ORIGINAL ARTICLE

Precision of manual landmarkidentification between as-received andoriented volume-rendered cone-beamcomputed tomography images

Abhishek Gupta,a Om Prakash Kharbanda,b Rajiv Balachandran,c Viren Sardana,d Shilpa Kalra,c

Sushma Chaurasia,c and Harish Kumar Sardanae

Chandigarh and New Delhi, India

aAcadScienbCentDentocDivisEducadCSIReCSIRand In

118

Introduction: The objective of this study was to evaluate the effect of the orientation of cone-beam computedtomography (CBCT) images on the precision and reliability of 3-dimensional cephalometric landmarkidentification. Methods: Ten CBCT scans were used for manual landmark identification. Volume-renderedimages were oriented by aligning the Frankfort horizontal and transorbital planes horizontally, and themidsagittal plane vertically. A total of 20 CBCT images (10 as-received and 10 oriented) were anonymized,and 3 random sets were generated for manual landmark plotting by 3 expert orthodontists. Twenty-fivelandmarks were identified for plotting on each anonymized image independently. Hence, a total of 60 imageswere marked by the orthodontists. After landmark plotting, the randomized samples were decoded andregrouped into as-received and oriented data sets for analysis and comparison. Means and standarddeviations of the x-, y-, and z-axis coordinates were calculated for each landmark to measure the centraltendency. Intraclass correlation coefficients were calculated to analyze the interobserver reliability oflandmark plotting in the 3 axes in both situations. Paired t tests were applied on the mean Euclidean distancecomputed separately for each landmark to evaluate the effect of 3-dimensional image orientation.Results: Inter-observer reliability (intraclass correlation coefficient,.0.9) was excellent for all 25 landmarks for the x-, y-, and z-axes on both before and after orientation of the images. Paired t test results showed insignificant differences forthe orientation of volume-rendered images for all landmarks except 3: R1 left (P 5 0.0138), sella (P 5 0.0490),and frontozygomatic left (P 5 0.0493). Also midline structures such as Bolton and nasion were plotted moreconsistently or precisely than bilateral structures. Conclusions: Orientation of the CBCT image does notenhance the precision of landmark plotting if each landmark is defined properly on multiplanar reconstructionslices and rendered images, and the clinician has sufficient training. The consistency of landmark identificationis influenced by their anatomic locations on the midline, bilateral, and curved structures. (Am J OrthodDentofacial Orthop 2017;151:118-31)

Three-dimensional craniofacial imaging such ascomputed tomography (CT) and cone-beam CT(CBCT) offers great potential in diagnosis and

treatment planning of complex skeletal deformities

emy of Scientific and Innovative Research (AcSIR), CSIR-Centraltific Instruments Organisation (CSIO) Campus, Chandigarh, India.re for Dental Education and Research; Division of Orthodontics andfacial Deformities, All India Institute of Medical Sciences, New Delhi, India.ion of Orthodontics and Dentofacial Deformities, Centre for Dentaltion and Research, All India Institute of Medical Sciences, New Delhi, India.-Central Scientific Instruments Organisation, Chandigarh, India.-Central Scientific Instruments Organisation (CSIO); Academy of Scientificnovative Research (AcSIR-CSIO), Chandigarh, India.

and assessment of growth and treatment effects.1-8

Conventionally, craniofacial analyses based on 2-dimensional (2D) cephalometry have several limitations:magnification, distortion, overlapping of craniofacial

All authors have completed and submitted the ICMJE Form for Disclosure ofPotential Conflicts of Interest, and none were reported.Address correspondence to: Om Prakash Kharbanda, Division of Orthodonticsand Dentofacial Deformities, Centre for Dental Education and Research, All IndiaInstitute of Medical Sciences, New Delhi 110029, India; e-mail, opk15@hotmail.com; dr.opk15@gmail.com.Submitted, April 2015; revised and accepted, June 2016.0889-5406/$36.00� 2017 by the American Association of Orthodontists. All rights reserved.http://dx.doi.org/10.1016/j.ajodo.2016.06.027

Gupta et al 119

structures, difficulty in locating hidden anatomic struc-tures, and so on.9-15 These limitations were alsohighlighted during comparisons of measurementsbetween dry skulls and those on 2D cephalogramswhile searching for the anatomic truth.16 With the ad-vancements in 3-dimensional (3D) imaging modalitiesin the last decade, these limitations of 2D analysishave been addressed to a certain extent.5,17

Synchronal literature in this decade has emphasizedthe pivotal role of 3D CT and CBCT imaging modalitiesin the 3D cephalometric analysis.2,5,10,17-20 But thechallenge for clinicians at present is to understand andinterpret 3D imaging.11 Conventionally, cephalometricanalysis is based on the landmark identification and plot-ting on 2D images. Training and familiarization with thelocation of landmarks on 3D images is essential becauselandmark identification errors are amajor source of ceph-alometric errors.15 Therefore, the need for new guidelinesfor 3D landmark identification is warranted.

Many studies in recent years have evaluated the reli-ability of landmark identification on CT and CBCTdata.21-26 Apart from the errors due to the lack ofexperience,27-31 the perceptions of the observer inlocalizing the anatomic landmarks on 3D images and thehead orientation also may influence landmark plottingon 3D images.32 A few studies have evaluated the effectof head orientation in CBCT synthesized posteroanteriorand lateral cephalograms22,24 and 3D CBCT imagingmodality21,23,25 on cephalometric measurements. Asignificant difference was found between dry skullcephalometric measurements and CBCT synthesizedlateral cephalogram measurements22,24 in different headpositions, whereas the differences in measurements on3D CBCT images were found to be statisticallyinsignificant from dry skull measurements.21-23,25 Studiesby Tomasi et al23 and Berco et al25 have shown statisticallyinsignificant differences between nonoriented CBCT im-ages and dry skulls; the data were derived using only a sin-gle skull for measurements. Similarly, Ludlow et al21 andHassan et al22 have also shown insignificant measurementdifferences between nonoriented CBCT data and dry skullswith 4 and 10 linear measurements, respectively.

These studies have tried to provide insight into theeffect of head orientation on the accuracy of linear mea-surements but not on the anatomic landmark positionsin 3 dimensions with a change in orientation21-26

(Table I). To authenticate the accuracy of plotted land-marks, baseline data (gold standard) were required tobe derived from the markings on the actual skull modelsfor comparison. Since it is not possible to obtain goldstandard measurements directly from living subjects,data can possibly be derived by repeated landmark plot-ting on CBCT images. The precision of landmark plotting

American Journal of Orthodontics and Dentofacial Orthoped

could be influenced by the orientation of the volume-rendered image. Direct evaluation of precision and con-sistency of 3D landmark plotting with and withoutorientation has not been investigated.32 In the light ofsuch data with uncertain standpoints on the effect oforientation on landmark identification, this study wasconducted to evaluate the effect of orientation on theprecision of 3D landmark identification vis-�a-vis as-received CBCT images.

MATERIAL AND METHODS

Ten CBCT images were collected randomly from anorthodontic clinic database retrospectively irrespectiveof age, sex, and ethnicity. The ethics committee of AllIndia Institute of Medical Sciences, New Delhi, Indiaapproved this study, and no patients were recruited forthis study separately. The CBCT scans were obtainedusing an i-CAT next generation machine (ImagingSciences International, Hatfield, Pa) with a field ofview of 17 3 22 cm and a scan time of 26 seconds.The data were saved in DICOM (version 1.7) formatwith an isometric voxel size of 0.25 to 0.30 mm. CBCTscans had been taken with the subject sitting uprightand in natural head position.

Three experienced orthodontists (R.B., S.K., S.C.)were asked to plot the landmarks on the CBCT imagesand were called O1, O2, and O3. Furthermore, 2 ob-servers (A.G., V.S.) separately were asked to perform ori-entations of the CBCT images and randomization of thedata for blind marking for the experiment. The observerwho had performed the orientations was called O4, andthe observer who had randomized the data for blindmarking was called O5. Observer O5 generated 3 randomsets of CBCT data referred to as SI, SII, and SIII for land-mark plotting of the 3 observers O1, O2, and O3.

Twenty-five commonly used cephalometric land-marks were selected, and operational definitions ofeach landmark33,34 in the 3 planes (axial/xy plane,coronal/xz plane, and sagittal/yz plane) werederived.18,28,31 In addition to the cross-sectional slices,3D volume-rendered images were also used to confirmlandmark positions. Three axes were defined: x-axis inthe right-left direction, y-axis in the anteroposterior di-rection, and z-axis in the inferior-superior direction. The3 orthodontists who participated in the study werefamiliarized with each landmark and mutually agreedon the definitions of each landmark in 3 dimensions.They practiced on 5 anonymized CBCT images. Any am-biguity in locating landmarks was resolved throughmutual discussion, and the landmark definitions wererefined with the consensus of the expert orthodontists.It was decided to plot 11 landmarks in volume-rendered images directly and confirm them on

ics January 2017 � Vol 151 � Issue 1

Table I. Studies evaluating accuracy of cephalometric measurements for different head positions

Study Authors Objective Sample size Observers (n) Observations per observer (n) Landmarks (n) Measurements (n)1. Neiva et al26 (2014) Landmark identification

comparison between CBCTMPR and 3D reconstruction

12 CBCTs 3 3 (on 3D) 1 3 (on MPR) 5 6 28 -

2. Tomasi et al23 (2011) Influence of inclination of theobject on reliability andreproducibility of CBCTmeasurements

1 dry mandible 3 3 (physical) 1 4 (on optimalCBCT scan) 1 4 (on 45�

rotated CBCT scan) 5 11

- 10

3. Cevidanes et al24 (2009) Head orientation in CBCTgenerated 2D cephalograms

12 CBCTs 3 4 (2 on NHP 2D projection and2 on oriented 2D projection)

- 50

4. Hassan et al22 (2009) Influence of patient scanningposition

8 dry skulls 3 2 (physical) 1 2 (on 3D) 1 2(MPR) 1 2 (on lateral andPA projection) 5 8

8 10

5. Berco et al25 (2009) Accuracy and reliability of 3Dcraniofacial measurements

1 dry skull 2 4 (physical) 1 4 (on optimalCBCT scan) 1 4 (on 45�

rotated CBCT scan) 5 12

17 29

6. Ludlow et al21 (2007) Influence of rotation onmeasurements made over2D and 3D images

30 dry skulls 11111 2 (physical) 1 1 (physical) 1(2 [on panoramic] and 2 [on3D])

- 4

7. Present study Influence of orientation 10 CBCTs 3 1 (before orientation) 1 1(after orientation) 5 2

25 -

MPR, Multiplanar reconstruction; NHP, natural head position; PA, posteroanterior.

120Gupta

etal

January2017�

Vol151�

Issue1

American

JournalofOrthodontics

andDentofacialO

rthopedics

Table II. Definitions of 25 landmarks used in the study and procedure of identification on different sectional slices

Landmark Abbreviation Definition on skull Sagittal slice Axial slice Coronal slice Remarks1. Nasion N Most anterior point of the

frontonasal suture inthe midsagittal plane

Anterior-most point Middle-anterior mostpoint on the anteriorcontour

Middle point Plotted on 3D volumetricdata and confirmed onMPR views

2. Orbitale left OrL The lowest point in theinferior margin of theleft orbit

Anterior-superior mostpoint

- - Plotted on 3D volumetricdata and confirmed onMPR views

3. Orbitale right OrR The lowest point in theinferior margin of theright orbit

Anterior-superior mostpoint

- - Plotted on 3D volumetricdata and confirmed onMPR views

4. A-point A-point The point at the deepestmidline concavity onthe maxilla between theanterior nasal spine andthe dental alveolus

Posterior-most point Middle-anterior mostpoint on the anteriorcontour

Middle point determinedby the sagittal and axialslices

Plotted through MPRviews and confirmed on3D volumetric data

5. Anterior nasal spine ANS Most anterior midpoint ofthe anterior nasal spineof maxilla

Most anterior point Anterior point and middlepoint

Middle point Plotted through MPRviews and confirmed on3D volumetric data

6. Posterior nasal spine PNS The sharp posteriorextremity of the nasalcrest of the hard palate

Most posterior point Posterior point andmidpoint

- Plotted through MPRviews and confirmed on3D volumetric data

7. B-point B-point Most posterior point in theconcavity along theanterior border of themandibular symphysis

Posterior-most point Middle-anterior mostpoint on the anteriorcontour

Middle point determinedby the sagittal and axialslices

Plotted through MPRviews and confirmed on3D volumetric data

8. Pogonion Pog/Pg Most anterior point onmandibular symphysis

Anterior-most point Middle-anterior-mostpoint on the anteriorcontour

Middle point determinedby the sagittal and axialslices

Plotted through MPRviews and confirmed on3D volumetric data

9. Menton Me Most inferior point on themandibular symphysis

Inferior-most point Middle most point Middle inferior-most point Plotted through MPRviews and confirmed on3D volumetric data

10. Gnathion Gn Midpoint of the curvatureof pogonion andmenton

Anterior-inferior mostpoint

Middle-anterior mostpoint

Middle inferior-most point Plotted through MPRviews and confirmed on3D volumetric data

11. Gonion left GoL Most inferior and posteriorpoint on leftmandibular corpus

Inferior and posterior mostpoint

Posterior most point Inferior-most point Plotted on 3D volumetricdata and confirmed onMPR views

12. Gonion right GoR Most inferior and posteriorpoint on rightmandibular corpus

Inferior and posterior mostpoint

Posterior most point Inferior-most point Plotted on 3D volumetricdata and confirmed onMPR views

13. Condylion left CoL Most superior point on theleft mandibular condyle

Superior-most point Midpoint determined bythe sagittal and coronalslices

Middle superior-mostpoint

Plotted through MPRviews and confirmed on3D volumetric data

Gupta

etal

121

American

JournalofOrthodontics

andDentofacialO

rthopedicsJanuary

2017�Vol151�

Issue1

Table II. Continued

Landmark Abbreviation Definition on skull Sagittal slice Axial slice Coronal slice Remarks14. Condylion right CoR Most superior point on the

right mandibularcondyle

Superior-most point Midpoint determined bythe sagittal and coronalslices

Middle superior-mostpoint

Plotted through MPRviews and confirmed on3D volumetric data

15. R1 left R1L The deepest point on thecurve of the anteriorborder of the left ramus

Deepest point Anterior point - Plotted on 3D volumetricdata and confirmed onMPR views

16. R1 right R1R The deepest point on thecurve of the anteriorborder of the rightramus

Deepest point Anterior point - Plotted on 3D volumetricdata and confirmed onMPR views

17. Sella S Midpoint of sella turcica Middle point of thepituitary fossa

Middle point of theanteroposterior andlateral width of thepituitary fossa

Middle point of the lateralwidth of the fossadetermined by thesagittal and axial slices

Plotted through MPRviews and confirmed on3D volumetric data

18. Basion B The most anterior point onthe anterior margin ofthe foramenmagnum inthe midsagittal plane

Inferior-posterior mostpoint

Anterior-most point Middle point Plotted through MPRviews and confirmed on3D volumetric data

19. Zygomatic point left ZyL The most lateral point onthe left outline of leftzygomatic arch

- Anterior lateral point Most lateral and superiorpoint

Plotted on 3D volumetricdata and confirmed onMPR views

20. Zygomatic point right ZyR The most lateral point onthe right outline of rightzygomatic arch

- Anterior lateral point Most lateral and superiorpoint

Plotted on 3D volumetricdata and confirmed onMPR views

21. Frontozygomatic left FzL The most medial andanterior point of leftfrontozygomatic sutureat the level of the lateralorbital rim

Most anterior Most anterior Medial point Plotted on 3D volumetricdata and confirmed onMPR views

22. Frontozygomatic right FzR The most medial andanterior point of rightfrontozygomatic sutureat the level of the lateralorbital rim

Most anterior Most anterior Medial point Plotted on 3D volumetricdata and confirmed onMPR views

23. Jugal point left JL The deepest midpoint ofleft jugal process ofmaxilla

Inferior-most point - Deepest point Plotted through MPRviews and confirmed on3D volumetric data

24. Jugal point right JR The deepest midpoint ofright jugal process ofmaxilla

Inferior-most point - Deepest point Plotted through MPRviews and confirmed on3D volumetric data

25. Bolton Bo The most posterior pointof foramen magnum inmidsagittal plane

Anterior point of posteriorborder of foramanmagnum

Mid and posterior mostpoint

Middle point Plotted through MPRviews and confirmed on3D volumetric data

MPR, Multiplanar reconstruction.

122Gupta

etal

January2017�

Vol151�

Issue1

American

JournalofOrthodontics

andDentofacialO

rthopedics

Gupta et al 123

multiplanar reconstruction slices; the remaining 14landmarks were plotted in reverse order (Table II).

For the purpose of orientation, DICOM data were im-ported into Dolphin 3D software (version 11.5; DolphinImaging and Management Systems, Chatsworth, Calif).Volume-rendered images were oriented by aligning theFrankfort horizontal plane and the transorbital plane hor-izontally, and the midsagittal plane vertical (Fig 1) by anobserver (O4), who was not involved in landmark plot-ting. After satisfactory orientation, new DICOM image sli-ces/series were saved. A total data set of 20 CBCT images(10 as-received and 10 after orientation) was thuscreated. These data sets were then anonymized for blindlandmark plotting using Mimics software (Materialise,Leuven, Belgium) by the same observer (O4). To prevent

Fig 1. Orientation of volume-rendered image; Top rrespectively, of as-received volume-rendered CBCTof the same CBCT image after orientation. Columndoes not pass through the left and right orbitale landmin the same horizontal line. Column (b)-(b'): the horizthrough the orbitale and porion landmarks. Rotatedhorizontal line. Column (c)-(c'): the vertical line in toplandmarks of the cranial base such as crista galli andsame vertical line.

American Journal of Orthodontics and Dentofacial Orthoped

bias, all as-received and oriented datasets were kept in1 folder and renamed randomly with numbers from 1to 20 in the order of plotting by an observer (O5) whohad not performed the orientation of the CBCT images.Three random sets (SI, SII, SIII) of data were generatedin the same manner for each of the 3 observers (O1,O2, O3). Hence, the observers were neither aware of theorientation of the CBCT data sets nor of the order ofthe CBCT images for the other observers.

Three observers (O1, O2, O3) independently plotted25 cephalometric landmarks on 20 CBCT images over4 weeks. After landmark plotting, the randomized sam-ples were decoded and regrouped into as-received andoriented data sets for analysis and comparison. Theadopted methodology of landmark plotting is depicted

ow shows frontal, right lateral, and top views,image. Second row is volume-rendered images(a)-(a'): the horizontal line in frontal view (a)arks. Rotated image in (a') shows both orbitalesontal line in right lateral view (b) does not passimage in (b') shows orbitale and porion in theview (c) does not pass through the midsagittalmidsella. Rotated image in (c') shows it in the

ics January 2017 � Vol 151 � Issue 1

Fig 2. Flowchart for the methodology.

124 Gupta et al

January 2017 � Vol 151 � Issue 1 American

in Figure 2. Three-dimensional coordinates (x, y, and z)of each landmark from all 20 data sets were exportedinto Excel (Microsoft, Redmond, Wash) sheets for theas-received and oriented datasets.

The mean of all 3 observers marking for each coordi-nate of the landmarks was used as the centroid for thatlandmark. Three-dimensional mean Euclidean distances(MED), mean deviations from centroid, and standard de-viations were calculated separately for statistical analysis.The centroid of the 3 observers' markings for each land-mark was considered as the gold standard. The mean ofthe Euclidean distances from the centroid of the ob-servers' markings to each observer's marking representedthe error in the detection of that landmark. The variationsof MED before and after orientation of the 3D imageswere analyzed. MED analyzed the overall error (ratherthan in each axis) in landmark identification. Therefore,the mean deviation was computed to analyze the varia-tion in each axis of all landmarks. Standard deviationsrepresented the variations among all samples of eachlandmark. The standard deviations were computed forboth MED and mean deviation to demonstrate the distri-bution of errors over the number of samples.

MED was calculated as shown in Figure 3. To obtainthe MED, following steps were carried out.

Fig 3. Flowchart for calculating MED.

Journal of Orthodontics and Dentofacial Orthopedics

Gupta et al 125

Step 1: A centroid was calculated with the 3D coordi-nates of all 3 observers for every landmark in each data set

Centroid ðCLÞ

5

O1x1O2

x1O3x

3;O1y1O2

y1O3y

3;O1z1O2

z1O3z

3

!

(1)

where CL (L is any of the 25 landmarks) represents thecoordinate of the centroid, and Oi

j (j 5 x, y, z;i 5 first, second, and third observers) is the coordinatesin x-, y- and z-axes for the 3 observers, respectively.

Step 2: Euclidean distance for each observer from thecentroid for all 3 observers was calculated by using thefollowing Euclidean distance formula.

Distance�Di

L

�5

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi�CLx � Oi

x

�21�CLy � Oi

y

�21�CLz � Oi

z

�2r(2)

where DiL is the Euclidean distance for each observer

from the centroid (i5 first, second, and third observers),and L is any of the 25 landmarks.

CL5 (CLx, CLy, CLz); CLx, CLy and CLz are x-, y- and z-axes coordinates of the centroid for the L landmark, andOij (j5 x, y, z; i5 first, second, and third observers) is the

coordinates in the x-, y- and z-axes for the observers,respectively.

Step 3: The mean distance of the 3 observers wascalculated separately for every landmark for each ofthe 10 data sets and called MED.

Mean Euclidean Distance�MEDs

L

�5

D1L1 D2

L1D3L

3(3)

where MEDsL (L is any of the 25 landmarks; S

[sample] 5 1, 2, 3, ., 10 for the data sets) is the MEDfor every landmark of each data set.

Step 4: Similarly, MED was calculated for all 25 land-marks in the 10 data sets obtained after orientation ofthe images.

Mean differences and standard errors were calculatedon the basis of MED for the 25 landmarks before and af-ter orientation of the images.

Mean difference ðMLÞ

5

�����P10

S5 1

�MEDS

L

�BO �

P10S5 1

�MEDS

L

�AO

10

�����(4)

American Journal of Orthodontics and Dentofacial Orthoped

where ðMEDSLÞBO (L is any of the 25 landmarks; S

[sample] 5 1, 2, 3,.,10 for data sets) is the MED forevery landmark of each data set before orientation. Simi-larly, ðMEDS

LÞAO is the MED for every landmark of eachdata set after orientation.

Standard error ðSELÞ5 standard deviation of XLffiffiffin

p

(5)

where XL 5 ½ðMEDSLÞBO � ðMEDS

LÞAO�, L is any of the25 landmarks; S (sample) 5 1, 2, 3,.,10 for thedifferent data sets of each landmark and n5 10 (numberof samples for a landmark).

Mean deviation and radial mean deviation werecalculated for the 25 landmarks before and after orien-tation of the images.

To evaluate mean deviation and radial mean devia-tion, the following steps were carried out.

Step 1: A centroid was calculated with the 3D coordi-nates of all 3 observers for every landmark for each dataset as shown in equation (1).

Step 2: The absolute mean difference from thecentroid in each coordinate axis was calculated sepa-rately for every landmark in each data set.

Mean Difference from Centroid�DLj

�

5

���CLj � O1j

���1���CLj � O2j

���1���CLj � O3j

���3

(6)

where DLj is mean difference from the centroid (j5 x, y,z) and L is any of the 25 landmarks. CLj is j

th axis centroidvalue for L landmark and Oi

j (j5 x, y, z; i5 first, second,and third observers) is the coordinate of ith observer in jth

axis.Step 3: The mean deviation was calculated as the

mean of the mean difference separately in each coordi-nate axis for all 10 samples.

Mean Deviation�MDLj

�5

D1Lj1 D2

Lj1D3Lj1.1D10

Lj

10(7)

where MDLj (L is any of the 25 landmarks; j 5 x, y, z) isthe mean deviation for every landmark in each axis. Ds

Lj(S [sample] 5 1, 2, 3,., 10 for data sets) is the differ-ence of the centroid of Sth sample in jth axis for Lth land-mark.

Step 4: The radial deviation was calculated by thesquare root of the sum of the squared mean differencefrom the centroid ðDLjÞ in the x-axis, y-axis, and z-axis for each data set.

ics January 2017 � Vol 151 � Issue 1

126 Gupta et al

Radial DeviationðRDLÞ5ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðDLxÞ21

�DLy

�21ðDLzÞ2

q(8)

where RDL is the radial deviation of each landmark foreach data set, and DLj (L is any of the 25 landmarks;j 5 x, y, z) is the difference from centroid for eachdata set.

Step 5: The radial mean deviation was calculated bythe sum of the radial deviations of all 10 data sets foreach landmark.

Radial Mean DeviationðRMDLÞ

5RD1

L1 RD2L1RD3

L1.1RD10L

10(9)

where RMDL is the radial mean deviation for Lth land-mark, and RDi

L is radial deviation of ith data set for Lth

landmark.Step 6: Similarly, the radial mean deviation was

calculated for all the 25 landmarks obtained after orien-tation of the images.

Landmarks were ranked based on their consistency ofmarkings (mean deviation). The rankings are given forthe x-, y-, and z-axes and their radial means separately(Table III).

Statistical analysis

Interobserver reliability among the 3 observers wasevaluated through intraclass correlation coefficients(ICC) using the Statistical Package for the Social Sciences(version 15; SPSS, Chicago, Ill) for the x-, y-, and z-axescoordinates before and after orientation of the CBCT im-ages. A paired t test was applied between the MED foreach of the 25 landmarks before and after orientation.

RESULTS

Interobserver reliability was found to be excellentboth before and after orientation for landmark plot-ting. The ICC values were greater than 0.9 for all land-marks in both situations. The overall mean deviationand standard deviation of landmark plotting areshown in Table III. Influence of orientation was alsoevaluated through a paired t test for Euclidean dis-tance between the as-received and oriented datasets. Corresponding P values for the paired t testwith mean differences and standard errors are listedin Table IV. The paired t test showed statistically insig-nificant differences for all landmarks except 3 land-marks: R1 left (P 5 0.0138), sella (P 5 0.0490), andfrontozygomatic left (P 5 0.0493). Ranking of land-marks based on the consistency in 3 axes and radially

January 2017 � Vol 151 � Issue 1 American

showed that midline landmarks are ranked highestcompared with bilateral landmarks and those oncurved structures (Table III).

DISCUSSION

Three-dimensional imaging modalities have severaladvantages over 2D cephalograms. Although differentCBCT machines have integrated head-positioning de-vices such as chin rests, head straps, and upper lip rests,no universally accepted standard guidelines are fol-lowed. Although natural head position has shownacceptable reproducibility in 2D and 3D cephalometricanalyses, head movement is likely to occur because ofthe long scanning time.35-38 The effect of headorientation has been studied by various authors.20-22,24

El-Beialy et al39 and Hassan et al22 showed that linearmeasurements on volume-rendered images are notaffected by changing the skull orientation.39 Moststudies have evaluated the reproducibility of linear mea-surements on CBCT images with different orienta-tions.22,39,40 The effect of head position on precisionand reproducibility of landmark identification wasstressed by Lou et al.32

Landmark definitions used for 2D cephalometric an-alyses may not apply as such to 3D image analyses. Theaddition of a third dimension leads to further uncer-tainties in landmark plotting. Most conventional 2Dlandmark definitions were defined as a superior-inferior or an anteroposterior point of the anatomicstructures in the sagittal/coronal view. The same land-mark in a 3D image may require a different definitionsince the 3D coordinates locate the exact anatomic lo-cus. Hence, the definition of each landmark needs tobe defined for each multiplanar slice (axial, sagittal,and coronal).17,27,30 This approach could lead tobetter reliability and reproducibility of landmarkidentification.41-43 The measurements made on slices,segmented surfaces, and volume-rendered imagesshowed good accuracy when compared with gold stan-dards (measurements using calipers on a dryskull).21,22,44-46

In 3D images, landmarks are difficult to plot, andthere are more chances of subjective error due to thecurved anatomic surfaces. The orientation of the skullmay influence landmark identification especially onthese curved structures. Sometimes, it may be necessaryto reorient and reformat the 3D image data before land-mark plotting. Apart from this, artifacts in CBCT imagingmay cause difficulty in detecting the landmarks in CBCTimages.47 The CBCT data used in this study were checkedfor major artifacts to reduce the uncertainty in landmarkplotting.

Journal of Orthodontics and Dentofacial Orthopedics

Table III. Ranking of landmarks based on radial and individual coordinate axis consistency

Landmark

Before orientation mean 6 SD (rank) After orientation mean 6 SD (rank)

Radial x y z Radial x y zNasion 0.58 6 0.21 (2) 0.33 6 0.18 (7) 0.23 6 0.11 (6) 0.35 6 0.23 (10) 0.53 6 0.25 (4) 0.20 6 0.07 (2) 0.21 6 0.12 (5) 0.40 6 0.29 (13)Orbitale left 1.19 6 0.60 (19) 1.06 6 0.61 (24) 0.35 6 0.19 (12) 0.32 6 0.17 (6) 1.21 6 0.93 (18) 1.12 6 0.92 (25) 0.31 6 0.19 (9) 0.28 6 0.15 (5)Orbitale right 1.26 6 0.62 (20) 1.16 6 0.63 (25) 0.31 6 0.22 (9) 0.29 6 0.16 (4) 1.23 6 0.79 (21) 1.10 6 0.82 (24) 0.33 6 0.24 (10) 0.30 6 0.11 (7)A-point 0.64 6 0.15 (3) 0.36 6 0.13 (9) 0.32 6 0.13 (10) 0.38 6 0.18 (11) 0.61 6 0.36 (7) 0.21 6 0.10 (3) 0.19 6 0.15 (3) 0.49 6 0.39 (18)Anterior nasal spine 0.70 6 0.25 (8) 0.42 6 0.14 (10) 0.40 6 0.27 (15) 0.31 6 0.18 (5) 0.54 6 0.18 (5) 0.19 6 0.11 (1) 0.38 6 0.22 (14) 0.26 6 0.13 (4)Posterior nasal spine 0.76 6 0.38 (9) 0.29 6 0.25 (5) 0.40 6 0.28 (14) 0.54 6 0.25 (18) 0.65 6 0.38 (10) 0.27 6 0.14 (7) 0.38 6 0.30 (13) 0.39 6 0.31 (12)B-point 0.99 6 0.43 (14) 0.51 6 0.31 (12) 0.16 6 0.13 (1) 0.75 6 0.46 (21) 0.78 6 0.32 (12) 0.29 6 0.23 (10) 0.19 6 0.14 (2) 0.64 6 0.34 (22)Pogonion 0.80 6 0.33 (11) 0.57 6 0.32 (16) 0.19 6 0.16 (5) 0.45 6 0.27 (17) 0.62 6 0.24 (9) 0.27 6 0.21 (9) 0.21 6 0.14 (6) 0.44 6 0.25 (15)Menton 0.84 6 0.39 (12) 0.59 6 0.40 (17) 0.47 6 0.29 (17) 0.21 6 0.06 (2) 0.61 6 0.25 (8) 0.33 6 0.19 (15) 0.43 6 0.29 (15) 0.15 6 0.10 (1)Gnathion 0.79 6 0.33 (10) 0.60 6 0.37 (18) 0.29 6 0.15 (8) 0.33 6 0.18 (7) 0.59 6 0.17 (6) 0.29 6 0.21 (12) 0.34 6 0.16 (12) 0.30 6 0.16 (6)Gonion left 1.10 6 0.43 (16) 0.56 6 0.24 (15) 0.49 6 0.27 (20) 0.74 6 0.42 (20) 1.38 6 0.43 (22) 0.56 6 0.25 (18) 0.65 6 0.27 (22) 1.01 6 0.45 (24)Gonion right 1.33 6 0.37 (22) 0.65 6 0.18 (19) 0.79 6 0.35 (23) 0.77 6 0.36 (23) 1.21 6 0.40 (19) 0.71 6 0.26 (21) 0.66 6 0.40 (23) 0.62 6 0.31 (21)Condylion left 1.14 6 0.51 (17) 0.89 6 0.36 (20) 0.48 6 0.34 (19) 0.45 6 0.34 (16) 1.11 6 0.60 (16) 0.64 6 0.44 (19) 0.62 6 0.43 (21) 0.53 6 0.42 (19)Condylion right 1.19 6 0.83 (18) 0.89 6 0.80 (21) 0.55 6 0.40 (22) 0.39 6 0.25 (12) 0.97 6 0.47 (14) 0.65 6 0.55 (20) 0.49 6 0.27 (17) 0.36 6 0.13 (11)R1 left 0.96 6 0.27 (13) 0.26 6 0.10 (3) 0.28 6 0.19 (7) 0.82 6 0.36 (25) 1.54 6 0.72 (23) 0.36 6 0.20 (16) 0.33 6 0.15 (11) 1.43 6 0.74 (25)R1 right 1.02 6 0.27 (15) 0.51 6 0.22 (13) 0.17 6 0.13 (2) 0.81 6 0.34 (24) 1.07 6 0.39 (15) 0.32 6 0.17 (13) 0.31 6 0.17 (8) 0.93 6 0.43 (23)Sella 0.68 6 0.22 (7) 0.52 6 0.27 (14) 0.18 6 0.13 (4) 0.34 6 0.13 (8) 0.50 6 0.17 (2) 0.27 6 0.13 (8) 0.20 6 0.13 (4) 0.33 6 0.17 (9)Basion 0.66 6 0.34 (5) 0.44 6 0.23 (11) 0.34 6 0.25 (11) 0.26 6 0.23 (3) 0.51 6 0.20 (3) 0.32 6 0.14 (14) 0.29 6 0.24 (7) 0.17 6 0.11 (2)Zygomatic point left 1.53 6 0.79 (24) 0.19 6 0.10 (1) 1.43 6 0.82 (24) 0.42 6 0.18 (14) 1.84 6 1.12 (25) 0.26 6 0.19 (5) 1.74 6 1.17 (25) 0.32 6 0.21 (8)Zygomatic point right 1.71 6 0.51 (25) 0.34 6 0.37 (8) 1.55 6 0.58 (25) 0.40 6 0.23 (13) 1.81 6 0.97 (24) 0.26 6 0.15 (4) 1.69 6 1.04 (24) 0.44 6 0.21 (14)Frontozygomatic left 0.66 6 0.18 (4) 0.30 6 0.15 (6) 0.41 6 0.20 (16) 0.34 6 0.19 (9) 0.80 6 0.20 (13) 0.46 6 0.19 (17) 0.49 6 0.18 (18) 0.34 6 0.23 (10)Frontozygomatic right 0.67 6 0.30 (6) 0.27 6 0.19 (4) 0.39 6 0.27 (13) 0.43 6 0.14 (15) 0.75 6 0.52 (11) 0.29 6 0.18 (11) 0.45 6 0.26 (16) 0.46 6 0.49 (16)Jugal point left 1.40 6 0.43 (23) 0.97 6 0.35 (23) 0.50 6 0.39 (21) 0.76 6 0.38 (22) 1.15 6 0.63 (17) 0.80 6 0.53 (22) 0.55 6 0.48 (19) 0.46 6 0.27 (17)Jugal point right 1.29 6 0.49 (21) 0.93 6 0.46 (22) 0.48 6 0.36 (18) 0.64 6 0.28 (19) 1.21 6 0.74 (20) 0.86 6 0.60 (23) 0.56 6 0.45 (20) 0.54 6 0.36 (20)Bolton 0.38 6 0.11 (1) 0.22 6 0.13 (2) 0.18 6 0.12 (3) 0.21 6 0.09 (1) 0.42 6 0.11 (1) 0.26 6 0.14 (6) 0.15 6 0.06 (1) 0.24 6 0.12 (3)

All measurements in millimeters; ranks 1 to 25 reflect descending order of consistency.

Gupta

etal

127

American

JournalofOrthodontics

andDentofacialO

rthopedicsJanuary

2017�Vol151�

Issue1

Table IV. MED and standard deviations before orientation (BO) and after orientation (AO) of CBCT images with cor-responding P values for t tests, mean differences, and standard errors

LandmarkMED 6 SD

for BOMED 6 SDfor AO

P value(t test)

Meandifference SE

Nasion 0.62 6 0.24 0.56 6 0.25 0.620893 0.05 0.11Orbitale left 1.23 6 0.60 1.24 6 0.95 0.980451 0.01 0.37Orbitale right 1.31 6 0.61 1.26 6 0.79 0.767966 0.05 0.16A-point 0.69 6 0.16 0.64 6 0.35 0.580744 0.05 0.09Anterior nasal spine 0.74 6 0.25 0.57 6 0.19 0.078452 0.16 0.08Posterior nasal spine 0.80 6 0.42 0.68 6 0.38 0.349206 0.12 0.12B-Point 1.04 6 0.43 0.81 6 0.33 0.311190 0.23 0.21Pogonion 0.83 6 0.35 0.66 6 0.24 0.271658 0.18 0.15Menton 0.89 6 0.39 0.64 6 0.25 0.100030 0.25 0.14Gnathion 0.83 6 0.33 0.62 6 0.18 0.173552 0.20 0.14Gonion left 1.14 6 0.46 1.45 6 0.46 0.234335 0.31 0.24Gonion right 1.38 6 0.38 1.26 6 0.41 0.443102 0.12 0.14Condylion left 1.19 6 0.53 1.16 6 0.60 0.803575 0.02 0.09Condylion right 1.25 6 0.84 1.01 6 0.47 0.307026 0.23 0.21R1 left 1.00 6 0.26 1.57 6 0.72 0.013887* 0.57 0.19R1 right 1.05 6 0.28 1.14 6 0.39 0.683621 0.08 0.20Sella 0.71 6 0.22 0.53 6 0.18 0.049053* 0.18 0.08Basion 0.69 6 0.35 0.53 6 0.20 0.200080 0.16 0.11Zygomatic point left 1.57 6 0.77 1.87 6 1.13 0.493591 0.30 0.42Zygomatic point right 1.76 6 0.51 1.87 6 0.94 0.735286 0.12 0.33Frontozygomatic left 0.69 6 0.18 0.84 6 0.20 0.049376* 0.15 0.07Frontozygomatic right 0.69 6 0.29 0.78 6 0.52 0.704458 0.09 0.23Jugal point left 1.43 6 0.43 1.22 6 0.69 0.474210 0.21 0.29Jugal point right 1.31 6 0.50 1.26 6 0.79 0.882382 0.05 0.33Bolton 0.40 6 0.11 0.45 6 0.12 0.269118 0.04 0.04

*Statistically significant difference (P\ 0.05).

128 Gupta et al

The CBCT data sets used in this study were anony-mized and blinded to prevent bias in landmark plotting.A dual verification approach was used for landmarkidentification. Landmark plotting was carried out onmultiplanar reconstruction slices and verified throughvolume-rendered images and vice versa. If any observerhad difficulty in locating any landmark in a specific mul-tiplanar reconstruction view such as sagittal, he or sheused another multiplanar reconstruction view: ie,coronal or axial with the volume-rendered image forplotting.

The results of this study showed high interobserverreliability (ICC .0.9) in 3D landmark plotting. Parket al48 reported similar results with 3D imaging, and alllandmarks were reproduced with zero interobserver er-ror. de Oliveira et al28 reported good interobserver reli-ability with ICC values greater than 0.9 for 65.55% ofthe landmarks studied.39 Excellent agreement wasobserved among observers (ICC .0.9) for both beforeand after orientation of 3D images. It shows that if thelandmarks are properly defined in 3 dimensions andare marked on the most appropriate multiplanar recon-struction slices/volume-rendered images, the headorientation may not be a significant factor affecting

January 2017 � Vol 151 � Issue 1 American

the precision of landmark plotting. Of 25 landmarks,11 landmarks were plotted on volume-rendered imagesdirectly and confirmed through multiplanar reconstruc-tion slices, and the remaining 14 landmarks were plottedin reverse order. Landmarks plotted directly on renderedimages showed good reproducibility in both situations;this was contrary to the studies by Periago et al46 andNeiva et al.26

MED was calculated before and after reorientation ofthe images, and Euclidean distance (mean deviation) ofeach axis was also calculated from the centroid.49 A sta-tistically significant difference was found for R1 left,Sella and frontozygomatic left landmarks. Midline land-marks showed greater consistency and precision than didbilateral landmarks (Table III). The mean radial errors of6 bilateral landmarks—orbitale, gonion, condylion, R1,zygomatic point, and jugal point—were greater than1 mm, and all others were within 1 mm. This error wasfound for both as-received and oriented images exceptfor before-orientation R1 left and after-orientation con-dylion right landmarks, where the numeric values weremarginally below 1 mm (Table III). It appears that refer-ence points located on a prominence or curvature havegreater variability compared with landmarks defined at

Journal of Orthodontics and Dentofacial Orthopedics

Gupta et al 129

plane positions.9 Since, orbitale is the inferior-mostpoint on the curvature of the inferior orbital margin,the perception of landmark identification may changewith the orientation of the CBCT image. There are chan-ces for subjective errors in landmark plotting. Also, chan-ces of errors in plotting gonion, condylion, R1, andzygomatic and jugal points are high for similar reasons.

For a more thorough evaluation of consistency andprecision, overall error of each landmark is shownwith the individual axis error: x-, y-, and z-axes(Table III). These 6 bilateral landmarks, which showedgreater than a 1-mm radial error, did not follow thecommon trend. A landmark with more error in 1 axiswas more precisely identified in another axis. However,a common pattern observed was that an axis that wasparallel to the curvature of the anatomic structurehad greater errors than the others. For orbitale, the x-axis, which is parallel to the inferior orbital margin,had a higher error than the other 2 axes. In case ofthe zygomatic landmark, it was the y-axis; for R1, itwas the z-axis. Gonion and jugal landmarks do nothave any parallel axes passing through their curvaturedue to complex anatomic structures; hence, gonionhad an error distributed evenly in all 3 axes, whereas ju-gal had a higher error value in the x-axis. B-point hada near 1-mm radial error, although the z-axis had ahigher contribution for a similar reason. Sella is afloating point that can only be plotted through multi-planar reconstruction slices by estimating themidsagittal plane of the pituitary fossa. Thus, sella isdifficult to plot on as-received images compared withoriented images. Therefore, precision of this landmarkis improved after orientation.

The slice thickness used in this study was equal to thevoxel size of that particular CBCT image. Landmark plot-ting onmultiplanar reconstruction slices was more accu-rate when a smaller slice thickness was used.21,22,48

Bilateral anatomic landmarks cannot be visualized onthe same slice (axial) when a small slice thickness isused. Scrolling back and forth between anteroposteriorand right or left through slices is necessary sometimesand time-consuming.22 Landmarks on the midsagittalplane were easy to plot after orientation. Before orienta-tion, midsagittal plane structures were found in differentsagittal slices even if there was no skeletal deformity forthe same reason. Care should be taken when plottingmidsagittal structures in case of a skeletal deformity.Plotting of landmarks first on volume-rendered imagesand then verified on the multiplanar reconstruction sli-ces may solve this problem partially. Plotting onvolume-rendered images is limited to surface landmarksonly; interior anatomic landmarks must be marked onmultiplanar reconstruction slices.

American Journal of Orthodontics and Dentofacial Orthoped

In 3D images, bilateral landmarks are more difficultto plot because of varied perceptions while marking oncurved anatomic surfaces. To improve the accuracy ofthese landmarks, proper anatomic definitions for eachlandmark should be followed while marking on multi-planar reconstruction views as well as volume-rendered images. Landmark location should also bereconfirmed on volume-rendered images if marked onmultiplanar reconstructions and vice versa.

During the calibration and training process, the ob-servers were reminded of the importance of checkingall 3 multiplanar reconstruction slices with rendered im-ages before moving for the identification of the nextlandmarks. In the oriented images, both midline andbilateral structures were easy to mark. When the imagewas properly oriented, the effort of scrolling back andforth through the slices was reduced, and it was lesstime-consuming. Special precautions should be takenwhile marking points on the prominences or broadcurved surfaces. We did not evaluate the effect of orien-tation on the efficiency of landmark plotting withrespect to the time factor; this seems to be another inter-esting area of investigation through a similar kind ofexperimentation.

CONCLUSIONS

1. The orientation of CBCT image has a negligible ef-fect on the precision of landmark plotting.

2. Midline landmarks are more consistently identifiedthan bilateral landmarks both before and afterorientation.

3. Landmarks on broad curved structures have moreerrors than others.

REFERENCES

1. Farronato G, Perillo L, Bellincioni F, Briguglio F, Farronato D,Dominici AD. Direct 3D cephalometric analysis performed onCBCT. J Inform Technol Softw Eng 2012;2:107.

2. Scarfe WC, Farman AG, Sukovic P. Clinical applications of cone-beam computed tomography in dental practice. J Can Dent Assoc2006;72:75-80.

3. Yu SH, NahmDS, Baek SH. Reliability of landmark identification onmonitor-displayed lateral cephalometric images. Am J OrthodDentofacial Orthop 2008;133:790.e1-6.

4. Holberg C, Steinhauser S, Geis P, Rudzki-Janson I. Cone-beamcomputed tomography in orthodontics: benefits and limitations.J Orofac Orthop 2005;66:434-44.

5. Cevidanes LH, Oliveira AE, Grauer D, Styner M, Proffit WR. Clinicalapplication of 3D imaging for assessment of treatment outcomes.Semin Orthod 2011;17:72-80.

6. Moshiri M, Scarfe WC, Hilgers ML, Scheetz JP, Silveira AM,Farman AG. Accuracy of linear measurements from imaging plateand lateral cephalometric images derived from cone-beamcomputed tomography. Am J Orthod Dentofacial Orthop 2007;132:550-60.

ics January 2017 � Vol 151 � Issue 1

130 Gupta et al

7. Cattaneo PM, Bloch CB, Calmar D, Hjortshoj M, Melsen B. Com-parison between conventional and cone-beam computedtomography-generated cephalograms. Am J Orthod DentofacialOrthop 2008;134:798-802.

8. Liedke GS, Delamare EL, Vizzotto MB, da Silveira HL, Prietsch JR,Dutra V, et al. Comparative study between conventional and conebeam CT-synthesized half and total skull cephalograms. Dento-maxillofac Radiol 2012;41:136-42.

9. Baumrind S, Frantz RC. The reliability of head film measurements.1. Landmark identification. Am J Orthod 1971;60:111-27.

10. Chien P, Parks E, Eraso F, Hartsfield J, Roberts W, Ofner S. Com-parison of reliability in anatomical landmark identification usingtwo-dimensional digital cephalometrics and three-dimensionalcone beam computed tomography in vivo. Dentomaxillofac Radiol2009;38:262-73.

11. Kragskov J, Bosch C, Gyldensted C, Sindet-Pedersen S. Comparisonof the reliability of craniofacial anatomic landmarks based oncephalometric radiographs and three-dimensional CT scans. CleftPalate Craniofac J 1997;34:111-6.

12. Tulunoglu O, Esenlik E, Gulsen A, Tulunoglu I. A comparison ofthree-dimensional and two-dimensional cephalometric evalua-tions of children with cleft lip and palate. Eur J Dent 2011;5:451-8.

13. Park CS, Park JK, Kim H, Han SS, Jeong HG, Park H. Comparison ofconventional lateral cephalograms with corresponding CBCT ra-diographs. Imaging Sci Dent 2012;42:201-5.

14. Nur M, Kayipmaz S, Bayram M, Celikoglu M, Kilkis D, Sezgin OS.Conventional frontal radiographs compared with frontal radio-graphs obtained from cone beam computed tomography. AngleOrthod 2012;82:579-84.

15. Olmez H, Gorgulu S, Akin E, Bengi AO, Tekdemir I, Ors F. Measure-ment accuracy of a computer-assisted three-dimensional analysisand a conventional two-dimensional method. Angle Orthod 2011;81:375-82.

16. Adams GL, Gansky SA, Miller AJ, Harrell WE Jr, Hatcher DC.Comparison between traditional 2-dimensional cephalometryand a 3-dimensional approach on human dry skulls. Am J OrthodDentofacial Orthop 2004;126:397-409.

17. Cattaneo PM, Melsen B. The use of cone-beam computed tomog-raphy in an orthodontic department in between research and dailyclinic. World J Orthod 2008;9:269-82.

18. Naji P, Alsufyani NA, Lagrav�ere MO. Reliability of anatomic struc-tures as landmarks in three-dimensional cephalometric analysisusing CBCT. Angle Orthod 2014;84:762-72.

19. Rossini G, Cavallini C, Cassetta M, Barbato E. 3D cephalometricanalysis obtained from computed tomography. Review of the liter-ature. Ann Stomatol (Roma) 2011;2:31-9.

20. Lagrav�ere MO, Major PW. Proposed reference point for 3-dimensional cephalometric analysis with cone-beam computerizedtomography. Am J Orthod Dentofacial Orthop 2005;128:657-60.

21. Ludlow JB, Laster WS, See M, Bailey LJ, Hershey HG. Accuracy ofmeasurements of mandibular anatomy in cone beam computedtomography images. Oral Surg Oral Med Oral Pathol Oral RadiolEndod 2007;103:534-42.

22. Hassan B, van der Stelt P, Sanderink G. Accuracy of three-dimensional measurements obtained from cone beam computedtomography surface-rendered images for cephalometric analysis:influence of patient scanning position. Eur J Orthod 2009;31:129-34.

23. Tomasi C, Bressan E, Corazza B, Mazzoleni S, Stellini E, Lith A.Reliability and reproducibility of linear mandible measurementswith the use of a cone-beam computed tomography and two ob-ject inclinations. Dentomaxillofac Radiol 2011;40:244-50.

January 2017 � Vol 151 � Issue 1 American

24. Cevidanes L, Oliveira AE, Motta A, Phillips C, Burke B, Tyndall D.Head orientation in CBCT-generated cephalograms. Angle Orthod2009;79:971-7.

25. Berco M, Rigali PH Jr, Miner RM, DeLuca S, Anderson NK, Will LA.Accuracy and reliability of linear cephalometric measurementsfrom cone-beam computed tomography scans of a dry humanskull. Am J Orthod Dentofacial Orthop 2009;136:17-8.

26. Neiva MB, Soares �AC, Lisboa Cde O, Vilella Ode V, Motta AT. Eval-uation of cephalometric landmark identification on CBCT multi-planar and 3D reconstructions. Angle Orthod 2015;85:11-7.

27. van Vlijmen OJ, Berge SJ, Swennen GR, Bronkhorst EM,Katsaros C, Kuijpers-Jagtman AM. Comparison of cephalometricradiographs obtained from cone-beam computed tomographyscans and conventional radiographs. J Oral Maxillofac Surg2009;67:92-7.

28. de Oliveira AE, Cevidanes LH, Phillips C, Motta A, Burke B,Tyndall D. Observer reliability of three-dimensional cephalometriclandmark identification on cone-beam computerized tomography.Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2009;107:256-65.

29. Damstra J, Fourie Z, Huddleston Slater JJ, Ren Y. Reliability andthe smallest detectable difference of measurements on 3-dimensional cone-beam computed tomography images. Am J Or-thod Dentofacial Orthop 2011;140:e107-14.

30. Oz U, Orhan K, Abe N. Comparison of linear and angular measure-ments using two-dimensional conventional methods andthree-dimensional cone beam CT images reconstructed from avolumetric rendering program in vivo. Dentomaxillofac Radiol2011;40:492-500.

31. Zamora N, Llamas JM, Cibrian R, Gandia JL, Paredes V. A study onthe reproducibility of cephalometric landmarks when undertakinga three-dimensional (3D) cephalometric analysis. Med Oral PatolOral Cir Bucal 2012;17:e678-88.

32. Lou L, Lagrav�ere MO, Compton S, Major PW, Flores-Mir C. Accu-racy of measurements and reliability of landmark identificationwith computed tomography (CT) techniques in the maxillofacialarea: a systematic review. Oral Surg Oral Med Oral Pathol Oral Ra-diol Endod 2007;104:402-11.

33. Gupta A, Kharbanda OP, Sardana V, Balachandran R, Sardana HK.A knowledge-based algorithm for automatic detection of cephalo-metric landmarks on CBCT images. Int J Comput Assist Radiol Surg2015;10:1737-52.

34. Gupta A, Kharbanda OP, Sardana V, Balachandran R, Sardana HK.Accuracy of 3D cephalometric measurements based on an auto-matic knowledge-based landmark detection algorithm. Int J Com-put Assist Radiol Surg 2016;11:1297-309.

35. Lundstrom A, Lundstrom F, Lebret LM, Moorrees CF. Natural headposition and natural head orientation: basic considerations incephalometric analysis and research. Eur J Orthod 1995;17:111-20.

36. Cooke MS, Wei SH. The reproducibility of natural head posture: amethodological study. Am J Orthod Dentofacial Orthop 1988;93:280-8.

37. Peng L, Cooke MS. Fifteen-year reproducibility of natural headposture: a longitudinal study. Am J Orthod Dentofacial Orthop1999;116:82-5.

38. Weber DW, Fallis DW, Packer MD. Three-dimensional reproduc-ibility of natural head position. Am J Orthod Dentofacial Orthop2013;143:738-44.

39. El-Beialy AR, Fayed MS, El-Bialy AM, Mostafa YA. Accuracy andreliability of cone-beam computed tomography measurements:Influence of head orientation. Am J Orthod Dentofacial Orthop2011;140:157-65.

Journal of Orthodontics and Dentofacial Orthopedics

Gupta et al 131

40. Sheikhi M, Ghorbanizadeh S, Abdinian M, Goroohi H, Badrian H.Accuracy of linear measurements of galileos cone beam computedtomography in normal and different head positions. Int J Dent2012;2012:214954.

41. Frongia G, Piancino MG, Bracco AA, Crincoli V, Debernardi CL,Bracco P. Assessment of the reliability and repeatability of land-marks using 3-D cephalometric software. Cranio 2012;30:255-63.

42. Lagrav�ere MO, Low C, Flores-Mir C, Chung R, Carey JP, Heo G,et al. Intraexaminer and interexaminer reliabilities of landmarkidentification on digitized lateral cephalograms and formatted3-dimensional cone-beam computerized tomography images.Am J Orthod Dentofacial Orthop 2010;137:598-604.

43. Ludlow JB, Gubler M, Cevidanes L, Mol A. Precision of cephalo-metric landmark identification: cone-beam computed tomographyvs conventional cephalometric views. Am J Orthod Dentofacial Or-thop 2009;136:312-3.

44. Ballrick JW, Palomo JM, Ruch E, Amberman BD, Hans MG. Imagedistortion and spatial resolution of a commercially available cone-

American Journal of Orthodontics and Dentofacial Orthoped

beam computed tomography machine. Am J Orthod DentofacialOrthop 2008;134:573-82.

45. Lascala CA, Panella J, Marques MM. Analysis of the accuracy oflinear measurements obtained by cone beam computed tomogra-phy (CBCT-NewTom). Dentomaxillofac Radiol 2004;33:291-4.

46. Periago DR, Scarfe WC, Moshiri M, Scheetz JP, Silveira AM,Farman AG. Linear accuracy and reliability of cone beam CTderived 3-dimensional images constructed using an orthodonticvolumetric rendering program. Angle Orthod 2008;78:387-95.

47. Schulze R, Heil U, Gross D, Bruellmann DD, Dranischnikow E,Schwanecke U, et al. Artefacts in CBCT: a review. DentomaxillofacRadiol 2011;40:265-73.

48. Park SH, Yu HS, Kim KD, Lee KJ, Baik HS. A proposal for a newanalysis of craniofacial morphology by 3-dimensional computedtomography. Am J Orthod Dentofacial Orthop 2006;129:e23-34.

49. Schlicher W, Nielsen I, Huang JC, Maki K, Hatcher DC, Miller AJ.Consistency and precision of landmark identification in three-dimensional cone beam computed tomography scans. Eur J Or-thod 2012;34:263-75.

ics January 2017 � Vol 151 � Issue 1

本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。

学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，

提供一站式文献检索和下载服务”的24 小时在线不限IP

图书馆。

图书馆致力于便利、促进学习与科研，提供最强文献下载服务。

图书馆导航：

图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具