ORIGINAL ARTICLE

Factors affecting the accuracy of buccal alveolar

bone height measurements from cone-beam
computed tomography images
Ryan Wood,a Zongyang Sun,b Jahanzeb Chaudhry,c Boon Ching Tee,d Do-Gyoon Kim,b Binnaz Leblebicioglu,e
and Greg Englandf
Columbus, Ohio

Introduction: The reasons for inaccuracies in alveolar bone measurement from cone-beam computed tomography
(CBCT) images might be multifactorial. In this study, we investigated the impact of software, the presence or absence
of soft tissues, the voxel size of the scan, and the regions in the jaws on buccal alveolar bone height measurements in
pigs at an age equivalent to human adolescents. Methods: Marker holes, apical to the maxillary and mandibular mo-
lar roots, and mesiodistal molar occlusal reference grooves were created in 6 fresh pig heads (12 for each jaw), fol-
lowed by CBCT scans at 0.4-mm and 0.2-mm voxel sizes under soft-tissue presence and soft-tissue absence
conditions. Subsequently, buccolingual sections bisecting the marker holes were cut, from which the physical
alveolar bone height and thickness were measured. One blinded rater, using Dolphin (version 11.5 Premium;
Dolphin Imaging, Chatsworth, Calif) and OsiriX (version 3.9; www.osirix-viewer.com) software, independently
collected alveolar bone height measurements from the CBCT images. Differences between the CBCT and the
physical measurements were calculated. The mean differences and the limit of agreement (LOA, 61.96 SD) for
every jaw, voxel-size, soft-tissue, and software condition were depicted. Each measurement was then assessed
for clinical inaccuracy by using 2 levels of criteria (absolute differences between CBCT and physical
measurements $1 mm, or absolute differences between CBCT and physical measurements $0.5 mm), and the
interactions between soft-tissue and voxel-size factors for every jaw and software condition were assessed by
chi-square tests. Results: Overall, the mean differences between the CBCT and the physical measurements for ev-
ery jaw, voxel-size, soft-tissue, and software condition were near 0. With all other conditions kept equal, the accuracy
of the maxillary CBCT measurements was inferior (larger limit of agreement ranges and higher frequencies of clinical
inaccuracy) to the mandibular measurements. The physical thickness of the maxillary alveolar crestal bone was less
than 1 mm and significantly thinner than the mandibular counterparts. For every jaw and software condition, the
accuracy of measurements from the 0.2-mm soft-tissue presence CBCT images was consistently superior
(smaller limit of agreement ranges and lower frequencies of clinical inaccuracy) to those from the 0.4-mm soft-
tissue presence, the 0.4-mm soft-tissue absence, and the 0.2-mm soft-tissue absence images; all showed similar
accuracies. Qualitatively, the soft-tissue absence images demonstrated much brighter enamel and alveolar bone
surface contours than did the soft-tissue presence images. Conclusions: At an adolescent age, the buccal alveolar
bone height measured from the maxillary molar region based on 0.4-mm voxel-size CBCT images can have
relatively large and frequently inaccurate measurements, possibly due to its thinness. By using 0.2-mm voxel-
size scans, measurement accuracy might be improved, but only when the overlying facial and gingival tissues
are kept intact. (Am J Orthod Dentofacial Orthop 2013;143:353-63)

W
ith increasing popularity of cone-beam question the appropriateness of using it to measure
computed tomography (CBCT) imaging in the height and thickness of the alveolar bone.5-8
orthodontics,1-4 researchers continue to Undoubtedly, compared with the basal bone in the

From the College of Dentistry, Ohio State University, Columbus, Ohio. The authors report no commercial, proprietary, or financial interest in the prod-
a
Resident, Division of Orthodontics. ucts or companies described in this article.
b
Assistant professor, Division of Orthodontics. Reprint requests to: Zongyang Sun, Room 4088 Postle Hall, 305 W 12th Ave, Co-
c
Assistant professor, Division of Oral Pathology and Radiology. lumbus, OH 43210; e-mail, sun.254@osu.edu.
d
Research assistant, Division of Orthodontics. Submitted, March 2012; revised and accepted, October 2012.
e
Associate professor, Division of Periodontology. 0889-5406/$36.00
f
Student. Copyright Ó 2013 by the American Association of Orthodontists.
This project was partially supported by the Delta Dental Master's Thesis Award http://dx.doi.org/10.1016/j.ajodo.2012.10.019
program (to R.W.).

353
354 Wood et al

Fig 1. A, CBCT image showing marker holes adjacent to molar roots (arrows); B, 4 equal layout views
of the CBCT image, showing image orientation before measuring buccal bone height in the coronal
view (detailed in the main text); C, illustration of alveolar bone height in the coronal view (enlarged)
from the CBCT image. The physical measurements were obtained similarly from buccolingual sections.

jaws, the alveolar bone is more difficult to visualize and
measure because of its thinness and proximity to the
teeth and the periodontal ligament.9-11 On the other
hand, when measuring the alveolar bone, it is critical
to obtain measurements close to physical truth, since
these measurements are often directly used to interpret
periodontal health or sequelae from orthodontic
treatment.12,13 For instance, underestimation of actual
alveolar bone height would lead to a misdiagnosis of
bone loss.
To date, reports on the accuracy of CBCT alveolar
bone measurements have been somewhat inconsis-
tent.5-7 Overall submillimeter limits of agreement were
reported by Timock et al,6 indicating excellent accuracy
for clinical use, but Patcas et al5 found up to 2.1 mm of
limits of agreement in the mandibular incisor region,
and Sun et al7 found that bone height was systematically
underestimated by 0.9 to 1.2 mm when the bone thick-
ness was near the voxel size. Although such inconsis-
tencies are most likely due to variations in research
designs and materials, they also suggest that the accu-
racy of alveolar bone measurements from CBCT images
might depend on various factors.
Among many possible factors, several are often
different in the studies, but they have been inadequately
investigated. These include variations in scan voxel size,
image analysis software, soft-tissue conditions (presence
or absence), and alveolar bone regions in the jaws. Dry
skulls without soft tissues have been commonly used
for accuracy studies until recently, when researchers
started using embalmed human cadavers.5,6,9,14
Although 1 group suggested that soft-tissue presence
tends to reduce CBCT measurement precision in
the mandibular incisor region,5 others found that
soft-tissue presence did not compromise measurement
accuracy.9 Similarly, decreasing voxel size has been
found to increase the accuracy of alveolar bone mea-
surements, but the interplay between this parameter
and other factors (eg, soft tissues) remains unknown.5,7
In addition, buccal bone thickness generally differs
between the 2 jaws, with a thinner maxillary bone.15,16
Whether such differences between the jaws affect the
accuracy of CBCT measurements of the alveolar bone
is currently unknown.
This study was aimed at understanding whether the
variations in scanning voxel sizes, software for imaging
analysis, soft-tissue presence or absence, and regions in
the jaws impact the accuracy of alveolar bone height
measurements from CBCT images. The discrepancies be-
tween CBCT and physical measurements were analyzed
for each factor for mean differences, limits of agreement
for each measurement, and proportions of clinically in-
accurate measurements.
MATERIAL AND METHODS
Six fresh domestic pig (Sus scrofa) heads, aged 3 to 6
months and equivalent to early adolescent humans, were
used.17 Specimen preparation is shown in Figure 1, A.
Reference marker holes were drilled in the buccal alveo-
lar bone at the apices of the first and second maxillary
and mandibular molars with a slow-speed dental hand
piece (Volvere VMax; Brasseler USA, Savannah, Ga).
Marker holes were made intraorally through the buccal
gingiva. Reference grooves bisecting the occlusal tables
of all molars in a mesiodistal direction were also created.
All 4 holes on the maxilla and the 3 mesial holes on the
mandible were used in this study, and each hole was
considered a sample for subsequent analyses. Three

March 2013
Vol 143
Issue 3 American Journal of Orthodontics and Dentofacial Orthopedics
Wood et al
instead of 4 holes from each mandibular half were used,
since the most distal hole involved the alveolar crest in
some specimens.
Each head was scanned with an iCAT 17-19 Platinum
CBCT machine (120 kVp, 5 mA) (Imaging Sciences Inter-
national, Hatfield, Pa) at 0.4-mm voxel-size (scanning
time, 8.9 seconds) and 0.2-mm voxel-size (scanning
time, 26.9 seconds) levels. These constituted the soft-
tissue presence CBCT images. Subsequently, superficial
soft tissues, including skin, muscles, and gingival tissues
covering the buccal alveolar bone, were dissected and re-
moved without damaging the bone. Internal soft tissues,
including the tongue, mucosa, and lingual gingivae,
were not removed. The specimens were then rescanned
at 0.4- and 0.2-mm voxel sizes with the same protocols
as before soft-tissue removal. These constituted the
soft-tissue absence CBCT images. Combined, a total of
24 sets of CBCT images were obtained for analyses.
After the soft-tissue absence scans, each maxillary or
mandibular quadrant was separated from the head by
using a Ryobi 9-in band saw (Techtronic Industries
North America, Anderson, SC). The specimens were fur-
ther cut into sections that bisected the marker holes and
in a direction perpendicular to the molar occlusal
groove, by using an IsoMet low-speed saw (Buehler,
Lake Bluff, Ill) under water lubrication. This resulted in
2 mirror sections for each marker hole, which were
used to obtain the physical measurements of alveolar
bone height and thickness. Using a digital caliper, 2
raters (R.W. and Z.S.) took these measurements indepen-
dently according to the same protocol. Bone height was
measured from the occlusal border of the marker hole to
the crest of the alveolar bone. Bone thickness was mea-
sured at 2 locations along the buccal surface: 0.5 and 1.0
mm from the alveolar bone crest. The distance (perpen-
dicular to the root's long axis) between the buccal sur-
face and the buccal bundle bone surface was
measured. The measurements from the 2 mirror sections
of each marker hole were averaged. These measurements
were tested for interrater reliability before further com-
bination (detailed below). No intrarater reliability was
tested for these physical measurements.
One rater (R.W.) measured the CBCT images. The
CBCT data (DICOM, Rosslyn, Va) were relabeled and re-
ordered randomly to blind the rater. The rater was
trained by an experienced oral and maxillofacial radiol-
ogist (J.C.) on how to take linear measurements from
CBCT images and then independently measured the
height of the buccal alveolar bone using the following
protocol.
After setting the resolution of the display monitor
at 1600 3 1200 pixels, the CBCT data were imported
into the software, Dolphin Imaging (version 11.5
American Journal of Orthodontics and Dentofacial Orthopedics
355
Premium; Dolphin Imaging, Chatsworth, Calif) or
OsiriX (version 3.9; www.osirix-viewer.com). Slice
thickness was set at 0.5 mm in coronal, sagittal, and
axial views. Then the specimen's occlusal groove was
made parallel to the axial and sagittal planes; this ori-
ented the coronal views perpendicular to the groove
(Fig 1, B). In the coronal view, which is equivalent
to the sections used for physical measurement, buccal
alveolar bone height at each marker hole was mea-
sured. The linear measurement tool, available in
both software programs, was used to measure alveolar
bone height, which was the distance from the occlusal
border of the marker hole to the alveolar crest (Fig 1,
C). During analysis, the rater was allowed to enlarge
the images but not to digitally change the images in
any other way.
Eight repeated CBCT measurements were obtained
from each marker hole under varied conditions as a result
of the 2 3 2 3 2 design of 3 factors, with each having 2
subcategories (software, Dolphin or Osirix; soft tissues,
presence or absence; voxel size, 0.4 or 0.2 mm). The rater
was blinded to the conditions when taking the measure-
ments.
To assess intrarater reliability, 8 sets of CBCT images
were randomly chosen for remeasurement 6 weeks after
the initial measurements. The same protocols were fol-
lowed for the remeasurements.
Statistical analysis
Intrarater reliability values (for CBCT measurements
only) and interrater reliability (for physical measure-
ments only) were assessed by intraclass correlation tests.
The differences between each set of CBCT and physical
measurements were calculated and used as a parameter,
DCBCT-Phy. The variances of DCBCT-Phy caused by the fac-
tors of software, soft tissue, and voxel size were tested by
repeated-measures analysis of variance (ANOVA) with
the maxilla and the mandible separated. The mean
DCBCT-Phy and the limit of agreement (LOA, 61.96 SD)
for each condition were compared with the Bland-Alt-
man method.18
Subsequently, to further examine the clinical impli-
cations of the discrepancies between the CBCT and the
physical measurements, the continuous DCBCT-Phy data
were converted to 1 of 2 categorical values: clinically ac-
curate or clinically inaccurate. Two levels of criteria (ab-
solute DCBCT-Phy value $1 or $0.5 mm) were used to
determine the clinical inaccuracy, from which the pro-
portion of clinically inaccurate measurements was calcu-
lated for every condition. The interactions between the
voxel-size and soft-tissue factors under varied jaw, inac-
curacy criteria, and software conditions were tested with
chi-square tests.
March 2013
Vol 143
Issue 3
356
Table I. Mean differences between CBCT and physical measurements (DCBCT-Phy) 6 limit of agreement (LOA, 1.96 SD)
for each testing condition
0.4-mm STP
Maxilla
Dolphin 0.08 6 2.23
OsiriX 0.05 6 2.16
Mandible
Dolphin 0.05 6 1.29
OsiriX 0.10 6 1.26
LOA: Mx-Mn
Dolphin 0.94
OsiriX 0.90
STP, Soft-tissue presence; STA, soft-tissue absence; Mx, maxilla; Mn, mandible.
The physical measurements of alveolar bone thick-
ness were compared between the maxilla and the man-
dible by 2-sample t tests.
RESULTS
Interrater reliability of the alveolar bone height
physical measurements was excellent (correlation coeffi-
cients: maxilla, 0.98; mandible, 0.95), so the measure-
ments from the 2 raters were averaged and used for
comparisons with the CBCT measurements.
Intrarater reliability values of the CBCT alveolar bone
height measurements were excellent (correlation coeffi-
cients: maxilla, 0.98; mandible 0.97).
The mean differences between CBCT and physical
measurements and limit of agreement range for every
jaw and software condition are listed in Table I. In
the maxilla, ANOVA showed that no factors signifi-
cantly changed the mean DCBCT-Phy, which were in
the range of 0.04 and 0.14 mm, respectively. In the
mandible, although ANOVA showed a significant inter-
action among software, soft-tissue, and voxel-size fac-
tors, even the greatest differences between the CBCT
and physical measurements among them were less
than 0.2 mm (Table I). All limit of agreement ranges
were greater in the maxilla (all .2 mm except for the
0.2-mm soft-tissue presence condition) than in the
mandible (all .1 mm except for the 0.2-mm soft-
tissue presence condition). With all other factors
kept equal, the limit of agreement values were nearly
identical between the Dolphin and Osirix software con-
ditions.
The Bland-Altman plots18 of the maxilla and the
mandible with the Dolphin software are shown in
Figure 2. The patterns of the Osirix plots are not shown
because they were similar to those of Dolphin. In both
jaws, with the other factors kept equal, the soft-tissue
presence 0.2-mm voxel-size subcategory demonstrated
a substantially smaller limit of agreement range and
March 2013
Vol 143
Issue 3
Wood et al
0.4-mm STA 0.2-mm STP 0.2-mm STA
0.14 6 2.24 0.03 6 1.03 0.04 6 2.22
0.08 6 2.27 0.04 6 1.06 0.03 6 2.11
0.22 6 1.26 0.04 6 0.64 0.02 6 1.01
0.06 6 1.23 0.09 6 0.64 0.06 6 1.23
0.98 0.39 1.21
1.04 0.42 0.88
less scatter of data points than those of the other 3 sub-
categories, which were generally similar.
Examples of buccolingual CBCT images of the max-
illary and mandibular alveolar bones from the same an-
imal with the same protocols and settings, but under
varied voxel-size and soft-tissue conditions, are shown
in Figure 3. Overall, the 0.2-mm voxel-size images ap-
peared clearer than did the 0.4-mm voxel-size images,
and the dentoalveolus in the soft-tissue absence images
showed greater surface contour brightness and contrast
than those in the soft-tissue presence images. This dif-
ference in brightness was especially prominent at the
enamel and buccal alveolar bone surfaces between the
0.2-mm soft-tissue presence and the 0.2-mm soft-
tissue absence images. These patterns did not change
with the software programs.
The proportions of clinically inaccurate measure-
ments are shown in Table II. First, the proportions of in-
accurate measurements were lower in the mandible than
in the maxilla, and lower at the 1-mm inaccuracy crite-
rion (absolute DCBCT-Phy value $1 mm) than at the
0.5-mm inaccuracy criterion (absolute DCBCT-Phy value
$0.5 mm). Next, with the jaw, inaccuracy criterion
and software conditions kept equal, the soft-tissue pres-
ence 0.2-mm subcategory demonstrated a substantially
lower value than those of the other 3 categories, which
were generally similar. The interaction between the
voxel-size and soft-tissue factors was not statistically
significant when tested separately for every condition
but reached significance when the measurements from
both software conditions were pooled and compared
at the 1-mm inaccuracy criterion. At the 0.5-mm inac-
curacy criterion, the interaction was nearly significant
for the maxilla but not for the mandible.
The physical measurements of alveolar bone thick-
ness are summarized in Figure 4. Overall, it was signifi-
cantly thinner in the maxilla than in the mandible (t
tests; P \0.001).
American Journal of Orthodontics and Dentofacial Orthopedics
Wood et al 357

Fig 2. Bland-Altman plots18 of CBCT measurements obtained under varied jaw, voxel-size, and soft-
tissue conditions. Only plots from the Dolphin software are shown because those from Osirix were
similar (Table I). Each data point, defined by x and y coordinates (average of CBCT and physical mea-
surements, difference between CBCT and physical measurements), represents the position of a CBCT
measurement relative to a physical measurement obtained from the same marker hole. The blue bro-
ken lines indicate the limit of agreement defined as 1.96 SD from the physical measurements, and the
green solid lines indicate the mean differences between the CBCT and physical measurements of all
points included in that plot. STP, Soft-tissue presence; STA, soft-tissue absence.

DISCUSSION a critical issue. Overall, these results confirmed our hy-

Among the factors investigated in this study, the
regions in the jaws, soft-tissue condition, and scan-
ning voxel size showed significant impacts on the ac-
curacy of alveolar bone height measurements from
CBCT images, whereas the software choice was not
pothesis that the accuracy of alveolar bone height
measurements from CBCT images depends on multiple
factors. These factors, as well as their clinical implica-
tions and the limitations of this study, are discussed
below.

American Journal of Orthodontics and Dentofacial Orthopedics March 2013
Vol 143
Issue 3
358 Wood et al

Fig 3. Examples of CBCT images. All images were captured from the left side of a pig with varied voxel-
size and soft-tissue conditions in the Dolphin software (images from Osirix software were similar). All
images were captured by using the same protocol and display resolution, brightness, and contrast set-
tings as those used for the CBCT measurements; the 4 images from the maxilla or the mandible were
from the same marker hole level. All images also have the same height and width ratio as that of the
actual CBCT images (long roots were a species-specific feature) and were resized with the same cal-
iper. STP, Soft-tissue presence; STA, soft-tissue absence.

Several previous studies have examined the accuracy
of alveolar bone height measurements from CBCT
images in various regions, and mixed results were
reported. More specifically, clinically significant inac-
curacies were found at the mandibular incisor5 and
maxillary molar regions,7 whereas overall accurate
measurements (limit of agreement range, \1 mm)
were reported by Timock et al6 when the maxillary
and mandibular molar regions were combined. Clearly,
the differences between the maxillary and mandibular
molar regions found in this study were in line with
the findings from the first 2 studies,5,7 but not with
the third one.6 These comparisons prompted us to
ask 2 questions. What is the underlying cause for less
accurate measurements from the maxillary than from
the mandible molar regions in our study, and why
was a similar difference between these regions not
found by Timock et al?
Less accurate measurements from the maxillary mo-
lar region can be largely attributed to thin bones in this
region. On average, maxillary buccal alveolar bones were
less than 1 mm thick, significantly thinner than the man-
dibular alveolar bones, which were above 1 mm (Fig 4).
The impact of alveolar bone thickness on the accuracy of
alveolar bone height has been well documented.7,8 More
specifically, thin bones (near voxel size) tend to become
indistinguishable from adjacent cementum on CBCT
images; thus, they are significantly less visible for
measurements. In this study, the average bone
thickness in the maxilla (Fig 4) was 2- to 3-fold of the
voxel size; this prevented a systemic underestimation
of bone height measurements, since the mean differ-
ences between CBCT and physical measurements was
near 0. However, it clearly was not thick enough to elim-
inate the difficulty involved in identifying bone bound-
aries, because a substantial proportion (Fig 3) of the

March 2013
Vol 143
Issue 3 American Journal of Orthodontics and Dentofacial Orthopedics
Wood et al 359

Table II. Proportion of potentially clinically inaccurate CBCT measurements

Interaction of voxel Interaction of voxel
Jaw, inaccuracy size and soft tissues size and soft tissues
criterion, software 0.4-mm STP 0.4-mm STA 0.2-mm STP 0.2-mm STA (software separated) (software values pooled)
Mx: jDCBCT-Phyj $1 mm P 5 0.03
Dolphin 25.00% 22.92% 8.33% 25.00% P 5 0.09
OsiriX 27.08% 22.92% 10.42% 22.92% P 5 0.15
Mn: jDCBCT-Phyj $1 mm P 5 0.03
Dolphin 13.90% 11.11% 0.00% 5.66% P 5 0.15
OsiriX 8.33% 11.11% 0.00% 11.11% P 5 0.13
Mx: jDCBCT-Phyj $0.5 mm P 5 0.09
Dolphin 39.58% 39.58% 22.92% 37.50% P 5 0.33
OsiriX 43.75% 35.42% 22.92% 37.50% P 5 0.16
Mn: jDCBCT-Phyj $0.5 mm P 5 0.25
Dolphin 22.22% 27.87% 5.56% 11.11% P 5 0.63
OsiriX 22.22% 19.44% 8.33% 19.44% P 5 0.25
STP, Soft-tissue presence; STA, soft-tissue absence; jDCBCT-Phyj, absolute value of differences between CBCT and physical measurements; Mx,
maxilla; Mn, mandible.

et al6 used cadaver specimens from elderly people

Fig 4. Comparison of alveolar bone thickness between
the maxilla and mandible obtained by physical measure-
ments. Error bars indicate standard deviations. P values
were based on 2-sample t tests.
measurements were different from the physical truth by
over 1 mm, with some even by over 2 mm (Fig 2; Table I).
In addition to bone thickness, the difference of bone
mineral density between the 2 jaws might also contrib-
ute to measurement differences between the molar
regions in the 2 jaws. Specifically, the maxillary posterior
bone generally has less mineral density than the mandib-
ular counterparts.19 Further confirmation for this latter
point requires more studies.
The inconsistency between our findings and those re-
ported by Timock et al6 is likely to be due to a difference
in specimen ages. We used young pigs at an age equiv-
alent to adolescent humans,17,20,21 whereas Timock
(mean age, 77 years). As a result, our specimens were still
growing and changing with tooth eruption and move-
ment, but their specimens had reached adult thickness
and mineralization. Two independent studies recently
measured alveolar bone thickness from CBCT images.
One included subjects aged 13 to 48 years and found
maxillary and mandibular bone thickness of 1.1 to 2.2
mm and 1.6 to 3.0 mm, respectively.15 The other in-
cluded subjects aged 18 to 58 years and found maxillary
and mandibular buccal bone thicknesses of 1.1 and 2.2
mm, respectively.16 Despite the differences in methodol-
ogies, overall these values were somewhat higher than
our bone-thickness measurements, suggesting that the
maxillary buccal bone might gain another 0.5 to 1.0
mm in thickness before reaching adulthood. Therefore,
we speculate that the specimens used by Timock et al,
who did not report maxillary and mandibular bone
thicknesses separately, had thicker maxillary buccal
bones than did our specimens; this might explain their
better accuracy.
Both the Bland-Altman parameters18 (Table I; Fig 2)
and the proportions of potentially clinically inaccurate
measurements (Table II) demonstrated consistent inter-
actions between the soft-tissue and voxel-size factors.
This interaction was not statistically significant when
tested for individual jaw and software conditions, most
likely because of a relatively small sample size of each
condition. After pooling the measurements from both
software programs, which were similar, statistical signif-
icance was reached for 1-mm inaccuracy criterion (abso-
lute DCBCT-Phy value ≥$1 mm) at the 1-mm inaccuracy
criterion in both jaws (Table II).
This interaction can be interpreted as follows: a 0.2-
mm voxel size improves the measurement accuracy, but

American Journal of Orthodontics and Dentofacial Orthopedics March 2013
Vol 143
Issue 3
360
the improvement is cancelled when the soft tissues are
removed. Based on the consistent findings reported by
others that a higher scanning resolution (smaller voxel
size) produces more accurate alveolar bone5,7 and
tooth measurements,22 it is reasonable to expect that
measurement accuracy improves when scanning voxel
size decreases. In this study, for consistency among test-
ing conditions (no voxel-size slice thickness was avail-
able in Osirix), a 0.5-mm slice thickness was used for
all image reconstructions before measurements. In mi-
crocomputed tomography studies, it has been found
that reconstructing the images at a thickness greater
than the scanning voxel size can coarsen the image res-
olution, but this effect did not eliminate the resolution
differential of the varied scanning voxel size.23 This
was clearly the case for our reconstructed CBCT images,
too (Fig 3). An improved measurement accuracy of the
0.2-mm soft-tissue presence condition, therefore, is
not surprising and can be explained as an outcome of
the higher scanning resolution.
On the other hand, few studies have investigated the
effect of soft-tissue removal on alveolar bone measure-
ment accuracy. Although it has been speculated that
soft-tissue presence produces inferior accuracy5 and
soft-tissue absence might increase tissue contrast and
reduce scatter and subsequently improve linear mea-
surement accuracy, these speculations have not been
substantiated.9 Ganguly et al9 recently found that the
measurement accuracy of soft-tissue presence (with em-
balmed human cadavers) was comparable with data pre-
viously obtained from soft-tissue absence conditions
(dry skulls); this indirectly and partially refutes rather
than supports these speculations.9 Additionally, despite
a lack of a report on the scanning voxel size by Ganguly
et al, based on the relatively short scanning time (15 sec-
onds) reported, their specimens were most likely scanned
with a relatively large voxel size (0.4 rather than 0.2 mm).
In our study, we directly compared measurement accu-
racy between the soft-tissue presence and absence con-
ditions for the 0.4-mm scans and found no differences
(Tables I and II); this further confirmed that soft-tissue
condition probably does not affect alveolar bone mea-
surement accuracy when the specimens are scanned at
a 0.4-mm voxel size.
For the 0.2-mm scans, no previous studies have com-
pared alveolar bone measurement accuracy between the
soft-tissue presence and absence conditions. Our study,
therefore, was the first that directly addressed this issue.
Unlike the 0.4-mm scans, the 0.2-mm scans clearly dem-
onstrated inferior accuracy associated with soft-tissue
absence (Tables I and II). On one hand, this finding
further refutes the speculations that soft-tissue removal
improves the accuracy of CBCT alveolar bone
March 2013
Vol 143
Issue 3 American Journal of Orthodontics and Dentofacial Orthopedics
Wood et al
measurements; on the other, this finding suggests
a yet-to-be recognized problem associated with soft-
tissue removal when specimens are scanned at a small
voxel size, such as 0.2 mm.
Clearly, it is important to confirm that the inferior
accuracy associated with 0.2-mm soft-tissue absence
relative to 0.2-mm soft-tissue presence is not caused
by experimental errors. One potential error is alveolar
bone damage that might have occurred during soft-
tissue removal and specimen sectioning. Undoubtedly,
to completely rule out alveolar bone damage, one
must compare the physical measurements of alveolar
bone height before and after soft-tissue removal.
Such a comparison, however, is impractical because
physical measurement of the alveolar bone before
soft-tissue removal is extremely difficult and highly in-
accurate. Therefore, the possibility of systematic alveo-
lar bone damage can only be ruled out by using
indirect evidence. In this study, care was taken during
soft-tissue removal; all sections for the physiologic
measurements were obtained by using a low-speed di-
amond saw with water lubrication; no alveolar damage
was observed on the specimens. From a different per-
spective, the mean differences between the CBCT and
physical measurements of the soft-tissue presence
and absence 0.2-mm conditions were essentially iden-
tical (Table I). Assuming that systematic damage of the
alveolar bone had happened during soft-tissue removal
and specimen sectioning, the physical measurements of
the alveolar bone height would become smaller; this
would result in a systematic overestimation of the
mean DCBCT-Phy value (positive values) for the soft-
tissue presence images but not for soft-tissue absence
images. These results were not found (Table I), so the
assumption of systematic bone damage is probably in-
correct. Furthermore, the differences in measurement
accuracy between the soft-tissue presence and absence
0.2-mm conditions were reflected in the limit of agree-
ment ranges and percentages of clinical inaccuracy
rather than the mean DCBCT-Phy values (Tables I and
II; Fig 2); this strongly indicates increased variability
of measurements associated with soft-tissue absence.
Because the same set of physical measurements was
used for both soft-tissue presence and absence condi-
tions, it is not reasonable that the same physical mea-
surement error (caused by potential alveolar bone
damage or any other unidentified errors) increases the
variability of only soft-tissue absence but not soft-
tissue presence. A more reasonable explanation, there-
fore, is that some differences inherent in the soft-tissue
presence and absence 0.2-mm CBCT images, rather
than the physical measurements, caused the differences
in data variability between these 2 conditions.
Wood et al
This latter explanation is supported by the qualitative
features of the CBCT images (Fig 3). Overall, the dentoal-
veolus in the soft-tissue absence images demonstrated
greater contrast and surface brightness than did the
soft-tissue presence images. This distinction was espe-
cially prominent at the enamel and alveolar bone sur-
faces in the 0.2-mm images. Since all images were
reconstructed to the computer screen from DICOM
data by using the same protocols and display, contrast,
and brightness settings, most likely these differences
must be attributed to factors involved in radiographic
physics or the primary reconstruction process from raw
CBCT data.24 Although the exact mechanisms for these
differences remain uncertain, the brighter surface con-
tours shown in the soft-tissue absence images strongly
indicate that they were affected by greater beam-
hardening artifacts.24 Normally, when the overlying
soft tissues are present, the low energy portion of the
polychromatic beam is infiltrated (probably mainly by
the facial skins and muscles, and slightly by the gingival
tissues) before reaching the alveolar bone and tooth
crowns. After soft-tissue removal, this infiltration no
longer existed, and the lower-energy photons were di-
rectly attenuated at the tooth and bone surfaces, result-
ing in an artificially high attenuation coefficient at the
enamel and bone surfaces and subsequently brighter
contours after primary reconstruction with the same al-
gorithm designed for clinical patients. Reasonably, this
artifact, which may or may not be the only one involved
in soft-tissue absence conditions, can increase the diffi-
culty in identifying the alveolar crest and measurement
inaccuracy.
Evidently, one would expect that a higher beam-
hardening artifact to show in 0.4-mm soft-tissue ab-
sence than in the 0.4-mm soft-tissue presence images,
too. Based on the image features (Fig 3), this was indeed
the case. The reason that the 0.4-mm soft-tissue absence
images did not show an even greater measurement inac-
curacy than did the 0.4-mm soft-tissue presence images
is not immediately clear, but, most likely, the measure-
ment inaccuracy caused by the blurriness of the alveolar
crest as a result of a large voxel size (0.4 mm) tends to
override the differences caused by the beam-hardening
artifacts. This point, and the interaction between voxel
sizes and soft-tissue conditions, can be further con-
firmed by adding more voxel sizes in future studies.
Various software programs have been used to
quantitatively analyze bone in CBCT and computed to-
mography images.5,6,9,25,26 Developed by different
persons or groups, these programs can differ in image
reconstruction, display, and analysis, and potentially
affect the accuracy of alveolar bone measurements.
Practically, however, it is extremely hard to compare all
American Journal of Orthodontics and Dentofacial Orthopedics
361
of them in 1 study. We compared Dolphin with Osirix.
Osirix is the most widely used DICOM data viewer in
the medical field and runs on a Macintosh operating
system (Apple, Cupertino, Calif); Dolphin is currently
one of the most commonly used programs in dentistry,
especially in orthodontics, and runs on a Windows
operating system (Microsoft, Redmond, Wash). Our
data, in a number of ways, demonstrated that the use
of either software program resulted in an identical
degree of accuracy for alveolar bone height
measurements. These results indicate that software
choice, at least between 2 of the most widely used
ones, is probably not a major concern for alveolar
bone linear measurements from CBCT images,
although generalization of this finding to other
DICOM viewer programs might require further
confirmation.
For the clinical implications, overall, pigs have similar
craniofacial anatomy and function to humans,27,28 with
highly comparable alveolar bone morphology, size, and
growth features; therefore, the use of a pig model is
clinically relevant.29 The pigs we used were equivalent
in age to human adolescents. Many human cadaver-
based CBCT studies had to use specimens from elderly
subjects (60-90 years) because of availability issues,
some with extensive dental work and missing teeth.6,9,14
Because of the differences of maxillary bone thickness
between adolescents and adults, findings based on
adults might not be directly applied to adolescents, at
least to the maxillary molar region of adolescents.6 In
comparison, our experimental materials were more com-
parable to adolescent humans, and the findings are
therefore more relevant and useful for the adolescent or-
thodontic patient population.
In this study, in addition to the commonly studied
parameters of differences between CBCT and physical
measurements and limits of agreement, we compared
the proportions of clinically inaccurate measurements
for each factor. The purpose of this parameter was to
provide a more practical way to understand the fre-
quency and severity of the inaccuracies involved in al-
veolar bone measurements from CBCT images. Two
criteria, 1 and 0.5 mm, were used to determine whether
a CBCT measurement was clinically inaccurate. A mea-
surement error up to 1 mm has been considered ac-
ceptable for the assessment of the periodontium,
especially alveolar bone, during clinical examinations.
For orthodontic patients, a major reason that we
want to measure their alveolar bones is to know the
periodontal sequelae after orthodontic treatment,
such as rapid palatal expansion.12 With no accuracy
criterion currently established specific to radiographic
alveolar bone readings, we reason that, to justify the
March 2013
Vol 143
Issue 3
362
use of CBCT for alveolar bone measurements, with its
associated additional cost and radiation, it should at
least have an accuracy level comparable with clinical
examinations. Therefore, a 1-mm criterion is probably
more reasonable and practical. Based on this criterion
(inaccurate, absolute DCBCT-Phy value $1 mm), our
data clearly demonstrated that, when a 0.4-mm voxel
size (commonly used for clinical patients) was adopted,
approximately 23% to 27% and 8% to 14% (ranges of
the 0.4-mm conditions in Table II) of measurements
from the maxillary and mandibular molar regions, re-
spectively, were clinically inaccurate. With strong rele-
vance of our specimens to adolescent humans as
discussed above, the high inaccuracy rate in the maxil-
lary molar region warrants caution when using CBCT
measurements from this region to interpret periodontal
sequelae of orthodontic treatment in adolesecents.12,13
Clearly, one can argue that the frequency of inaccu-
rate measurements would be different if we had used a
more stringent criterion (inaccuracy, absolute DCBCT-Phy
value $0.5 mm). Indeed, more CBCT measurements
would be considered inaccurate (35%-44% and 19%-
28% measurements from the maxillary and mandibular
molar regions, respectively; Table II). Notably, the
same patterns of inaccuracy (maxilla worse than mandi-
ble, soft-tissue presence 0.4 mm equal to soft-tissue ab-
sence 0.4 mm, and soft-tissue absence 0.2 mm worse
than soft-tissue presence 0.2 mm) remained, although
statistically the differences were insignificant (Table II).
On the other hand, because 0.5 mm is a much more
stringent criterion than what is used in clinical examina-
tions, whether it should be used to justify or oppose the
use of CBCT images for alveolar bone measurements re-
quires further investigation and discussion; this was be-
yond the scope of our study.
Based on our findings, this inaccuracy problem
can be improved to a certain extent by using small
voxel-size scans (0.2 mm), while keeping facial and
gingival soft tissues intact. Although the latter is al-
ready the case for most patients, because of radiation
concerns, whether small voxel-size scans should be
used on patients for this purpose should be carefully
considered.
As with many other studies on this topic, this study
had limitations. Several more factors, such as patient
movement during scanning and metal artifacts, were
not included.14,30 The former factor is a limitation in
many studies on this topic involving the use of dead
experimental materials. The latter factor was examined
recently in 2 studies. One found that CBCT is less
affected by metal artifacts than multidetector
computed tomography,14 whereas the other found that
regions near metal rods were substantially affected.30
March 2013
Vol 143
Issue 3 American Journal of Orthodontics and Dentofacial Orthopedics
Wood et al
Based on the latter study, caution is needed when apply-
ing our findings to patients who are still wearing metallic
orthodontic appliances. Additionally, in the soft-tissue
absence samples, the lingual soft tissues were not re-
moved to keep the jaws steady for CBCT scanning. Al-
though these tissues might increase overall image
degradation, the data from this study (Fig 2) and
others5,9,14 suggest that this is not a critical issue
because buccal alveolar bone measurement appears to
be primarily affected by the soft tissues on the buccal
side (facial and gingival). This is particularly the case
for the beam-hardening artifacts as explained above.
Lastly, only 1 computed tomography machine and 1
rater were used; both could increase the bias of the find-
ings. Blinding techniques were used in this study to re-
duce rater bias, and future studies should make
comparisons with more machines and raters.
However, through this study, overall, 2 points are
clear about whether one can accurately measure alveolar
bone height from CBCT images. First, the answer is not
simply yes or no, and it might depend on multiple fac-
tors. Second, to accurately assess alveolar bone changes
resulting from orthodontic treatment in adolescent pa-
tients, improvements in CBCT technology and measur-
ing techniques are needed.
CONCLUSIONS
1. In adolescent pigs (comparable with adolescent
humans), when measuring the buccal alveolar bone
height in the maxillary molar region from CBCT im-
ages by 0.4-mm voxel-size scans, relatively large (up
to approximately a 2-mm discrepancy from the phys-
ical truth) and frequent (over 20%) clinical inaccura-
cies ($1 mm different from the physical truth)
occurred. This is most likely because the buccal alve-
olar bone in the maxilla is thin (on average \1 mm).
2. For 0.4-mm voxel-size scans, measurements from
the mandibular molar regions were generally within
1 mm from the physical truth, with only about
a 10% frequency in obtaining clinically inaccurate
measurements ($1 mm different from the physical
truth).
3. Alveolar bone height measurement accuracy varied
with voxel-size and soft-tissue conditions. For
0.4-mm voxel-size scans, removing the soft tissues
did not change alveolar bone height measurement
accuracy. Compared with the 0.4-mm voxel-size
scans, 0.2-mm voxel-size scans showed improved
measurement accuracy only when the soft tissue
was present. Soft-tissue removal cancelled the im-
provement from using a smaller voxel size, probably
because of increased beam-hardening artifacts.
Wood et al
4. The different imaging software programs had little
impact on the accuracy of the CBCT measurements.
5. Overall, the accuracy of alveolar bone height mea-
surements from CBCT images in adolescent patients
is determined by multiple factors. Improvement of
its accuracy, especially for 0.4-mm voxel-size scans
of the maxillary molar region, is needed for its better
clinical use.
REFERENCES
1. Oberoi S, Knueppel S. Three-dimensional assessment of impacted ca-
nines and root resorption using cone beam computed tomography.
Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2011;113:260-7.
2. Bouwens DG, Cevidanes L, Ludlow JB, Phillips C. Comparison of
mesiodistal root angulation with posttreatment panoramic radio-
graphs and cone-beam computed tomography. Am J Orthod Den-
tofacial Orthop 2011;139:126-32.
3. Caglayan F, Tozoglu U. Incidental findings in the maxillofacial re-
gion detected by cone beam CT. Diagn Interv Radiol 2012;18:
159-63.
4. Corbridge JK, Campbell PM, Taylor R, Ceen RF, Buschang PH.
Transverse dentoalveolar changes after slow maxillary expansion.
Am J Orthod Dentofacial Orthop 2011;140:317-25.
5. Patcas R, Muller L, Ullrich O, Peltomaki T. Accuracy of cone-beam
computed tomography at different resolutions assessed on the
bony covering of the mandibular anterior teeth. Am J Orthod Den-
tofacial Orthop 2012;141:41-50.
6. Timock AM, Cook V, McDonald T, Leo MC, Crowe J, Benninger BL,
et al. Accuracy and reliability of buccal bone height and thickness
measurements from cone-beam computed tomography imaging.
Am J Orthod Dentofacial Orthop 2011;140:734-44.
7. Sun Z, Smith T, Kortam S, Kim DG, Tee BC, Fields H. Effect of bone
thickness on alveolar bone-height measurements from cone-beam
computed tomography images. Am J Orthod Dentofacial Orthop
2011;139(2):e117-27.
8. Leung CC, Palomo L, Griffith R, Hans MG. Accuracy and reliability
of cone-beam computed tomography for measuring alveolar bone
height and detecting bony dehiscences and fenestrations. Am J Or-
thod Dentofacial Orthop 2010;137(Supp):S109-19.
9. Ganguly R, Ruprecht A, Vincent S, Hellstein J, Timmons S, Qian F.
Accuracy of linear measurement in the Galileos cone beam com-
puted tomography under simulated clinical conditions. Dento-
maxillofac Radiol 2011;40:299-305.
10. Lascala CA, Panella J, Marques MM. Analysis of the accuracy of
linear measurements obtained by cone beam computed tomogra-
phy (CBCT-NewTom). Dentomaxillofac Radiol 2004;33:291-4.
11. Suomalainen A, Vehmas T, Kortesniemi M, Robinson S, Peltola J.
Accuracy of linear measurements using dental cone beam and
conventional multislice computed tomography. Dentomaxillofac
Radiol 2008;37:10-7.
12. Rungcharassaeng K, Caruso JM, Kan JY, Kim J, Taylor G. Factors
affecting buccal bone changes of maxillary posterior teeth after
rapid maxillary expansion. Am J Orthod Dentofacial Orthop
2007;132:428.e1-8.
13. Lee KM, Kim YI, Park SB, Son WS. Alveolar bone loss around lower
incisors during surgical orthodontic treatment in mandibular
prognathism. Angle Orthod 2012;82:637-44.
American Journal of Orthodontics and Dentofacial Orthopedics
363
14. Patcas R, Markic G, Muller L, Ullrich O, Peltomaki T,
Kellenberger CJ, et al. Accuracy of linear intraoral measurements
using cone beam CT and multidetector CT: a tale of two CTs. Den-
tomaxillofac Radiol 2012;41:637-44.
15. Ono A, Motoyoshi M, Shimizu N. Cortical bone thickness in the
buccal posterior region for orthodontic mini-implants. Int J Oral
Maxillofac Surg 2008;37:334-40.
16. Silvestrini Biavati A, Tecco S, Migliorati M, Festa F, Marzo G,
Gherlone E, et al. Three-dimensional tomographic mapping re-
lated to primary stability and structural miniscrew characteristics.
Orthod Craniofac Res 2011;14:88-99.
17. Wang S, Liu Y, Fang D, Shi S. The miniature pig: a useful large an-
imal model for dental and orofacial research. Oral Dis 2007;13:
530-7.
18. Bland JM, Altman DG. Statistical methods for assessing agreement
between two methods of clinical measurements. Lancet 1986;1:
307-10.
19. Devlin H, Horner K, Ledgerton D. A comparison of maxillary and
mandibular bone mineral densities. J Prosthet Dent 1998;79:
323-7.
20. Weaver ME, Jump EB, McKean CF. The eruption pattern of
permanent teeth in miniture swine. Arch Oral Biol 1969;14:
323-31.
21. Gedrange T, Kobel C, Harzer W. Hard palate deformation in an
animal model following quasi-static loading to stimulate that
of orthodontic anchorage implants. Eur J Orthod 2001;23:
349-54.
22. Wenzel A, Haiter-Neto F, Frydenberg M, Kirkevang LL. Variable-
resolution cone-beam computerized tomography with enhance-
ment filtration compared with intraoral photostimulable phosphor
radiography in detection of transverse root fractures in an in vitro
model. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2009;
108:939-45.
23. Kim DG, Christopherson GT, Dong XN, Fyhrie DP, Yeni YN. The ef-
fect of microcomputed tomography scanning and reconstruction
voxel size on the accuracy of stereological measurements in hu-
man cancellous bone. Bone 2004;35:1375-82.
24. Scarfe WC, Farman AG. What is cone-beam CT and how does it
work? Dent Clin North Am 2008;52:707-30.
25. Fourie Z, Damstra J, Gerrits PO, Ren Y. Accuracy and reliability of
facial soft tissue depth measurements using cone beam computer
tomography. Forensic Sci Int 2010;199:9-14.
26. Milchteim C, Yu WD, Ho A, O'Brien JR. Anatomical parameters of
subaxial percutaneous transfacet screw fixation based on the anal-
ysis of 50 computed tomography scans. J Neurosurg Spine 2012;
16:573-8.
27. Herring SW. The dynamics of mastication in pigs. Arch Oral Biol
1976;21:473-80.
28. Strom D, Holm S, Clemensson E, Haraldson T, Carlsson GE.
Gross anatomy of the mandibular joint and masticatory mus-
cles in the domestic pig (Sus scrofa). Arch Oral Biol 1986;31:
763-8.
29. Yeh K, Popowics T, Rafferty K, Herring S, Egbert M. The effects of
tooth extraction on alveolar bone biomechanics in the miniature
pig, Sus scrofa. Arch Oral Biol 2010;55:663-9.
30. Pauwels R, Stamatakis H, Bosmans H, Bogaerts R, Jacobs R,
Horner K, et al. Quantification of metal artifacts on cone beam
computed tomography images. Clin Oral Implants Res 2011 Dec
15 [Epub ahead of print].
March 2013
Vol 143
Issue 3