You are on page 1of 13

ORIGINAL ARTICLE

Comparison between traditional 2-dimensional


cephalometry and a 3-dimensional approach on
human dry skulls
Gregory L. Adams, DDS, MS,a Stuart A. Gansky, DrPH,b Arthur J. Miller, PhD,a William E. Harrell, Jr, DMD,c
and David C. Hatcher, DDS, MSc, MRCD(c)d
San Francisco, Calif, Alexander City, Ala, and Sacramento, Calif

The cephalogram is the standard used by orthodontists to assess skeletal, dental, and soft tissue
relationships. This approach, however, is based on 2-dimensional (2D) views used to analyze 3-dimensional
(3D) objects. The purpose of this project was to evaluate and compare a 3D imaging system and traditional
2D cephalometry for accuracy in recording the anatomical truth as defined by physical measurements with
a calibrated caliper. Thirteen skeletal landmarks were located by both radiographic methods on 9 dry human
skulls. Intraclass correlation (0.995), variance (0.054 mm2), and standard deviation (SD) (0.237 mm) were
averaged over 76 measurements and derived from precision calipers to establish these physical measure-
ments as a reliable gold standard to make comparisons of the 2D and 3D radiographic methods. The results
showed great variability of the 2D from the gold standard, with the range varying from ⫺17.68 mm
(underestimation of Gn-Zyg R) to ⫹15.52 mm (overestimation of Zyg L-Zyg R). In contrast, the 3D method
(Sculptor, Glendora, Calif) indicated a range of the SD from ⫺3.99 (underestimation) mm to ⫹2.96 mm
(overestimation). The 3D evaluation was much more precise, within approximately 1 mm of the gold standard.
These results indicate that, when the actual distance is measured on a human skull in its true dimensions of
3D space, the Sculptor program, by using a 3D method, is more precise and 4 to 5 times more accurate than
the 2D approach. Evaluating distances in 3D space with a 2D image grossly exaggerates the true measure
and offers a distorted view of craniofacial growth. There is an inherent problem of representing a linear
measure occupying a 3D space with a 2D image. (Am J Orthod Dentofacial Orthop 2004;126:397-409)

ephalometry was defined by Moyers1 as a

C
proach— errors of projection and errors of identifica-
radiographic technique for abstracting the hu- tion.2
man head into a measurable geometric scheme. Errors of projection are caused because the head-
Cephalometric radiography is used to describe the film is a 2-dimensional (2D) representation of a 3-di-
morphology and growth of the craniofacial skeleton, mensional (3D) object. X-ray beams are nonparallel
predict growth, plan treatment, and evaluate treatment and originate from a small source, leading to radio-
results. Most of these tasks require identifying specific graphs that are imperfect enlargements affected by the
landmarks and calculating various angular and linear distances between the focus, the object, and the film.3-7
variables. Two types of errors occur with this ap- Landmarks and structures not situated in the midsagittal
plane are usually bilateral, unfortunately giving a dual
image on the radiograph. Bilateral structures in the
a
Division of Orthodontics, Department of Growth and Development, Univer- symmetric head do not superimpose on the lateral
sity of California at San Francisco. cephalogram, because the fan of the x-ray beam ex-
b
Division of Oral Epidemiology and Dental Public Health, Department of
Preventive and Restorative Dental Sciences, Center for Health and Community, pands as it passes through the head, causing a diver-
University of California at San Francisco, San Francisco, Calif. gence between the images of all bilateral structures.
c
Private practice, Alexander City, Ala. Misalignment of the cephalostat and rotation of the
d
Owner, Diagnostic Digital Imaging, Sacramento, Calif.
William E. Harrell, Jr, is cofounder and chairman of the board of Acuscape patient’s head in the cephalostat in any plane will also
International, Inc, and David C. Hatcher is a key Acuscape contributor. induce errors of projection.
Reprint requests to: Dr A. J. Miller, Division of Orthodontics, Department of Errors of identification are the errors of identifying
Growth and Development, School of Dentistry, University of California at San
Francisco, Dental Clinics Building, 707 Parnassus St, San Francisco, CA specific landmarks on the headfilms and are considered
94143-0438; e-mail, amiller@itsa.ucsf.edu. by many investigators as the major sources of error in
Submitted, February 2003; revised and accepted, March 2004. cephalometrics.8-11 Several factors are involved and
0889-5406/$30.00
Copyright © 2004 by the American Association of Orthodontists. include the quality of the radiographic image, the
doi:10.1016/j.ajodo.2004.03.023 precision of landmark definition, the reproducibility of
397
398 Adams et al American Journal of Orthodontics and Dentofacial Orthopedics
October 2004

Table I. Landmarks, abbreviations, and operational definitions for 9 radiopaque markers on human skulls
Abbreviation Landmark Operational definition

S Sella Point representing midpoint of pituitary fossa on skull


Na Nasion Most anterior point of frontonasal suture in median plane
A A Point Deepest midline concavity on maxilla between anterior nasal spine and prosthion
B B Point Deepest midline concavity on mandibular symphysis between pogonion and infradentale
Gn Gnathion Most anteroinferior point on symphysis of chin
Zyg R Zygomaticotemporal - right Point of junction between right zygomatic and right suture temporal bone
Zyg L Zygomaticotemporal - left Point of junction between left zygomatic and left suture temporal bone
Go R Gonion - right Point of intersection between right ramal plane and right mandibular plane
Go L Gonion - left Point of intersection between left ramal plane and left mandibular plane
Co R Condylion - right Most central point of dorsal surface of right condylar head
Co L Condylion - left Most central point of dorsal surface of left condylar head
ZMX R Zygomaxillary suture - right Central point of right zygomatic suture
ZMX L Zygomaxillary suture - left Central point of left zygomatic suture

the landmark location, the operator, and the registration position and the desired 3D point coordinates. Re-
procedure. The magnitude of error varies greatly from cently, a methodology for identifying landmarks in
landmark to landmark and is usually an ellipse or ovoid calibrated and digitized x-rays and optical images
envelope of error. The major problem is that such errors (Acuscape and Sculptor, Glendora, Calif) has been
lead to missing subtle changes and allow only the under development and will soon be commercially
grossest changes to be observed clearly. available.24,25 In this study, we used this new program
Broadbent’s introduction of the cephalostat stressed and a traditional 2D evaluation to evaluate 9 dry skulls.
the importance of using and coordinating both the These methods are compared to measures on skulls in
lateral and frontal headfilms to define the craniofacial which the landmarks are located by external markers,
form.12 Although the method is sound, it has often been and the actual linear measures between 2 points (eg,
reported as difficult to apply and yields less accurate condyle to symphysis) are made externally with a
measurements than true anatomic values.13 In recent precision caliper. Earlier studies have evaluated the
years, computed tomography (CT) has provided 3D human skull with a 3D digitizing system and compared
reconstruction of the entire craniofacial skeleton14-17 that to physical caliper measures.26 Other investigators
from axial slices allowing methods to evaluate all have evaluated skulls using CT but often only with the
internal structures including the muscles.18 CT pro- transverse image, and not with a 3D reconstruction, or
vides a 3D method to assess some of the same measures not with an interest in orthodontic landmarks but with
as the 2D cephalogram and evaluate asymmetrical a physical anthropological focus.27-32 Some recent
mandibles.19,20 Magnetic resonance imaging recon- studies have compared cephalometric radiographs or
structed into 3D forms might provide a relatively CT images of skulls to physical measures. Those
accurate method to effectively determine growth and studies indicated differences between the 2D cephalo-
change, but it has not been studied. Unfortunately, both metric radiographic technique and the gold standard33
methods are too expensive and have limited availability but relatively good agreement between the 3D CT
to be used consistently by practicing orthodontists. image and the actual measures.34-39 One study com-
However, the conical x-ray approach, which effectively pared the traditional 2D cephalometric measures of 9
offers 3D reconstruction and equipment that is much dry human skulls to the same measures done in 3D CT
more affordable for a radiographic laboratory, is now a images, but only in terms of reliability.40 The latest
reality.21 approaches in 3D cephalometry have applied a matrix
The potential tools that can provide 3D evaluation transformation of the 3D coordinate system by using
of a patient seeking orthodontic treatment have been CT for evaluating human subjects or by applying a grid
used in stereophotogrammetry.22,23 Stereophotogram- to 3D images.41,42
metry is a technique developed by civil engineers to
measure 3D relationships of objects and terrains. A MATERIAL AND METHODS
second technique that complements this approach has Nine undocumented dry human skulls were ob-
been the “bundle adjustment method.” This method tained from the Anatomy Department at the University
self-calibrates the system, by using many measurement of California at San Francisco. Intact skulls of variable
points simultaneously, to calculate the imaging device morphology were selected. Radiopaque Beekley spots
American Journal of Orthodontics and Dentofacial Orthopedics Adams et al 399
Volume 126, Number 4

Table II. Radiographic image used to identify radiopaque landmarks and linear measures from these landmarks on
skulls
S Na A B Gn ZMX R Zyg R Co R Go R ZMX L Zyg L Co L

Go L Frontal Lateral Lateral Lateral Lateral Frontal Frontal Frontal Frontal Lateral Lateral Lateral
Co L Frontal Lateral Lateral Lateral Lateral Frontal Frontal Frontal Frontal Lateral Lateral
Zyg L Frontal Frontal Frontal Frontal Lateral Frontal Frontal Frontal Frontal Lateral
ZMX L Frontal Frontal Frontal Lateral Lateral Frontal Frontal Frontal Frontal
Go R Frontal Lateral Lateral Lateral Lateral Lateral Lateral Lateral
Co R Frontal Lateral Lateral Lateral Lateral Lateral Lateral
Zyg R Frontal Frontal Frontal Frontal Lateral Lateral
ZMX R Frontal Frontal Frontal Lateral Lateral
Gn Frontal Lateral Lateral Lateral
B Lateral Lateral Lateral
A Lateral Lateral
Na Lateral

(Beekley Corp, Bristol, Conn) were placed on 13 images into a 3D matrix to obtain a true Cartesian space
skeletal landmarks (Table I) as targets for both physical (x, y, z) reference system assigned to the patient. This
and radiographic measurements. The targets were lo- calibration began with imaging the calibration frame
cated based on criteria commonly used in 2D cephalo- and the skull in the same field of view from 2 or more
metrics and represented midsagittal and bilateral land- points of view. The locations of the bearings in the
marks. framework were known and used to assign a reference
The physical distance between landmarks was de- system to the calibration frame. The unknowns in-
termined by manual measurement with digital calipers cluded the skull and the source reference system. To
with a resolution of 0.03 mm (Pro Quality Precision compute the location of the source (7 degrees of
Dial Caliper, LS Starrett, Athol, Mass). To measure the freedom [DF]: x ⫽ width, y ⫽ height, z ⫽ depth, yaw,
error of the method, 3 skulls were selected for 2 pitch, roll, and focal length), the image sets are placed
additional measurement sessions. The mean distances, into a 3D matrix. The reference system from the
standard deviations, and variances were evaluated for calibration frame is then set or transmitted to the skull.
76 measurements to determine intraoperator reliability. When the known geometry of calibration markers is
Reliability (intraclass correlation) was calculated from fitted to the image of the calibration frame, there is a
the sums of squares in a 2-way analysis of variance geometric transformation of the known geometry of the
(ANOVA) test. calibration markers. This transformation is the key to
A special proprietary calibration framework mathematically locating the source with 7 DF. The next
(Acuscape, Glendora, Calif) was placed on the skull for geometric transformation locates the source relative to
the imaging sessions for taking x-rays. The framework the calibration frame with 7 DF and is repeated for all
included a headpiece with previously surveyed ra- images. After this calibration, all images have been
diopaque metal bearings. Each skull was placed in a placed into the same 3D matrix with the calibrated
cephalostat, and 3 images were acquired: standard frame reference. The reference system can then be
lateral (90°), frontal (0°), and oblique (45°). The adjusted and set to the skull anatomy to become the
standard lateral projections oriented the midsagittal “skull-specific reference.” This skull- or patient-spe-
plane parallel to the film. For frontal projections, the cific reference system will apply to all anatomy in the
midsagittal plane was set perpendicular to the film. 3D space that is represented (Fig 1).
Care was taken to avoid changes in the orientation of The 3 radiographic images represent 3 cephalomet-
the calibrated framework. The most appropriate radio- ric points of view and locations of the 3 sources. The
graphic projection for a given measurement was chosen resultant images are merged into the same 3D matrix,
(Table II). and landmarks are located on the images from 2 or
The films were scanned with a flatbed scanner more different points of view. Triangulation mathemat-
(ScanMaker 9600XL, Microtek, Carson, Calif) cali- ical calculations locate and record each landmark in
brated to determine precise pixels/mm in the horizontal space (x, y, z) with the skull or patient reference
(x) and vertical (y) directions. A transparency adapter system.
was used, and the scanner was set at a resolution of 150 Locating a landmark from 2 views will lead to an
dots per inch. The next procedure established all error envelope around the true anatomic landmark (Fig
400 Adams et al American Journal of Orthodontics and Dentofacial Orthopedics
October 2004

Fig 1. External reference system (blue box) becomes Fig 2. Changing from 2 to 3 views to locate anatomic
associated with source (head) and calibration frame (not landmark (black dot) in 3D space affects error envelope
shown but appears as multiple points across forehead). (outer line as circle or other shape).
A, Source (patient’s head) located with 7 DF (x, y, z,
yaw, pitch, roll, focal length) relative to calibration frame
first point (x1, y1, z1) to the second point (x2, y2, z2)
(not shown) by taking x-ray or photograph of patient. B,
When known geometry of calibration markers is fitted to
is registered as an identified landmark in the program.
image of calibration frame, geometric transformation of The notation (x1, y1, z1) represents values in 3D space.
virtual frame occurs. Geometric transformation then All measurements are then downloaded from Sculptor
locates source relative to calibration frame so that all into a spreadsheet for statistical analysis.
images are placed into 3D matrix. C, Patient’s specific A 2-way ANOVA without replication was per-
reference system will apply to all anatomy in 3D space. formed to evaluate the similarity of the 3 samples
separately (physical gold standard, conventional 2D,
and Sculptor 3D), resulting in intraclass correlations
2). In an orthogonal view, the error envelope is circular. (ICC) for each measure. The following equation was
In a coplanar view, the error margin perpendicular to used:
the beam angle reduces, while the error envelope
increases in the direction parallel to the beam angle. ICC ⫽ MSsubject ⫺ MSerror/MSsubject ⫹ 共r ⫹ 1兲 MSerror
The use of 3 planes of view reduces the error envelope where MSsubject is the mean square of the skulls from
to the best or smallest of all 3 points of view. the ANOVA, MSerror is the mean square error from the
During a Sculptor session, the 3D distance from the ANOVA, and r is the number of replicates.43
American Journal of Orthodontics and Dentofacial Orthopedics Adams et al 401
Volume 126, Number 4

Fig 3. Bland-Altman plot for measurement Go L-Go R. Thickest line (black) represents absolute
measure for that skull for that measurement set at 0. Mean (red dotted) and 2 SD (red solid) for
conventional 2D cephalometric measurement, and mean (blue dotted) and 2 SD (blue solid) for 3D
Sculptor measurement are shown.

The hypothesis was that the identification of land- cephalometric measures and then for the 3D Sculptor
marks and linear measurements, by using 3 planes of measures. In this example, all points were within the 2
space derived from 3 separate full-head x-rays taken at SD bounds for both the 2D and 3D methods.
different angles on dry human skulls with the Sculptor
system, would be closer in accuracy to the gold RESULTS
standard reference of physical measurements than the Repeated measures from the sums of squares in a
traditional 2D method. First, the differences in sample 2-way ANOVA without replication indicated excellent
means and standard error of the difference in sample reliability with a mean ICC of 0.995, with all 63
means were obtained for each measurement pair. Be- measurable pairs ⱖ 0.972 for the physical measure-
cause the ICC and variance indicated that the physical ments (Table III). This gives a mathematical value of
measurements were highly reliable, the paired t test was the similarity between measurements; this reflects the
used to assess the difference between Sculptor and the usefulness of this model to serve as a reference gold
conventional 2D measurement system.43 Second, a standard by which the remaining experiments can be
modified Bland-Altman method44 was applied, graph- completed.43 Values near 0 (or negative) indicate poor
ing the differences between physical (gold standard) reliability; values close to 1 indicate high reliability.
and 2D and 3D measurements (y-axis) versus physical The mean value of 0.995 indicated high reliability of
measurement (x-axis). This method illustrates the mag- the physical measures, as they were repeated on the
nitude of the differences (accuracy, bias) from anatom- landmarks and the measures among the landmarks.
ical truth and the distribution of points (precision, Two measures (S-Gn, S-B) were not completed be-
variability) of the measurement variables (Fig 3). The cause the anatomy of the skulls prevented using the
differences of the 2D cephalometric measures and the physical caliper. In comparison, the conventional 2D
3D Sculptor measures from the physical measures are method also indicated a high reproducibility margin
shown as individual points compared with the physical (mean ICC: 0.983) as did the 3D evaluation (mean ICC:
measurement for that skull. The mean ⫾ 2 SD is shown 0.998).
for the 9 measures from the 9 skulls for the 2D In comparing each skull measure between the
402 Adams et al American Journal of Orthodontics and Dentofacial Orthopedics
October 2004

Table III. Intraclass correlation values for physical measurements


S Na A B Gn ZMX R Zyg R Co R Go R ZMX L Zyg L Co L

Go L 0.998 0.997 0.996 0.998 0.998 0.999 0.999 0.997 0.999 0.998 0.996 0.972
Co L 0.999 0.997 0.998 0.996 0.996 0.999 0.999 0.999 0.999 0.992 0.994
Zyg L 0.997 0.999 0.994 0.999 0.999 0.999 0.999 0.999 0.999 0.996
ZMX L 0.999 0.997 0.995 0.987 0.999 0.999 0.999 0.999 0.999
Go R 0.994 0.998 0.997 0.999 0.998 0.999 0.999 0.987
Co R 0.998 0.992 0.992 0.998 0.999 0.996 0.994
Zyg R 0.982 0.976 0.992 0.983 0.992 0.972
ZMX R 0.998 0.999 0.997 0.995 0.997
Gn cnm 0.999 0.995 0.999
B cnm 0.998 0.984
A 0.997 0.997
Na 0.998

cnm, Could not be measured.

conventional 2D approach and the physical measure- sures significantly varied from the physical measures
ments, the conventional approach ranged from under- (Table IV, Fig 4; Na-B). Other 2D cephalometric
estimating (Gn-Zyg R; ⫺17.68 mm) to overestimating measures were relatively near the physical measures
(Zyg L-Zyg R; ⫹15.52 mm) the true values of the gold (Fig 5; Gn-Co L). Some 2D cephalometric measures
standard. The paired t test indicated that 70 of the 76 also had much more range in their differences from
measures were significantly different from the physical their physical measures, suggesting that this measure
measurements at ␣ ⫽ 0.05 (Table IV). By chance and its orientation in space could markedly affect the
alone, only 3 or 4 significant measurements were 2D cephalometric image and length more than would
expected. Five of the 6 measurements that did not be expected (Fig 6; A point-Co L).
indicate a significant difference have a coefficient of In comparing each measure from the 3D approach
variation (SD/mean) greater than 1, indicating that the with the physical measurements, the 3D approach had a
SD is greater than the mean, exhibiting excessive much smaller range from underestimating (A-Co R;
variability. Ranking the measurements according to ⫺3.99 mm) to overestimating (A-ZMX; ⫹2.96 mm).
mean difference (Table V) showed a pattern that Comparison with the paired t test indicated that 67 of
depended on several factors (lateral position of the the 76 measures were significantly different between
anatomical point, plane of measure in the frontal the 3D and physical measures (Table VI). Of the 9
cephalometric image). The 2D cephalometric measures measures that did not differ significantly, only 1 had a
of lateral anatomical points in the frontal plane dem- coefficient of variation greater than 1, resulting in
onstrated some of the greatest differences from the excessive variation in the measure. The mean 3D
physical measure. Measures with 1 point lateral (Zyg, differences from true for 75 of 76 measures were
Go, Co) and the second near the midline (Na, Gn) had negative. Ranking the measurements according to mean
much larger differences from the physical measure with difference showed that measures from a midline point
the 2D cephalometric analysis. to a much more lateral anatomical point were more
The Bland-Altman plot—illustrating the magni- variable when the lateral point was to the right (Table
tude of the difference in the anatomic truth and VII).
distribution of points—showed low precision (high The Bland-Altman plots showed high precision
variability) for the conventional 2D method and low with the Sculptor method, yet with some minor inac-
accuracy (high bias) compared with the physical curacies with a mean of ⫺1.05 ⫾ 0.54 mm difference
measurements (Figs 4 and 5). Some plots for indi- from the physical measures indicating slight underesti-
vidual measures showed that the 2D method had high mation of the anatomical truth (Fig 7). The mean
variability and high bias, sometimes overestimating difference ranged from ⫺2.49 (Na-Co R) to ⫹0.30 mm
and sometimes underestimating. The measurement (A-ZMX L); 75 of 76 mean differences were negative
points were not in a pattern (linear or quadratic (67 were statistically significant), and the 1 positive
errors) in each plot or between plots, signifying no mean difference was not statistically significant from 0.
relationship between actual length of the measure- The Sculptor 3D approach was usually 4 or 5 times
ment and mean difference between 2D measures and closer to the physical measures than the conventional
the physical measures. Some 2D cephalometric mea- 2D approach (Table VIII) and demonstrated much less
American Journal of Orthodontics and Dentofacial Orthopedics Adams et al 403
Volume 126, Number 4

Table IV.76 measures that compared conventional 2D Table IV. continued


cephalographic measures to gold standard of physical
measurements Mean difference SD Minimum Maximum
Measurement (mm) (mm) (mm) (mm)
Mean difference SD Minimum Maximum
Measurement (mm) (mm) (mm) (mm) ZMX R-Zyg R ⫺2.38 1.93 ⫺6.32 0.41
Zyg R-Go L 6.93 3.48 1.84 13.32
S-Go L 5.50 2.58 0.16 8.77 Zyg R-Co L 7.35 2.43 4.40 12.08
S-Co L 4.66 0.93 3.55 5.84 Zyg R-Zyg L 13.61 2.10 8.33 15.52
S-Zyg L 2.78 2.06 ⫺1.78 4.77 Zyg R-ZMX L 10.79 1.22 9.18 12.57
S-ZMX L ⫺10.97 2.83 ⫺14.27 ⫺6.32 Zyg R-Go R 5.32 0.91 4.24 6.73
S-Go R 8.36 1.65 6.05 10.92 Zyg R-Co R 3.83 0.44 3.05 4.65
S-Co R 4.93 0.96 4.02 5.93 Co R-Go L 10.81 1.29 9.32 12.86
S-Zyg R 4.00 2.60 1.17 10.07 Co R-Co L 10.75 0.91 9.27 12.35
S-ZMX R ⫺11.24 3.91 ⫺16.84 ⫺4.33 Co R-Zyg L 8.44 0.95 7.09 9.91
S-Gn cnm Co R-ZMX L ⫺2.60 2.17 ⫺5.26 1.54
S-B cnm Co R-Go R 4.24 1.49 2.57 5.90
S-A 7.84 1.15 6.50 9.46 Go R-Go L 9.01 0.97 7.57 10.14
S-Na 6.22 0.84 5.51 7.41 Go R-Co L 9.48 5.22 ⫺2.66 14.42
Na-Go L ⫺0.54 1.81* ⫺3.00 1.92 Go R-Zyg L 9.01 2.36 4.45 12.30
Na-Co L ⫺8.36 2.23 ⫺11.67 ⫺5.32 Go R-ZMX L ⫺2.19 3.66* ⫺6.39 5.68
Na-Zyg L ⫺4.49 1.93 ⫺7.15 ⫺1.32 ZMX L-Go L 3.85 0.82 2.26 4.82
Na-ZMX L 5.42 1.44 3.47 8.65 ZMX L-Co L 2.26 1.27 0.56 3.98
Na-Go R 1.20 1.60* ⫺1.55 2.69 ZMX L-Zyg L ⫺2.04 1.46 ⫺4.24 0.07
Na-Co R ⫺11.38 2.19 ⫺15.50 ⫺7.94 Zyg L-Go L 1.60 1.28 ⫺0.73 3.40
Na-Zyg R ⫺3.09 3.10* ⫺9.48 1.29 Zyg L-Co L 1.17 0.66 0.18 2.08
Na-ZMX R 5.32 1.82 1.32 6.97 Co L-Go L 2.11 1.19 ⫺0.46 3.26
Na-Gn 9.82 0.51 8.88 10.40
Na-B 8.32 0.32 7.73 8.74 Positive mean values indicate differences (biases) from physical caliper
Na-A 4.99 0.25 4.70 5.35 (gold standard), measure overestimate; negative mean values indicate
A-Go L ⫺3.96 1.47 ⫺6.37 ⫺2.05 differences (biases) from underestimate; cnm, could not be measured.
A-Co L ⫺7.39 3.42 ⫺13.20 ⫺3.72 *Coefficient of variation (SD/mean) ⬎1; all were significantly different
A-Zyg L ⫺5.04 2.49 ⫺9.58 ⫺1.46 from 0 (biased) with paired t test (P ⬍ .05) except Na-Go L/R, B-Co L,
A-ZMX L 3.76 1.70 0.83 6.41 Gn-Co L, ZMX R-Go L, ZMX L-Go R.
A-Go R ⫺6.25 2.70 ⫺11.83 ⫺3.60
A-Co R ⫺10.80 1.79 ⫺12.47 ⫺7.58
A-Zyg R ⫺3.20 3.60* ⫺8.88 2.11 variability (mean SD: 0.54 mm) than the 2D (mean SD:
A-ZMX R 3.89 1.86 1.14 6.95
A-Gn 4.63 0.62 3.51 5.49 6.94 mm).
A-B 3.11 0.55 1.94 3.69
B-Go L ⫺5.28 2.60 ⫺8.21 ⫺1.22 DISCUSSION
B-Co L ⫺1.90 2.50* ⫺3.98 3.84 This is the first study that has compared physical
B-Zyg L 4.31 4.78* ⫺5.45 10.22 measurements of the human skull to both the standard
B-ZMX L ⫺6.83 2.76 ⫺10.79 ⫺2.24
B-Go R ⫺9.01 2.74 ⫺12.34 ⫺4.35 2D cephalometric analysis and a newly developed 3D
B-Co R ⫺8.75 2.83 ⫺13.22 ⫺3.45 system that also uses radiographs. The landmarks were
B-Zyg R 5.81 5.04 ⫺6.00 10.29 chosen by an expert craniofacial radiologist (D.C.H.),
B-ZMX R ⫺10.09 2.44 ⫺13.89 ⫺6.21 who has spent over 20 years evaluating x-rays, CT, and
B-Gn 1.54 0.36 0.93 2.02 magnetic resonance imaging of the head. An additional
Gn-Go L ⫺4.32 2.51 ⫺7.60 ⫺0.53
Gn-Co L ⫺2.31 3.04* ⫺7.38 1.15 strength of the project is that this investigator chose
Gn-Zyg L ⫺7.36 2.55 ⫺11.79 ⫺4.69 landmarks based on years of determining information
Gn-ZMX L ⫺3.17 2.46 ⫺6.67 0.59 from radiographs. By identifying each landmark di-
Gn-Go R ⫺10.19 3.56 ⫺15.52 ⫺4.81 rectly on the skull, errors of identification were elimi-
Gn-Co R ⫺6.25 2.90 ⫺11.17 ⫺1.21 nated. By applying physical measurements to these
Gn-Zyg R ⫺12.13 2.65 ⫺17.68 ⫺8.58
Gn-ZMX R ⫺6.12 2.68 ⫺11.23 ⫺2.02 known and marked sites, the actual anatomical dis-
ZMX R-Go L ⫺3.59 6.37* ⫺10.71 7.29 tances could be established as a gold standard with
ZMX R-Co L ⫺5.23 3.01 ⫺10.07 ⫺1.10 which the 2D and the 3D evaluations of the radiographs
ZMX R-Zyg L 9.80 1.40 7.90 11.70 could be compared to determine their true accuracy.
ZMX R-ZMX L 12.15 1.16 9.38 13.60 In the Acuscape program, the most important aspect
ZMX R-Go R 7.97 1.02 6.83 9.89
ZMX R-Co R 4.78 1.70 1.88 7.21
to enhance its optimal accuracy is the calibration
framework. It must not move on the subject during the
404 Adams et al American Journal of Orthodontics and Dentofacial Orthopedics
October 2004

Table V. Example of 25 measures that compared greatest differences when compared with the true phys-
conventional 2D cephalographic measures to physical ical measures. This would be expected in the lateral 2D
measures ranked by largest to smallest mean differences cephalometric image, but the great differences were
Mean particularly evident in measures from the frontal 2D
difference SD image (eg, Zyg L-Zyg R, Co L-Co R). These large
Measurement (mm) (mm) differences can be explained because the actual plane
Zyg L-Zyg R 13.61 2.10 for each measure was not at the midpoint of the frontal
Gn-Zyg R 12.13 2.65 plane projected for the 2D x-ray image. When the range
Na-Co R 11.38 2.19 of measures was considered instead of the difference
Co L-Co R 10.75 0.91 from the physical measure, some 2D cephalometric
Gn-Go R 10.19 3.56
measures had much greater variation (eg, Na-Zyg R),
Na-Gn 9.82 0.51
B-Go R 9.01 2.74 suggesting that a particular measure and its orientation
Go L-Go R 9.01 0.97 to the film could be much more effective in providing
B-Co R 8.75 2.83 a spurious length measurement. The concept of linear
Na-Co L 8.36 2.23 projection transformation is important to deciphering
Na-B 8.32 0.32
why 2D cephalometric measures can be erroneous.45,46
Gn-Zyg L 7.36 2.55
Gn-Co R 6.25 2.90 The Sculptor 3D approach had 2 key advantages.
S-Na 6.22 0.84 First, it had high precision with a greatly reduced bias
B-Go L 5.28 2.60 to the gold standard. The consistency of the bias
Na-A 4.99 0.25 (slightly negative) suggests that engineering the soft-
A-Gn 4.63 0.62
ware or calibration framework, or a simple mathemat-
Na-Zyg L 4.49 1.93
Gn-Go L 4.32 2.51 ical correction factor, can correct the bias. Second, the
A-B 3.11 0.55 Sculptor 3D approach had much less variability in its
Na-Zyg R 3.09 3.10 measure compared with the conventional 2D approach:
Gn-Co L 2.31 3.04 it is more consistent. The greater variability of the 2D
B-Co L 1.90 2.50
measure reflects the inherent problems of representing
Na-Go R 1.20 1.60
Na-Go L 0.54 1.81 a linear measure in 3D space along a 2D axis. Some
measures markedly underestimate or overestimate the
actual distance. For example, measuring Zyg R-Zyg L
record taking, must remain intact during nonuse, and from the frontal x-ray overestimates the actual distance
must be sturdy enough to handle repetitive use. Any across the skull. The magnitude of enlargement or
movement of the framework while taking pictures and overestimation is related to the distances between the
x-rays can alter the geometric relationship between the focus, the object, and the film. The closer these points
same points on various images. Normally, this should are to the source and the greater the distance between
be a random, nondirectional bias. these points (at a 90° plane), the greater the enlarge-
Most 2D cephalometric measures were not accu- ment from the anatomic truth. The rotation of the head
rate, but some measures were relatively close to the can also alter the measurement between 2 anatomical
physical measure. The cephalometric image of the left sites as seen in the 2D cephalogram. In addition,
side will be magnified 5%-10%, depending on the traditional 2D cephalometric analysis usually does not
distance of the anatomical site from the film plane. correct for magnification error. Most clinicians analyze
When a measurement extends from the midline (eg, Na, a lateral headfilm without adjusting for the magnifica-
Gn) to an anatomical site nearer the film plane (eg, Go tion that can vary from 7% to 12% in most units. We
L, Zyg L, Co L) in the 2D method, the inaccuracy of the used the 2D cephalometric approach as a clinician
measurement is compensated by the magnification, normally would. Even if we adjusted each 2D cepha-
resulting in a value closer to the real measure. When lometric measure by 10%, on a 10-mm difference from
bilateral landmarks are considered, the magnification a true measure defined by the physical caliper, it would
becomes an average of the 2 sides (left, 3%-5%; right, correct the measure only to a 9-mm difference. In
12%-15%; average, 7%-9%). An unexpected finding contrast, the calibration head unit of the Sculptor
was that some measures of 2 midline points (Na-Gn, system corrects for both head rotation and magnifica-
Na-B) showed a large difference from the physical tion on both sides, and does not just correct for an
caliper measure. It was expected that the anatomical average of 2 superimposed images as is done with the
sites in the most lateral positions of the anatomy (eg, 2D cephalometric approach. The 3D Sculptor system
zygomatic arch, gonial angle, condyles) would have the corrects for both bilateral structures. Effective evalua-
American Journal of Orthodontics and Dentofacial Orthopedics Adams et al 405
Volume 126, Number 4

Fig 4. Bland-Altman plot for measurement Na-B. 2D cephalometric measure (green circle) is
markedly different from physical measure of same skull, but 3D Sculptor measure (yellow diamond)
is relatively near true measure. Thickest line (black) represents physical measurement for each of 9
skulls and is set at 0. Mean and 2 SDs for the conventional 2D measurements (red), and mean and
2 SDs for 3D measurement (purple) are shown.

Fig 5. Bland-Altman plot for measurement Gn-Co L. 2D cephalometric measure (green circle) is
relatively near both physical measure of same skull and 3D Sculptor measure (yellow diamond) for
all but 2 skulls. Thickest line (black) represents physical measurement for each of 9 skulls and is set
at 0. The mean and 2 SD for conventional 2D measurements (red), and mean and 2 SDs for 3D
measurement (purple) are shown.
406 Adams et al American Journal of Orthodontics and Dentofacial Orthopedics
October 2004

Fig 6. Bland-Altman plot for measurement A-Co L. 2D cephalometric measure (green circle) had
large range from physical measure of same skull, but 3D Sculptor measure (yellow diamond) was
relatively near true measure. Thickest line (black) represents physical measurement for each of 9
skulls and is set at 0. Mean and 2 SDs for conventional 2D measurements (red), and mean and 2
SDs for 3D measurement (purple) are shown.

Fig 7. Scatter plot with smoothed lines shows how mean the 76 measurements differ from physical
measurements with 2D cephalometric (red) or 3D Sculptor (blue) approach. Mean of physical
measurements set at 0, and mean of same measurements taken with 2 approaches is compared.
Note that means for Sculptor values are much closer to physical measurements.
American Journal of Orthodontics and Dentofacial Orthopedics Adams et al 407
Volume 126, Number 4

Table VI.76 measures that compared Sculptor 3D Table VI. continued


cephalographic measures to gold standard of physical
measurements Mean
difference SD Minimum Maximum
Mean Measurement (mm) (mm) (mm) (mm)
difference SD Minimum Maximum
Measurement (mm) (mm) (mm) (mm) Zyg R-Co L ⫺1.06 0.33 ⫺1.58 ⫺0.69
Zyg R-Zyg L ⫺1.01 0.32 ⫺1.61 ⫺0.51
S-Go L ⫺1.27 0.34 ⫺1.72 ⫺0.91 Zyg R-ZMX L ⫺1.45 0.55 ⫺2.51 ⫺0.63
S-Co L ⫺0.65 0.73* ⫺1.84 0.30 Zyg R-Go R ⫺0.76 0.46 ⫺1.34 0.28
S-Zyg L ⫺1.19 0.66 ⫺2.75 ⫺0.51 Zyg R-Co R ⫺0.81 0.54 ⫺1.74 ⫺0.18
S-ZMX L ⫺1.30 0.57 ⫺2.09 ⫺0.39 Co R-Go L ⫺0.81 0.49 ⫺1.53 ⫺0.20
S-Go R ⫺0.59 0.88 ⫺1.33 1.47 Co R-Co L ⫺0.99 0.57 ⫺2.30 ⫺0.40
S-Co R ⫺0.85 0.08 ⫺0.91 0.76 Co R-Zyg L ⫺1.75 0.37 ⫺2.47 ⫺1.33
S-Zyg R ⫺0.11 0.42* ⫺0.52 0.80 Co R-ZMX L ⫺2.09 0.61 ⫺2.86 ⫺1.24
S-ZMX R ⫺0.33 0.32* ⫺0.82 0.28 Co R-Go R ⫺0.38 0.33 ⫺0.87 0.06
S-Gn cnm Go R-Go L ⫺0.57 0.47 ⫺1.37 0.14
S-B cnm Go R-Co L ⫺0.93 0.47 ⫺1.95 ⫺0.40
S-A ⫺0.57 1.41* ⫺2.17 1.53 Go R-Zyg L ⫺1.45 0.56 ⫺2.56 ⫺0.80
Go R-ZMX L ⫺1.94 0.52 ⫺3.05 ⫺1.44
S-Na ⫺0.44 0.74* ⫺1.05 0.97
ZMX L-Go L ⫺1.50 0.65 ⫺2.49 ⫺0.72
Na-Go L ⫺1.72 0.41 ⫺2.27 ⫺1.07
ZMX L-Co L ⫺1.39 0.38 ⫺2.20 ⫺0.99
Na-Co L ⫺1.43 0.53 ⫺2.54 ⫺0.90
ZMX L-Zyg L ⫺0.55 0.35 ⫺1.36 ⫺0.22
Na-Zyg L ⫺1.03 0.26 ⫺1.49 ⫺0.78
Zyg L-Go L ⫺1.07 0.63 ⫺2.04 ⫺0.31
Na-ZMX L ⫺0.53 0.26 ⫺0.95 ⫺0.13
Zyg L-Co L ⫺1.01 0.38 ⫺1.54 ⫺0.36
Na-Go R ⫺2.17 0.80 ⫺3.36 0.55 Co L-Go L ⫺0.43 0.33 ⫺0.95 0.08
Na-Co R ⫺2.49 0.49 ⫺3.26 ⫺1.81
Na-Zyg R ⫺1.78 0.48 ⫺2.73 ⫺1.02 Positive mean values indicate differences (biases) from physical caliper
Na-ZMX R ⫺0.89 0.43 ⫺1.89 ⫺0.48 (gold standard) measure overestimate; negative mean values indicate
Na-Gn ⫺1.12 0.35 ⫺1.40 ⫺0.30 differences (biases) from underestimate; cnm, could not be measured.
Na-B ⫺0.81 0.22 ⫺1.08 ⫺0.45 *Coefficient of variation (SD/mean) ⬎1; all were significantly different
Na-A ⫺0.45 0.28 ⫺0.82 ⫺0.06 from 0 (biased) with paired t test (P ⬍ .05) except S-Go R, S-Zyg R,
A-Go L ⫺1.16 0.87 ⫺2.58 ⫺0.14 S-A, S-Na, A-Zyg L, A-ZMX L, B-ZMX R, B-Gn, Gn-ZMX L.
A-Co L ⫺1.15 0.62 ⫺2.00 ⫺0.15
A-Zyg L ⫺0.47 0.62* ⫺1.27 ⫺0.75
A-ZMX L 0.30 1.20* ⫺1.16 2.96 Table VII. Example of 25 measures that compared
A-Go R ⫺2.35 0.80 ⫺3.51 ⫺1.22 Sculptor 3D cephalographic measures to gold
A-Co R ⫺2.24 0.92 ⫺3.99 ⫺1.09 standard of physical measurements ranked by largest
A-Zyg R ⫺1.43 0.45 ⫺2.36 ⫺0.89 to smallest mean differences
A-ZMX R ⫺0.66 0.33 ⫺1.17 ⫺0.01
A-Gn ⫺0.97 0.40 ⫺1.75 ⫺0.43 Mean difference
A-B ⫺0.75 0.32 ⫺1.22 ⫺0.33 Measurement (mm) SD (mm)
B-Go L ⫺0.72 0.37 ⫺1.21 ⫺0.10
B-Co L ⫺0.82 0.29 ⫺1.21 ⫺0.52 Na-Co R 2.49 0.49
B-Zyg L ⫺0.59 0.32 ⫺1.06 ⫺0.19 Na-Go R 2.17 0.80
B-ZMX L ⫺0.66 0.31 ⫺1.13 ⫺0.35 Gn-Go R 2.01 0.36
B-Go R ⫺1.78 0.57 ⫺2.65 ⫺0.99 Gn-Co R 2.01 0.32
Na-Zyg R 1.78 0.48
B-Co R ⫺1.66 0.61 ⫺2.47 ⫺0.54
B-Go R 1.78 0.57
B-Zyg R ⫺0.79 0.65 ⫺1.49 0.33
Na-Go L 1.72 0.41
B-ZMX R ⫺0.31 0.44* ⫺0.91 ⫺0.26
B-Co R 1.66 0.61
B-Gn ⫺0.28 0.32* ⫺0.76 0.14
Na-Co L 1.43 0.53
Gn-Go L ⫺0.65 0.35 ⫺1.16 0.09
Gn-Zyg R 1.20 0.34
Gn-Co L ⫺0.97 0.31 ⫺1.53 ⫺0.51 Na-Gn 1.12 0.35
Gn-Zyg L ⫺0.85 0.48 ⫺1.69 ⫺0.11 Na-Zyg L 1.03 0.26
Gn-ZMX L ⫺0.60 0.86* ⫺2.07 1.17 Zyg L-Zyg R 1.01 0.32
Gn-Go R ⫺2.01 0.36 ⫺2.71 ⫺1.57 Co L-Co R 0.99 0.57
Gn-Co R ⫺2.01 0.32 ⫺2.50 ⫺1.60 A-Gn 0.97 0.40
Gn-Zyg R ⫺1.20 0.34 ⫺1.56 ⫺0.79 Gn-Co L 0.97 0.31
Gn-ZMX R ⫺0.65 0.42 ⫺1.17 ⫺0.03 Gn-Zyg L 0.85 0.48
ZMX R-Go L ⫺0.84 0.23 ⫺1.21 ⫺0.55 B-Co L 0.82 0.29
ZMX R-CoL ⫺0.92 0.26 ⫺1.33 ⫺0.55 Na-B 0.81 0.22
ZMX R-Zyg L ⫺0.67 0.23 ⫺0.97 ⫺0.35 A-B 0.75 0.32
ZMX R-ZMX L ⫺0.57 0.57 ⫺1.06 0.83 B-Go L 0.72 0.37
ZMX R-Go R ⫺1.65 0.55 ⫺2.70 ⫺0.84 Gn-Go L 0.65 0.35
ZMX R-Co R ⫺1.74 0.76 ⫺2.88 ⫺0.71 Go L-Go R 0.57 0.47
ZMX R-Zyg R ⫺0.93 0.36 ⫺1.45 ⫺0.41 Na-A 0.45 0.28
Zyg R-Go L ⫺0.98 0.36 ⫺1.40 ⫺0.34 S-Na 0.44 0.74
408 Adams et al American Journal of Orthodontics and Dentofacial Orthopedics
October 2004

Subset of absolute measures (mm) that compared conventional 2D cephalographic measurements and
Table VIII.
3D Sculptor measurements to gold standard of physical measurements from human skulls (n ⫽ 9)
Mean ⫾ SD Range

Measurement Physical 2D Ceph 3D Sculptor Physical 2D Ceph 3D Sculptor

Na-Go L 110.17 ⫾ 6.35 109.63 ⫾ 7.00 108.45 ⫾ 6.56 103–122 101–123 101–121
Na-Zyg L 73.22 ⫾ 3.73 68.73 ⫾ 3.64 72.19 ⫾ 3.78 70–81 63–75 69–80
Na-Gn 105.08 ⫾ 5.41 114.93 ⫾ 5.69 103.96 ⫾ 5.27 98–114 107–124 97–112
A-B 37.64 ⫾ 2.83 40.75 ⫾ 3.03 36.90 ⫾ 2.75 34–42 37–46 33–41
Gn-Go R 81.04 ⫾ 3.98 70.84 ⫾ 5.74 79.03 ⫾ 3.64 77–89 62–80 75–86
Gn-Co L 113.85 ⫾ 4.79 111.55 ⫾ 5.81 112.89 ⫾ 4.76 106–122 104–122 105–121
GoL-Go R 87.11 ⫾ 7.36 96.12 ⫾ 8.08 86.54 ⫾ 7.19 77–98 84–109 77–98
Zyg L-Zyg R 113.67 ⫾ 5.07 127.29 ⫾ 6.07 112.67 ⫾ 5.01 108–122 121–136 107–121

tion of craniofacial growth will require more accurate 9. Savara BS, Tracey WE, Miller PA. Analysis of errors in
measures in 3 dimensions to be anatomically cor- cephalometric measurements of three-dimensional distances on
the human mandible. Arch Oral Biol 1966;11:209-17.
rect.47,48 10. Gravely JF, Benzies PM. The clinical significance of tracing
error in cephalometry. Br J Orthod 1974;1:95-101.
CONCLUSIONS 11. Mitgaard J, Bjork A, Linder-Aronson S. Reproducibility of
When using a gold standard of physical measure- cephalometric landmarks and errors of measurement of cepha-
ments from the dry skulls, cephalometric evaluation of lometric cranial distances. Angle Orthod 1974;44:56-61.
12. Broadbent BH. A new x-ray technique and its application to
radiographic images taken in the conventional 2D orthodontia. Angle Orthod 1931;1:45.
system often renders both inaccurate and imprecise 13. Grayson B, Cutting C, Bookstein FL, Kim H, McCarthy JG. The
measurements. In contrast, the relatively new 3D ceph- three-dimensional cephalogram: theory, technique, and clinical
alometric system (Sculptor) provides a much more application. Am J Orthod Dentofacial Orthop 1988;94:327-37.
precise evaluation of linear measures and only slightly 14. Maki K, Shibasaki Y, Fukuhara T. A new approach to mandib-
ular growth employing x-ray CT. Dent Jpn (Tokyo) 1989;26:77-
inaccurate measures that are underestimated by about
80.
1.0 mm. The cephalometric measures from the 3D 15. Maki K, Okano T, Morohashi T, Yamada S, Shibasaki Y. The
Sculptor provide a much more valuable tool in assess- application of three-dimensional quantitative computed tomog-
ing growth and the effects of orthodontic treatment. raphy to the maxillofacial skeleton. J Dent Maxillofac Radiol
The program is used on personal computers (DOS 1997;26:39-44.
16. Maki K, Miller AJ, Okano T, Shibasaki Y. Changes in cortical
format) and requires a calibrated headpiece to be worn
bone mineralization in the developing mandible: a three-dimen-
by a subject while the radiographs and pictures are sional quantitative computed tomography study. J Bone Miner
collected. Its relative ease of use, with its powerful Res 2000;15:700-9.
calculation and self-calibrating software, means that it 17. Vannier MW, Marsh JL, Warren JO. Three-dimensional CT
is a readily accessible program for the clinician. reconstruction images for craniofacial surgical planning and
evaluation. Radiol 1984;150:179-85.
18. Maki K, Miller AJ, Okano T, Shibasaki Y. A three-dimensional,
REFERENCES quantitative computed tomographic study of changes in distribu-
1. Moyers RE, Bookstein FL, Hunter WS. Analysis of the cranio- tion of bone mineralization in the developing human mandible.
facial skeleton: Cephalometrics. In: Moyers RE, editor. Hand- Arch Oral Biol 2001;46:667-78.
book of orthodontics. Chicago: Yearbook; 1988. 247-309. 19. Maki K, Miller AJ, Okano T, Hatcher D, Yamaguchi T,
2. Baumrind S, Frantz TC. The reliability of head film measure- Kobayashi H, et al. Cortical bone mineral density in asymmet-
ments. 1. Landmark identification. Am J Orthod 1971;60:111-27. rical mandibles: a three-dimensional quantitative computed to-
3. Adams JW. Correction of error in cephalometric roentgeno- mography (QCT) study. Eur J Orthod 2001;23:217-32.
grams. Angle Orthod 1940;10:3-13. 20. Hatcher DC, McEvoy SP, Mah RT, Faulkner MG. Distribution
4. Brodie AG. Cephalometric roentgenology: history, technique of local and general stresses in the stomatognathic system. In:
and uses. J Oral Surg 1949;7:185-98. McNeill C, editor. Science and practice of occlusion. Chicago:
5. Hixon EH. Cephalometrics and longitudinal research. Am J Quintessence; 1997. 259-70.
Orthod 1960;46:36-42. 21. Mozzo P, Procacci C, Tacconi A, Martini PT, Andreis IA. A new
6. Salzmann JA. Limitations of roentgenographic cephalometrics. volumetric CT machine for dental imaging based on the cone-
Am J Orthod 1964;50:169-88. beam technique: preliminary results. Eur Radiol 1998;8:1558-64.
7. Bjork A, Solow B. Measurements on radiographs. J Dent Res 22. Burke PH, Banks P, Beard LF, Tee JE, Hughes C. Stereophoto-
1962;41:672-83. graphic measurement of change in facial soft tissue morphology
8. Hixon EH. The norm concept and cephalometrics. Am J Orthod following surgery. Br J Oral Surg 1983;21:237-45.
1956;42:898-919. 23. Burke PH, Beard LF. Stereophotogrammetry of the face: A
American Journal of Orthodontics and Dentofacial Orthopedics Adams et al 409
Volume 126, Number 4

preliminary investigation into the accuracy of a simplified system experimental validation using phantom objects. Changgeng Yi
evolved for contour mapping by photography. Am J Orthod Xue Za Zhi 2000;23:354-9.
1967;53:769-82. 37. Cavalcanti MG, Haller JW, Vannier MW. Three-dimensional
24. Quintero J-C. Validation of a computer-aided approach to computed tomography landmark measurement in craniofacial
three-dimensional craniofacial imaging [thesis]. San Francisco: surgical planning: experimental validation in vitro. J Oral Max-
University of California at San Francisco; 1998. illofac Surg 1999;57:690-4.
25. Adams GL. A comparison between traditional two-dimensional 38. Kim DO, Kim HJ, Jung H, Jeong HK, Hong SI, Kim KD.
cephalometry and a three-dimensional approach [thesis]. San Quantitative evaluation of acquisition parameters in the three-
Francisco: University of California at San Francisco; 2000. dimensional imaging with multidetector computed tomography
26. Hildebolt CF, Vannier MW. Three-dimensional measurement using human skull phantom. J Digit Imaging 2002;15(Suppl
accuracy of skull surface landmarks. Am J Phys Anthropol 1):254-7.
1988;76:497-503. 39. Jung H, Kim HJ, Kim DO, Hong SI, Jeong HK, Kim KD, et al.
27. Waitzman AA, Posnick JC, Armstrong DC, Pron GE. Craniofa- Quantitative analysis of three-dimensional rendered imaging of
cial skeletal measurements based on computed tomography: part the human skull acquired from multi-detector row computed
I. Accuracy and reproducibility. Cleft Palate Craniofac J 1992; tomography. J Digit Imaging 2002;15:232-9.
29:112-7. 40. Kragskov J, Bosch C, Gyldensted C, Sindet-Pedersen S. Com-
28. Richtsmeier JT, Paik CH, Elfert PC, Cole TM, Dahlman HR. parison of the reliability of craniofacial anatomic landmarks
Precision, repeatability, and validation of the localization of
based on cephalometric radiographs and three-dimensional CT
cranial landmarks using computed tomography scans. Cleft
scans. Cleft Palate Craniofac J 1997;34:111-6.
Palate Craniofac J 1995;32:217-27.
41. Kitaura H, Yonetsu K, Kitamori H, Kobayashi K, Nakamura T.
29. Nagashima M, Inoue K, Sasaki T, Miyasaka K, Matsumura G,
Standardization of the 3-D CT measurement for length and
Kodama G. Three-dimensional imaging and osteometry of adult
angles by matrix transformation in the 3-D coordinate system.
human skulls using helical computed tomography. Surg Radiol
Cleft Palate Craniofac J 2000;37:349-56.
Anat 1998;20:291-7.
42. Haffner CL, Pessa JE, Zadoo VP, Garza JR. A technique for
30. Fuhrmann RA, Schnappauf A, Diedrich PR. Three-dimensional
three-dimensional cephalometric analysis as an aid in evaluating
imaging of craniomaxillofacial structures with a standard per-
changes in the craniofacial skeleton. Angle Orthod 1999;69:345-8.
sonal computer. Dentomaxillofac Radiol 1995;24:260-63.
31. Furst IM, Austin P, Pharoah M, Mahoney J. The use of computed 43. Landis JR, Koch GG. A review of statistical methods in the
tomography to define zygomatic complex position. J Oral Max- analysis of data arising from observer reliability studies (part I).
illofac Surg 2001;59:647-54. Statistica Neerlandica 1975;29:101-23.
32. Berlis A, Putz R, Schumacher M. Direct and CT measurements 44. Bland JM, Altman DG. Statistical methods for assessing agree-
of canals and foramina of the skull base. Br J Radiol 1992;65: ment between two methods of clinical measurement. Lancet
653-61. 1986;1:307-10.
33. Kusnoto B, Evans CA, BeGole EA, de Rijk W. Assessment of 45. Tsao DH, Kazanoglu A, McCasland JP. Measurability of radio-
3-dimensional computer-generated cephalometric measure- graphic images. Am J Orthod 1983;84:212-6.
ments. Am J Orthod Dentofacial Orthop 1999;116:390-9. 46. Tsao DH. Graphic description of figure representation of the
34. Chidiac JJ, Shofer FS, Al-Kutoub A, Laster LL, Ghafari J. teeth and mandible in centric occlusion. J Prosthet Dent 1982;
Comparison of CT scanograms and cephalometric radiographs in 47:95-100.
craniofacial imaging. Orthod Craniofac Res 2002;5:104-13. 47. Harrell WE, Hatcher DC, Bolt RL. In search of anatomic truth:
35. Hildebolt CF, Vannier MW, Knapp RH. Validation study of skull 3-dimensional digital modeling and the future of orthodontics.
three-dimensional computerized tomography measurements. Am J Orthod Dentofacial Orthop 2002;122:325-30.
Am J Phys Anthropol 1990;82:283-94. 48. Harrell WE. The future of imaging for orthodontics. In: Riolo M,
36. Lo LJ, Lin WY, Wong HF, Lu KT, Chen YR. Quantitative Avery J, editors. Essentials of orthodontic practice. Chicago:
measurements on three-dimensional computed tomography: an EFOP Press; 2003. 405-13.

You might also like