reproducibility and accuracy of linear measurements on dental models derived from cone-beam computed...

ORIGINAL ARTICLE

Reproducibility and accuracy of linearmeasurements on dental models derived fromcone-beam computed tomography comparedwith digital dental casts

Olivier de Waard,a Frits Andreas Rangel,b Piotr Stanislaw Fudalej,c Ewald Maria Bronkhorst,d

Anne Marie Kuijpers-Jagtman,e and Karel Hero Breuningf

Nijmegen, The Netherlands, and Olomouc, Czech Republic

aPostgRadbobJuniboudcSenioSwitzdStatiRadboeProfeboudfAssisboudAll autentiaAddreCranioHB NSubm0889-Copyrhttp:/

328

Introduction: The aim of this studywas to determine the reproducibility and accuracy of linearmeasurements on2 types of dental models derived from cone-beam computed tomography (CBCT) scans: CBCT images, andAnatomodels (InVivoDental, San Jose, Calif); these were compared with digital models generated from dentalimpressions (Digimodels; Orthoproof, Nieuwegein, The Netherlands). The Digimodels were used as the refer-ence standard. Methods: The 3 types of digital models were made from 10 subjects. Four examiners repeated37 linear tooth and arch measurements 10 times. Paired t tests and the intraclass correlation coefficient wereperformed to determine the reproducibility and accuracy of the measurements. Results: The CBCT imagesshowed significantly smaller intraclass correlation coefficient values and larger duplicate measurement errorscompared with the corresponding values for Digimodels and Anatomodels. The average difference betweenmeasurements on CBCT images and Digimodels ranged from�0.4 to 1.65 mm, with limits of agreement valuesup to 1.3 mm for crown-width measurements. The average difference between Anatomodels and Digimodelsranged from �0.42 to 0.84 mm with limits of agreement values up to 1.65 mm. Conclusions: Statistically sig-nificant differences between measurements on Digimodels and Anatomodels, and between Digimodels andCBCT images, were found. Although the mean differences might be clinically acceptable, the random errorswere relatively large compared with corresponding measurements reported in the literature for both Anatomo-dels and CBCT images, and might be clinically important. Therefore, with the CBCT settings used in this study,measurements made directly on CBCT images and Anatomodels are not as accurate as measurements on Dig-imodels. (Am J Orthod Dentofacial Orthop 2014;146:328-36)

In orthodontics, study model analysis is an essentialpart of the diagnosis, treatment planning, and evalu-ation of treatment progress.1-4 When digitalization

raduate student, Department of Orthodontics and Craniofacial Biology,ud University Medical Centre, Nijmegen, The Netherlands.or researcher, Department of Orthodontics and Craniofacial Biology, Rad-University Medical Centre, Nijmegen, The Netherlands.r lecturer, Department of Orthodontics, University of Bern, Bern,erland; associate professor, Palacky University, Olomouc, Czech Republic.stician/methodologist, Department of Preventive and Curative Dentistry,ud University Medical Centre, Nijmegen, The Netherlands.ssor and chair, Department of Orthodontics and Craniofacial Biology, Rad-University Medical Centre, Nijmegen, The Netherlands.tant professor, Department of Orthodontics and Craniofacial Biology, Rad-University Medical Centre, Nijmegen, The Netherlands.thors have completed and submitted the ICMJE Form for Disclosure of Po-l Conflicts of Interest, and none were reported.ss correspondence to: Olivier de Waard, Department of Orthodontics andfacial Biology, Radboud University Medical Centre, Postbox 9101, 6500ijmegen, The Netherlands; e-mail, [email protected], December 2012; revised and accepted, May 2014.5406/$36.00ight � 2014 by the American Association of Orthodontists./dx.doi.org/10.1016/j.ajodo.2014.05.026

was introduced in the orthodontic world, digital modelsbecame available to replace traditional plaster casts. Themost frequently used method to obtain digital dentalmodels is to digitize plaster models or dentalimpressions. The technology used to generate digitalmodels from dental models or impressions variesconsiderably. Orthocad (Cadent, Carlstadt, NJ) uses“destructive scanning” with multiple scans of the plastermodel cut in thin slices. Emodels (GeoDigm, FalconHeights, Minn) scans the surface of a complete plastermodel. Impressions can also be scanned directly usingcone-beam computed tomography (CBCT) technology(Digimodels; Orthoproof, Nieuwegein, TheNetherlands).5

Digitized plaster models or digital models derivedfrom dental impressions have been shown to be a validtool for undertaking simple diagnostic measurementssuch as tooth size, arch width, overjet, overbite, archlength, and Bolton ratio.6 In a systematic review,Fleming et al5 found that overall, the mean differences

Delta:1_given name

Delta:1_surname

Delta:1_given name

Delta:1_surname

Delta:1_given name

Delta:1_surname

mailto:[email protected]

http://dx.doi.org/10.1016/j.ajodo.2014.05.026

de Waard et al 329

between linear measurements on plaster and digitalmodels were small (0.04-0.4 mm).

Digital dental models provide several advantages overplaster models. Digital models can be stored electroni-cally; this reduces storage space and the risk of damage.Furthermore, digital models can be shared easily over anetwork.6,7 A copy of the digital model can be securedat a second site for minimal or no costs. However,computer failure might mean that the models aretemporarily or forever inaccessible.7

Now, study model analysis can also be performeddirectly on radiographs, such as CBCT scans, without theneed for impressions. Study model analysis can be donedirectly on the radiograph of the headwith dedicated soft-ware. For example, Anatomage software (InVivoDental,San Jos�e, Calif) can be used to measure the dentition ona CBCT image. The files of a CBCT image can also besent by e-mail to InVivoDental for segmentation of thedentition to produce digital dentalmodels (Anatomodels).Studymodel analysis can be done directly on these Anato-models using the Anatomage measuring tools. The accu-racy of linear measurements of the dentition on theseAnatomodels for orthodontic purposes has been reportedin 4 articles.8-11 According to these publications, theaccuracy of measurements on Anatomodels is related tothe voxel size and spatial resolution of the CBCT images.Reduction of the spatial resolution can result in a lower-quality image, more noise and artifacts, and less anatomicinformation.12 Spatial resolution is lower at shorter scan-ning times and larger voxel sizes.12 Only in 1 study, ashorter scanning time, lower kilovoltage (129 kV), lowermilliamperage (47.74 mA), and a voxel size of 0.4 mmwere used.9 In all other studies, a smaller voxel size(0.3 mm) and longer scanning times were applied.8,10,11

A longer scanning time improves the spatial resolutionbut also increases the radiation dose.12

To our knowledge, no study has compared the accu-racy of measurements directly performed on CBCT im-ages (without segmentation of the dentition) withAnatomodels and Digimodels, and only 1 article usedCBCT scans with a voxel size of 0.4 mm and shorterscanning time for measuring the dentition. Therefore,the aim of this study was to examine the reproducibilityand accuracy of linear measurements on CBCT imagesand Anatomodels obtained from CBCT images with arelatively large voxel size of 0.4 mm and a relatively shortscanning time compared with the reference measure-ments on Digimodels.

MATERIAL AND METHODS

From the archives of the Department of Orthodonticsand Craniofacial Biology of the Radboud University Med-ical Centre in The Netherlands, pretreatment records were

American Journal of Orthodontics and Dentofacial Orthoped

selected of patients who met the following inclusioncriteria: (1) impressions to fabricate digital models and aCBCT image obtained on the same day; (2) permanentdentition from left first molar to right first molar in botharches; (3) normal dental crown morphology; (4) no fea-tures that would influence crownmorphology such as res-torations, caries, attrition, or fracture; (5) good-qualitydigital dental models without irregularities; and (6) themeasurements were not compromised by streaking arti-facts on the CBCT radiographs, caused by metal restora-tions or fixed appliances, because these patients had nometal restorations or orthodontic appliances. Twenty-four patients met these inclusion criteria. Ten patientsfrom this group were randomly selected for this study.

TheCBCT scanswere retrieved fromtheCBCTdatabaseof patients who had combined surgical and orthodontictreatment at the Radboud University Medical Centre.These extended-height scans were used for 3-dimensional planning for the orthognathic surgery. Allpatients signed the informed consent form for this treat-ment, including the CBCT documentation. The followingsettingswhere used for the i-CAT 3D Imaging System (Im-aging Sciences International, Hatfield, Pa): 129 kV,47.74 mA, 40 seconds with a resolution of 0.4 mm/voxel,and a field of view of 17 cm in diameter and 22 cm inheight.

For each patient, 3 types of dental models wereanalyzed: Digimodels, CBCT images, and Anatomodels.The Digimodels were used as the reference standard.To process the Digimodels (Fig 1), alginate dental im-pressions (Orthotrace; Cavex Holland BV, Haarlem, TheNetherlands) were sent to the OrthoProof company onthe same day as the impressions were taken. Within24 hours, the impressions were scanned with a computedtomography scanner with a voxel size of 0.15mm (Hytec,Los Alamos, NM). Subsequently, the scanned data wereused to fabricate digital models. These Digimodelswere imported through the Internet from the Ortho-Proof Web server into the patient management systemat the Radboud University Medical Centre. Studymodel analysis of the digital dental models was con-ducted with the Digimodels software.

Alternatively, study model analysis can be carried outdirectly on the digital image of the dentition on theCBCT image using dedicated software. The DICOM filesof the CBCT imageswere sent by e-mail to the Anatomagecompany for segmentationof the dentition from theCBCTimages to process the digital dental models, which arecalled Anatomodels. The company uses automatic thresh-olding logarithms for the segmentation process and doesnot share information on this rendering process.

Four dental students were trained and calibrated toperform the measurements on the 2 digital models and

ics September 2014 � Vol 146 � Issue 3

Fig 1. Digital model of themandibular dentition and somemeasurements on Digimodel.

330 de Waard et al

the CBCT image of each patient. The definitions of the37 measurements used in this study are given in TableI. Figures 1 to 3 show examples of the measurementson the 3 different models. The measurements onthe Digimodels were carried out with the Digimodelsoftware (version 2.2.1; Orthoproof) on a 15-in monitorwith a resolution of 12803 800 pixels and 32-bit color-screen setting.

The measurements on the CBCT images and Anato-models were carried out with the Anatomage measuringprogram on a 20-in computer monitor with the sameresolution and color depth setting as the 15-in monitor.The measurements on the CBCT images were taken onthe 3-dimensional rendered images. Enlargement, clip-ping, and rotation of the images to facilitate the mea-surements were routinely performed. There was notime limit for performing the measurements. All mea-surements were repeated 10 times by the 4 observers;this means that each of the 37 measurements wasrepeated 40 times. There was an interval of at least2 weeks between the series of measurements.

Statistical analysis

Interobserver reliability was assessed by calculatingthe intraclass correlation coefficient (ICC). Intraobserverreliability was assessed by the individual mean standarddeviations of the measurements for the 4 observers. AnICC of 0.90 or higher is considered excellent, between0.75 and 0.90 is good, and between 0.50 and 0.75 ismoderate. An ICC less than 0.50 is considered to bepoor.13 Statistical significance was set at P\0.05.

The reproducibility of the measurements was as-sessed by comparing the duplicate measurement errorsof the 3 study models.

Means and standard deviations were calculatedfor all variables, and limits of agreement values were

September 2014 � Vol 146 � Issue 3 American

calculated for all crown-width measurements to assessthe accuracy of the measurements on the 3 studymodels. The differences between the means of the vari-ables on the Digimodels, CBCT images, and Anatomo-dels for the 4 observers were calculated. The pairedt test was used to evaluate the differences between themean values of each variable. Statistical analysis wasperformed with the Statistical Package for Social Sci-ences software (version 18.2; SPSS, Chicago, Ill).

RESULTS

The interobserver reliability values were determinedfor all 37 variables of the 2 models and the CBCT imagesseparately. The mean ICC values were 0.88 (range, 0.62-0.99) for measurements on the Digimodels, 0.76 (range,0.20-0.98) for the CBCT images, and 0.92 (range, 0.76-0.99) for Anatomodels. The ICC values for the Digimod-els were considered good or excellent for 33 of the 37variables (higher than 0.75) and moderate for the other4 measurements. For the CBCT images, the ICC valueswere good for 22 measurements, moderate for 13 mea-surements (between 0.5 and 0.75), and poor for 2 mea-surements (\0.5). Especially, the intercanine andintermolar distances in the mandible had low ICC valueson the CBCT images. On the Anatomodels, the ICCvalues for all measurements were above 0.75.

To test the intraobserver reliability of the 4 observers,the individual standard deviations of the measurementswere determined. The intraobserver reliability of the ob-servers with the largest and the smallest standard devia-tions in measurements are shown in Figure 4. All 4observers showed larger standard deviations for theCBCT images compared with the measurements on theDigimodels and the Anatomodels. The intraobserver reli-ability values for measurements on these models werecomparable.

When comparing the duplicate measurement errorsfor Digimodels and CBCT images, 36 of the 37 duplicatemeasurement errors were significantly larger on theCBCT images.

Ten statistically significant and different duplicatemeasurement errors were found between the Anatomo-dels and the Digimodels: 8 measurements had largerduplicate measurement errors on the Anatomodels,and 2 measurements of segments had smaller duplicatemeasurement errors on the Anatomodels. Differencesbetween the duplicate measurement errors of the 3 dig-ital models are presented in Figure 5.

The mean values of the variables measured on Digi-models, CBCT images, and Anatomodels were compared.Statistically significant differences between measure-ments on the CBCT images models and the Digimodels

Journal of Orthodontics and Dentofacial Orthopedics

Fig 2. CBCT image of the mandibular dentition of the pa-tient of Figure 1 with some measurements.

Table I. Definitions of the 37 variables measured on the digital dental casts

Measurement DefinitionMesiodistal width of the crowns of incisors, canines,and premolars (maxilla and mandible)

Distance between the distal and mesial contact points of incisors,canines, premolars, respectively (mm) (20 distances)

Intermolar distance in maxilla and mandible Distance between the top of the mesiobuccal cusps of the firstpermanent molars (mm) (2 distances)

Intercanine distance in maxilla and mandible Distance between the cusps of the canines (mm) (2 distances)Interincisal distance Distance between the incisal edges of the maxillary and mandibular incisors

( mm) measured on a central point of the incisal edge of teeth 11 and 41Segment A Distal contact point of tooth 15 to the mesial contact point of 14Segment B Mesial contact point of tooth 14 to the distal contact point of 12Segment C Distal contact point of tooth 12 to the mesial contact point of 11Segment D Mesial contact point of tooth 11 to the distal contact point of 22Segment E Distal contact point of tooth 22 to the mesial contact point of 24Segment F Mesial contact point of tooth 24 to the distal contact point of 25Segment G Distal contact point of tooth 35 to the mesial contact point of 34Segment H Mesial contact point of tooth 34 to the distal contact point of 32Segment I Distal contact point of tooth 32 to the mesial contact point of 31Segment J Mesial contact point of tooth 31 to the distal contact point of 42Segment K Distal contact point of tooth 42 to the mesial contact point of 41Segment L Mesial contact point of tooth 44 to the distal contact point of 45

Tooth numbers are according to the F�ed�eration Dentaire Internationale (FDI) system.

Fig 3. Anatomodel derived from the CBCT scan of thepatient in Figure 1 with some measurements.

de Waard et al 331

(P values\0.05) were found in 32 of the 37 measure-ments. Of these 32 variables, 24 variables measured onthe CBCT images were smaller and 8 were largercompared with the measurements on the Digimodels.The mean difference for crown width between thesemodels ranged from 0.14 to �0.45 mm. The corre-sponding lower and upper limits of the 95% confidenceinterval of the difference were �0.55 and 0.33 mm,respectively (Fig 6, A; Table II). The mean difference inmeasurements for the transversal, segmental, and inter-incisal distances ranged from �0.2 to 1.65 mm. The


corresponding lower and upper limits of the 95% confi-dence interval of the difference were�0.35 and 2.71 mm, respectively (Fig 6, A). All statisti-cally significant crown-width measurements weresmaller on the CBCT images. All significant intermolar,intercanine, and interincisal distances were larger onthe CBCT images (Fig 6, A). Statistically significant dif-ferences between the Digimodels and the Anatomodelswere found for 22 measurements. Twelve measurementson Anatomodels were smaller, and 10 measurementswere larger. The mean difference in crown widths


Fig 4. Individual standard deviations for all measure-ments (mm): A, largest and B, smallest standard devia-tions among observers. The x-axis numbers representtooth types.

Fig 5. Duplicate measurement errors for A, CBCT im-ages and B, Anatomodels in comparison with the dupli-cate measurement errors of Digimodels. Statisticallysignificant errors for measurements on CBCT imagesand Anatomodels are marked in red if they were largerthan for Digimodels and in blue if they were smaller thanfor Digimodels. The x-axis numbers represent tooth types.

332 de Waard et al

between Digimodels and Anatomodels ranged from 0.36to�0.42 mm. The corresponding lower and upper limitsof the 95% confidence interval of the difference were�0.52 and 0.57 mm, respectively (Fig 6, B; Table II).The mean difference for measurements of the trans-versal, segmental, and interincisal distances on thesemodels had a range of �0.17 to 0.84 mm. The corre-sponding lower and upper limits of the 95% confidenceinterval of the difference were �0.36 and 1.09 mm,respectively (Fig 6, B). Compared with the measurementson the Digimodels, all significant intermolar, interca-nine, and interincisal distances were larger on the Ana-tomodels. All significant crown-width measurementson the Anatomodels were smaller compared with thesame measurements on the Digimodels, except for themaxillary right central incisor.


The spread of the differences for absolute measure-ments is described by the standard error of the mean.It can be concluded that the standard error of themean of transversal measurements for CBCT images issubstantially larger compared with the standard errorof the mean for measurements on the Anatomodels(Fig 6). Table II shows the limits of agreement of thecrown-width measurements. The range of the limits ofagreement values was larger for Anatomodels than forCBCT images in 18 of the 20 variables. The largest rangesof limits of agreement values were found for the maxil-lary left central incisor (�1.11 to 1.54 mm) in the Ana-tomodels and for the maxillary right central incisor(�1.02 to 1.3 mm) in the CBCT images.


Fig 6. Mean differences between A, Digimodels andCBCT images and B, Digimodels and Anatomodels withstandard error of the mean (SEM). Statistically significanterrors for measurements on CBCT images and Anatomo-dels are marked in red if they were larger than for Digimo-dels and in blue if they were smaller than for Digimodels.The x-axis numbers represent tooth types.

de Waard et al 333

DISCUSSION

In this study, the reproducibility and accuracy ofmeasurements performed on digital dental modelsproduced by scanning of the impressions (Digimodels)were compared with measurements on CBCT imagesand digital models derived from CBCT images of thesame patient (Anatomodels). Compared with the mea-surements on the reference models (Digimodels), almostall measurements on the CBCT images and Anatomodelshad good to excellent agreement, with high correlationcoefficients. Only the CBCT images showed ICC valuesbelow 0.5 for 2 measurements. Measurements on theCBCT images had the lowest interobserver reliability,


with a mean ICC value of 0.76 and a range of 0.20to 0.98.

Earlier studies comparing the accuracy betweenlinear measurements on plaster models and digitalmodels obtained from dental impressions or scannedplaster models found mean differences between 0.04and 0.62 mm.7,14-17 Studies that compared theaccuracy between linear measurements on CBCT-derived models and OrthoCad digital models foundmaximum mean differences between 0.44 and0.62 mm for linear measurements.8,10,11

For the CBCT images in our study, mean differencesbetween the mandibular intermolar distances (1.13 mm)and the intercanine distances (1.65 mm) on the CBCTimages with the corresponding distance measurementson the Digimodels were larger than the maximummean differences reported in the literature (0.44-0.62 mm).

For the Anatomodels, the mean differences for themandibular intermolar distance (0.84 mm) and 1segment in the mandibular arch (0.77 mm) were largerthan the above-mentioned maximum mean differencesin the literature (0.44-0.62 mm).

The limits of agreement values for both the CBCT im-ages and the Anatomodels were large, for Anatomodelsgoing up to 1.65 mm for the crown-width measurementof the maxillary right central incisor, and the largestlimits of agreement for the maxillary left central incisorranged from�1.11 to 1.54 mm. The limits of agreementvalues for the CBCT images were, in general, a bitsmaller, but the limits of agreement range for the maxil-lary right central incisor of �1.02 to 1.3 mm was stilllarge. A difference of 1 mm for a maxillary incisor meansa potential measurement error of more than 10%, andthis might be clinically important. The limits of agree-ment values in this study were much larger comparedwith those in the study of Lightheart et al.11 However,in that study, a different CBCT device was used, andthe voxel size differed from our study, and only trans-versal measurements were performed, with fewerrepeated measurements.

The relatively large standard deviations and absolutemean differences, especially for the transversal measure-ments as found in this study, could be due to the inac-curate anatomic reconstruction of the occlusal surfaceson the CBCT scans. The CBCT scans were made withthe teeth in habitual occlusion. Separation of theocclusal planes out of occlusion followed by reconstruc-tion of the occlusal surfaces of both arches during themanufacturing of the digital models by the techniciancould lead to less accurate occlusal surfaces, and thismight have caused inadequate outlines of the tips ofthe cusps used for measuring the transversal distances.10


Table II. Mean differences and differences for duplicate measurement errors between Digimodels and Anatomodels,and Digimodels and CBCT images, for teeth (FDI numbering) with measurements

Mean difference

Limits ofagreement

Duplicate measurement error

P Difference

95% CI of the difference

P Difference

95% CI of the difference

Lower Upper Lower UpperCrown-width measurements: Anatomodels-Digimodels15 0.001 �0.20 �0.31 �0.09 �0.86.0.46 0.003 0.03 0.01 0.0514 0.001 �0.33 �0.47 �0.19 �1.20...0.53 0.422 0.01 �0.01 0.0313 0.001 �0.29 �0.44 �0.14 �1.21...0.63 0.240 0.01 �0.01 0.0412 0.473 0.04 �0.07 0.16 �0.67...0.75 0.291 0.01 �0.01 0.0411 0.001 0.36 0.15 0.57 �0.93...1.65 0.723 �0.01 �0.05 0.0421 0.052 0.21 �0.02 0.43 �1.11...1.54 0.278 0.01 �0.01 0.0322 0.742 �0.03 �0.22 0.16 �1.19...1.13 0.365 0.01 �0.01 0.0323 0.024 �0.19 �0.36 �0.03 �1.21...0.82 0.069 0.02 0.001 0.0524 0.001 �0.31 �0.39 �0.23 �0.8...0.19 0.872 0.001 �0.02 0.0125 0.004 �0.17 �0.29 �0.06 �0.89...0.54 0.037 0.02 0.001 0.0435 0.015 �0.20 �0.36 �0.04 �1.16...0.77 0.114 0.02 0.001 0.0434 0.001 �0.24 �0.36 �0.12 �0.97...0.49 0.369 0.01 �0.01 0.0333 0.093 �0.10 �0.22 0.02 �0.82...0.62 0.001 0.06 0.03 0.0832 0.093 �0.14 �0.3 0.02 �1.14...0.86 0.002 0.03 0.01 0.0631 0.186 �0.09 �0.23 0.05 �0.93...0.74 0.867 0.001 �0.02 0.0241 0.226 �0.09 �0.23 0.06 �0.96...0.79 0.116 0.02 0.001 0.0442 0.006 �0.22 �0.38 �0.07 �1.18...0.73 0.008 0.03 0.01 0.0643 0.268 �0.07 �0.20 0.06 �0.87...0.72 0.003 0.04 0.01 0.0644 0.001 �0.42 �0.52 �0.31 �1.04...0.2 0.241 0.02 �0.01 0.0445 0.001 �0.24 �0.35 �0.13 �0.90...0.42 0.025 0.02 0.001 0.04

Crown-width measurements: CBCT images-Digimodels15 0.001 �0.29 �0.40 �0.18 �0.99...0.41 0.001 0.08 0.04 0.1114 0.001 �0.41 �0.52 �0.30 �1.06...0.24 0.001 0.07 0.04 0.1013 0.001 �0.41 �0.52 �0.29 �1.13...0.32 0.001 0.11 0.08 0.1512 0.001 �0.20 �0.31 �0.09 �0.88...0.49 0.001 0.09 0.06 0.1311 0.145 0.14 �0.05 0.33 �1.02...1.3 0.008 0.08 0.02 0.1421 0.281 0.09 �0.08 0.27 �0.97...1.16 0.001 0.07 0.05 0.0922 0.001 �0.19 �0.30 �0.08 �0.84...0.46 0.001 0.11 0.07 0.1523 0.001 �0.45 �0.55 �0.34 �1.09...0.19 0.001 0.10 0.07 0.1224 0.001 �0.41 �0.52 �0.31 �1.03...0.2 0.001 0.08 0.05 0.1125 0.001 �0.23 �0.33 �0.13 �0.85...0.38 0.001 0.08 0.05 0.1135 0.002 �0.17 �0.28 �0.06 �0.83...0.49 0.001 0.08 0.05 0.1034 0.001 �0.39 �0.50 �0.29 �1.05...0.26 0.001 0.11 0.07 0.1633 0.102 �0.09 �0.19 0.02 �0.71...0.54 0.001 0.13 0.08 0.1732 0.001 �0.31 �0.43 �0.19 �1.03...0.41 0.001 0.10 0.07 0.1331 0.005 �0.14 �0.23 �0.05 �0.71...0.43 0.001 0.08 0.05 0.1141 0.019 �0.13 �0.25 �0.02 �0.82...0.55 0.001 0.09 0.05 0.1242 0.001 �0.24 �0.35 �0.12 �0.94...0.46 0.001 0.10 0.06 0.1343 0.003 �0.17 �0.27 �0.06 �0.82...0.49 0.001 0.12 0.08 0.1644 0.001 �0.45 �0.53 �0.37 �0.93...0.03 0.001 0.09 0.06 0.1245 0.001 �0.18 �0.28 �0.08 �0.77...0.41 0.001 0.09 0.06 0.12

334 de Waard et al

Another reason for the mean differences, the rela-tively large standard deviations, and the limits of agree-ment values could have been the voxel size. Because theboundary of an object (eg, a tooth) is marked on theCBCT by a voxel, the dimension of the tooth on aCBCT scan depends on the dimensions of the voxel. Dur-ing measuring, it was assumed that the boundary be-tween the dentition and the surrounding mucosa or air


was between the contact points of 2 voxels. In reality,this is not always the case. Possibly, the actual boundarybetween the dentition and the air or the mucosa islocated in the center of the voxel. In our study, theCBCT scans were made with a voxel size of 0.4 mm.The maximum measurement error caused by the sizeof the voxel is half of the diagonal of the voxel sizeused (0.35 mm). Measurements on the dentition of


de Waard et al 335

digital models based on CBCT images require the selec-tion of 2 boundaries, thus 2 voxels. The maximum errorfor the measurements in this study caused by the voxelsize will then be about 0.7 mm.

A decrease in voxel size might enhance the quality ofthe scans and result in a better correlation between themeasurements on the dentition of the CBCT-derivedmodels and the measurements on plaster or digital studymodels based on scanning of plaster models.11 However,reduction of the voxel size only to enhance the accuracyof measurements of the dentition is not feasible, takinginto account that radiation exposure should be as low asreasonable achievable.

The automatic threshold selection procedure for thesegmentation process could be a source of error. Howev-er, a disadvantage associated with CBCT radiographs isthe inability to display actual Hounsfield units as inmedical computed tomography. A standard scaling sys-tem for gray values in a radiograph is necessary to enablecomparison of the gray values from different ma-chines.18 To date, such a system is not available forCBCT machines. Therefore, it is arbitrary to interpretthe gray levels used for segmentation procedures inthis study. Furthermore, the manufacturers do not sharemore information on their segmentation process.

The results indicate smaller crown-width measure-ments on CBCT images and Anatomodels comparedwith Digimodels. This could be related to the effectknown as the “partial averaging effect,” the effect thatdifferent densities within a voxel are averaged andrepresented by 1 number.19 Thus, the numbers of thinbony walls will tend to drop below the segmentationrange or threshold set for bone because the bone densityis averaged with that of surrounding air.20 An incorrectadjustment of threshold settings to visualize the differ-ence in density between tissues along the boundaryof objects, such as enamel, air, and mucosa, could pro-duce the smaller crown-width measurements on theCBCT images and Anatomodels compared with theDigimodels.20

The duplicate measurement errors in this study mea-sure the reproducibility. In general, the duplicate mea-surement errors on the Anatomodels were comparablewith those on the Digimodels. In contrast, we foundthat the reproducibility of measurements on the CBCTimages was lower than the reproducibility of measure-ments on the Digimodels and Anatomodels. The Anato-models were produced by the InVivoDental company,using thresholds for segmentation of the dentition. Aftersegmentation, the outline of the dentition is clearlyvisible. Possibly, this process increases the measurementaccuracy and reproducibility of measurements on theAnatomodels compared with the CBCT images.


In a systematic review, Fleming et al5 found thatoverall, the mean differences between linear measure-ments on plaster and digital models were small (0.04-0.4 mm). Because of the additional advantages of digitaldental models such as reducing storage space, risk ofdamage, transferability of these models, and using thesame digital measurement method, we used Digimodelsas a reference in this study instead of plaster models ordirect measurements on a patient's dentition. The voxelsize (0.15 mm) of the CBCT scanner, which produced thedigital image of the impression in this study, was smallerthan the voxel size used for the CBCT radiographs of thepatients. Dimensional changes of the impression mate-rial (alginate) after impression taking and during disin-fection and pouring of the plaster can occur.21,22

CONCLUSIONS

Statistical differences between measurements onDigimodels and Anatomodels, and between Digimodelsand CBCT images, were found. The reproducibility ofmeasurements on Anatomodels is comparable to Digi-models and better than measurements made directlyon the CBCT images.

Although the mean differences might be clinicallyacceptable, the random errors were relatively largecompared with corresponding measurements reportedin the literature for both Anatomodels and CBCT images,and might be clinically important. Therefore, with theCBCT settings used in this study, measurements directlyperformed on CBCT images and Anatomodels are not asaccurate as measurements on Digimodels.

REFERENCES

1. Han UK, Vig KW, Weintraub JA, Vig PS, Kowalski CJ. Consistencyof orthodontic treatment decisions relative to diagnostic records.Am J Orthod Dentofacial Orthop 1991;100:212-9.

2. Baumgaertel S, Palomo JM, Palomo L, Hans MG. Reliability andaccuracy of cone-beam computed tomography dental measure-ments. Am J Orthod Dentofacial Orthop 2009;136:19-25.

3. El-ZanatyHM, El-Beialy AR, AbouEl-EzzAM,Attia KH, El-Bialy AR,Mostafa YA. Three-dimensional dental measurements: an alterna-tive to plaster models. Am J Orthod Dentofacial Orthop 2010;137:259-65.

4. Bell A, Ayoub AF, Siebert P. Assessment of the accuracy of a three-dimensional imaging system for archiving dental study models.J Orthod 2003;30:219-23.

5. Fleming PS, Marinho V, Johal A. Orthodontic measurements ondigital study models compared with plaster models: a systematicreview. Orthod Craniofac Res 2011;14:1-16.

6. Veenema AC, Katsaros C, Boxum SC, Bronkhorst EM, Kuijpers-Jagtman AM. Index of Complexity, Outcome and Need scoredon plaster and digital models. Eur J Orthod 2009;31:281-6.

7. Quimby ML, Vig KW, Rashid RG, Firestone AR. The accuracy andreliability of measurements made on computer-based digitalmodels. Angle Orthod 2004;74:298-303.


http://refhub.elsevier.com/S0889-5406(14)00534-4/sref1






















336 de Waard et al

8. Kau CH, Littlefield J, Rainy N, Nguyen JT, Creed B. Evaluation ofCBCT digital models and traditional models using the Little's In-dex. Angle Orthod 2010;80:435-9.

9. Tarazona B, Llamas JM, Cibrian R, Gandia JL, Paredes V. A com-parison between dental measurements taken from CBCT modelsand those taken from a digital method. Eur J Orthod 2013;35:1-6.

10. Creed B, Kau CH, English JD, Xia JJ, Lee RP. A comparison of theaccuracy of linear measurements obtained from cone beamcomputerized tomography images and digital models. Semin Or-thod 2011;17:49-56.

11. Lightheart KG, English JD, Kau CH, Akyalcin S, Bussa HI Jr,McGrory KR, et al. Surface analysis of study models generatedfrom OrthoCAD and cone-beam computed tomography imaging.Am J Orthod Dentofacial Orthop 2012;141:686-93.

12. Ballrick JW, Palomo JM, Ruch E, Amberman BD, Hans MG. Imagedistortion and spatial resolution of a commercially available cone-beam computed tomography machine. Am J Orthod DentofacialOrthop 2008;134:573-82.

13. Portney LG, WatkinsMP. Foundations of clinical research: applica-tions to practice. 2nd ed. Upper Saddle River, NJ: Prentice Hall;1999.

14. Keating AP, Knox J, Bibb R, Zhurov AI. A comparison of plaster,digital and reconstructed study model accuracy. J Orthod 2008;35:191-201.

15. Asquith J, Gillgrass T, Mossey P. Three-dimensional imaging of or-thodontic models: a pilot study. Eur J Orthod 2007;29:517-22.


16. Bootvong K, Liu Z, McGrath C, H€agg U, Wong RW, Bendeus M,et al. Virtual model analysis as an alternative approach to plastermodel analysis: reliability and validity. Eur J Orthod 2010;32:589-95.

17. Watanabe-Kanno GA, Abr~ao J, Miasiro Junior H, Sanchez-Ayala A,Lagrav�ere MO. Reproducibility, reliability and validity of measure-ments obtained from Cecile3 digital models. Braz Oral Res 2009;23:288-95.

18. Mah P, Reeves TE, McDavid WD. Deriving Hounsfield units usinggrey levels in cone beam computed tomography. DentomaxillofacRadiol 2010;39:323-35.

19. Halazonetis DJ. Commentary. (Baumgaertel S, Palomo JM, PalomoL, Hans MG. Reliability and accuracy of cone-beam computedtomography dental measurements. Am J Orthod DentofacialOrthop 2009;136:19-25.) Am J Orthod Dentofacial Orthop2009;136:25-8.

20. Spoor F, Jeffery N, Zonneveld F. Using diagnostic radiology in hu-man evolutionary studies. J Anat 2000;197(Pt 1):61-76.

21. Coleman RM, Hembree JH Jr, Weber FN. Dimensional stability ofirreversible hydrocolloid impression material. Am J Orthod 1979;75:438-46.

22. Stevens DR, Flores-Mir C, Nebbe B, Raboud DW, Heo G, Major PW.Validity, reliability, and reproducibility of plaster vs digital studymodels: comparison of peer assessment rating and Bolton analysisand their constituent measurements. Am J Orthod Dentofacial Or-thop 2006;129:794-803.




















































reproducibility and accuracy of linear measurements on dental models derived from cone-beam computed...

Documents