Computer Methods and Programs in Biomedicine 172 (2019) 95–101

Contents lists available at ScienceDirect

Computer Methods and Programs in Biomedicine

journal homepage: www.elsevier.com/locate/cmpb

Semi-automatic and robust determination of dental arch form in

dental cone-beam CT with B-spline approximation
Myungsoo Bae a,1, Jae-Woo Park c,1,∗, Namkug Kim a,b,1,∗∗
a
Department of Convergence Medicine, University of Ulsan College of Medicine, Asan Medical Center, 88 Olympic-Ro 43-Gil Songpa-Gu, Seoul 05505,
Republic of Korea
b
Department of Radiology, University of Ulsan College of Medicine, Asan Medical Center, 88 Olympic-Ro 43-Gil Songpa-Gu, Seoul 05505, Republic of Korea
c
Department of Orthodontics, Kooalldam Dental Hospital, 1418 Gyeongwon-daero, Bupyeong-gu, Incheon, Republic of Korea

a r t i c l e i n f o a b s t r a c t

Article history: Background and objective: The dental arch form is generally used as a base for planning orthodontic
Received 12 December 2018 treatments. It is, therefore, vital to determine the proper individual dental arch form for more accurate
Revised 9 February 2019
orthodontic treatment. We aimed to develop and validate a robust algorithm for semi-automatic deter-
Accepted 23 February 2019
mination of the dental arch form in dental cone-beam CT (CBCT) images with the cubic B-spline approx-
imation.
Keywords: Methods: Our algorithm consists of tooth segmentation, determination of an occlusal plane, and genera-
B-spline curve approximation tion of intersection points between the teeth and the offset plane from the occlusal plane in CBCT images.
Cone-beam CT (CBCT) By fitting a curve to the intersection points using the cubic B-spline curve approximation, the dental arch
Dental arch form
form was finally determined. The accuracy of the dental arch forms was evaluated by comparison with
Occlusal plane
gold standards determined by an expert orthodontist.
Mandible
Maxilla Results: Thirteen dental CBCT scans from nine subjects were enrolled in this study. From the CBCT scans,
13 maxillary arch forms and 11 mandibular arch forms with Class I occlusion were determined by our
proposed algorithm and evaluated for validation. The mean error between the dental arch forms of gold
standards and our method using the cubic B-spline was 0.413 ± 0.092 mm (range, 0.264–0.587 mm).
Conclusions: Our proposed method showed reliable accuracy of determining the dental arch forms for the
maxilla and mandible. These results suggested that this method might be used for planning automatic
tooth setup for individual patients.
© 2019 Elsevier B.V. All rights reserved.

1. Intruduction Subsequently, various kinds of individualized treatments based

Orthodontics depends on an inherently 3D space, which em-
ploys the recent three-dimensional (3D) dental imaging technolo-
gies for clinical applications. In the late 1990s, the method for
planning orthodontic treatment using 3D digital models and fab-
ricating orthodontic appliances was first commercialized [1]. Us-
ing this method, the degree of orthodontic tooth movement could
be more accurately controlled than with the traditional manual or-
thodontic treatment, due to the availability of a digital orthodontic
setup using the digital model.
∗
Corresponding author.
∗∗
Corresponding author at: Department of Convergence Medicine, University of
Ulsan College of Medicine, Asan Medical Center, 88 Olympic-Ro 43-Gil Songpa-Gu,
Seoul 05505, Republic of Korea.
E-mail addresses: jeuspark@gmail.com (J.-W. Park), namkugkim@gmail.com
(N. Kim).
1
These two authors contributed equally to this paper.
on tooth and bone features were researched and developed. Some
manufacturers provided information of treatment planning, indi-
vidualized brackets, and archwires by considering the tooth shape
of each individual patient [2,3]. These services involve simulation
and planning of the clinical treatment process in orthodontics with
digital tooth models and a virtual tooth setup.
For simulating these orthodontic treatment processes, the den-
tal arch form is generally used as a base. In previous studies,
standard arch forms including a semicircle, ellipse, parabola, hy-
perbola, and catenary curve were reported. Although the square,
ovoid, and tapered dental arch forms have not yet been accurately
defined, they do exist in nature [4]. Many mathematical functions
have been proposed as models of the human dental arch; these
include the cubic spline function [5], Bezier cubic equation [6],
Euclidian distance matrices [7], polynomial functions [8,9], conic
sections [10], beta function [11], and Fourier series [12]. However,
there have been reported more normal variations from these spe-
cific arch forms [10]. Therefore, we should consider how to deter-

https://doi.org/10.1016/j.cmpb.2019.02.013
0169-2607/© 2019 Elsevier B.V. All rights reserved.
96 M. Bae, J.-W. Park and N. Kim / Computer Methods and Programs in Biomedicine 172 (2019) 95–101

mine the individual dental arch form and apply it to orthodontic

treatment.
Ahmad et al. [13] proposed a method of dental arch generation
by interpolating the teeth landmarks in a 2D image using a cubic
spline curve. Kondo et al. [14] introduced a two-step curve fitting
method for determining the dental arch in the range image using
ridge points. Majstorovic et al. [15] proposed a method for defin-
ing the dental arch on a 3D digital model obtained by scanning
impression. They defined the dental arches by fitting splines of dif-
ferent degrees through 12 marked points, and analyzed the fitting
errors. Current intraoral scanning technologies offer relatively ac-
curate dental imaging. However, dental cone-beam computed to-
mography (CBCT) is still commonly used in various dental applica-
tions such as treatment planning for orthodontic issues or dental
implants since it provides anatomical information [16–18]. Sa-ing
et al. [19] presented a method of detecting the dental arch in a
dental CT slice. The method detects an initial curve from the jaw
area of the selected slice using morphological thinning, and gener-
ates the dental arch by curve fitting to the curve points. Luo et al.
[20] proposed an algorithm that automatically places the dental
arch in dental CBCT images. The arch form is generated by fitting
a parabolic curve to the dental arch area that is isolated from the
maximum intensity projection (MIP) image of CBCT images.
Here, we proposed a new algorithm that automatically deter-
mines individual dental arch forms using segmented teeth in den- Fig. 1. Overview of our proposed method for determining a dental arch form.
tal CBCT images. The algorithm consists of the determination of an
occlusal plane, the generation of intersection points between the
teeth and the offset plane from the occlusal plane, and a curve fit-
ting to the intersection points using the cubic B-spline approxima- maxillary (or mandibular) teeth were selected in Fig. 3(b). An ini-
tion. The B-spline is one of the most popular methods for curve tial plane (pln1) was obtained by fitting a plane to the sample
modeling in many industrial applications because of its proper- points in Fig. 3(b). Among the points of the #1 and #6 teeth (e.g.
ties such as affine invariance and continuity [21]. In particular, the #11, #21, #16, and #26 teeth for the upper jaw and #31, #41, #36,
shape of B-spline curve can be locally modified by moving the con- and #46 teeth for the lower jaw in Fig. 2), the farthest point from
trol points. Our algorithm was validated by comparison with gold the initial plane for each tooth was selected in Fig. 3(c). Finally,
standards determined by an expert orthodontist with >20 years of the best-fitting plane to these farthest points was identified as the
experience. occlusal plane in Fig. 3(d).
The rest of this paper is organized as follows. First, we provide If both of the #1 teeth are not available, the points of the #2
an overview of our proposed algorithm for determining the dental teeth are chosen as alternatives. If a #6 tooth is not available, the
arch form. Next, we describe our proposed method in detail includ- alternative tooth points are chosen from the #5, #7, or #4 tooth
ing the determination of an occlusal plane and the curve fitting. (in that order).
And finally, we present and discuss the test results.
2.3. Determination of dental arch form
2. Methods
Once the occlusal plane was determined, the dental arch form
2.1. Overview was generated by cubic B-spline curve fitting based on a set of
sample points which were obtained from the intersecting points
As shown in Fig. 1, after reading dental CBCT images to ob- between a plane and a set of voxels associated with the (maxil-
tain the dental arch form, our proposed algorithm for determining lary or mandibular) teeth. The intersecting planes were determined
the dental arch form involves four steps including: tooth segmen- by moving the maxillary occlusal plane along the direction of its
tation, determination of an occlusal plane, generation of tooth in- normal vector by 5 mm in maxillary teeth superiorly and moving
tersection points, and curve fitting using B-spline approximation. the mandibular occlusal plane along the opposite direction of the
The first step, tooth segmentation from dental CBCT images, is a normal vector by 3 mm in mandibular teeth inferiorly as shown
pre-processing step of generating the dental arch form in our algo- in Fig. 4. In our study, a 20% sampling rate was used, and we as-
rithm. The teeth can be segmented with or without the roots. Here, sumed that the teeth of each subject should contain at least one
we performed tooth segmentation using our in-house software #7 tooth. If only one #7 tooth was available, the other one was
called A-View (Asan medical center, Seoul, South Korea) [22] with then temporarily generated using the symmetrical tooth points to
basic image processing algorithms including the watershed method determine the complete dental arch form.
and manual correction. In addition, tooth numbering for individual The next step is curve fitting based on the sample points for
tooth separation from the dental CBCT images was performed us- determining the dental arch form. In the curve fitting, parametriza-
ing in-house A-View software as shown in Fig. 2. tion is critical, which affects the resulting curve approximation.
The parametrization was performed based on angle and distance
2.2. Determination of occlusal plane in Fig. 5.
Let c1 and c2 be two center points between #17 and #27 teeth
Once the segmented teeth were obtained in Fig. 3(a), the oc- in the maxilla and #37 and #47 teeth in the mandible and be-
clusal plane of the maxillary or mandibular teeth was determined tween teeth #11 and #21 in the maxilla and #31 and #41 in the
as following procedure: Sample points (5% sampling rate) of the mandible, respectively. Let ln be the line formed by c1 and c2 (cyan
 M. Bae, J.-W. Park and N. Kim / Computer Methods and Programs in Biomedicine 172 (2019) 95–101 97

Fig. 2. Tooth segmentation and numbering (by the Federation Dentaire Internationale numbering system): (a) maxillary teeth and (b) mandibular teeth. (#18, #28, #38, and
#48 teeth were excluded from the study.).

Fig. 3. The procedure for finding an occlusal plane from a set of upper teeth: (a) a set of upper teeth; (b) sample points of the teeth (magenta) and the initial fitting plane
(pln1, green); (c) the selected farthest points (blue); and (d) the occlusal plane (red). (For interpretation of the references to colour in this figure legend, the reader is referred
to the web version of this article.)

Fig. 5. The parametrization of a curve using angle and distance.

Fig. 4. Finding the intersection points for the arch form: (a) determining the inter-
secting plane (maxillary teeth (top) and mandibular teeth (bottom)), and (b) inter-
secting points between the plane and the set of teeth (green). (For interpretation
of the references to colour in this figure legend, the reader is referred to the web
version of this article.)
line in Fig. 5). The parametrization was first performed using an-
gles between ln and the lines formed by c1 and the sample inter-
section points. The angle for each sample point was mapped to the
range of 0–1.
However, the parametrization using only the angle might cause
an incorrect result at the end parts of the fitting curve. To avoid
this problem, the sample intersection points in the ranges 0.0–
0.2 and 0.8–1.0, as shown in enclosing rectangles in Fig. 5, were
reparameterized using distances. For sample points in the ranges
0.0–0.2 (i.e., in the left rectangle in Fig. 5), the shortest distance
between each sample point and one of two short sides of the rect-
angle (i.e., along the corresponding major principal axis (red) of
the sample points in Fig. 5) was computed, and the point was
98 M. Bae, J.-W. Park and N. Kim / Computer Methods and Programs in Biomedicine 172 (2019) 95–101
Fig. 6. The process of determining the dental arch form from the segmented teeth: (a) upper teeth segmentation, (b) intersection points, (c) curve fitting to the intersection
points using B-spline approximation (red) with the control polygon (blue), and (d) the final dental arch form. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)
assigned to the parameter value between 0.0 and 0.2 using the dis-
tance. Reparametrization of the sample points in the ranges 0.8–1.0
(i.e., in the right rectangle in Fig. 5) was performed in the same
way.
Once the parametrization was complete, curve fitting to the
sample points was performed using least squares cubic B-spline
curve approximation. A cubic B-spline curve is defined as follows:
[21]
L+2

P (u ) = di Ni3 (u ),
i=0
where point di is the control point, L is the number of curve seg-
ments of the B-spline curve, and Ni3 (u ) is the basis function for the
cubic case. The basis function is defined recursively as follows:

1, i f ui−1 ≤ u < ui
Ni0 (u ) =
0, else
and
u − ui−1 ui+n − u n−1
Nin (u ) = N n−1 (u ) + N (u ),
ui+n−1 − ui−1 i ui+n − ui i+1
where n > 0 and u is the parameter value, and U = u0 , …, uL + 4 is
the knot sequence with the L domain knots.
We then obtained the fitting curve to the sample intersection
points by minimizing the distances between intersection points
Xi and the least squares approximation of a cubic B-spline curve,
P(wi ), where wi is the associated parameter value as follows:

m −1
minimize Xi − P (wi )2,
i=0
where m is the number of the intersection points. In our study, we
used the uniform cubic B-spline curve approximation with 10 con-
trol points (i.e., seven curve segments of the cubic B-spline curve).
Fig. 6 shows the process to generate a fitting curve from the seg-
mented teeth using the cubic B-spline approximation.
3. Results
We obtained 13 dental CBCT scans from nine subjects aged 20–
39 years who were enrolled for this study with informed con-
sents. All samples showed Class I occlusion with minor crowding
(<2.0 mm). Each CBCT scan contained a 768 × 768 matrix with
a 0.3-mm pixel size and 0.3-mm slice thickness with 576 slices.
From the CBCT scans, 13 maxillary arch forms and 11 mandibular
arch forms were determined by our proposed algorithm and eval-
uated for validation. In addition, an expert orthodontist with >20
years’ experience determined gold standard arch forms in the same
CBCT scans for comparison using in-house A-View software that
performed the cubic Hermite spline interpolation with seven man-
ually selected points (i.e., six curve segments) from the (maxillary
or mandibular) teeth points on the intersecting plane.
Generating the gold standard was performed twice in each
CBCT scan. The second determination of the gold standard was per-
formed three months after the first gold standard was generated.
Thereafter, for each scan, we measured the error between the arch
forms of the first gold standard and our method and the error be-
tween the arch forms of the second gold standard and our method.
The average of the two measured errors for each scan was used for
the accuracy evaluation in our study.
To measure the error between the two dental arch forms in our
study, we computed the mean error between the two curves bi-
directionally [23]. The error E between the two curves C1 and C2 is
defined as follows:
1 
E1 = min p2 ∈C2 D( p1 , p2 ),
NC1
p1 ∈C1
1 
E2 = min p1 ∈C1 D( p2 , p1 ),
NC2
p2 ∈C2
1
E= (E1 + E2 ),
2
where NC1 and NC2 denote the number of points on curves C1 and
C2 respectively, and D is the Euclidean distance.
The two curves were coarsely aligned using known points such
as the first, middle, and end points of each curve. The ICP algo-
rithm [24] was then applied to achieve fine curve alignment. Fig. 7
provides examples of the gold standards (blue) in Fig. 7(a), the
dental arch forms (red) approximated by the cubic B-spline curve
in Fig. 7(b), and the aligned arch forms in Fig. 7(c).
Table 1 lists the errors between the gold standards and dental
arch forms determined by our method at the maxilla and mandible
from the experiment. The table also includes errors with curve ap-
proximations by three different numbers of curve segments (five,
seven, or nine) of the cubic B-spline curve (i.e., different numbers
of knots). Fig. 8 shows these three different errors for each den-
tal arch form. The measured errors were 0.403–0.880 mm, 0.264–
0.587 mm, and 0.224–0.730 mm for the B-spline approximations
with five, seven, and nine curve segments, respectively. Table 1 and
Fig. 8 indicate that the errors for the curve approximations with
seven and nine curve segments are lower than those with five
curve segments. Meanwhile, there was no significant difference
in the curve errors between the maxillary and mandibular arch
forms.
Majstorovic et al. [15] defined the dental arch by fitting a spline
through 12 marked points on a 3D digital model and measured an
average fitting error using Software Siemens NX10. They defined the
dental arches in 11 different phases of the orthodontics treatment,
and generated the dental arches by fitting splines of 6 different
degrees (i.e. 3th–8th) for each phase. They reported that the av-
erage fitting errors for 3th, 4th, 5th, 6th, 7th, and 8th degrees of
splines were 0.799, 0.845, 0.358, 0.274, 0.223, and 0.222 mm, re-
spectively.
 M. Bae, J.-W. Park and N. Kim / Computer Methods and Programs in Biomedicine 172 (2019) 95–101 99

Fig. 7. Examples of the dental arch forms and their alignment with the maxilla (upper) and mandible (lower): (a) gold standards, (b) dental arch forms approximated by the
cubic B-spline curve, and (c) the aligned arch forms.

Table 1
Experimental results: mean, standard deviation (SD), minimum (Min) and maximum (Max) errors between
the manually generated arch form and the arch form approximated by our method using our test data set.

Error (mm)

Maxilla Mandible Total

5 7 9 5 7 9 5 7 9

Mean 0.678 0.421 0.429 0.679 0.405 0.373 0.679 0.413 0.404
SD 0.153 0.112 0.130 0.105 0.061 0.069 0.133 0.092 0.110
Max 0.875 0.587 0.730 0.880 0.505 0.511 0.880 0.587 0.730
Min 0.403 0.264 0.224 0.492 0.321 0.252 0.403 0.264 0.224

Fig. 8. Experimental results of the curve errors with the 24 dental arch forms approximated by three different numbers of curve segments: the maxillary arch forms in 1–13
and the mandibular arch forms in 14–24.

We implemented and tested our method developed by using 4. Discussion

C++ programming language (Visual C++ 2010) under the Mi-
crosoft Windows 8.1 operating system on a PC with an Intel Core
i7 processer (3.4 GHz) and 16 GB of memory. The average process-
ing time for generating the dental arch form of each jaw using our
method excluding the tooth segmentation step was 0.25 ± 0.05 s,
while the orthodontist spent 15.42 ± 0.97 s for the same process
(i.e., determining the occlusal plane and the arch form using in-
house A-View software).
Here we proposed a method that could automatically and reli-
ably generate the patient specific dental arch form using individ-
ually segmented teeth. Our proposed method was compared and
validated by the error measurement between dental arch forms of
gold standards and our method. The results showed excellent co-
ordination between the dental arch forms that could be applied to
clinical situations. In the results, however, the largest errors origi-
100 M. Bae, J.-W. Park and N. Kim / Computer Methods and Programs in Biomedicine 172 (2019) 95–101
Fig. 9. Dental arch forms of (a) the gold standard and (b) our method at the maxilla, and (c) their alignment with error of 0.582 mm.
nated from the discrepancy of the endpoints. In this case, we may
solve this problem by simply manually adjusting the end control
points to generate the more accurate arch form. Fig. 9 shows larger
errors at the end parts of the aligned dental arch forms (Fig. 9(c)).
This result may be because the curve determined by our method
passes closer the middle of the teeth region than the gold stan-
dard, especially in the #16 and #26 teeth.
These results could have encouraged us to take a step forward
to automatic tooth setup. All digital orthodontic treatments started
from the tooth setup, which was labor intensive and time consum-
ing. Once we had the segmented tooth data and the orthodontist
chose the arch form for the subject, we could automatically align
the segmented teeth to the determined arch form.
In general, alignment of teeth is the first step of the digital or-
thodontic treatment. Although we should consider the torque and
inclination of individual teeth for tooth setup, there were some
cases in which we could achieve the treatment goal by aligning
the teeth. Tooth inclination and torque should then be retouched
to improve dental aesthetics.
Recently, there were attempts to use 3D printed brackets for
orthodontic tooth movement [25,26]. Bae et al. [25] reported an
approach of anterior teeth alignment using 3D printed brackets, in
which case the treatment goal was simply aligning the teeth into
the arch form. In such a simple case, our proposed method could
help to set the treatment goal. Moreover, our method could reduce
the labor cost in aligning teeth, and make it easier for tooth setup.
There are several limitations. At first, the number of enrolled
subjects is relatively small. Therefore, we need to extend this study
with larger number of subjects with various types of occlusion and
compare results of arch forms determined by our method with
gold standards. In addition, there are missing teeth and teeth with
metal artifacts in some cases, which make difficult to generate
proper dental arch forms. For further study, we need to extend this
study to consider various kinds of missing teeth and teeth with
metal artifacts, which could be considered as missing teeth.
5. Conclusion
Here, we proposed a semi-automatic algorithm for determin-
ing a dental arch form with cubic B-spline curve approximation in
dental CBCT images. Our proposed method consists of tooth seg-
mentation, the determination of an occlusal plane, the generation
of intersection points, and curve fitting using the B-spline approx-
imation with the parametrization using angle and distance. The
method could reliably and automatically determine the dental arch
forms with Class I occlusion for the maxilla and mandible. We val-
idated our proposed method by measuring the error between the
gold standard and the dental arch form determined by our method.
Based on these results, we found that our proposed method could
be applied to digital orthodontic treatment for individual patients,
and used for other digital dentistry applications.
Ackonwledgements
This work was supported by the Industrial Technology Innova-
tion Program (10052753) funded by the Ministry of Trade, Industry
and Energy (MOTIE, Korea) and a grant of the Korea Health Tech-
nology R&D Project through the Korea Health Industry Develop-
ment Institute (KHIDI) funded by the Ministry of Health & Welfare,
Republic of Korea (HI18C1638).
Conflicts of interest
The authors report no conflicts of interest.
Ethical standard
All procedures performed in studies involving human partici-
pants were in accordance with the ethical standards of the institu-
tional and/or national research committee and with the 1975 Dec-
laration of Helsinki, as revised in 2008(5). For this type of study,
formal consent is not required.
References
[1] R.L. Boyd, R.J. Miller, V. Vlaskalic, The Invisualign system in adult orthodontics:
mild crowding and space closure cases, J. Clin. Orthod. 34 (20 0 0) 203–212.
[2] R.P. Scholz, D.M. Sarver, Interview with an Insignia doctor: David M Sarver,
Am. J. Orthod. Dentofacial Orthop. 136 (2009) 853–856.
[3] M. Magali, C. Fauquet, C. Galletti, C. Palot, D. Wiechmann, J. Mah, Digital design
and manufacturing of the lingualcare bracket system, J. Clin. Orthod. 39 (2005)
375–382.
[4] E.A. BeGole, D.L. Fox, C. Sadowsky, Analysis of change in arch form with pre-
molar expansion, Am. J. Orthod. Dentofacial Orthop. 113 (1998) 307–315.
[5] E.A. BeGole, R.C. Lyew, A new method for analyzing change in dental arch
form, Am. J. Orthod. Dentofacial Orthop. 113 (1998) 394–401.
[6] T.U. Taner, S. Ciger, H. El, D. Germec, A. Es, Evaluation of dental arch width and
form changes after orthodontic treatment and retention with a new comput-
erized method, Am. J. Orthod. Dentofacial Orthop. 126 (4) (2004) 464–475.
[7] S. Mutinelli, M. Cozzani, M. Manfredi, M. Bee, G. Siciliani, Dental arch changes
following rapid maxillary expansion, Eur. J. Orthod. 30 (2008) 469–476.
[8] J.M. Felton, P.M. Sinclair, D.L Jones, R.G. Alexander, A computerized analysis of
the shape and stability of mandibular arch form, Am. J. Orthod. Dentofacial
Orthop. 92 (1987) 478–483.
[9] T. Tirvino, D.F. Siqueira, M.A. Scanavini, A new concept of mandibular den-
tal arch forms with normal occlusion, Am. J. Orthod. Dentofacial Orthop. 133
(10.e) (2008) 15–22.
[10] J. Henrikson, M. Persson, B. Thilander, Long-term stability of dental arch form
in normal occlusion from 13 to 31 years of age, Eur. J. Orthod. 23 (2001) 51–61.
[11] S. Braun, W.P. Hnat, D.E. Fender, H.L. Legan, The form of the human dental
arch, Angle Orthod. 68 (1998) 29–36.
[12] A.P. Valenzuela, M.A. Pardo, S. Yezioro, Description of dental arch form using
the Fourier series, Int. J. Adult Orthod. Orthognath. Surg. 17 (2002) 59–65.
[13] R.R. Ahmad, N. Ghazali, A.S. Rambely, U.K.S. Din, N. Hassan, Application of cu-
bic spline in the implementation of braces for the case of a child, J. Math. Stat.
8 (1) (2012) 144–149.
[14] T. Kondo, S.H. Ong, K.W.C. Foong, Tooth segmentation of dental study models
using range images, IEEE Trans. Med. Imaging 23 (3) (2004) 350–362.
[15] N.V. Majstorovic, S.P. Živković, B.R. Glisic, Dental arch monitoring by splines
fitting error during orthodontic treatment using 3D digital models, Vojnosanit.
Pregl. (2017), doi:10.2298/VSP161212067M.
[16] H.M. Alamri, M. Sadrameli, M.A. Alshalhoob, M. Sadrameli, M.A. Alshehri, Ap-
plications of CBCT in dental practice: a review of the literature, Gen. Dent. 60
(5) (2012) 390–400.
 M. Bae, J.-W. Park and N. Kim / Computer Methods and Programs in Biomedicine 172 (2019) 95–101
[17] E. Venkatesh, S.V. Elluru, Cone beam computed tomography: basics and appli-
cations in dentistry, J. Istanb. Univ. Fac. Dent. 51 (3) (2017) S102–S121.
[18] V. Manavella, F. Romano, L. Corano, C. Bignardi, M. Aimetti, Three-dimensional
volumetric changes in severely resorbed alveolar sockets after ridge augmenta-
tion with bovine-derived xenograft and resorbable barrier: a preliminary study
on CBCT imaging, Int. J. Oral Maxillofac. Implants 33 (2) (2018) 373–382.
[19] V. Sa-ing, K. Wangkaoom, S.S. Thongvigitmanee, Automatic dental arch detec-
tion and panoramic image synthesis from CT images, in: Conf. Proc. IEEE Eng.
Med. Biol. Soc., 2013, pp. 6099–6102.
[20] T. Luo, C. Shi, X. Zhao, Y. Zhao, J. Xu, Automatic synthesis of panoramic radio-
graphs from dental cone beam computed tomography data, PLoS One 11 (6)
(2016) e0156976.
[21] G.E. Farin, Curve and Surface for Computer Aided Geometric Design, fourth ed.,
Academic Press, Boston, 1996.
101
[22] H.W. Kim, S.H. Oh, N. Kim, E. Nakazawa, I.J. Rhyu, Rapid method for electron
tomographic reconstruction and three-dimensional modeling of the murine
synapse using an automated fiducial marker-free system, Microsc. Microanal.
19 (S5) (2013) 182–187.
[23] L. Zhang, A. Razdan, G. Farin, J. Femiani, M. Bae, C. Lockwood, 3D face authen-
tication and recognition based on bilateral symmetry analysis, Vis. Comput. 22
(1) (2006) 43–55.
[24] P. Besl, H. McKay, A method for registration of 3-D shapes, IEEE Trans, Pattern
Anal. Mach. Intell. 14 (2) (1992) 239–256.
[25] G.S. Bae, Y.I. Kim, S.S. Kim, S.B. Park, W.S. Son, 3D-printed double-wire bracket
for anterior alignment, J. Clin. Orthod. 51 (2017) 377–381.
[26] K.F. Krey, N. Darkazanly, R. Kühnert, S. Ruqe, 3D-printed orthodontic brackets
– proof of concept, Int. J. Comput. Dent. 19 (2016) 351–362.