IDEAS AND INNOVATIONS
Determination of Facial Symmetry in Unilateral Cleft Lip and Palate
Patients From Three-Dimensional Data: Technical Report and Assessment
of Measurement Errors
Emeka Nkenke, M.D., D.M.D., Ph.D., Bernhard Lehner, M.D., Manuel Kramer, Gerd Haeusler, Dr.Ing., Ph.D.,
Stefanie Benz, Maria Schuster, M.D., Friedrich W. Neukam, M.D., D.M.D., Ph.D., Eleftherios G. Vairaktaris, M.D.,
D.M.D., Ph.D., Jochen Wurm, M.D.
Objective: To assess measurement errors of a novel technique for the threedimensional determination of the degree of facial symmetry in patients suffering from unilateral cleft lip and palate malformations.
Design: Technical report, reliability study.
Setting: Cleft Lip and Palate Center of the University of Erlangen-Nuremberg,
Erlangen, Germany.
Patients: The three-dimensional facial surface data of five 10-year-old unilateral cleft lip and palate patients were subjected to the analysis. Distances,
angles, surface areas, and volumes were assessed twice.
Main Outcome Measures: Calculations were made for method error, intraclass correlation coefficient, and repeatability of the measurements of distances, angles, surface areas, and volumes.
Results: The method errors were less than 1 mm for distances and less than
1.58 for angles. The intraclass correlation coefficients showed values greater
than .90 for all parameters. The repeatability values were comparable for cleft
and noncleft sides.
Conclusion: The small method errors, high intraclass correlation coefficients, and comparable repeatability values for cleft and noncleft sides reveal
that the new technique is appropriate for clinical use.
KEY WORDS: measurement error, optical 3D imaging, plane of symmetry, 3D
cephalometry, unilateral cleft lip and palate
It is generally accepted that differences between the dimensions of the left and right halves of the human face are common even in aesthetically pleasing faces. Differences of up to
2 mm between the two facial halves are considered to be within a normal range (Nkenke et al., 2003a). Studies on facial
laterality report a larger right side hemiface in a normal population (Burke, 1971, 1979; Shah and Joshi, 1978; Farkas and
Cheung, 1981; Koff et al., 1981, 1985; Peck et al., 1990).
Congenital anomalies such as cleft lip and palate malformations are causal factors that predispose to the development of
facial asymmetry with a larger noncleft side (Bishara et al.,
1994).
It has been stated that a greater degree of asymmetry hinders
unobtrusive participation in the social life of cleft lip and palate patients. This seems to be especially true for the nasolabial
appearance (Tan and Pigott, 1993). Therefore, one of the major
goals of cleft lip and palate surgery is the establishment of
facial symmetry.
Dr. Nkenke is with the Department of Oral and Maxillofacial Surgery, University of Erlangen-Nuremberg, Erlangen, Germany. Dr. Lehner is with the
Department of Oral and Maxillofacial Surgery, Kantonsspital Lucerne, Switzerland. Mr. Kramer is with the Department of Oral and Maxillofacial Surgery,
University of Erlangen-Nuremberg, Erlangen, Germany. Dr. Haeusler is Chair
for Optics, University of Erlangen-Nuremberg, Erlangen, Germany. Ms. Benz
is with the Department of Oral and Maxillofacial Surgery, University of Erlangen-Nuremberg, Erlangen, Germany. Dr. Schuster is with the Department
of Phoniatrics and Pedaudiology, University of Erlangen-Nuremberg, Erlangen,
Germany. Dr. Neukam is with the Department of Oral and Maxillofacial Surgery, University of Erlangen-Nuremberg, Erlangen, Germany. Dr. Vairaktaris
is with the Department of Oral and Maxillofacial Surgery, University of Athens,
Athens, Greece. Dr. Wurm is with the Department of Otorhinolaryngology,
University of Erlangen-Nuremberg, Erlangen, Germany.
The ‘‘Deutsche Forschungsgemeinschaft’’ supported the study (Special Research Sector 603, Model-Based Analysis and Visualization of Complex Scenes
and Sensor Data—Subproject C4).
Submitted September 2004; Accepted July 2005.
Address correspondence to: Dr. Emeka Nkenke, Department of Oral and
Maxillofacial Surgery, University of Erlangen-Nuremberg, Glueckstr. 11, 91054
Erlangen, Germany. E-mail [email protected].
129
130
Cleft Palate–Craniofacial Journal, March 2006, Vol. 43 No. 2
Each cleft lip repair technique has its own advantages and
disadvantages. The surgical options available have advanced
the care of affected children to a point where new techniques
and developments are likely to bring about only small changes
(Atack et al., 1997). Consequently, more sophisticated methodology is required to detect improvements (Roberts et al.,
1991). It is difficult to assess facial appearance in a valid and
reliable way. In particular, the quantification of asymmetry is
often of a subjective nature (Bearn et al., 2002a, 2002b). Because it is a three-dimensional (3D) phenomenon with transverse, vertical, and sagittal components, a comprehensive assessment of asymmetry requires a method to investigate all
three components simultaneously (Ras et al., 1994a, 1994b;
Ras et al., 1995).
In recent years, several studies have been carried out that
adopted 3D imaging techniques for the analysis of the facial
surface of cleft lip and palate patients. Unfortunately, the determination of the plane of symmetry often was confined to a
limited number of landmarks, not taking advantage of all
points on the facial surface (Ferrario et al., 1994; Nkenke et
al., 2003a). Moreover, few studies have quantified surface areas and facial volumes in unilateral cleft lip and palate (UCLP)
patients as a part of the symmetry analysis (Russell et al.,
2000; Ferrario et al., 2003; Nkenke et al., 2003b).
Therefore, it was the aim of the present study to introduce
a more comprehensive technique of 3D analysis of facial symmetry of patients with UCLP malformations and to assess the
measurement errors of the different parameters that were determined.
MATERIAL
AND
FIGURE 1 Method of assessment of the plane of symmetry that bisects
the distance (horizontal line) between Pi (original point) and P9i (mirror
point after rigid transformation).
METHODS
An optical 3D sensor (CAM3D, 3D-shape GmbH, Erlangen,
Germany; http://www.3D-shape.com/) was used for acquisition of the facial surfaces. The sensor is based on a modification of the phase-measuring triangulation method. A sequence of phase-shifted fringe patterns of structured light is
projected on the region of interest. Two CCD cameras record
the data from different directions. Subsequently, the images of
the phase-shifted patterns are evaluated by means of a fourshift algorithm to receive the 3D shape of the object’s surface.
The 3D sensor takes advantage of an astigmatic optical device
for the projection of precise sinusoidal intensity coded fringe
patterns instead of a commonly used Ronchi-grating fringe
projection (Häusler and Gruber, 1992).
The optical 3D sensor was calibrated for a measurement
volume of 300 mm3, adapted to the average dimensions of a
human head. The required measurement time for data acquisition was 640 milliseconds.
During data assessment, the position of the patients was adjusted reproducibly using a cephalometric head holder to prevent movement artifacts and soft tissue distortion caused by
altered inclination of the head. The optical data were assessed
with the patient’s lips at rest. The acquired 3D images could
be viewed immediately on the computer screen from any angle
desired.
The Frankfort horizontal plane and the plane of symmetry
were used as reference for anthropometric analysis of the facial
surface data. The Frankfort horizontal plane was defined by a
line connecting tragion and orbitale landmarks of both hemifaces.
Determination of the plane of symmetry of the facial surface
began with generation of mirror images of the optical 3D surface data (Fig. 1). Thereby, point Pi of the original facial surface received a corresponding point P*i in the mirror data set.
Coarse and fine registrations were performed to superimpose
original and mirror images. With the registration algorithms, a
rigid transformation was calculated that minimized the distance between the two data sets. During this procedure, for
each point of the original data set, the closest point in the
superimposed mirror data set was determined. If the distance
between these two points was larger than 2 mm, both were
excluded from calculation of the rigid transformation. Subsequently, the calculated rigid transformation was applied to all
points P*i of the mirror data set, leading to a modified mirror
data set consisting of points P9i. Each original data point Pi
now corresponded with a point P9i in the modified mirror data
set subjected to the rigid transformation. By bisecting the distances between all corresponding pairs Pi and P9i, the plane of
symmetry was assessed. Further detail on the procedure has
been provided previously (Benz et al., 2002).
Nkenke et al., MEASUREMENT ERRORS OF 3-D FACIAL ANALYSIS
131
FIGURE 3 Nostril landmarks and angles (for the definition of the landmarks, see Table 2; a 5 angle of the long axis of the noncleft side nostril
with the plane of symmetry, b 5 angle of the long axis of the cleft side
nostril with the plane of symmetry, green 5 noncleft side, red 5 cleft side).
FIGURE 2 Facial landmarks (for the definition of the landmarks see Table 2; green 5 noncleft side, red 5 cleft side).
Anthropometric facial landmarks that were determined are
shown in Figures 2 and 3, and Table 1. Lamed is the medial
intersection of the long axis of the nostril with the nostril rim.
Lalat is the lateral intersection of the long axis of the nostril
with the nostril rim. The short axis of the nostril was defined
as a line perpendicular to the long axis of the nostril showing
the largest distance between two opposing points on the nostril
rim. Sasup is the superior intersection of the short axis of the
nostril and the nostril rim. Sainf is the inferior intersection of
the short axis of the nostril and the nostril rim. The definitions
of the axes of the nostril and the adjacent landmarks were
chosen according to Yamada et al. (2002).
The distances of the different landmarks from the plane of
symmetry were calculated when projected on the plane of
symmetry in a coronal plane that was perpendicular to the
plane of symmetry. Differences in vertical height of corresponding landmarks of the cleft and noncleft sides were determined by projecting them on an axial plane through the
columella base (Col) parallel to the Frankfort horizontal plane.
Moreover, the distance between corresponding landmarks of
the two hemifaces projected on the plane of symmetry was
determined in an anterior-posterior direction parallel to the
Frankfort horizontal plane. Additionally, the angle between the
long axis of the nostril and the plane of symmetry was cal-
culated (Fig. 3). Nostril areas and vermillion areas were assessed separately for cleft and noncleft sides (Figs. 4 and 5).
The visible, virtual volumes of the midface, the upper lip,
and the nose were determined without considering underlying
voids and bony structures of the oronasal cavity (Fig. 6). As
a border for the volume determination of each half of the nose
of the two hemifaces, a line along nasion, endocanthion, nasal
sulcus, and columella base was used. For the upper lip volume,
the line proceeded along the labial fissure, cheilion, alar base,
and columella base. For each midfacial half, the difference in
volume between cleft and noncleft sides was determined with
a border along nasion, exocanthion, cheilion, and stomium. By
TABLE 1 Definition of Landmarks Extracted From the Optical
Three-Dimensional Images
Abbreviation
FH
N
En
Prn
Col
Ch
Lb
Lt
Sto
Gbase
Glat
Gsup
Lalat
Lamed
Sainf
Sasup
Nostril length
Nostril width
Nostril angle
Nostril area
Landmark
Frankfort horizontal plane, plane through tragion and orbitale landmarks
Soft tissue nasion
Endocanthion
Pronasale
Columella base, point with maximum vertical curvature bisecting the columella caudally
Cheilion, point located at each labial commissure
Bottom of the cupid’s bow
Top of the cupid’s bow
Stomium, point bisecting the distance between the cheilion
points of both hemifaces
Most inferior point of the alar groove
Most lateral point of the alar groove
Most superior point of the alar groove
Lateral point of the long axis of the nostril
Medial point of the long axis of the nostril
Inferior point of the short axis of the nostril
Superior point of the short axis of the nostril
Distance between Lamed and Lalat (long axis of the nostril)
Distance between Sainf and Sasup (short axis of the nostril)
Angle of the long axis of the nostril formed with the plane
of symmetry
Area of the nostril in a plane through the most anterior
points of the nostril
132
Cleft Palate–Craniofacial Journal, March 2006, Vol. 43 No. 2
FIGURE 4 Nostril areas (green 5 noncleft side, red 5 cleft side).
projecting all single data points of the border lines of the volumes onto the plane of symmetry in frontal planes perpendicular to the Frankfort horizontal plane and the plane of symmetry, the posterior boundary areas of the volumes of interest
were generated.
This facial analysis technique was applied to the facial surface data of five 10-year-old patients suffering from complete
UCLP malformations. The data assessment was repeated after
a time interval of 6 months by the same examiner.
Statistics
Method errors of the different parameters were determined
using the Dahlberg formula
Se 5
! Od
2
2n
where Se is the standard deviation of the differences of each
of the replicates from its mean, n is the number of patients,
and d is the difference between the first and the second recordings (Dahlberg, 1940). Intraclass correlation coefficients
of the repeated measurements were calculated according to
Bland and Altman (1996b). The higher the correlation coefficient between repeated measurements of a parameter, the better
this parameter discriminates between individuals. The repeatability of the determination of the different parameters and the
95% limits of agreement were assessed according to Bland and
Altman (1996a). Of the differences between the first and second measurements, 95% are expected to be less than the repeatability. All calculations were done using the StatsDirect
Version 1.9.14 (http//www.statsdirect.com).
FIGURE 5 Vermillion area.
Method errors calculated that the Dahlberg formula showed
values less than 1 mm for linear measurements and values less
than 1.58 for angular measurements. The intraclass correlation
coefficients were greater than .90 for all parameters. The repeatability was comparable for cleft and noncleft sides for all
parameters.
DISCUSSION
Satisfactory functional results can be achieved after primary
repair of the malformations in the majority of the UCLP patients. From this point of view, there is only a limited need for
corrections and repeat operations. However, morphologic results tend to be less satisfactory (Breitsprecher et al., 1999).
Despite the many advances in surgery, the cleft lip nose continues to be a stigma of cleft surgery. Many patients require
secondary repair because of nasolabial asymmetries (Kane et
al., 2000).
It is well known that the method used to reproduce facial
characteristics (two-dimensional [2D] or 3D) and the method
RESULTS
Determination of the plane of symmetry of the facial surface
of the five patients yielded exactly the same coordinates of the
plane for first and second measurements. The results of the
determination of distances, angles, areas, and volumes are given in Tables 2 through 7.
FIGURE 6 Facial volumes (yellow 5 plane of symmetry, blue 5 surface
bordering the midfacial volume, red 5 surface bordering the nasal volume,
green 5 surface bordering the lip volume).
Nkenke et al., MEASUREMENT ERRORS OF 3-D FACIAL ANALYSIS
133
TABLE 2 Facial Landmarks (Distance of Facial Landmarks From the Plane of Symmetry Determined in an Axial Plane Perpendicular
to the Plane of Symmetry and Parallel to the Frankfort Horizontal Plane; Negative Values Reveal a Deviation to the Noncleft Side,
Positive Values Reveal a Deviation to the Cleft Side; for Definition of the Landmarks, See Table 1)
Pat. 1/Meas. 1
Pat. 2/Meas. 1
Pat. 3/Meas. 1
Pat. 4/Meas. 1
Pat. 5/Meas. 1
Pat. 1/Meas. 2
Pat. 2/Meas. 2
Pat. 3/Meas. 2
Pat. 4/Meas. 2
Pat. 5/Meas. 2
Method error
Intraclass correlation
coefficient
95% limits of
agreement
Repeatability
Endocanthion
Noncleft*
(mm)
Endocanthion
Cleft
(mm)
Pronasale
(mm)
Columella Base
(mm)
Cheilion Noncleft
(mm)
Cheilion Cleft
(mm)
15.5
15.1
15.9
14.4
13.8
15.0
15.5
15.1
14.9
13.2
0.22
14.5
14.7
16.8
15.3
13.2
14.2
14.3
16.9
15.2
13.7
0.36
22.6
23.9
22.7
23.8
22.4
22.9
23.2
22.8
23.2
22.7
0.31
22.2
23.1
22.7
22.0
21.1
22.0
23.3
22.5
22.2
21.4
0.15
21.0
19.9
24.2
21.0
19.7
21.2
19.3
24.0
20.2
19.1
0.39
20.5
20.7
23.0
20.9
15.7
20.0
21.1
22.7
21.3
15.1
0.33
0.91
0.96
0.95
0.95
0.95
0.98
21.04–1.42
1.16
20.62–0.74
0.62
20.34–1.19
1.07
20.86–1.11
0.92
21.05–0.81
0.87
20.40–0.51
0.42
* Noncleft 5 noncleft side; cleft 5 cleft side; Pat. 5 patient; Meas. 5 measurement; Method error 5 method error calculated by the Dahlberg formula.
applied to quantify asymmetry (global or localized analysis)
may play a major role in the predominant site of asymmetry
(Ferrario et al., 1994). Most studies involving the quantitative
assessment of facial asymmetry in living persons have been
performed on 2D reproductions of hard tissue (radiographs) or
soft tissue (photographs) morphology. Both methods project a
complex 3D structure onto a 2D plane, thus causing one of
the facial dimensions (usually facial depth) to be lost. It is
obvious that a correct and comprehensive evaluation of the
surface of the face should involve all three spatial planes simultaneously (Ras et al., 1994b; O’Grady and Antonyshyn,
1999; Ferrario et al., 2001).
To evaluate the efficacy of the many different treatment
methods with a standardized and quantitative method and to
decide whether a surgical technique is adequate to improve the
appearance of patients with cleft lip and palate deformities, it
is desirable to objectively assess different facial features threedimensionally (Coghlan et al., 1987; Laitung et al., 1993; Russell et al., 2000). Therefore, it was the aim of the present study
to introduce a new comprehensive method for the 3D analysis
of facial symmetry of patients with UCLP malformations and
to assess the measurement errors that occur during the determination of different parameters such as distances, areas, angles, and volumes.
Although some studies have been carried out with different
3D imaging techniques, most of them have been confined to
the determination of anthropometric landmarks. When single
landmark measurements are used for symmetry analysis, endless comparisons between the two facial halves can be made.
The position of a localized asymmetry can be easily appreci-
TABLE 3 Nasal Landmarks (Distance of Nasal Landmarks From the Plane of Symmetry Determined in an Axial Plane Perpendicular
to the Plane of Symmetry and Parallel to the Frankfort Horizontal Phase; Negative Values Reveal a Deviation to the Noncleft Side,
Positive Values Reveal a Deviation to the Cleft Side; for the Definition of the Landmarks, See Table 1)
Pat. 1/Meas. 1
Pat. 2/Meas. 1
Pat. 3/Meas. 1
Pat. 4/Meas. 1
Pat. 5/Meas. 1
Pat. 1/Meas. 2
Pat. 2/Meas. 2
Pat. 3/Meas. 2
Pat. 4/Meas. 2
Pat. 5/Meas. 2
Method error
Intraclass correlation
coefficient
95% limits of
agreement
Repeatability
Glat
Noncleft*
(mm)
Glat
Cleft
(mm)
Gsup
Noncleft
(mm)
Gsup
Cleft
(mm)
Gbase
Noncleft
(mm)
Gbase
Cleft
(mm)
18.4
18.7
18.0
19.3
16.2
19.3
18.9
17.2
18.6
16.7
0.44
16.8
14.7
19.0
16.1
17.7
15.9
14.3
19.7
15.8
17.8
0.40
13.0
13.8
12.6
15.3
15.5
13.7
13.4
12.3
15.6
15.1
0.32
9.7
11.2
11.7
11.0
11.3
9.4
11.8
12.4
10.4
10.9
0.39
13.2
15.4
11.6
12.9
10.1
13.0
16.1
12.5
12.5
9.9
0.40
13.9
13.2
15.0
15.2
14.3
14.1
13.8
14.6
15.0
13.4
0.37
0.91
0.94
0.93
0.91
0.95
0.92
21.35–1.40
1.23
21.02–1.33
1.10
20.93–1.02
0.88
21.21–1.21
1.08
1.36–1.01
1.12
20.94–1.23
1.01
* Noncleft 5 noncleft side; cleft 5 cleft side; Pat. 5 patient; Meas. 5 measurement; Method error 5 method error calculated by the Dahlberg formula.
134
Cleft Palate–Craniofacial Journal, March 2006, Vol. 43 No. 2
TABLE 4 Nostril Landmarks (Distance of Nasal Landmarks From the Plane of Symmetry Determined in an Axial Plane Perpendicular
to the Plane of Symmetry and Parallel to the Frankfort Horizontal Plane; Negative Values Reveal a Deviation to the Noncleft Side,
Positive Values Reveal a Deviation to the Cleft Side; for the Definition of the Landmarks, See Table 1)
Pat. 1/Meas. 1
Pat. 2/Meas. 1
Pat. 3/Meas. 1
Pat. 4/Meas. 1
Pat. 5/Meas. 1
Pat. 1/Meas. 2
Pat. 2/Meas. 2
Pat. 3/Meas. 2
Pat. 4/Meas. 2
Pat. 5/Meas. 2
Method error
Intraclass correlation
coefficient
95% limits of
agreement
Repeatability
Lalat
Noncleft*
(mm)
Lalat
Cleft
(mm)
Lamed
Noncleft
(mm)
Lamed
Cleft
(mm)
13.9
14.3
10.9
10.5
10.5
13.3
14.8
11.2
9.5
10.8
0.42
13.5
12.1
9.4
11.7
13.2
13.1
12.7
9.5
11.0
13.9
0.39
4.9
9.1
5.2
8.3
9.3
4.6
9.6
5.4
8.5
9.0
0.23
2.0
20.5
21.7
0
3.0
1.9
20.7
21.6
0.1
2.3
0.24
0.94
0.93
0.99
0.98
21.21–1.38
1.17
21.25–1.17
1.09
20.74–0.64
0.62
20.48–0.81
0.66
* Noncleft 5 noncleft side; cleft 5 cleft side; Pat. 5 patient; Meas. 5 measurement; Method error 5 method error calculated by the Dahlberg formula.
ated, but a global evaluation of the face is lost (Ferrario et al.,
2001). Recently, a new technique of symmetry analysis has
been introduced and has been validated, experimentally as well
as clinically (Benz et al., 2002; Nkenke et al., 2003a). Instead
of using single landmarks for the determination of the plane
of symmetry, the technique takes advantage of all data points
of a 3D facial surface image. This plane of symmetry is not
dependent on manually defined landmarks and, therefore, is
highly reproducible (Benz et al., 2002). In the present study,
the determination of the plane of symmetry in five patients at
two different points of time yielded exactly the same coordinates for each pair of corresponding planes of symmetry. The
new method for the assessment of symmetry is available now
in daily routine.
In addition to the symmetry analysis, a global evaluation of
the face should include the assessment of differences in facial
surface areas and visible soft-tissue volumes between cleft and
noncleft sides. These data allow an improved determination of
the influence of surgery on facial growth (Russell et al., 2000;
Ferrario et al., 2003). To date, there are no studies taking advantage of all the different aforementioned parameters during
the evaluation of discrepancies between the two hemifaces of
UCLP patients. In the present study, such a comprehensive
analysis has been performed for the first time.
The determination of well-defined facial surfaces areas for
symmetry analysis has been described previously by the use
of photographs or video imaging (Hurwitz et al., 1999; Russell
et al., 2001). With both techniques, 3D structures are projected
on 2D planes. Therefore, it seems that the areas assessed in
the present study are more precise, because the curvature of
TABLE 5 Nostril Dimensions, Nostril Area, and Nostril Angle Formed With the Plane of Symmetry (for the Definitions of the
Parameters, See Table 1)
Pat. 1/Meas. 1
Pat. 2/Meas. 1
Pat. 3/Meas. 1
Pat. 4/Meas. 1
Pat. 5/Meas. 1
Pat. 1/Meas. 2
Pat. 2/Meas. 2
Pat. 3/Meas. 2
Pat. 4/Meas. 2
Pat. 5/Meas. 2
Method error
Intraclass correlation
coefficient
95% limits of
agreement
Repeatability
Nostril Length
Noncleft*
(mm)
Nostril Length
Cleft
(mm)
Nostril Width
Noncleft
(mm)
Nostril Width
Cleft
(mm)
Nostril Angle
Noncleft
(degrees)
Nostril Angle
Cleft
(degrees)
Nostril Area
Noncleft
(mm22)
Nostril Area
Cleft
(mm2)
15.3
11.8
14.5
13.0
15.5
15.7
11.2
14.1
13.9
15.4
0.37
17.7
16.5
15.0
17.0
16.2
17.2
16.9
15.2
16.2
16.9
0.38
11.3
8.7
7.0
6.7
10.3
11.9
8.2
7.0
6.8
10.0
0.27
12.4
7.6
7.4
9.6
11.7
13.1
7.2
6.7
9.1
12.3
0.42
97.7
75.9
39.4
29.1
33.8
97.0
75.9
38.8
29.9
33.1
0.43
99.8
92.9
77.9
85.5
119.7
100.4
93.5
77.1
84.8
118.7
0.54
135
89
88
69
153
144
83
91
67
149
3.61
146
91
100
114
152
139
96
102
118
146
3.31
0.94
0.98
0.98
0.97
0.99
0.99
0.99
0.99
21.19–1.12
1.04
21.17–1.17
1.05
20.80–0.85
0.74
21.21–1.34
1.15
20.97–1.46
1.18
21.29–1.84
1.50
* Noncleft 5 noncleft side; cleft 5 cleft side; Pat. 5 patient; Meas. 5 measurement; Method error 5 method error calculated by the Dahlberg formula.
24.16–4.16
3.72
23.16–3.96
3.28
Nkenke et al., MEASUREMENT ERRORS OF 3-D FACIAL ANALYSIS
135
TABLE 6 Labial Landmarks (Lb and Ltlat, Respectively 5 Distance of the Labial Landmarks Lb and Lt From the Plane of Symmetry
Determines in an Axial Plane Perpendicular to the Plane of Symmetry and Parallel to the Frankfort Horizontal Plane; Negative Values
Reveal a Deviation to The Noncleft Side, Positive Values Reveal a Deviation to the Cleft Side; Ltvert 5 Shortest Distance of the Labial
Landmark Ltvert From an Axial Plane Through Col Parallel to the Frankfort Horizontal Plane; Ltap 5 Shortest Distance of the Labial
Landmark Lt From a Coronar Plane Through Col Perpendicular to the Plane of Symmetry; Negative Values Reveal That the Landmark
Lies Behind the Coronal Plane, Positive Values Reveal That the Landmark Lies in Front of the Coronal Plane; for the Definitions of the
Landmarks, See Table 1)
Pat. 1/Meas. 1
Pat. 2/Meas. 1
Pat. 3/Meas. 1
Pat. 4/Meas. 1
Pat. 5/Meas. 1
Pat. 1/Meas. 2
Pat. 2/Meas. 2
Pat. 3/Meas. 2
Pat. 4/Meas. 2
Pat. 5/Meas. 2
Method error
Intraclass correlation
coefficient
95% limits of
agreement
Repeatability
Lb
(mm)
Ltlat
Noncleft*
(mm)
Ltlat
Cleft
(mm)
Ltvert
Noncleft
(mm)
Ltvert
Cleft
(mm)
Ltop
Noncleft
(mm)
Ltap
Cleft
(mm)
0.2
21.3
23.4
22.2
1.7
0.6
21.1
23.7
22.8
1.3
0.28
7.2
7.0
10.0
7.7
5.7
7.7
7.3
9.2
7.1
5.9
0.37
7.0
3.6
3.4
4.6
9.1
6.7
3.4
3.7
4.2
8.1
0.37
10.4
11.6
13.1
13.2
11.4
10.0
11.2
13.7
13.4
11.8
0.31
13.2
11.8
15.0
12.4
10.6
13.4
11.3
14.5
12.1
10.9
0.26
2.9
0.3
0.8
2.2
20.5
2.6
0.5
0.7
2.6
20.7
0.20
0.7
0.2
3.2
2.3
21.0
0.9
0.4
2.9
2.6
20.7
0.20
0.98
0.91
0.97
0.94
0.97
0.98
0.98
21.05–1.20
1.02
20.57–1.21
1.02
21.01–0.85
0.84
20.60–0.88
0.71
20.67–0.95
0.77
20.62–0.60
0.55
20.67–0.33
0.54
* Noncleft 5 noncleft side; cleft 5 cleft side; Pat. 5 patient; Meas. 5 measurement; Method error 5 method error calculated by the Dahlberg formula.
the planes are preserved by the 3D imaging technique, whereas
it is lost when photographs or video imaging are used. Because
of their 2D origin, photographs, as well as video images, are
susceptible to parallax errors when the images are not assessed
from the correct direction. It is not possible to correct later for
these parallax errors. When 3D surface data are used, as in the
present study, parallax errors do not play any role during image acquisition and the real information is not obscured.
Unfortunately, previous studies of Hurwitz et al. (1999) and
Russell et al. (2001) did not provide a complete metrical analysis of the assessed surfaces; they did not give absolute data
in mm2 for the areas, but counted only the number of pixels
or gave only the ratio of the cleft side area to the noncleft side
area to characterize the dimension of the measured areas.
Therefore, a relevant comparison of these area data to the data
of the present study, where quantification in mm2 has been
completed, is not possible.
In the current literature, techniques for the determination of
facial volumes have been proposed (Ferrario et al., 2003).
However, they give only rough approximations of the volume,
because they do not take advantage of the complete facial surface bordering the volume of interest (Ferrario et al., 1998).
The volume determination is based on the generation of geometrical structures (tetrahedra) using only single landmarks
from the facial surface as their vertices (Ferrario et al., 1998).
The lateral surfaces of the tetrahedra are defined by these ver-
TABLE 7 Area of the Vermillion and Virtual Facial Volumes
Pat. 1/Meas. 1
Pat. 2/Meas. 1
Pat. 3/Meas. 1
Pat. 4/Meas. 1
Pat. 5/Meas. 1
Pat. 1/Meas. 2
Pat. 2/Meas. 2
Pat. 3/Meas. 2
Pat. 4/Meas. 2
Pat. 5/Meas. 2
Method error
Intraclass correlation
coefficient
95% limits of
agreement
Repeatability
Areavermillion
Noncleft*
(mm2)
Areavermillion
Cleft
(mm2)
Volumemidf.
Noncleft
(cm3)
Volumemidf.
Cleft
(cm3)
Volumenose
Noncleft
(cm3)
Volumenose
Cleft
(cm3)
Volumelip
Noncleft
(cm3)
Volumelip
Cleft
(cm3)
126
134
170
56
97
133
134
166
59
95
2.79
144
114
136
85
154
132
110
131
91
163
4.40
43.6
34.1
54.2
38.6
53.9
42.9
34.3
53.8
38.1
54.4
0.34
53.1
39.2
53.9
44.6
52.4
53.5
39.1
53.1
43.9
51.7
0.42
12.8
8.4
8.1
11.4
16.1
13.2
8.9
8.0
11.2
14.9
0.44
8.5
3.8
6.1
5.2
11.6
7.9
4.2
5.9
5.6
11.0
0.33
5.3
4.2
7.4
6.1
6.7
5.5
4.6
7.0
5.7
6.2
0.28
5.4
5.4
7.3
5.1
9.1
5.0
5.8
7.9
4.9
8.4
0.35
0.99
0.95
0.99
0.99
0.99
0.99
0.95
0.99
29.27–7.78
7.74
215.6–18.2
15.23
21.26–1.30
1.15
21.59–1.36
1.34
22.04–1.92
1.78
22.02–2.62
2.16
20.19–0.17
0.16
20.18–0.18
0.16
* Noncleft 5 noncleft side; cleft 5 cleft side; Pat. 5 patient; Meas. 5 measurement; Method error 5 method error calculated by the Dahlberg formula.
136
Cleft Palate–Craniofacial Journal, March 2006, Vol. 43 No. 2
tices. They do not follow the convex and concave facial surface. Therefore, a large amount of individual information on
the facial surface of the single patient is lost in the course of
this kind of volume determination. In the present study, all
data points of the surfaces surrounding the volumes of interest
were used for the volume determination. The technique of volume determination had been validated (Nkenke et al., 2003b).
Therefore, the new technique seems to provide better basic
conditions to assess minimal differences between cleft and
noncleft sides than the methods proposed previously.
Previous studies revealed that for the calculation of measurement errors, a sample of five patients is sufficient to gain
relevant results (Nkenke et al., 2003b). In the present study,
the calculation of measurement errors has shown that they
were small for all the different parameters. The method errors
calculated by the Dahlberg formula were less than 1 mm for
distances and less than 1.58 for angles. Although relative errors
calculated from these data can vary distinctively, depending
on the landmark that has been chosen, it has been stated that
absolute errors in this range do not have any clinical importance independent of the relative deviation (Houston, 1983;
Sandler, 1988). Therefore, the method introduced in the present study seems to be appropriate for the assessment of distances and angles in clinical use.
Intraclass correlation coefficients of greater than .90 indicate
a good relationship between the two sets of measurements,
with little random error (Murphy and Willmot, 2005). All repeatability data fell between a narrow 95% limit of agreement,
which means that the method of assessing distances, angles,
areas, and volumes can detect useful clinical differences
(Bland and Altman, 1996b). Because the 95% limits of agreement showed a broad overlapping for the different corresponding measurements of cleft and noncleft sides, the technique
can be used safely in patients affected by cleft lip and palate
malformations who potentially suffer from facial asymmetry.
In conclusion, it seems that with the new method of facial
analysis, planes of symmetry of the face can be determined
reproducibly without having to depend on manually determined reference points. With the new technique that respects
the three dimensions during facial analysis, the data of interest
can be determined without loss of information. Small measurement errors help to collect clinically relevant information.
Future trials on larger populations of patients will allow a more
comprehensive and consistent evaluation of the consequence
of different operation methods.
REFERENCES
Atack NE, Hathorn ISH, Semb G, Dowell T, Sandy JR. A new index for assessing surgical outcome in unilateral cleft lip and palate subjects aged five:
reproducibility and validity. Cleft Palate Craniofac J. 1997;34:242–246.
Bearn DR, Sandy JR, Shaw WC. Cephalometric soft tissue profile in unilateral
cleft lip and palate patients. Eur J Orthod. 2002a;24:277–284.
Bearn DR, Sandy JR, Shaw WC. Photogrammetric assessment of the soft tissue
profile in unilateral cleft lip and palate. Cleft Palate Craniofac J. 2002b;39:
597–603.
Benz M, Laboureux X, Maier T, Nkenke E, Seeger S, Neukam FW, Hausler G.
The symmetry of faces. In: Greiner G, Niemann H, Ertl T, Girod B, Seidel
HP (eds.). Vision, Modeling, and Visualization. Amsterdam: IOS Press,
2002;332–339.
Bishara SE, Burkey PS, Kharouf JG. Dental and facial asymmetries: a review.
Angle Orthod. 1994;64:89–98.
Bland JM, Altman DG. Statistics notes: measurement error. BMJ. 1996a;313:
744.
Bland JM, Altman DG. Statistics notes: measurement error and correlation coefficients. BMJ. 1996b;313:41–42.
Breitsprecher L, Fanghanel J, Metelmann HR, Mlynski G, Wurfel F, Freise K,
Knape U. The influence of the muscles of facial expression on the development of the midface and the nose in cleft lip and palate patients. A reflection of functional anatomy, facial esthetics and physiology of the nose.
Anat Anz. 1999;181:19–25.
Burke PH. Stereophotogrammetric measurement of normal facial asymmetry in
children. Hum Biol. 1971;43:536–548.
Burke PH. Growth of the soft tissues of the middle third of the face between
9 and 16 years. Eur J Orthod. 1979;1:1–13.
Coghlan BA, Matthews B, Pigott RW. A computer-based method of measuring
facial asymmetry. Results from an assessment of the repair of cleft lip deformities. Br J Plast Surg. 1987;40:371–376.
Dahlberg G. Statistical Methods for Medical and Biological Students. New
York: Interscience Publications, 1940.
Farkas LG, Cheung G. Facial asymmetry in healthy North American Caucasians. An anthropometrical study. Angle Orthod. 1981;51:70–77.
Ferrario VF, Ciusa V, Tartaglia GM. The effect of sex and age on facial asymmetry in healthy subjects: a cross-sectional study from adolescence to midadulthood. J Oral Maxillofac Surg. 2001;59:382–388.
Ferrario VF, Sforza C, Poggio CE, Schmitz JH. Facial volume changes during
normal human growth and development. Anat Rec. 1998;250:480–487.
Ferrario VF, Sforza C, Poggio CE, Tartaglia GM. Distance from symmetry: a
three-dimensional evaluation of facial asymmetry. J Oral Maxillofac Surg.
1994;52:1126–1132.
Ferrario VF, Sforza C, Tartaglia GM, Sozzi D, Caru A. Three-dimensional lip
morphometry in adults operated on for cleft lip and palate. Plast Reconstr
Surg. 2003;111:2149–2156.
Häusler G, Gruber M. Simple, robust and accurate phase measuring triangulation. Optik. 1992;89:18–24.
Houston WJB. The analysis of errors in orthodontic measurements. Am J Orthod. 1983;83:382–390.
Hurwitz DJ, Ashby ER, Llull R, Pasqual J, Tabor C, Garrison L, Gillen J,
Weyant R. Computer-assisted anthropometry for outcome assessment of
cleft lip. Plast Reconstr Surg. 1999;103:1608–1623.
Kane AA, Pilgram TK, Moshiri M, Marsh JL. Long-term outcome of cleft lip
nasal reconstruction in childhood. Plast Reconstr Surg. 2000;105:1600–
1608.
Koff E, Borod JC, Strauss E. Development of hemiface size asymmetry. Cortex.
1985;21:153–156.
Koff E, Borod JC, White B. Asymmetries for the hemiface size and mobility.
Neuropsychologica. 1981;19:825–830.
Laitung JK, Coghlan BA, Pigott RW. A comparison of computer versus panel
assessment of two groups of patients with cleft lip and palate. Br J Oral
Surg. 1993;46:18–22.
Murphy TC, Willmot DR. Image analysis of oronasal fistulas in cleft palate
patients acquired with an intraoral camera. Plast Reconstr Surg. 2005;115:
31–37.
Nkenke E, Benz M, Maier T, Wiltfang J, Holbach LM, Kramer M, Häusler G,
Neukam FW. Relative eno- and exophthalmometry in zygomatic fractures
comparing optical non-contact, non-ionizing 3D imaging to the Hertel instrument and computed tomography. J Craniomaxillofac Surg. 2003a;31:
362–368.
Nkenke E, Langer A, Laboureux X, Benz M, Maier T, Kramer M, Häusler G,
Kessler P, Wiltfang J, Neukam FW. Validation of in vivo assessment of facial
soft tissue volume changes and clinical application in midfacial distraction:
a technical report. Plast Reconstr Surg. 2003b;112:367–380.
O’Grady KF, Antonyshyn OM. Facial asymmetry: three-dimensional analysis
using laser surface scanning. Plast Reconstr Surg. 1999;104:928–937.
Nkenke et al., MEASUREMENT ERRORS OF 3-D FACIAL ANALYSIS
Peck S, Peck L, Kataja M. Skeletal asymmetry in esthetically pleasing faces.
Angle Orthod. 1990;61:43–46.
Ras F, Habets LLMF, Van Ginkel FC, Prahl-Andersen B. Facial left-right dominance in cleft lip and palate: three-dimensional evaluation. Cleft Palate
Craniofac J. 1994a;31:461–465.
Ras F, Habets LLMF, Van Ginkel FC, Prahl-Andersen B. Three-dimensional
evaluation of facial asymmetry in cleft lip and palate. Cleft Palate Craniofac
J. 1994b;31:116–121.
Ras F, Habets LLMF, Van Ginkel FC, Prahl-Andersen B. Logitudinal study on
three-dimensional changes of facial asymmetry in children between 4 to 12
years of age with unilateral cleft lip and palate. Cleft Palate Craniofac J.
1995;32:463–468.
Roberts CT, Semb G, Shaw WC. Strategies for the advancement of surgical
methods in cleft lip and palate. Cleft Palate Craniofac J. 1991;28:141–149.
Russell KA, Waldman SD, Lee JM. Video-imaging assessment of nasal mor-
137
phology in individuals with complete unilateral cleft lip and palate. Cleft
Palate Craniofac J. 2000;37:542–550.
Russell KA, Waldman SD, Tompson B, Lee JM. Nasal morphology and shape
parameters as predictors of nasal esthetics in individuals with complete unilateral cleft lip and palate. Cleft Palate Craniofac J. 2001;38:476–485.
Sandler PJ. Reproducibility of cephalometric measurements. Br J Orthod. 1988;
15:105–110.
Shah SM, Joshi MR. An assessment of asymmetry in the normal craniofacial
complex. Angle Orthod. 1978;48:141–148.
Tan KK, Pigott RW. A morbidity review of children with complete unilateral
cleft lip nose at 10 6 1 years of age. Br J Plast Surg. 1993;46:1–6.
Yamada T, Mori Y, Minami K, Mishima K, Sugahara T. Three-dimensional
facial morphology following cleft lip repair using the triangular flap with
or without rotation advancement. J Craniomaxillofac Surg. 2002;30:337–
342.