Comparison of Cone Beam Computed Tomography Imaging With

8/3/2019 Comparison of Cone Beam Computed Tomography Imaging With

1/14

RESEARCH

Comparison of cone beam computed tomography imaging with

physical measures

SA Stratemann1, JC Huang1, K Maki2, AJ Miller1 and DC Hatcher1

1Division of Orthodontics, Department of Orofacial Sciences, School of Dentistry, University of California at San Francisco, USA;2Department of Orthodontics, Showa University, Tokyo, Japan

Objectives: The goal of this study was to determine the accuracy of measuring lineardistances between landmarks commonly used in orthodontic analysis on a human skull usingtwo cone beam CT (CBCT) systems.Methods: Measurements of length were taken using volumetric data from two CBCT

systems and were compared with physical measures using a calliper applied to one humanadult skull. Landmarks were identified with chromium steel balls embedded at 32 cranial and33 mandibular landmarks and the linear measures were taken with a digital calliper. The skullwas then scanned with two different CBCT systems: the NewTomH QR DVT 9000 (AperioInc, Sarasota, FL) and the Hitachi MercuRay (Hitachi Medico Technology, Tokyo, Japan).CT data including the landmark point data were threshold segmented using CyberMeds CBWorks software (CB Works 1.0, CyberMed Inc., Seoul, Korea). The resulting segmentationswere exported from CB Works as VRML (WRL) files to Amira software (Amira 3.1,Mercury Computer Systems GmbH, Berlin, Germany).Results: The error was small compared with the gold standard of the physical callipermeasures for both the NewTom (0.070.41 mm) and CB MercuRay (0.000.22 mm)generated data. Absolute error to the gold standard was slightly positive, indicating minorcompression relative to the calliper measurement. The error was slightly smaller in the CB

MercuRay than in the NewTom, probably related to a broader greyscale range for describingbeam attenuation in 12-bit vs 8-bit data.Conclusions: The volumetric data rendered with both CBCT systems provided highlyaccurate data compared with the gold standard of physical measures directly from the skulls,with less than 1% relative error.Dentomaxillofacial Radiology (2008) 37, 8093. doi: 10.1259/dmfr/31349994

Keywords: cone beam computed tomography; mandible; craniofacial

Introduction

With the emergence of cone beam CT (CBCT) and its

application to the craniofacial region,15 including thediscipline of orthodontics,69 the concept of evaluatingCBCTs accuracy for completing three-dimensional(3D) measurements from 3D-rendered volumetric andradiographic images10 needs to be ascertained, as hasbeen done with earlier and traditional CT systems.CBCT provides an alternative to traditional CT

systems1113 using reduced radiation and shorter

acquisition scan times, as well as lower billingcosts.1416 In CBCT, a conical beam of X-rays that issized to encompass a region of interest rotates about thepatient in a circular path.17 The CBCT system acquiresimage data in a single revolution of a paired source andset of detector arrays18 and collects a volume ofinformation, as opposed to a stack of multiple slicesof the scanned object as in conventional CT.19 Digitalimages consist of a set of ordered picture elements(pixels) arranged in a planar grid. Pixels are the smallestcomponents of the image and have specific rectangulardimensions. In digital radiography, the pixel colour isgrey, such that intensities from black to white are

*Correspondence to: Arthur J Miller, PhD Professor, Division of

Orthodontics, Department of Orofacial Sciences, School of Dentistry,

University of California at San Francisco, San Francisco, CA 941430438, USA;

E-mail: [email protected]

Received 5 December 2006; revised 17 April 2007; accepted 19 April 2007

Dentomaxillofacial Radiology (2008) 37, 8093 2008 The British Institute of Radiology

http://dmfr.birjournals.org


2/14

divided between a spectrum of greyscale values with 28

(256) shades in 8-bit data or 212 (4096) shades in12-bit data. Image sharpness or resolution is ultimatelyrelated to pixel size and the number of shades of grey.18

Error associated with volumetric data fundamentallystems from voxel size and data quality (8-bit vs 12-bit).This is because the intensity value assigned to eachvoxel depends on the mean intensity (or beam atten-uation in Hounsfield units (HU) of the materialcontained within its volume. With a smaller voxel size,less material is averaged, leading to a more accurateassignment of an attenuation value. Likewise, whenthere are more possible intensity values as in 8-bit vs12-bit data (256 vs 4096 shades of grey), there is abroader palette of choices from which the actualintensity value can be chosen.

Radiological studies to test the resolution, distortionand background noise of the CB MercuRayTM andother CBCT systems have been performed. Geometrical

calibration standards for the CBCT system provideadditional approaches to maintain the accuracy of sucha system from day to day.20 Using standardized testcharts and phantoms, Yamamoto et al21 found that theCB MercuRayTM has a spatial resolution of2.0 Lp mm21 and a 3D image resolution of 1.25 Lp mm21. Image distortion as measured bycircularity of a 7 mm radius acrylic phantom was94.53% of the ideal value with a small standarddeviation (SD). Araki et al,22 examining characteristicsof the same system, reported a resolving power of over2.0 Lp mm21 and circularity of a 4 mm pipe phantomat 99% of ideal. Image noise was found to be greater

than in standard CT, with a SD of the CT value ofwater of approximately 80 HU. Both of these studieswere conducted in the high-resolution D (dental) modepresent on an alternate version of the CB MercuRayTM,suggesting that our system, with the full head, mighthave a slightly lower resolution.

Other contributions to error in CT data relate toimaging artefacts and the error inherent in landmarkidentification. Artefacts in CT data can result from thebeam scatter of radiodense objects such as metallicdental restorations and even cortical bone. Backgroundartefacts can be related to the geometry of the path ofthe beam detector system. Landmark identification isconsidered to be, in the two-dimensional (2D) cephalo-

metric literature, the major source of error, with eachlandmark having a systematic pattern of the error.23

Similar error in landmark identification is to beexpected in 3D, but patterns of the error may differbecause of different types of problems that may beencountered in spatially locating landmarks. Forexample, the landmark nasion (Na), located at theintersection of the internasal suture with the fronto-nasal suture in the midsagittal plane in 2D cephalo-metry, shows a pattern primarily of vertical error inlocation on a plain film radiograph,23 partly due tofrequent burnout of the nasal bones. In 3D, theproblem may be more related to the difficulty in

establishing suture location due to loss of detail at thelevel of sutures in rendered volumes and in establishingthe midsagittal plane without the introduction of amarking plane. Therefore, the scatter pattern of error innasion identification in 2D, a vertically orientatedellipse, might be more mushroom-shaped in 3D withlittle error in finding the bony cortical surface and moredifficulty associated with centring the landmark withinthe confines of the two sutures.

Based on measurements made on two separate CTscans (1.5 mm slice thickness) using conventional CTsystems and physically digitized 3D coordinates takenon a set of 10 dry skulls, Richtsmeier et al24 reportedthat the mean error in landmark position was alwaysless than 0.5 mm and negligible for some landmarks.Repeatability studies demonstrated that measurementinaccuracy was responsible for less than 2% of the totalvariance in the data collected. The authors noted thatwhile coordinate data taken from a set of CT slices were

internally consistent, data from two separate scanscould not be directly compared because of differencesin translational and rotational components of thescanned object relative to the axes set up by the scan.Therefore, coordinate systems have to be aligned beforetwo different scans of the same object can be compared.Richtsmeier et al24 also indicated that locating landmarkson rendered volumes may be less labour intensive thanusing slice data for such purposes, and that usingreconstructions may lead to reduction in landmarkplacement because such 3D renderings more accuratelyreflect the biological object being studied, which wassupported by other investigators.12,25 Linear measures

taken from 28 mandibular landmarks, defined from3D reconstructions of CT axial slices using aparticular imaging program and a spiral CT systemand compared with a 3D space digitizer, indicated thatabsolute differences ranged from 0.001 mm to3.9 mm.26 Validation of obtaining 3D measures froma traditional CT system, scanning phantom objects(cube, sphere, cylinder) and an adult human skull,compared with physical measurements with a calliper,have shown differences ranging from 0% to 2.57%,validating the accuracy of using traditional CTsystems for 3D measures.27 Methods to calibrate theCBCT systems are also now being developed (multipleprojection images using rotating point-like objects

such as metal ball bearings)20 and will include definingthe parameters for quantitating the mineralization ofthe craniofacial bones.28

Studies comparing 3D linear measures commonlyused in orthodontics attained from CBCT systems ofskulls have not been done as they have with conven-tional CT. The goal of this present study is to evaluatethe accuracy of the CBCT-generated images from twodifferent systems and to compare them, using linearmeasures of different size and orientation, with the goldstandard of physical measurements. We have reducedpart of the problem in this type of study by defining thelandmarks with known metal markers. Once the

CBCT imaging errorSA Stratemann et al

Dentomaxillofacial Radiolo


3/14

accuracy of the various CBCT systems is establishedusing 3D measures, the application of these systems toevaluating craniofacial growth, changes in the airway,adaptation in the condyle and temporomandibular joint, position of impacted teeth and asymmetricalmandibles becomes possible and will lead to their

diagnostic value.4,2934

Materials and methods

In the present study, CBCT data from two systems wereused. Initially, imaging data were acquired from theNewTomH QR DVT 9000 (Aperio Inc, Sarasota, FL) atDiagnostic Digital Imaging, Sacramento, CA. This waslater followed by data collected on the CBMercuRayTM (Hitachi Medico Technology, Tokyo,Japan), located in the Division of Orthodontics at theUniversity of California at San Francisco. These CBCTsystems are described in the following sections.

Cone beam CT systems

NewTom: The NewTomH QR DVT 9000 CBCT is avolume imaging machine designed with the intent ofcapturing the anatomy of the maxillofacial region. Inthis device, the patient lies supine, as in conventionalCT, with the head centred within the mechanism. Asingle 75 s scan acquires 360 images (one per degree) ina 13 cm high cylindrical field of view with 0.29 mm 3

voxels (in high-resolution mode) and 8 bits per voxel(285 256 shades of grey).14,18 In the typical patient

scans, a 1 mm slice thickness was used, yielding 0.29mm261 mm voxels.

The CB MercuRayTM is another cone beam system alsodeveloped specifically for the purpose of craniofacialimaging.21,35 With the patient sitting upright, a rotatingsource/detector gantry captures a volumetric image ofthe patients head, a process similar in nature topanoramic radiography. The version of the system usedin this study has a scalable 12 inch charge-coupled devicedetector that can be set in several field-of-view modes.

Imaging error

Physical measurements: plastic skull: A preliminarystudy of a plastic model was used to establish themethod and measurements that would be used with thehuman skull, and the plastic skull was evaluated onlywith the NewTom CBCT. A plastic model of a humanskull (Classic Skull Version A20, 3B Scientific GmbH,Hamburg, Germany) was marked with size 7K leadshot (diameter 5 2.4 mm) at 40 cranial and 33mandibular landmarks (Figure 1a). By preparing eachsite with a round dental burr, landmarks could beplaced in the skull such that they were submerged to thediameter of each spherical shot pellet (Tables 1 and 2).Pellets were affixed to the skull with cyanoacrylatecement.

A set of 136 linear measurements among a subsetof 53 of the points, comprising 49 cranial and87 mandibular point pairs, was made on the modelskull using a digital calliper with a nominal resolutionof 0.01 mm (Mitutoyo CD-6C No. 500-171,Mitutoyo America Corp., Aurora, IL). Distancesbetween the members of the landmark pairs wererecorded as inside or outside measurements, dependingon the nature of the physical approach to the points assituated on the surface of the skull. Using physicalcallipers with some landmarks was more difficultbecause of the mechanics of applying the two calliperend points. Each distance was recorded by a singleinvestigator (SAS) on five separate occasions, separatedby 1 week. The mean diameter and standard deviationof the lead shot pellets were established at2.400.02 mm by measuring each of 30 sample shotpellets in 5 repetitions. By addition or subtraction ofone pellet diameter to inside or outside measurements,

respectively, the magnitude of the mean centre-to-centre distance and its SD for each of the 136-pointpairs was determined to the nearest 0.01 mm. Thecalliper measurement for each point pair was set as agold standard.

Segmentation measurements: plastic skull: The plasticskull was submerged in a thin, custom built, acrylictank (Tap Plastics, Sacramento, CA, Figure 2) filled withwater for X-ray beam attenuation and imaged in aNewTomH QR DVT 9000 CBCT (Aperio Inc.) with a1 mm slice thickness. Voxel dimensions were0.360.361.0 mm. The CT voxels comprising the point

data were identified and rendered into a surface mesh bythreshold segmentation techniques using Amira software(Amira 3.1, Mercury Computer Systems GmbH, Berlin,Germany). Linear centre-to-centre measurements of the136-point pairs were recorded to the nearest 0.01 mmusing the Measuring Tool within the software by asingle investigator (SAS) at 5 separate intervals, 1 weekapart. The mean and standard deviations for eachmeasurement were determined.

Physical measurements: human skull: A human skull,obtained from the private collection of one of the authors(DCH), was then examined after establishing theprotocol with the plastic skull and the human skull was

scanned with the two different CBCT systems. Thehuman skull was marked with chromium steel balls(1/16 inch chromium steel ball, part # N57803448;Penn Tool Co., Maplewood, NJ) at 32 cranial and33 mandibular landmarks (Figure 1b, Tables 1 and 2).Nominal specifications for the balls were as follows:grade, 25; diameter deviation, 0.00005 inches; spheri-city, 0.000025 inches; hardness, Rockwell C 63. Each sitewas prepared with a round dental burr to allow themarkers to be placed in the skull such that they weresubmerged to the diameter of each spherical ball.Markers were affixed to the skull with cyanoacrylatecement. To eliminate mobility effects for the dry skull,


ntomaxillofacial Radiology


4/14

cyanoacrylate cement was placed in the periodontalligament (PDL) space surrounding the landmarked teethto affix the reference teeth to the adjacent alveolar bone.

A set of 125 linear measurements, comprising43 cranial and 82 mandibular point pairs, was madeon the skull using a digital calliper with a nominalresolution of0.01 mm. Distances between the membersof the landmark pairs were recorded as inside or outsidemeasurements, depending on the nature of the physicalapproach to the points as situated on the surface of the

skull. Each distance was recorded by a single investigator(SAS) on five separate occasions, separated by 1 week. Themean diameter andSD of the chromium steel bearings wasestablished at 1.580.00 mm by measuring each of 30samplebearingsin 5 repetitions.By additionor subtractionof 1 spherical diameter to inside or outside measurements,respectively, the magnitude of the mean centre-to-centredistance and its SD for each of the 125 point pairs wasdetermined to the nearest 0.01 mm. The calliper measure-ment for each point pair was set as a gold standard.

a b

Figure 1 (a) Plastic human skull model. (b) Adult human skull

Table 1 Landmarks, abbreviations and operational definitions for maxillary landmarks on the plastic human skull and real human skull

Landmark Description Bilateral points

U1 Centre of the labioincisal edge, maxillary central incisor (only in human plastic skull) YesU3 Cusp tip, maxillary canine YesU6 Mesiobuccal cusp tip, maxillary first molar (only in human plastic skull) YesTub Most posterior inferior point centred mediolaterally on the maxillary tuberosity YesGPF Greater palatine foramen YesPNS Posterior nasal spine NoOr Orbitale, centred anteroposteriorly at the most inferior point on the lateral border of the

orbital rimYes

IOF Infraorbital foramen YesPR Most lateroinferior point on the outer border of the piriform rim YesANS Anterior nasal spine NoPr Prosthion, highest point on the alveolar crest at the midline between the maxillary central

incisorsNo

A A point, greatest concavity on the maxillary midline NoNPF Nasopalatine foramen, centred at the midline (only human plastic skull) No




5/14

Segmentation measurements: human skull: The humanskull was submerged in a custom-made acrylic tank

(Figure 2), filled with water for beam attenuation andimaged in the NewTomH QR DVT 9000 CBCT systemwith a 0.3 mm slice thickness. The resulting cubicvoxels were 0.3 mm3 in size. CT data including thelandmark point data were threshold-segmented usingCB Works software (CB Works 1.0; CyberMed Inc.,Seoul, Korea). The resulting segmentations wereexported from CB Works as VRML (.wrl) files toAmira software (Amira 3.1, Mercury ComputerSystems GmbH, Berlin, Germany) and converted intoan Amira hexadecimal binary surface (.surf) format.Linear centre-to-centre measurements of 120 of the125 point pairs were recorded to the nearest 0.01 mm

using the Measuring Tool within the Amira softwareby a single investigator (SAS) at five separate intervals,1 week apart. The point pairs Or-Tub (left and right),Or-IOF (left and right) and Or-Or could not be madebecause the vertical size of the skull was larger than the6 inch high scan volume of the NewTomH QR DVT 9000

CBCT, such that landmark point Or was not included inthe CT data. The mean and standard deviations for eachmeasurement were determined.

Next, the human skull, placed in the same tankwithout water for the purposes of beam attenuation,was imaged in a CB MercuRayTM CBCT (HitachiMedico Technology) at 120 kVp, 15 mA in the F modewith a 0.376 mm slice thickness. Voxels in this systemwere 0.376 mm3. CT data were then evaluated as withthe NewTom data.

Statistical methods

Human skull trial: For the human skull trial in theNewTom 9000 and CB MercuRay, the values for eachof the five repetitions of the 125 calliper measurements(corrected appropriately for the 1.58 mm sphericaldiameter for inside and outside measurements) andcentre-to-centre segmentation measurements, wereentered into a Microsoft Excel spreadsheet (Table 3;Microsoft Excel 2002, Microsoft Co., Redmond, WA).For each machine, analysis of the results was identical.Mean values and their standard deviations (SDs) werecalculated for each linear distance. Absolute error, themean difference between the calliper measurement andthe segmentation measurement (C S) and its SD, werealso determined for each landmark pair. Relative error(RE) for each measurement was calculated as thepercent difference of the gold standard calliper mea-surement (C) to the segmentation measurement (RE 5(C S)/C6 100%). Values for the absolute error, theSDs of the absolute error and the relative error weregraded from 1 (least error) to 125 (greatest error) usingthe Excel RANK function. Plots of the rankedmeasurement differences (absolute error) and theirSDs, as well as the relative error, were made in Excelto depict the range of each value. Histograms of theabsolute and relative measurement error and the SDs ofthe absolute error were made using the ExcelHistogram function.

One-way ANOVA (a5 0.05) in the software pro-gram StatView (StatView for Windows 5.0.1; SASInstitute Inc., Cary, NC) was used to detect thepresence of statistically different groups in the absoluteand relative error estimates when considered bymeasurement magnitude, principal planar dimension,or jaw. Post hoc Fishers PSLD (a5 0.05) tests wereused when ANOVA found significant differences toidentify which of the compared groups were different.

Comparison of error in the CBCT systemsImaging error in the NewTom and CB MercuRaysystems was compared with the overall estimate of error

Table 2 Landmarks, abbreviations and operational definitions formandibular landmarks on the plastic human skull and real humanskull

Landmark DescriptionBilateralpoints

L1 Centre of the labioincisal edge, mandib-

ular central incisor (only in human plasticskull)

Yes

L3 Cusp tip, mandibular canine YesL6 Mesiobuccal cusp tip, mandibular first

molarYes

R1 Greatest concavity centred mediolaterallyon the anterior border of the ascendingramus

Yes

Cor Tip of coronoid process YesLatConPol Lateral condylar pole, most lateral point

centred anteroposteriorly and superioin-feriorly on the mandibular condyle

Yes

MedConPol Medial condylar pole, most medial pointcentred anteroposteriorly and superioin-feriorly on the mandibular condyle

Yes

Co Condylion, most superior point centred

mediolaterally and anteroposteriorly onthe mandibular condyle

Yes

R2 Greatest concavity centred mediolaterallyon the posterior border of the ascendingramus

Yes

R3 Most inferior point centred mediolater-ally on the sigmoid notch

Yes

MnFor Mandibular foramen YesR4 Greatest concavity centred mediolaterally

on the antegonial notchYes

Go Gonion, greatest convexity of the angularprocess centred mediolaterally

No

Mf Mental foramen YesMe Menton, most inferior point on the

mental symphysis centred anteroposter-iorly

No

Pg Pogonion, greatest midline convexity on

the anterior border of the mental sym-physis

No

Gn Gnathion, midline point on the ante-roinferior border of the mental symphy-sis, centred between menton andpogonion

No

Id Infradentale, highest point on the alveo-lar crest at the midline between themandibular central incisors

No

B B point, greatest concavity on the man-dibular midline

No




6/14

in the 120 of the 125 landmark pairs that were viewablein the scan volumes of both machines (the point pairsOr-Tub (left and right), Or-IOF (left and right) and Or-Or were not available from the NewTom scan). Theoverall estimate of error for each measurement wastaken to be the arithmetic mean of the calliper-derived

and segmentation means from the NewTom and CBMercuRay images with five repetitions. Absolute andrelative error to the overall mean were calculated forthe calliper, NewTom and CB MercuRay. StatViewwas used to perform one-way ANOVA (a5 0.05) todetect the presence of statistically significant groups inthe absolute and relative error estimates. Post hocBonferroni/Dunn tests (a5 0.05) were performed whenappropriate to identify which of the examined pairswere actually different. The software was also used toconstruct scattergrams of the error.

Results

NewTomThe mean absolute error to the gold standard callipermeasurements (2 SD) for the human skull was0.070.41 mm, with a median of 0.07 mm and a range

of error, as indicated by the magnitude of two SDs, forthe NewTom is slightly greater than three voxel widths.The distribution of the absolute error indicates that themeasurements derived from the NewTom scan were, onaverage, slightly smaller than the calliper-derivedmeasurements. SDs of the individual measurements

remain clustered relatively tightly around 0.020.03 mm.

The mean relative error (2 SD) for the human skullwas 0.191.56%, with a median of 0.24% and a rangeof21.60% to 1.79%. As in the corresponding absoluteerror, the relative error measurements from the scan

volume were distributed with a positive shift. Plottingthe absolute error against the overall measurementmagnitude indicated that larger error appeared to occurwith smaller measurements (Figure 4a). This phenom-enon speaks of a comparatively constant absolute error.Considering the absolute error by measurement magni-tude indicated the possibility of small differences in theerror with different lengths across the scan volume ofthe skull. ANOVA on absolute error by measurementsize did not demonstrate the presence of statisticallydifferent groups.

Absolute error was then considered in the context ofthe principal spatial plane, jaw and measurement size.In a plot of error rank vs absolute error, it is notapparent that any spatial plane or either jaw has apredominately different error, although the sagittalmeasurements in both jaws appear to span a greaterrange of ranks than those in the vertical or transverseplanes. Subdividing the absolute error by plane alone(Figure 5a) gave a sense that few differences werepresent in the error among the planes. There were nostatistically significant differences in the absolute errorby ANOVA, and Fishers PSLD showed no statisticallysignificant differences in the absolute error between thesagittal and transverse measurements, with the meanerror (2 SD) standing at 0.080.45 mm for thesagittal plane, 0.080.27 mm for the vertical plane and

Figure 2 Custom acrylic tank for imaging the human skull with the NewTom CBCT system



of20.430.48 mm (Figure 3b). Therefore, the envelope


7/14

0.060.31 mm for the transverse plane. This informa-tion confirms that little dimensional distortion ispresent in the scan volume.

An ANOVA analysis of the absolute error by jaw(Figure 5b) indicated the presence of statisticallydifferent groups, confirmed by Fishers PSLD, withthe mean error (2 SD) standing at 0.000.42 mm forthe maxilla and 0.110.40 mm for the mandible.Because the evaluation involved an almost threefolddifference in group sizes, with 38 maxillary measure-ments being compared with 92 mandibular measure-ments, and because almost all of the largermeasurements involved the mandible, the conclusionsmust be considered with caution. A significant differ-ence was observed between the error in the maxilla andmandible. Considering the absolute error by measure-ment magnitude (Figure 5c) indicated the possibility ofsmall differences in the error with different lengthsacross the scan volume of the skull.

CB MercuRayThe mean absolute error to the gold standard callipermeasurements (2 SD) for the human skull was0.000.22 mm, with a median of 0.01 mm and a rangeof 20.39 mm to 0.24 mm (Figure 3B). Therefore, theenvelope of error, as indicated by the magnitude of 2 SD,for the CB MercuRay, is slightly less than two voxelwidths. The distribution of the absolute error indicatesthat the measurements derived from the CB MercuRayscan were, on average, very slightly smaller than thecalliper-derived measurements. SDs of the individualmeasurements remain clustered relatively tightly around

0.010.04 mm. The mean relative error (2 SD) for thehuman skull was 0.010.06%, with a median of 0.01%and a range of20.06% to 0.11%. As in the correspondingabsolute error, the relative error measurements from thescan volume were distributed with a positive shift.

Plotting the absolute error against the overallmeasurement magnitude indicates fairly uniform erroracross all measurement magnitudes (Figure 4b). Asimilar plot of relative error against the overallmeasurement magnitude concurs. Relative error tapersdown somewhat from negative and positive extremes of20.40% to 0.30% at smaller measurements towards20.10% to 0.10% as measurement size increases. Thisphenomenon agrees with a comparatively constantabsolute error.

Absolute error was then considered in context of theprincipal spatial plane, jaw and measurement size. In aplot of error rank vs absolute error, it is not apparentthat any spatial plane or either jaw has a predominatelydifferent error, although the sagittal measurements inboth jaws appear to span a greater range of ranks thanthose in the vertical or transverse planes. Breakingdown the absolute error by plane alone (Figure 5d)gave a sense that some differences were present in theerror among the planes. ANOVA on absolute error byplane indicated the presence of statistically differentgroups. Fishers PSLD showed a statistically significant

difference in the absolute error between the sagittal andtransverse measurements, with the mean error (2 SD)standing at 0.010.06 mm for the sagittal plane and0.020.08 mm for the transverse plane. Given thatthere were many more sagittal measurements thantransverse measurements (96 compared with 19,

respectively) and that the means 2 SD do overlap,the actual significance of the statistical result isdiminished. Fishers PSLD also showed a statisticallysignificant difference in the absolute error between thevertical and transverse measurements, with a meanerror (2 SD) of 20.010.03 mm for the verticalplane. As there were fewer vertical measurements thantransverse measurements (10 compared with 19, respec-tively) and as the means 2 SD do overlap, thesignificance of the statistical result is lessened.Significant differences were observed in the error inthe sagittal and transverse planes as well as in thevertical and transverse planes.

Examining the absolute error by jaw for the humanskull trial was not so illuminating (Figure 5e). ANOVAon absolute error by jaw indicated no statisticallydifferent groups, confirmed by Fishers PSLD, with themean error (2 SD) standing at 0.000.04 mm for themaxilla and 0.010.07 mm for the mandible.

Considering the absolute error by measurementmagnitude (Figure 5f) indicated the possibility of smalldifferences in the error with different lengths across thescan volume of the skull. ANOVA on absolute error bymeasurement size demonstrated the presence of statisti-cally different groups. Fishers PSLD showed statisticallysignificant differences in the absolute error betweenseveral sets of measurements. Several measurement sizeranges differed from 7080 mm, 8090 mm, 90100 mm,100110 mm and 120130 mm measurements, but thesedifferences cannot be seriously considered because onlyone or two measurements fell into each of these ranges.No interaction between jaw and spatial plane were foundin multifactorial ANOVA. Differences relating to trans-verse error described previously were apparent.

Comparison of error between the systemsError was small in both CBCT systems. Analysis ofabsolute error to the gold standard indicated that allmeasurements averaged 0.070.41 mm (mean2 SD)for the NewTom (Table 4). Relative error was0.191.56% (mean 2 SD). For the CB MercuRay,the mean absolute error to the gold standard was0.000.22 mm (mean 2 SD) for the 120 measure-ments. The estimate for relative error came to0.010.06% (mean 2 SD). An ANOVA statisticalanalysis indicated the presence of statistically significantdifferences among the calliper-, NewTom- and CBMercuRay-derived measurements compared with theoverall estimate of the measurement for both absoluteand relative error (P, 0.0001). The Bonferroni/Dunntests also indicated that NewTom was statisticallydifferent from both the calliper and CB MercuRay forabsolute and relative errors (P# 0.0001).




8/14

Table 3 Measurements from physical caliper and using volumetric renderings by two different CBCT systems

Caliper Measures (mm) NewTom 9000 (mm) CB MercuRay (mm)

Measurement Mean SD Mean SD Mean SD

RightU3U7 28.42 0.01 28.53 0.07 28.38 0.04

U7Tub 21.31 0.01 21.19 0.08 21.31 0.08TubGPF 15.16 0.02 15.10 0.08 15.14 0.02GPFPNS (IS) 18.25 0.02 18.54 0.12 18.28 0.04TubPNS 23.97 0.02 24.32 0.17 23.96 0.03OrTub 53.62 0.01 N/A N/A 53.64 0.04IOFTub 47.85 0.01 47.43 0.03 47.82 0.07OrIOF 20.69 0.02 N/A N/A 20.71 0.07IOFU3 46.02 0.01 45.80 0.05 46.05 0.02IOFU7 46.75 0.02 46.54 0.04 46.74 0.04IOFPR 22.70 0.02 22.92 0.08 22.74 0.03PRANS 14.73 0.02 14.77 0.03 14.72 0.03PRPr 26.17 0.01 25.75 0.04 26.15 0.02PRU3 31.13 0.02 30.87 0.03 31.15 0.04IOFA 38.07 0.01 38.32 0.08 38.07 0.04IOFPr 47.77 0.02 47.65 0.07 47.77 0.02LeftU3U6 18.70 0.02 18.95 0.12 18.71 0.03

U6Tub 29.25 0.02 29.31 0.13 29.21 0.03TubGPF 15.08 0.02 15.31 0.07 15.09 0.02GPFPNS (IS) 18.28 0.01 18.29 0.04 18.24 0.05TubPNS 25.23 0.01 25.34 0.05 25.26 0.05OrTub 53.64 0.01 N/A N/A 53.61 0.03IOFTub 47.66 0.02 47.82 0.09 47.63 0.03OrIOF 16.37 0.02 N/A N/A 16.38 0.08IOFU3 47.92 0.02 47.94 0.09 47.93 0.04IOFU6 43.47 0.02 43.90 0.02 43.50 0.01IOFPR (IS) 26.00 0.01 25.84 0.17 26.01 0.02PRANS 12.58 0.01 12.63 0.09 12.55 0.04PRPr 23.76 0.01 23.50 0.05 23.75 0.02PrU3 20.14 0.02 20.34 0.09 20.15 0.03IOFA (IS) 39.17 0.01 39.18 0.05 39.10 0.04IOFPr (IS) 48.92 0.01 48.53 0.03 48.92 0.03RightLeftU3 33.63 0.01 33.82 0.03 33.62 0.04U7U6 56.01 0.02 56.02 0.08 56.00 0.03Tub 45.34 0.01 45.41 0.12 45.34 0.07PR 19.82 0.02 19.75 0.05 19.83 0.03Or 87.70 0.02 N/A N/A 87.69 0.06IOF 53.47 0.02 53.56 0.06 53.46 0.07MidlineANSA 5.89 0.02 5.98 0.03 5.91 0.02APr 12.90 0.01 12.83 0.07 12.91 0.04ANSPr 18.42 0.01 18.19 0.04 18.43 0.03NPFPNS (IS) 45.34 0.02 45.23 0.07 45.35 0.03ANSPNS 53.26 0.02 53.30 0.02 53.23 0.03RightL3L6 20.91 0.01 21.16 0.03 20.88 0.05L6R1 31.76 0.02 31.99 0.04 331.72 0.01R1Cor 32.94 0.01 32.73 0.04 32.90 0.04CorR3 22.89 0.01 23.05 0.03 22.86 0.05

R3LatConPol 24.89 0.01 24.85 0.03 24.91 0.03R3MedConPol (IS) 25.40 0.01 25.22 0.02 25.40 0.04R3Co 25.82 0.02 25.71 0.04 25.87 0.05CoLatConPol 8.74 0.02 8.75 0.03 8.76 0.03CoMedConPol 16.16 0.01 16.11 0.02 16.17 0.02LatConPolR2 41.80 0.01 41.88 0.04 41.77 0.04MedConPolR2 40.41 0.02 40.46 0.03 40.47 0.03R3MnFor 18.41 0.01 18.64 0.02 18.41 0.04MnForR1 18.89 0.01 18.77 0.04 18.85 0.04MnForR2 19.28 0.01 19.58 0.04 19.27 0.04MnForR4 33.80 0.01 33.55 0.02 33.79 0.06MnForGo 28.18 0.01 28.18 0.02 28.20 0.04MnForCor 37.27 0.01 36.99 0.04 37.25 0.05MnForCo 41.93 0.01 41.50 0.03 41.86 0.02CoR2 45.39 0.01 45.07 0.02 45.35 0.04R2Go 19.58 0.01 19.30 0.04 19.61 0.05




9/14

Caliper Measures (mm) NewTom 9000 (mm) CB MercuRay (mm)

Measurement Mean SD Mean SD Mean SD

GoR4 21.23 0.02 21.06 0.04 21.21 0.03R4Me 64.39 0.02 64.42 0.04 64.37 0.03GoMe 85.08 0.01 84.82 0.02 85.08 0.04

GoGn 88.61 0.02 88.35 0.03 88.62 0.04IdL3 17.01 0.01 17.18 0.03 17.01 0.02MfL3 29.59 0.01 29.20 0.01 29.58 0.04MfL6 21.49 0.02 21.75 0.02 21.50 0.02MfId 30.38 0.02 30.29 0.03 30.38 0.02MfB 27.98 0.01 27.96 0.04 27.97 0.04MfPg 32.41 0.01 32.24 0.02 32.45 0.06MfGn 32.70 0.01 32.68 0.03 32.75 0.03MfMe 31.62 0.01 31.61 0.03 31.66 0.07LeftL3L6 18.96 0.01 18.99 0.01 18.95 0.02L6R1 32.31 0.02 32.70 0.04 32.29 0.03R1Cor 31.65 0.01 31.94 0.05 31.65 0.04CorR3 21.19 0.01 21.07 0.03 21.17 0.02R3LatConPol 24.38 0.01 24.23 0.03 24.43 0.03R3MedConPol (IS) 26.48 0.02 26.26 0.03 26.54 0.04R3Co 27.77 0.01 27.31 0.03 27.75 0.02CoLatConPol 8.82 0.01 8.66 0.05 8.83 0.02CoMedConPol 16.66 0.01 16.47 0.02 16.64 0.03LatConPolR2 40.83 0.02 40.60 0.03 40.76 0.03MedConPolR2 40.47 0.02 40.57 0.03 40.43 0.02R3MnFor 16.77 0.02 16.92 0.04 16.78 0.02MnForR1 19.94 0.02 19.88 0.01 19.96 0.02MnForR2 17.63 0.01 17.64 0.02 17.66 0.01MnForR4 34.43 0.01 34.45 0.02 34.45 0.05MnForGo 27.06 0.02 26.93 0.04 27.04 0.02MnForCor 35.25 0.02 35.11 0.03 35.28 0.02MnForCo 40.53 0.01 40.10 0.01 40.58 0.01CoR2 44.66 0.02 44.38 0.03 44.62 0.02R2Go 18.12 0.02 17.88 0.03 18.13 0.01GoR4 21.08 0.02 20.88 0.05 21.04 0.03R4Me 64.97 0.02 64.68 0.04 64.94 0.06GoMe 84.75 0.02 84.28 0.04 84.77 0.04

GoGn 87.87 0.01 87.47 0.01 87.81 0.05IdL3 13.67 0.02 13.45 0.03 13.65 0.03MfL3 31.74 0.01 31.69 0.05 31.72 0.01MfL6 24.03 0.01 24.13 0.03 24.04 0.02MfId 31.77 0.02 31.44 0.03 31.75 0.03MfB 30.91 0.01 30.50 0.04 30.91 0.04MfPg 33.67 0.01 33.40 0.04 33.68 0.02MfGn 33.47 0.02 33.11 0.04 33.43 0.02MfMe 32.41 0.02 32.03 0.04 32.35 0.02RightLeftL3 22.50 0.02 22.23 0.04 22.50 0.03L6 48.41 0.02 48.39 0.01 48.43 0.02R1 79.44 0.01 79.27 0.01 79.37 0.03Cor 94.40 0.02 94.17 0.04 94.28 0.01R3 93.84 0.02 93.91 0.01 93.80 0.08LatConPol 123.14 0.02 123.06 0.02 123.07 0.03MedConPol (IS) 81.45 0.01 81.58 0.05 81.45 0.03

Co 110.22 0.02 110.21 0.03 110.18 0.03MnFor (IS) 87.73 0.02 87.76 0.02 87.70 0.06R2 103.82 0.02 103.63 0.05 103.88 0.02Go 101.73 0.01 101.49 0.04 101.73 0.02R4 85.24 0.02 85.21 0.03 85.18 0.05Mf 50.66 0.01 50.29 0.04 50.64 0.05MidlineMeGn 5.24 0.02 5.26 0.04 5.24 0.02GnPg 3.98 0.01 3.94 0.03 3.98 0.01PgId 22.90 0.02 22.53 0.03 22.91 0.04BId 7.70 0.01 7.67 0.03 7.73 0.02PgB 15.33 0.01 15.25 0.02 15.34 0.04

For explanation of abbreviations, see Table 2

Table 3 Continued




10/14

Discussion

CBCT provides a valuable tool for evaluating thecraniofacial region. Effective radiation dose from thescan of a maxillomandibular volume has been mea-sured in phantom studies to be 50.27 mSv,14 which issignificantly less than that of other CT modalities andin the range of conventional dental radiological meth-ods.14,36,37 Segmentation approaches continuallyimprove as the CBCT software develops,38 allowingstructural analysis of the parts of the scan volume andfacilitating analyses more complex than simple linear andangular measurements.

In this study, the error was small compared with thegold standard for both the NewTom and CBMercuRay. Absolute error to the gold standard wasslightly positive, indicating minor compression relativeto the calliper measurement. The error was slightlysmaller in the CB MercuRay than in the NewTom,probably related to a broader greyscale range fordescribing beam attenuation in 12-bit vs 8-bit data. Theeffect of voxel size did not seem to be an important

factor, as the NewTom had 0.3 mm3 voxels vs the0.376 mm3 voxels in the F mode of the CB MercuRay.Although statistical differences were evident among thecalliper, NewTom and CB MercuRay measurements,the observed differences were below the level of clinicalsignificance for orthodontic evaluations and were at thelimit of the nominal resolution of the gold standard.However, as diagnostic approaches become morerefined for evaluating the local structure of trabecular

bone in fields such as periodontics and oral surgery, theresolution will become a more critical issue.

In the preliminary study using the plastic skull withthe NewTom to establish the protocol, the sagittalmeasurements had greater absolute and relative errorsthan the transverse measurements. Aboudara39 founddistortion at the peripheral portions of the NewTomscan volume using a plastic airway phantom and it ispossible that a similar effect has been observed here,with some the sagittal plastic skull landmarks sitting atthe outer portions of the scan. There is also the issuethat the plastic skull was scanned with 1 mm slicethickness leading to an anisotropic voxel compared

a

b

Figure 3 Ranked absolute error and standard deviations of absolute error of all measurements with the human skull scanned with (a) theNewTom CBCT system and (b) the CB MercuRay system for the same measurements




11/14

with the isotropic voxels used with the human skull. Noeffects were seen by plane for the human skull whenscanned with the NewTom CBCT system.

In the CB MercuRay, absolute error in the transverseplane was greater than both the sagittal and the verticalplanes, possibly because there were many transversemandibular measurements that approached the edge ofthe scan volume. Relative error in the CB MercuRaywas greater in the vertical plane than in either thesagittal or the transverse planes, probably because the

vertical measurements tended to be smaller in magni-tude than those of the other two spatial planes. Whencomparing the measurements in the maxilla with themandible, a significant difference was seen only withthe NewTom CBCT, in which the maxillary error ofboth types was significantly less than mandibular error.Again, many of the maxillary landmarks would havetended to be confined to the centre of the scan volumecompared with those in the mandible.

The design of this experiment inherently maximizedthe possibility of finding the optimal measurement errorbecause it minimized error associated with landmarkidentification by forcing the user to pick the centre of a

tiny sphere. Moreover, the radiodense marker materialwas easily separated during the segmentation processby attenuation coefficient from the surrounding back-ground of air, bone, enamel and dentin. Minimal levelsof interpretation and decision-making accompany sucha specific task. It is expected that landmark identifica-tion and placement may introduce greater error,probably within the realm of 0.51.0mm, as estimatedby Yoo et al.40

In conclusion, the data indicate that, for a single user,

distortion across the scan volumes for both CBCTsystems is small. This is supported by other studiesapplying CBCT to analysing the depth and diameter ofsimulated bone defects in both an acrylic block and ahuman mandible.10 Therefore, the hypothesis thatsegmented volumes of craniofacial structures from conebeam CT data are accurate models of the objects theyrepresent with respect to linear measures can beaccepted. This finding, which supports studies madein traditional CT systems, helps to validate the tool tobe used in establishing diagnostic evaluations of thecraniofacial region including the dentition, temporo-mandibular joint and airway,41 and provides the first

a

b

Figure 4 Absolute error vs measurement size for the human skull scanned with (a) the NewTom CBCT system and (b) the CB MercuRay system




12/14

steps to superimposition of 3D-constructed craniofacialvolumetric data.34,42 Generation of 3D coloured dis-placement maps derived from segmented data willprovide a much more vibrant and accurate indicationof how the cranioskeleton grows and develops than hasbeen possible for the past 70 years using two-dimensional

full head radiographs and cephalometric analysis.However, at present, it is evident that CBCT still haslimitations, particularly defining the details of theenameldentin interfaces, pulp chamber and the trabe-cular bone surrounding the teeth, but this will improve asdoes the imaging technology.4345

c f

b e

a d

Figure 5 Absolute error vs gold standard measurement size by (a,d) plane, (b,e) jaw and (c,f) measurement size for the human skull in (ac) theNewTom CBCT system and (df) the CB MercuRay system




13/14

References

1. Mozzo P, Procacci C, Tacconi A, Martini PT, Andreis IA. A newvolumetric CT machine for dental imaging based on cone-beamtechnique: preliminary results. Eur Radiol 1998; 8: 15581564.

2. Fullmer JM, Scarfe WC, Kushner GM, Alpert B, Farman AG.Cone beam computed tomographic findings in refractory chronicsuppurative osteomyelitis of the mandible. Br J Oral MaxillofacSurg 2007; 45: 364371.

3. Guerrero ME, Jacobs R, Loubele M, Schutyser F, Suetens P, vanSteenberghe D. State-of-the-art on cone beam CT imaging forpreoperative planning of implant placement. Clin Oral Investig2006; 10: 17.

4. Honda K, Larheim T, Maruhashi K, Matsumoto K, Iwai K.Osseous abnormalities of the mandibular condyle: diagnosticreliability of cone beam computed tomography compared withhelical computed tomography based on an autopsy material.Dentomaxillofac Radiol 2006; 35: 152157.

5. Tantanapornkul W, Okouchi K, Fujiwara Y, Yamashiro M,Maruoka Y, Obhayashi N, et al. A comparative study of cone-beam computed tomography and conventional panoramic radio-graphy in assessing the topographic relationship between themandibular canal and impacted third molars. Oral Surg Oral MedOral Pathol Oral Radiol Endod 2007; 103: 253259.

6. Kau CH, Richmond S, Palomo JM, Hans MG. Three-dimen-

sional cone beam computerized tomography in orthodontics. JOrthod 2005; 32: 282293.

7. Walker L, Enciso R, Mah J. Three-dimensional localization ofmaxillary canines with cone-beam computed tomography. Am JOrthod Dentofacial Orthop 2005; 128: 418423.

8. Farman AG, Scarfe WC. Development of imaging selectioncriteria and procedures should precede cephalometric assessmentwith cone-beam computed tomography. Am J Orthod DentofacialOrthop 2006; 130: 257265.

9. Swennen GR, Schutyser F. Three-dimensional cephalometry:spiral multi-slice vs. cone-beam computed tomography. Am JOrthod Dentofacial Orthop 2006; 130: 410416.

10. Pinsky HM, Dyda S, Pinsky RW, Misch KA, Sarment DP.Accuracy of three-dimensional measurements using cone-beamCT. Dentomaxillofac Radiol2006; 35: 410416.

11. Kitaura H, Yonetsu K, Kitamori H, Kobayashi K, Nakamura T.

Standardization of 3-D CT measurements for length and anglesby matrix transformation in the 3-D coordinate system. CleftPalate Craniofac J 2000; 37: 349356.

12. Cavalcanti MG, Rocha SS, Vannier MW. Craniofacial measure-ments based on 3D-CT volume rendering: implications forclinical applications. Dentomaxillofac Radiol2004; 33: 170176.

13. Krarup S, Darvann TA, Larsen P, Marsh JL, Kreiborg S. Three-dimensional analysis of mandibular growth and tooth eruption. JAnat 2005; 207: 669682.

14. Mah JK, Danforth RA, Bumann A, Hatcher D. Radiationabsorbed in maxillofacial imaging with a new dental computedtomography device. Oral Surg Oral Med Oral Pathol Oral RadiolEndod 2003; 96: 508513.

15. Tsikiakis K, Donta C, Gavala S, Karayianni K, KamenopoulouV, Hourdakis CJ. Dose reduction in maxillofacial imaging usinglow dose Cone Beam CT. Eur J Radiol 2005; 56: 413417.

16. Ludlow JB, Davies-Ludlow LE, Brooks SL, Howerton WB.Dosimetry of 3 CBCT devices for oral and maxillofacialradiology: CB MercuRay, NewTom 3G and i-CAT.Dentomaxillofac Radiol2006; 35: 219226.

17. Carlsson C. Imaging modalities in x-ray computerized tomogra-phy and in selected volume tomography. Phys Med Biol1999; 44:R2356.

18. Hatcher DC, Aboudara CL. Diagnosis goes digital. Am J OrthodDentofac Orthop 2004; 125: 512515.

19. Sukovic P. Cone beam computed tomography in craniofacialimaging. Orthod Craniofacial Res 2003; 6 (Suppl. 1):3136.

20. Yang K, Kwan AL, Miller DF, Boone JM. A geometriccalibration method for cone beam CT systems. Med Phys 2006;33: 16951706.

21. Yamamoto K, Ueno K, Seo K, Shinohara D. Development ofdento-maxillofacial cone-beam x-ray computed tomographysystem. Orthod Craniofacial Res 2003; 6 (Suppl. 1):160162.

22. Araki K, Maki K, Seki K, Sakamaki K, Harata Y, Sakaino R, etal. Characteristics of a newly developed dentomaxillofacial X-raycone beam CT scanner (CB MercuRayTM): system configurationand physical properties. Dentomaxillofac Radiol2004; 33: 5159.

23. Baumrind S, Frantz RC. The reliability of head film measurements.1. Landmark identification. Am J Orthod1971; 60: 111127.

24. Richtsmeier JT, Paik CH, Elfert PC, Cole TM III, Dahlman HR.Precision, repeatability and validation of the localization ofcranial landmarks using computed tomography scans. CleftPalate Craniofac J 1995; 32: 217227.

25. Cavalcanti MG, Haller JW, Vannier MW. Three-dimensionalcomputed tomography landmark measurement in craniofacialsurgical planning: experimental validation in vitro. J OralMaxillofac Surg 1999; 57: 690694.

26. Williams FL, Richtsmeier J. Comparison of mandibular land-marks from computed tomography and 3D digitizer data. ClinAnat 2003; 16: 494500.

27. Lo LJ, Lin WY, Wong HF, Lu KT, Chen YR. Quantitativemeasurement of three-dimensional computed tomography: anexperimental validation using phantom objects. Chag Gun Med J2000; 23: 354359.

28. Aranyarachkul P, Caruso J, Gantes B, Schulz E, Riggs M, Dus I,

et al. Bone density assessments of dental implant sites: 2.Quantitative cone-beam computerized tomography. Int J OralMaxillofac Implants 2005; 20: 416424.

29. Mah J, Hatcher D. Current status and future needs in craniofacialimaging. Orthod Craniofacial Res 2003; 6 (Suppl. 1):1016.

30. Maki K, Inou N, Takanishi A, Miller AJ. Computer-assistedsimulations in orthodontic diagnosis and the application of a newcone beam x-ray computed tomography. Orthod Craniofacial Res2003; 6 (Suppl. 1):95101.

31. Maki K, Inou N, Takanishi A, Miller AJ. Modeling of structure,quality and function in the orthodontic patient. OrthodCraniofacial Res 2003; 6 (Suppl. 1):5258.

32. Maki K, Miller AJ, Okano T, Hatcher D, Yamaguchi H,Shibasaki Y. Cortical bone mineral density in asymmetricalmandibles: a three-dimensional quantitative computed tomogra-phy study. Eur J Orthod 2001; 23: 217232.

Table 4 Summary of error relative to the caliper measurements

Trial Plastic skull NewTom Human skull NewTom Human skull CB MercuRay

Error Type Absolute Relative Absolute Relative Absolute Relative

Error 0.210.38 mm 0.861.87% 0.070.41 mm 0.191.56% 0.000.22 mm 0.010.06%Plane S . T S.T NE NE T . V V.T

T.

S V.

SJaw NE NE Mn . Mx Mn . Mx NE NE

S, sagittal; T, transverse; NE, no effect; V, vertical; Mn, mandible; Mx, maxilla




14/14

33. Maki K, Miller AJ, Okano T, Shibasaki Y. A three-dimensional,quantitative computed tomographic study of changes in distribu-tion of bone mineralization in the developing human mandible.Arch Oral Biol 2001; 46: 667678.

34. Cevidanes LH, Bailey LJ, Tucker GR Fr, Styner MA, Mol A,Phillips CL, et al. Superimposition of 3D cone-beam CT modelsof orthognathic surgery patients. Dentomaxillofac Radiol 2005;

34: 369375.35. Seo K, Ueno K, Matsuoka M, Kanou M, Yamamoto K, Tanaka

K, et al. Development of dentomaxillofacial conebeam X-ray CTsystem model CB MercuRay. MEDIX 2003; 37: 4045.

36. Schulze D, Heiland M, Thurman H, Adam G. Radiation exposureduring midfacial imaging using 4- and 16-slice computed tomo-graphy, cone beam computed tomography systems and conven-tional radiography. Dentomaxillofac Radiol 2004; 33: 8386.

37. Ludlow JB, Davies-Ludlow LE, Brooks SL. Dosimetry of twoextraoral direct digital imaging devices: NewTom cone beam CTand Orthophos Plus DS panoramic unit. Dentomaxillofac Radiol2003; 32: 229234.

38. Loubele M, Maes F, Schutyser F, Marchal G, Jacobs R, SuetensP. Assessment of bone segmentation quality of cone-beam CTversus multislice spiral CT: a pilot study. Oral Surg Oral MedOral Pathol Oral Radiol Endod 2006; 102: 225234.

39. Aboudara CL. Nasopharyngeal airway size determined on lateral

cephalometric headfilms and cone-beam CT scans. Growth and

development. San Francisco, CA: University of California, SanFrancisco; 2003, p. 60.

40. Yoo SK, Kim YO, Kim H-J, Kim NH, Jang YB, Kim K-D,et al. Alignment of CT images of skull dysmorphology usinganatomy-based perpendicular axes. Phys Med Biol 2003; 48:26812695.

41. Ogawa T, Enciso R, Shintaku WH, Clark GT. Evaluation of

cross-section airway configuration of obstructive sleep apnea.Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2007; 103:102108.

42. Cevidanes LH, Bailey LJ, Tucker SF, Styner MA, Mol A, PhillipsCL, et al. Three-dimensional cone-beam computed tomographyfor assessment of mandibular changes after orthognathic surgery.Am J Orthod Dentofacial Orthop 2007; 131: 4450.

43. Kobayashi K, Shimoda S, Nakagawa Y, Yamamoto A. Accuracyin measurement of distance using limited cone-beam computer-ized tomography. Int J Oral Maxillofac Implants 2004; 19:228231.

44. Holberg C, Steinhauser S, Geis P, Rudzki-Janson I. Cone-beamcomputed tomography in orthodontics: benefits and limitations. JOrofac Orthop 2005; 66: 434444.

45. Lagravere MO, Fang Y, Carey J, Toogood RW, Packota GV,Maior PW. Density conversion factor determined using a cone-beam computed tomography unit NewTom QR-DVT 9000.

Dentomaxillofac Radiol2006; 35: 407409.



Documents

Comparison of Cone Beam Computed Tomography Imaging With