Eur J Oral Sci 2009; 117: 300305 Printed in Singapore.

All rights reserved

2009 The Authors. Journal compilation 2009 Eur J Oral Sci

European Journal of Oral Sciences

A comparison between twodimensional and three-dimensional cephalometry on frontal radiographs and on cone beam computed tomography scans of human skulls
van Vlijmen OJC, Maal TJJ, Berge SJ, Jagtman AM. A comparison between cephalometry on frontal radiographs and on human skulls. Eur J Oral Sci 2009; 117: compilation 2009 Eur J Oral Sci Bronkhorst EM, Katsaros C, Kuijperstwo-dimensional and three-dimensional cone beam computed tomography scans of 300305. 2009 The Authors. Journal

Olivier J. C. van Vlijmen1,2, Thomas J. J. Maal2,3, Stefaan J. Berg2,3, Ewald M. Bronkhorst4, Christos Katsaros1,5, Anne Marie KuijpersJagtman1,2
1

Department of Orthodontics and Oral Biology, Radboud University Nijmegen Medical Centre, Nijmegen, the Netherlands; 23D-Facial Imaging Research Group Nijmegen-Bruges (3D-FIRG), the Netherlands; 3Department of Oral and Maxillofacial surgery, Radboud University Nijmegen Medical Centre, Nijmegen, the Netherlands; 4Department of Preventive and Curative Dentistry, Radboud University Nijmegen Medical Centre, Nijmegen, the Netherlands; 5Department of Orthodontics and Dentofacial Orthopedics, University of Bern, Bern, Switzerland

The aim of this study was to evaluate whether measurements performed on conventional frontal radiographs are comparable to measurements performed on threedimensional (3D) models of human skulls derived from cone beam computed tomography (CBCT) scans and if the latter can be used in longitudinal studies. Cone beam computed tomography scans and conventional frontal cephalometric radiographs were made of 40 dry human skulls. From the CBCT scan a 3D model was constructed. Standard cephalometric software was used to identify landmarks and to calculate ratios and angles. The same operator identied 10 landmarks on both types of cephalometric radiographs, and on all images, ve times with a time interval of 1 wk. Intra-observer reliability was acceptable for all measurements. There was a statistically signicant and clinically relevant dierence between measurements performed on conventional frontal radiographs and on 3D CBCT-derived models of the same skull. There was a clinically relevant dierence between angular measurements performed on conventional frontal cephalometric radiographs, compared with measurements performed on 3D models constructed from CBCT scans. We therefore recommend that 3D models should not be used for longitudinal research in cases where there are only two-dimensional (2D) records from the past.

Anne Marie Kuijpers-Jagtman, Department of Orthodontics and Oral Biology, Radboud University Nijmegen Medical Centre, PO Box 9101, 6500 HB Nijmegen, the Netherlands Telefax: +31243540631 E-mail: orthodontics@dent.umcn.nl Key words: cephalometry; cone beam computed tomography; maxillofacial; orthodontics; three-dimensional imaging Accepted for publication February 2009

Since the early 1930s, both frontal and lateral cephalometry has been utilized for analyzing both maxillofacial and orthodontic deformities, especially to evaluate growth and/or treatment changes (1, 2). The frontal cephalogram, however, has not routinely been used in orthodontic treatment planning (3), possibly because of limitations and diculties in analysis of the frontal cephalogram, such as identifying landmarks of superimposed structures, errors in reproducing head posture, and relatively low added value of frontal radiographs because they only provide additional information about asymmetries and width of the jaws (4). For treatment planning of surgical cases, frontal radiographs can be benecial (e.g. in patients treated with surgically assisted rapid maxillary expansion). During the last years cone beam computed tomography (CBCT) technology has been increasing in popularity. A CBCT scan exposes the patient to less radiation than a multislice computed tomography (MSCT) scan (5, 6) and can therefore be used for a wider range of

patients. Cone beam computed tomography uses a different type of acquisition compared with conventional MSCT. The X-ray source produces a cone-shaped X-ray beam. This makes it possible to capture the image in one rotation, instead of capturing every individual slice separately, as is the case in MSCT. Although a CBCT scan has reduced radiation in comparison to a MSCT scan, a CBCT scan is still not suitable for the standard orthodontic patient (7). However, for patients with craniofacial anomalies and orofacial clefts, or patients requiring orthognatic surgery, CBCT has many benets. Recently it has been shown (811) that conventional lateral cephalometric radiographs, which may be considered as the gold standard, can be compared with constructed lateral cephalometric radiographs from CBCT scans and thus the latter can be used for longitudinal research in cases where there are conventional radiographic records from the past. In another study from our group that investigated whether the same holds true for

A comparison between 2D and 3D cephalometry

301

frontal cephalometric radiographs, some dierences between the two methods were found. These dierences were caused by dierent positioning of the skulls in both devices. When constructing two-dimensional (2D) images from three-dimensional (3D) data, the positioning of the patient is of utmost importance, but when performing 3D cephalometry (12) the positioning of the patient is of no importance. As 3D scans, and therewith 3D cephalometry, are rapidly gaining popularity, there can be a problem in longitudinal studies because older records will be conventional 2D radiographs. Therefore, the aim of this study was to evaluate whether measurements on conventional frontal radiographs are comparable to measurements on 3D models of human skulls derived from CBCT scans and in addition whether the latter can be used in longitudinal studies.

Material and methods

The sample consisted of 40 dry human skulls obtained from the collection of the Department of Orthodontics and Oral Biology of the Radboud University Nijmegen Medical Centre. The skulls were selected from a larger sample according to the following criteria: the presence of permanent upper and lower incisors; the presence of rst permanent upper and lower molars; and the presence of a reproducible, stable occlusion. The mandible was related to the skull based on the position of the condyle in the fossa and maximum occlusal interdigitation. The mandibular position was xed with broad tape from the ipsilateral temporal bone around the horizontal ramus of the mandible to the contralateral temporal bone. Radiography Each skull was positioned in the cephalostat (Cranex Tome Ceph; Soredex, Tuusula, Finland) by xing it between the ear rods. The ear rods were placed in the pori acoustici externi and the Frankfurt Horizontal plane was placed horizontally, parallel to the oor. Cephalometric radiographs were taken according to the following radiographic settings. For larger skulls (n = 30), the adult settings were chosen: 70 kV, 10 mA, 0.6 s. For the smaller skulls (n = 10), the child settings were chosen: 70 kV, 10 mA, 0.5 s (Fig. 1A). The viewbox software (dHAL Software, Kissia, Greece) was used to identify conventional cephalometric hard tissue landmarks and to calculate distances and angles. The same skulls were placed in the I-Cat CBCT scanner (Imaging Sciences, Hateld, PA, USA) on a foam platform with the Frankfurt Horizontal plane parallel to the oor. The skulls were placed in the centre of the CBCT scanner using the midline light beam to coincide with the midsagittal plane. The CBCT scan was taken for all skulls in the extended height mode: 129 kVp, 47.74 mA, 40 s with a resolution of 0.4 voxel. A 3D model of each skull was constructed (Fig. 1B) from the CBCT data using maxilim (Medicim, Sint-Niklaas, Belgium). This same software was used to analyse the constructed 3D models cephalometrically. Cephalometry For the cephalometric analysis, 10 conventional hard-tissue cephalometric landmarks (Fig. 2 and Table 1) were identi-

Fig. 1. Cephalometric frontal radiographs of the same skull. (A) Conventional radiograph. (B) Cone beam computed tomography (CBCT)-constructed three-dimensional (3D) model.

ed. Twelve widely used cephalometric variables (nine angles and three linear ratios) were calculated in viewbox for the 2D measurements and in maxilim for the 3D measurements (Table 2). All images were measured in random order. Statistical analysis For both the conventional frontal cephalometric radiographs and the CBCT-constructed frontal cephalometric radiographs, the same operator (OV) marked the landmarks on all 80 images ve times, each time with a time interval of 1 wk. The mean value and standard deviation of these ve measurements were used for the statistical analysis. The intra-observer reliability was calculated by means of the Pearson correlation coecient for the rst and second measurements. For each measurement, the standard error was calculated and compared with the standard error of the same measurement in the other group. Paired t-tests were performed to compare the means of corresponding

302

van Vlijmen et al.

A

measurements on the two cephalometric radiographs of the same skull.

Results
Intra-observer reliability for both the conventional frontal cephalometric radiographs and the CBCTconstructed 3D model was acceptable for most measurements. The correlation coecient between the rst and second measurements ranged from 0.23 to 0.99 (average = 0.71) for the conventional frontal radiographs and from 0.42 to 0.93 (average = 0.79) for the 3D models (see Table 3). The duplicate measurement error for the CBCTconstructed 3D model was signicantly smaller for six measurements (R CR-L, R J-ANS, Jl/Jr/Ll, Jl/Jr/Lr, Ll-CR-Lr, ME/AGl/CR; see Table 2 for denitions of these anatomical variables) compared with the standard error of the conventional frontal cephalometric radiographs. For the conventional frontal radiographs this was the case for one measurement (ME/AGr/CR). For ve measurements there were no statistically signicant dierences between both types of radiographs (AGlAGr/Ll-Lr, R CR-L, R ME-AG, AGl/AGr/Ll, AGl/

Fig. 2. (A) Schematic representation of the landmarks that were identied. (B) Angles that were measured. The landmarks are dened in Table 1.

Table 1 Anatomical landmarks used in this study

AGl AGr AR Antegonion left Antegonion right Articulare The antoganial notch at the lateral inferior margin of the antegonial protuberances on the left side The antoganial notch at the lateral inferior margin of the antegonial protuberances on the right side The point of intersection of the dorsal contours of the processus articularis mandibulae and the pharyngeal part of the clivus Both on the left and on the right side At the jugal process the intersection of the outline of the maxillary tuberosity and the zygomatic buttress on the left side At the jugal process the intersection of the outline of the maxillary tuberosity and the zygomatic buttress on the right side The most lateral point of the orbital cavity on the left side The most lateral point of the orbital cavity on the right side The tip of the anterior nasal spine Most superior point at its intersection with the sphenoid The most inferior point of the symphysis of the mandible, as seen on the lateral jaw projection

Jl Jr Ll Lr ANS CR ME

Jugale left Jugale right Lateral orbital margin left Lateral orbital margin right Anterior nasal spine Crista Galli Menton

Table 2 Cephalometric variables

R ME-AG R J-ANS R CR-L ME/AGl/CR ME/AGr/CR AGl/ME/AGr Jl/Jr/Ll Jl/Jr/Lr AGl/AGr/Ll AGl/AGr/Lr AGl-AGr/Ll-Lr Ll/CR/Lr Ratio between line ME-AG left and line ME-AG right Ratio between line J left-ANS and line J right-ANS Ratio between line CR-L left and line CR-L right Angle between line ME-left AG and line left AG-CR Angle between line ME-right AG and line right AG-CR Angle between line ME-left AG and line ME-right AG Angle between the line J left-J right and line J left-L left Angle between the line J left-J right and line J right-L right Angle between the line AG left-AG right and line AG left-L left Angle between the line AG left-AG right and line AG right-L right Angle between the line AG left-AG right and line left L-right L Angle between the line left L-CR and line right L-CR

A comparison between 2D and 3D cephalometry

Table 3

303

Intra-observer reliability expressed as Pearsons correlation coecient for rst and second measurements and the duplicate measurement error for 40 cases with 95% condence interval (CI)
Conventional Duplicate measurement error Reliability R CR-L: R J-ANS: R ME-AG: AGl/AGr/Ll-Lr () AGl/AGr/Ll () AGl/AGr/Lr () AGl/ME/AGr: () Jl/Jr/LL () Jl/Jr/Lr () Ll/CR/Lr () ME/AGl/CR () ME/AGr/CR () 0.23 0.49 0.84 0.67 0.96 0.89 0.99 0.51 0.64 0.34 0.99 0.99 Error 0.02 0.04 0.03 0.95 0.72 0.72 1.20 2.28 2.20 4.18 0.89 0.66 95% CI SE 0.020.02 0.040.04 0.030.04 0.841.05 0.640.80 0.640.80 1.061.33 2.032.54 1.952.44 3.714.64 0.790.99 0.590.73 Reliability 0.88 0.62 0.42 0.44 0.81 0.82 0.92 0.89 0.87 0.93 0.93 0.91 3D model Duplicate measurement error Error 0.02 0.03 0.03 0.94 0.63 0.72 1.16 1.09 0.92 1.22 0.76 0.88 95% CI SE 0.010.20 0.030.04 0.030.03 0.831.04 0.560.70 0.640.80 1.031.28 0.961.21 0.821.02 1.091.36 0.670.84 0.780.98 P-value for error 0.021 0.009 0.105 0.439 0.052 0.489 0.333 < 0.001 < 0.001 < 0.001 0.032 < 0.001

The P-value relates to the test for statistically signicant dierence between the two methods. 3D, three dimensional; SE, standard error. The anatomical variables are dened in Table 2.

Table 4 Average dierence between corresponding measurements taken on the conventional frontal radiographs and the CBCTconstructed 3D model
Dierence 2D-CBCT R CR-L: R J-ANS: R ME-AG: AGl/AGr/Ll-Lr () AGl/AGr/Ll () AGl/AGr/Lr () AGl/ME/AGr: () Jl/Jr/LL () Jl/Jr/Lr () Ll/CR/Lr () ME/AGl/CR () ME/AGr/CR () 0.02 )0.03 0.12 )0.93 2.03 0.15 39.66 )1.73 1.30 33.04 16.34 11.76 95% CI 0.000.03 )0.05 )0.01 0.100.14 )1.37 )0.49 1.552.50 )0.340.64 34.7144.61 )2.47 )0.98 0.651.94 31.0335.04 14.4018.28 10.4913.02 P-value 0.010 0.002 0.0001 0.0001 0.533 0.0001 0.0001 0.0001 0.0002 0.0001 0.0001 0.0001 Reliability )0.150 )0.302 )0.234 0.273 0.789 0.683 )0.130 0.549 0.655 0.174 0.716 0.820

< <

< < < < < <

95% condence interval (CI), corresponding P-value, and reliability. 3D, three dimensional; CBCT, cone beam computed tomography; 2D-CBCT, conventional/two dimensional cephalometry; SE, standard error. The anatomical variables are dened in Table 2.

AGr/Lr, Gonial Angle l; see Table 2 for denitions of these anatomical variables). The reproducibility of the measurements on the CBCT-constructed 3D model was therefore higher than the reproducibility of the measurements on conventional frontal radiographs. There was a statistically signicant dierence between the conventional frontal cephalometric radiographs and the CBCT-constructed 3D model for 11 out of 12 measurements (Table 4). For the measurement AGl/ AGr/Ll, no statistically signicant dierence was found. Although the average dierence for those measurements between the two methods was clearly statistically signi-

cant (all P values < 0.01), for most of them the actual mean average dierence for repeated measurements ranged from 0.02 to 0.12 for the ratios. For the angles these average dierences ranged from 0.15 to 39.66.

Discussion
In this study, twelve widely used cephalometric measurements on conventional frontal radiographs and CBCT-constructed 3D models of dry human skulls were compared. We used dry skulls because it is not

304

van Vlijmen et al.

considered to be ethical to expose patients twice to radiation from both a conventional radiograph and a CBCT. In this way, we obtained two images that were completely comparable. The largest error in cephalometric studies is the error in landmark identication (13) and each landmark exhibits a characteristic pattern of error that contributes to measurement inaccuracy (14). Images from dry skulls do not suer from distortion caused by soft tissues. This reduces the chance of errors in landmark identication because it makes accurate localization of bony landmarks easier. Testing the reliability of soft tissue measurements is, of course, not possible. In a previous study our group showed that the position of the patient in frontal radiology plays an important role in the outcome of a cephalometric analysis because the measurements are inuenced by tilt or rotations of the head (11). The position of the patient in a CBCT scanner is, for 3D angular measurements, not that important because rotation along the long axis or the sagittal axis is not of any inuence on angles measured. In this study each skull was only placed in the cephalostat once and the same image was traced ve times. This approach was chosen because the biggest error in cephalometric studies is the error in landmark identication (13) and we wanted to eliminate other factors that could inuence the identication of landmarks. In a clinical situation the repeated positioning of the patient in a cephalostat or a CBCT scanner is probably more variable than the single positioning of a skull in a study design. This may have resulted in an underestimation of the error. Further research must show whether this is of signicant inuence. For all measurements, except one (AGl/AGr/Ll), statistically signicant dierences were found between the conventional frontal cephalometric radiographs compared with CBCT-constructed 3D models. For four of these measurements (AGl/ME/AGr, Ll/CR/Lr, ME/ AGl/CR, ME/AGr/CR) these dierences ranged from 11.76 to 39.66 and are well beyond any reasonable clinically acceptable range, as described by Hajeer et al. (15). The question arises: What caused these large differences? This is most obvious in the AGl/ME/AGr angle and can best be explained if we take a closer look at this angle. In 2D the landmarks needed to measure this angle are all in one tomographic plane and the angle measured is around 120. In 3D these three landmarks are in dierent tomographic planes and therefore the angle measured is completely dierent, on average around 80. This angle clearly shows that comparing 2D and 3D cephalometric measurements should be performed with great caution and one should be very careful when interpreting the results. Comparable results were found in a similar study when comparing 2D lateral radiographs with 3D models. Nevertheless, the mean dierences for the AP radiographs were much larger than for lateral radiographs. The reason for this is that landmarks for AP cephalometry are located in more dierent tomographic planes compared with those for lateral cephalometry. In this study, all measurements were performed by one observer. The question remains whether this observer

made a systematic error. The standard errors in Table 3 are acceptable. Systematic errors in the identication of landmarks are the same for both types of radiographs and therefore do not have any inuence on reproducibility. It is therefore justied to have one observer for this type of study. Five repetitions of the measurements were undertaken to allow estimation of the variability of each landmark for each skull. That is where the number of repetitions of the measurements pays o. For determining the intraobserver reliability, however, per denition only two measurements are needed. Statistically there is no reason to suspect a dierent intra-observer reliability between the rst and second measurement series compared with, for example, the second and third or the third and fourth measurement series. Therefore, it does not give relevant additional information. This is why we choose to mention only the reliability of the rst two series of measurements in our results. The mean value of all ve repeated measurements and their variance, however, was used for further statistical analysis. In general, orthodontists and maxillofacial surgeons are experienced in the use and interpretation of conventional frontal cephalometric radiographs (Fig. 1A). However, tracing a 3D model (Fig. 1B) is very dierent and landmarks have to be well dened in all three dimensions. Conventional cephalometry is a 2D representation of a 3D structure. This has certain disadvantages. Now that it is possible to make a 3D model of the skull, it is also possible to perform a 3D cephalometric analysis (12, 16). In such an analysis, the actual anatomical structures can be identied more realistically than in a 2D projection. In 2D anterior posterior radiology, positioning of the patient can result in major diculties. As long as a 2D projection of 3D scans is used for analysis, positioning of the patient remains an important factor. If a 3D model is made and a 3D analysis is performed, it is likely that positioning of the patient has no eect on the measurements made during a 3D cephalometric analysis. Future research needs to conrm this hypothesis. Therefore, there is a need to further develop and test the measurements made in new 3D-cephalometric analyses. It can be concluded that there is a clinically relevant dierence between angular measurements performed on conventional frontal cephalometric radiographs, compared with measurements on CBCT-constructed 3D models. Tracings of 3D models are therefore not suitable for longitudinal research in cases where there are only 2D records from the past.

References
1. Broadbent BH. A new X-ray technique and its application to orthodontia. Angle Orthod 1931; 1: 4566. 2. Hofrath H. Bedeutung der ro ntgenfern und abstands aufnahme fu r die diagnostik der kieferanomalien. Fortschr Orthod 1931; 1: 231258. 3. Gottlieb EL, Nelson AH, Vogels DS. 1990 JCO study of orthodontic diagnosis and treatment procedures. 1. Results and trends. J Clin Orthod 1991; 25: 145156.

A comparison between 2D and 3D cephalometry

4. Ghafari GG. Posteroanterior cepahalometry: craniofacialfrontal analysis. In: Jacobson A, ed. Radiographic cephalometry. Hanover Park: Quintessence Publishing Co Inc, 2006; 267 292. 5. Ludlow JB, Davies-Ludlow LE, Brooks SL, Howerton WB. Dosimetry of 3 CBCT devices for oral and maxillofacial radiology: CB Mercuray, NewTom 3G and i-CAT. Dentomaxillofac Radiol 2006; 35: 219226. 6. Ludlow JB, Ivanovic M. Comparative dosimetry of dental CBCT devices and 64-slice CT for oral and maxillofacial radiology. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2008; 106: 106114. 7. Silva MA, Wolf U, Heinicke F, Bumann A, Visser H, Hirsch E. Cone-beam computed tomography for routine orthodontic treatment planning: a radiation dose evaluation. Am J Orthod Dentofacial Orthop 2008; 133: 640.e1640.e5. 8. Greiner M, Greiner A, Hirschfelder U. Variance of landmarks in digital evaluations: comparison between CT-based and conventional digital lateral cephalometric radiographs. J Orofac Orthop 2007; 68: 290298. 9. Kumar V, Ludlow JB, Mol A, Cevidanes L. Comparison of conventional and cone beam CT synthesized cephalograms. Dentomaxillofac Radiol 2007; 36: 263269. 10. Moshiri M, Scarfe WC, Hilgers ML, Scheetz JP, Silveira AM, Farman AG. Accuracy of linear measurements from imaging plate and lateral cephalometric images derived from

305

11.

12. 13. 14.

15.

16.

cone-beam computed tomography. Am J Orthod Dentofacial Orthop 2007; 132: 550560. SJ, Swennen GRJ, Bronkhorst Van Vlijmen OJC, Berge EM, Katsaros C, Kuijpers-Jagtman AM. Comparison of cephalometric radiographs obtained from cone beam computed tomography scans and conventional radiographs. J Oral Maxillofac Surg 2009; 67: 9297. Swennen GRJ, Schutyser F, Hausamen JE, eds. Threedimensional cephalometry. A color atlas and manual. Berlin: Heidelberg Springer, 2005; 183226. Houston WJ. The analysis of errors in orthodontic measurements. Am J Orthod 1983; 83: 382390. Lou L, Lagravere MO, Compton S, Major PW, FloresMir C. Accuracy of measurements and reliability of landmark identification with computed tomography (CT) techniques in the maxillofacial area: a systematic review. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2007; 104: 402411. Hajeer MY, Ayoub AF, Millett DT, Bock M, Siebert JP. Three-dimensional imaging in orthognathic surgery: the clinical application of a new method. Int J Adult Orthodon Orthognath Surg 2002; 17: 318330. Swennen GRJ, Schutyser F. Three-dimensional virtual approach to diagnosis and treatment planning of maxillo-facial deformity. In: Bell WH, Guerrero CA, eds. Distraction osteogenesis of the facial skeleton. Hamilton: BC Decker, 2007; 5579.