sensitive in revealing similarity within and between genders. In applying Procrustes analysis, size is treated as a separate variable so difference in size does not influence the outcome. Variations within and between groups can be visualized in the space of the original data set which help in understanding the shape variation between and within males and females.

Conclusions

1. Soft tissue facial shape, form, and asymmetry are similar in 8- to 12-year-old children and any differences are related to size rather than shape. The larger facial measurements correlated significantly with body measurements.

2. 3D studies and shape analysis allow both qualitative and quantitative assessments of facial shape, form, and asymmetry.

Funding

Libyan Ministry of Higher Education to I.B.; Engineering and Physical Science Research Council grant for the development of MorphAnalyser to B.T.

ReferencesBožic M et al. 2009 Facial morphology of Slovenian and Welsh

white populations using 3-dimensional imaging. Angle Orthodontist 79: 640–645

Bugaighis I, O’Higgins P, Tiddeman B, Mattick C, Ben Ali O, Hobson R 2010 Three-dimensional geometric morphometrics applied to the study of children with cleft lip and/or palate from the North East of England. European Journal of Orthodontics 32: 514–521

Clinical Standards Advisory Group. 1998 Report of the clinical standards advisory group. Her Majesty’s Stationery Office, London

Farkas L G 1994 Examination. In: Farkas L G (ed). Anthropometry of the head and face Raven Press, New York, pp. 3–56.

Farkas L G, Cheung G 1981 Facial asymmetry in healthy North American Caucasians. An anthropometrical study. Angle Orthodontist 51: 70–77

Farkas L G, Posnick J C 1992 Growth and development of regional units in the head and face based on anthropometric measurements. Cleft Palate-Craniofacial Journal 29: 301–302

Farkas L G, Posnick J C, Hreczko T M 1992a Growth patterns of the face: a morphometric study. Cleft Palate-Craniofacial Journal 29: 308–315

Farkas L G, Posnick J C, Hreczko T M 1992b Anthropometric growth study of the head. Cleft Palate-Craniofacial Journal 29: 303–308

Ferrario V F, Sforoza C, Ciusa V, Serrae G, Tartagli G M 1999 Morphometry of the normal human ear: a cross-sectional study from adolescence to mid-adulthood. Journal of Craniofacial Genetics and Developmental Biology 19: 226–233

Ferrario V F, Sforza C, Poggio C E, Tartaglia G 1994 Distance from symmetry: a three-dimensional evaluation of facial asymmetry. Journal of Oral and Maxillofacial Surgery 52: 1126–1132

by Iman B

ugaighis on Decem

ber 28, 2011http://ejo.oxfordjournals.org/

Dow

nloaded from

I. BUGAIGHIS ET AL.10 of 10

Ferrario V F, Sforza C, Poggio C E, Schmitz J H 1997 Three-dimensional study of growth and development of the nose. Cleft Palate-Craniofacial Journal 34: 309–317

Ferrario V F, Sforza C, Schmitz J H, Ciusa V, Colombo A 2000 Normal growth and development of the lips: a 3-dimensional study from 6 years to adulthood using a geometric model. Journal of Anatomy 196: (Pt 3), 415–423

Ferrario V F, Sforza C, Ciusa V, Dellavia C, Tartaglia G M 2001 The effect of sex and age on facial asymmetry in healthy subjects: a cross-sectional study from adolescence to mid-adulthood. Journal of Oral and Maxillofacial Surgery 59: 382–388

Garrahy A 2002 3D assessment of dentofacial deformity in children with oral clefts. Thesis, University of Glasgow

Gor T, Kau C H, English J D, Lee R P, Borbely P 2010 Three-dimensional comparison of facial morphology in white populations in Budapest, Hungary, and Houston, Texas. American Journal of Orthodontics and Dentofacial Orthopedics 137: 424–432

Hajeer M 2003 3D-soft-tissue, 2D hard-tissue and psychological changes following orthognathic surgery. Thesis, University of Glasgow

Hamdan A M 2010 Soft tissue morphology of Jordanian adolescents. Angle Orthodontist 80: 80–85

Hwang H S, Kim W S, McNamara J A Jr 2002 Ethnic differences in the soft tissue profile of Korean and European-American adults with normal occlusions and well-balanced faces. Angle Orthodontist 72: 72–80

Kau C H H, Richmond S 2008 Three-dimensional analysis of facial morphology surface changes in untreated children from 12 to 14 years of age. American Journal of Orthodontics and Dentofacial Orthopedics 134: 751–760

Kau C H H, Zhurov A, Richmond S, Cronin A, Savio C, Mallorie C 2006 Facial templates: a new perspective in three dimensions. Orthodontics and Craniofacial Research 9: 10–17

Koff E, Borod J C, White B 1981 Asymmetries for hemiface size and mobility. Neuropsychologia 19: 825–830

Mitteroecker P, Gunz P, Weber G W, Bookstein F L 2004 Regional dissociated heterochrony in multivariate analysis. Annals of Anatomy 86: 463–470

Miyajima K, McNamara J A Jr, Kimura T, Murata S, Iizuka T 1996 Craniofacial structure of Japanese and European-American adults with normal occlusions and well-balanced faces. American Journal of Orthodontics and Dentofacial Orthopedics 110: 431–438

O’Higgins P, Jones N 1998 Facial growth in Cercocebus torquatus: an application of three-dimensional geometric morphometric techniques to the study of morphological variation. Journal of Anatomy 193: (Pt 2), 251–272

Ras F, Habets L L, van Ginkel F C, Prahl-Andersen B 1994 Three-dimensional evaluation of facial asymmetry in cleft lip and palate. Cleft Palate-Craniofacial Journal 31: 116–121

Ras F, Habets L L, van Ginkel F C, Prahl-Andersen B 1995 Method for quantifying facial asymmetry in three dimensions using stereophotogrammetry. Angle Orthodontist 65: 233–239

Shaner D J, Peterson A E, Beattie O B, Bamforth J S 2000 Assessment of soft tissue facial asymmetry in medically normal and syndrome-affected individuals by analysis of landmarks and measurements. American Journal of Medical Genetics 93: 143–154

Stauber I et al. 2008 Three-dimensional analysis of facial symmetry in cleft lip and palate patients using optical surface data. Journal of Orofacial Orthopaedics 69: 268–282

Tiddeman B, Duffy N, Rabey G 2000 Construction and visualisation of three dimensional facial statistics. Computer Methods and Programmes in Biostatistics 36: 9–20

Toma A M, Zhurov A, Playle R, Richmond 2008 A three-dimensional look for facial differences between males and females in a British-Caucasian sample aged 15½ years old. Orthodontic and Craniofacial Research 11: 180–185

White J E 2005 A longitudinal study of facial growth in infants using a three-dimensional imaging technique. Thesis, University of Glasgow by Im

an Bugaighis on D

ecember 28, 2011

http://ejo.oxfordjournals.org/D

ownloaded from