Dentomaxillofacial Radiology (2006) 35, 410416

q 2006 The British Institute of Radiology

http://dmfr.birjournals.org

RESEARCH
Accuracy of three-dimensional measurements using
cone-beam CT
HM Pinsky1, S Dyda1, RW Pinsky2, KA Misch3 and DP Sarment*,3
1
University of Michigan School of Dentistry, 1011 North University Avenue, Ann Arbor, MI 48109-1078 USA; 2Department
of Radiology, University of Michigan School of Medicine, 1011 North University Avenue, Ann Arbor, MI 48109-1078 USA;
3
Department of Periodontics and Oral Medicine, University of Michigan School of Dentistry, 1011 North University Avenue,
Ann Arbor, MI 48109-1078 USA

Objectives: Lesions causing intraosseous defects in the head and neck region are difficult to
diagnose using two-dimensional radiography, and three-dimensional (3D) data provided by CT is
useful but often difficult to obtain. Recently, cone-beam CT (CBCT) was made available, with the
potential to become a practical tool in dentistry. However, there is limited evidence to prove that
defect volume can be determined accurately. Therefore, this in vitro validation study aimed at
establishing whether linear and 3D CBCT, using volumetric measurements, is accurate for
determining osseous defect sizes.
Methods: Depth and diameter of simulated bone defects in (i) an acrylic block and (ii) a human
mandible were blindly measured electronically by five examiners using CBCT. Linear
measurements were compared with predetermined machined dimensions. Using software, volume
extraction was performed by another examiner on the acrylic phantom and compared with known
dimensions. Data were analysed using paired t-tests.
Results: Using the acrylic block, mean width accuracy was 2 0.01 mm (^ 0.02 SE) and mean
height difference was 2 0.03 mm (^ 0.01 SE; P . 0.05). For the human mandible, mean width
accuracy was 2 0.07 mm (^ 0.02 SE) and mean height accuracy was 2 0.27 mm (^ 0.02 SE;
P , 0.01). Volume accuracy was 2 6.9 mm3 (^ 4 SE) for automated calculations and 2 2.3 mm3
(^ 2.6 SE) for the manual measurements ( P , 0.001).
Conclusions: CBCT has the potential to be an accurate, non-invasive, practical method to reliably
determine osseous lesion size and volume. Further clinical validation will lead to a vast array of
applications in oral and maxillofacial diagnosis.
Dentomaxillofacial Radiology (2006) 35, 410416. doi: 10.1259/dmfr/20987648

Keywords: CT, osseous defect, periapical diseases, diagnostic imaging

Introduction

Observation of a radiographic lesion is usually a sign
which is suggestive of pathology. Together with medical
history and clinical testing, diagnosis and treatment
recommendations can be made. While the information
gleaned from a dental radiograph is substantial, there are
limitations associated with the use of a two-dimensional
image.1 They include localization and size of a lesion in a
buccolingual plane,2 surface characteristics of the lesion
(i.e. smooth versus rough)3 and changes that appear over
time when comparison of films to detect progression or
*Correspondence to: David Sarment, DDS, MS, Department of Periodontics and
Oral Medicine, University of Michigan School of Dentistry, 1011 North University
Avenue, Ann Arbor, MI 48109-1078 USA; E-mail: sarment@umich.edu
Received 9 November 2005; revised 27 December 2005; accepted 5 January 2006
healing is made.3,4 When surgical access is necessary,
quality imaging is critical to localize the lesion and its
proximity to important structures. Unfortunately, tra-
ditional radiographs cannot provide sufficient information
and multiple corrective actions may be necessary during
surgery. Therefore, an imaging modality with three-
dimensional (3D) capability is essential to enhance
diagnosis and treatment.
3D imaging using CT has been available for a number
of years. Although CT has been utilized for head and neck
diagnosis5 and various oral surgical procedures,6 its
limitations (hospital based, excessive radiation, cost and
general practicality) are one of the reasons for the paucity
of literature containing information about the use of this
technology for routine dental applications. Recently, a
 Cone beam volumetry
HM Pinsky et al 411

more practical new generation of CT, cone-beam CT

(CBCT) has become available.7 CBCT technology has
been exploited in machines specifically designed for head
and neck imaging.8 13 It involves a unit that can be
comparable in size with a conventional panoramic radio-
graphic machine. It is able to accomplish rapid volumetric
image acquisition in a time of exposure in the range of 20 s,
the amount of radiation exposure can be similar to that of a
full mouth series,14,15 image quality is as good as
conventional CT and7 it has a voxel (the smallest
detectable unit) size which may be as low as 0.2 mm (it
is 0.5 1 mm for most fan-beam units).
Consequently, CBCT has the potential to become a routine
non-invasive diagnostic instrument for various dental
applications in which defect characterization, localization
and volume are important. Unfortunately, these applications
have rarely been explored, and pre-clinical evaluation of
CBCT verifying linear and volume accuracy has scarcely
been performed. In addition, small defects such as a
periodontal or periapical lesion have not been studied.
Therefore, this study aims at establishing whether CBCT
linear and volume measurements of small defects represen-
tative of typical dental radiolucencies are accurate, when
compared with known standard direct measurements.

Materials and methods

In order to establish accuracy in controlled environments, two
consecutive pre-clinical studies were conducted. The first part
used a cast acrylic block with 64 engineered simulated
cavities. For the second part, with the University of Michigan
anatomical donations program approval, a human mandible
was obtained and 21 engineered simulated bone defects of
various sizes were prepared. Measurement techniques were
identical for both arms of this study.
Preparation of samples: First, to establish a gold
standard, simulated defects of known height and width
were prepared using flute end mills (Niagra Cutter,
Amherst, NY). Samples were mounted and defects created
using a vertical milling machine (Bridgeport, CT).
According to manufacturers, drill diameter was
^ 0.05 mm and depth accuracy was ^ 0.0254 mm.
For the cast acrylic model, 64 defects (depths 4 mm,
5 mm, 6 mm and 7 mm; diameters 4 mm, 6 mm, 8 mm and
10 mm) were arbitrarily placed to obtain four samples of
each size (Figure 1a). For the mandible, 21 simulated
defects were drilled: 4 mm and 6 mm diameters were
combined to depths varying from 2 mm to 4.5 mm (0.5 mm
increments). Since the surface of cortical bone is uneven, a
notch was positioned to precisely locate the measurement
positions (Figure 1b). The actual volume was subsequently
calculated using known diameters and depths.
Measurements: Two methods were used in this study.
The first method consisted of obtaining linear measure-
ments of the test holes. A volume was then mathematically
calculated. The second method utilized software for
automated volume extraction.
Figure 1 Measurements were performed on (a) a cast acrylic block and
(b) a human mandible, using simulated defects. Note that for the mandible
the exact location of CBCT measurements was identified with a notch
(arrow)
Linear measurements: Scanning of samples was per-
formed using a CBCT dental unit (I-CAT, ISI, PA,
98 mAs, 120 kVp). Files were transferred and read using
software with linear measurement capability (Xoran, Ann
Arbor, MI). The voxel size utilized was 0.2 mm. So the
accuracy was limited to the size of the voxel. Examiners
(dental students, dental residents and faculty) were briefly
trained to manipulate the software and perform measure-
ments of diameters and depths of the simulated defects
using CBCT scan images (Figure 2). Training was
performed until each examiner felt comfortable with the
use of electronic measurement tools, but there was no
attempt to calibrate them. Examiners were masked so that
the actual size was not known. All measurements of
diameter and depth were obtained for each defect. Using
these depth and diameter measurements, volumes were
mathematically calculated. Measurements were repeated
twice, at least 1 week apart.
Automated volume calculation: CBCT data were
imported into a separate software (Analyze; Analyze
Direct Inc., MN). Segmentation was performed by thresh-
olding densities and the area of interest was indicated
(Figure 3). Surface extraction was then automatically
repeated for each axial section, and sum calculation
algorithm was executed to obtain total volume. All
measurements were performed by the same trained
examiner and repeated at two separate time intervals.

Dentomaxillofacial Radiology
 Cone beam volumetry
412 HM Pinsky et al

Figure 2 Linear CBCT measurements. (a) Width and (b) height were measured on the cast acrylic block after scanning. (c) Similarly, the mandible was
scanned and a 3D reconstruction showed orientation (using Mimics, Materialise Inc., MD). (d) Defects were localized (insert). Width and height
measurements were recorded blindly

Statistical analysis will be greater. A 5% level of significance and an 80%

Sample size was calculated using pilot data. It was
estimated that measurements vs true measurements will
have a large size effect since a medium- or small-size effect
will likely be found in a clinical study in which deviations
power were utilized to calculate sample sizes.16 Measure-
ments were recorded in comparison with the gold standard,
with a negative number indicating underestimation and a
positive number indicating overestimation. Interclass and
intraclass correlation coefficients were calculated to

Figure 3 Automated volume evaluation was utilized using software. (a) The Plexiglas model was scanned and viewed using CBCT. (b) Segmentation
threshold was adjusted (arrow). (c) Once ideal threshold values were determined (arrow), (d) total volume was quantified by automated segmentation of
the artificial defect

Dentomaxillofacial Radiology
 Cone beam volumetry
HM Pinsky et al 413

determine variability within examiners, as well as amongst Table 1 Interclass correlations between examiners. Average (minimum-
techniques. Two-tailed paired t-tests were performed to maximum correlation) for each measurement
determine statistical differences between tested methods CT width CT height
and known sizes. Cast acrylic 0.97(0.95 0.99) 0.97(0.98 0.99)
Mandible 0.93(0.85 0.98) 0.96(0.95 0.97)

Results varied between 0.93 and 0.97 on average, with slightly

better reliability using the cast acrylic (Table 1). CBCT
Intraexaminer reliability varied from 0.75 to 0.99, with an measurements were compared with prepared cavities
average of 0.96 (data not shown). Interclass correlation (Figure 4). Using the acrylic block, mean width accuracy

Figure 4 Linear measurements using CBCT, for (a) the cast acrylic and (b) the mandible, are plotted against cavity height and width. Mean errors and
standard deviations for each sample are shown as inserts

Dentomaxillofacial Radiology
 Cone beam volumetry
414 HM Pinsky et al

was 2 0.01 mm (^ 0.02 SE) and mean height difference
was 2 0.03 mm (^ 0.01 SE) (P . 0.05, Figure 4a insert).
For the human mandible, mean width accuracy was
2 0.07 mm (^ 0.02 SE) and mean height accuracy was
2 0.27 mm (^ 0.02 SE) (P , 0.01, Figure 4b insert). There
was also statistically significant difference between acrylic
and mandibular mean widths (data not shown). Volume
accuracy was 2 6.9 mm 3 (^ 4 SE) for automated
calculations and 2 2.3 mm3 (^ 2.6 SE) for the manual
measurements (P , 0.001; Figure 5). This error rep-
resented 2% and 0.4% of total volume for the automated
and manual methods, respectively. When exploring
whether increasing height or width induced greater
measurement errors, no correlation was found (data not
shown).
Discussion
The purpose of this study was to verify whether CBCT
measurements were accurate for small simulated defects,
which were similar in size to radioluencies caused by
periodontal or periapical diseases. A machined acrylic
block and a mandibular specimen were utilized. Results
show that intraclass and interclass reliability were high for
CBCT digital linear measurements. It is important to note

Figure 5 Errors in volume calculations, using the acrylic block (mean ^ SE) and percentage of total volume. AV, automated volume calculation; LM,
volume calculated from linear measurements. Note that SE are comparatively high because mean errors are small

Dentomaxillofacial Radiology

that reliability was obtained, despite the fact that
examiners had no previous training in CT image analysis.
Since little effort was made to calibrate the examiners,
these results suggest that the methods are not examiner
dependent.
Mean linear accuracy was lower than 0.1 mm for cast
acrylic, and smaller than 0.3 mm for the mandible
specimen. This correlates with the chosen voxel size,
the smallest identifiable unit (similar to a pixel on a
personal printer) of 0.2 mm on this CBCT unit. Overall,
measurements were either smaller or within two voxels
(0.2 mm/voxel), but it is expected that such accuracy
cannot be repeated on patients because of other factors
such as presence of trabeculations, soft tissue or move-
ment. The comparison of cast acrylic and the single
mandible in the present study confirms these specu-
lations. Further research may confirm that variations in
bone density could impact on the results. The largest
error was found in height measurements on the mandible.
A likely explanation is that osseous surface at the apex of
the defect is more difficult to define in the presence of
trabeculations because CBCT densities vary with various
bone densities. This is particularly true while using a dry
mandible since trabeculations are wide, only contrasting
with air. Again, despite these potential limitations, CBCT
errors remained small and clinically insignificant. For
CBCT volume estimations, manual measurements were a
result of linear calculations. An automated volume
extraction method was explored in the hope that it will
apply to more complex defect shapes. Although there was
a trend of obtaining greater errors while using this
method, it was neither statistically nor clinically signifi-
cant because it only represented a small percentage of
total volume. At the outer limit of the material, a
software algorithm calculates the transition between the
material and air by averaging values, thereby producing a
grey voxel which is more difficult to identify and could
explain this trend.
There are only a few studies investigating whether
CBCT measures clinical accuracy. In a report using human
skulls, Lascala et al also found a non-statistical under-
estimation with CBCT linear measurements. They utilized
2 mm metal spherical markers and recorded large distances
(30 100 mm) which could account for greater (but lesser
percentage) errors.17 In a geometric accuracy study using
mathematical distance calculations and a different CBCT
unit, Marmulla et al found an image deviation of 0.13 mm
on average, which was below the voxel size.18 This study
involved a geometric cube with edge lengths of 12 mm,
which was in the range of the model we used. They were
investigating linear relationships to determine the accuracy
of known points in space as determined by the intersection
of geometric lines. In contrast, the present study evaluated
References
1. Quirynen M, Lamoral Y, Dekeyser C, Peene P, van Steenberghe D,
Bonte J, et al. CT scan standard reconstruction technique for reliable
jaw bone volume determination. Int J Oral Maxillofac Implants 1990;
5: 384 389.
Cone beam volumetry
HM Pinsky et al 415
small measurements and a new automated volume
calculation technique.
Other reports also suggested that CT scanning is a better
diagnostic tool for periodontal lesions located buccally,
lingually, or within furcations.19 In a study comparing
radiographs to CT images to diagnose the interproximal
periodontal lesions on human specimens, Fuhrmann et al
reported an underestimation of 0.6 mm with radiographs
and an overestimation of 0.2 mm using CT. However, for
infrabony defects, radiographic error increased to 2.2 mm
and was underestimated by 0.2 mm with CT.20 Interest-
ingly, CT error was similar to our results, although the
machines were very different.
A mandible was chosen as it is representative of the
anatomy imaged using the CBCT machine utilized in this
report. In addition, a buccal approach was chosen to obtain
direct measurements. This model is similar to Bender et al21
who used mechanically drilled holes, but differs from
Tirell et al22 who used decalcifying acid to create artificial
defects with random shapes. Artificial lesions could also be
prepared through an extraction socket,21 or by sectioning
mandibles,23,24 but standardized defects were necessary in
the present study to standardize measurements.
Overall, the present study demonstrates that clinically
acceptable accuracy can probably be obtained for oral
applications requiring evaluation of small osseous defects.
In addition, image segmentation and volume extraction has
the potential for use in clinical trials or other clinical
settings where accurate evaluation and volume changes
over time are important.
Since CBCT image quality on patients is decreased by
the presence of soft tissue and possible patient movement
during scanning, further studies are indicated to confirm
the present results clinically. In addition, the automated
volume methodology should be further investigated for
analysis of complex images associated with trabeculated
bone or multilocular radioleucencies. It is expected that
future software development will also simplify automated
determination of defect volume, regardless of their
anatomic localization and shape.
Acknowledgments
The authors would like to acknowledge the following people
for their invaluable advice and assistance: Patricia Anderson,
MS; Christopher Dackson; Ken Guire, MA; Jim Hamilton,
DDS; Anna Shafto, BS, MT. We also thank Xoran Inc. and
Materialise Inc. for the software support. This study was
funded in part by the University of Michigan Department of
Periodontics and Oral Medicine, and by Imaging Sciences
International, Inc.
2. Farman AG, Kushner GM, Gould AR. A sequential approach to
radiological interpretation. Dentomaxillofac Radiol 2002; 31:
291 298.
Dentomaxillofacial Radiology
 Cone beam volumetry
416 HM Pinsky et al
3. Andreasen JO, Rud J. Correlation between histology and radiography
in the assessment of healing after endodontic surgery. Int J Oral Surg
1972; 1: 161 173.
4. Friedman S, Lustmann J, Shaharabany V. Treatment results of apical
surgery in premolar and molar teeth. J Endod 1991; 17: 30 33.
5. Bou Serhal C, Jacobs R, Flygare L, Quirynen M, van Steenberghe D.
Perioperative validation of localisation of the mental foramen.
Dentomaxillofac Radiol 2002; 31: 39 43.
6. Sarment DP, Sukovic P, Clinthorne N. Accuracy of implant
placement with a stereolithographic surgical guide. Int J Oral
Maxillofac Implants 2003; 18: 571 577.
7. Mozzo P, Procacci C, Tacconi A, Martini PT, Andreis IA. A new
volumetric CT machine for dental imaging based on the cone-beam
technique: preliminary results. Eur Radiol 1998; 8: 1558 1564.
8. Ziegler CM, Woertche R, Brief J, Hassfeld S. Clinical indications for
digital volume tomography in oral and maxillofacial surgery.
Dentomaxillofac Radiol 2002; 31: 126 130.
9. Araki K, Maki K, Seki K, Sakamaki K, Harata Y, Sakaino R, et al.
Characteristics of a newly developed dentomaxillofacial X-ray cone
beam CT scanner (CB MercuRay): system configuration and physical
properties. Dentomaxillofac Radiol 2004; 33: 51 59.
10. Hashimoto K, Arai Y, Iwai K, Araki M, Kawashima S, Terakado M. A
comparison of a new limited cone beam computed tomography machine
for dental use with a multidetector row helical CT machine. Oral Surg
Oral Med Oral Pathol Oral Radiol Endod 2003; 95: 371377.
11. Arai Y, Tammisalo E, Iwai K, Hashimoto K, Shinoda K. Develop-
ment of a compact computed tomographic apparatus for dental use.
Dentomaxillofac Radiol 1999; 28: 245 248.
12. Danforth RA, Peck J, Hall P. Cone beam volume tomography: an
imaging option for diagnosis of complex mandibular third molar
anatomical relationships. J Calif Dent Assoc 2003; 31: 847 852.
13. Farman AG. Applying DICOM to dentistry. J Digit Imaging 2005;
18: 23 27.
Dentomaxillofacial Radiology
14. Mah JK, Danforth RA, Bumann A, Hatcher D. Radiation absorbed in
maxillofacial imaging with a new dental computed tomography
device. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2003;
96: 508 513.
15. Ludlow JB, Davies-Ludlow LE, Brooks SL. Dosimetry of two
extraoral direct digital imaging devices: NewTom cone beam CT and
Orthophos Plus DS panoramic unit. Dentomaxillofac Radiol 2003;
32: 229 234.
16. Cohen J. Statistical power analysis for the behavioral sciences, 2nd
edn. Hillsdale, NJ: Lawrence Erlbaum Associates; 1988.
17. Lascala CA, Panella J, Marques MM. Analysis of the accuracy of
linear measurements obtained by cone beam computed tomography
(CBCT-NewTom). Dentomaxillofac Radiol 2004; 33: 291 294.
18. Marmulla R, Wortche R, Muhling J, Hassfeld S. Geometric accuracy
of the NewTom 9000 Cone Beam CT. Dentomaxillofac Radiol 2005;
34: 28 31.
19. Pistorius A, Patrosio C, Willershausen B, Mildenberger P, Rippen G.
Periodontal probing in comparison to diagnosis by CT-scan. Int Dent
J 2001; 51: 339 347.
20. Fuhrmann RA, Bucker A, Diedrich PR. Assessment of alveolar bone
loss with high resolution computed tomography. J Periodontal Res
1995; 30: 258 263.
21. Bender IB, Seltzer S. Roentgenographic and direct observation of
experimental lesions in bone: I. J Am Dent Assoc 1961; 62: 152 160.
22. Tirrell BC, Miles DA, Brown Jr CE, Legan JJ. Interpretation of
chemically created lesions using direct digital imaging. J Endod
1996; 22: 74 78.
23. Shoha RR, Dowson J, Richards AG. Radiographic interpretation of
experimentally produced bony lesions. Oral Surg Oral Med Oral
Pathol 1974; 38: 294 303.
24. Lee SJ, Messer H. Radiographic appearance of artificially prepared
periapical lesions confined to cancellous bone. Int Endod J 1986; 19:
64 72.