- Email: [email protected]
Comparison of radiographic and photographic measurement of mandibular asymmetry
ORIGINAL ARTICLE
Comparison of radiographic and photographic measurement of mandibular asymmetry Raymond Edler, BDS, FDS, M.ORTH.RCS,a David Wertheim, MA, PhD, CEng, MIEE, CPhys MinstP,b and Darrel Greenhill, BSc, PhD, MIEEc London and Surrey, United Kingdom This study compared measurement of mandibular asymmetry by digitization of mandibular outlines from standardized facial photographs and posteroanterior cephalometric radiographs. Four ratios were used in calculating asymmetry: area (relative size of right and left mandibular segments), perimeter or length of outlines, compactness (shape), and moment. The records of 28 patients with varying degrees of asymmetry were used. A significant relationship was found for 3 of the ratios (area, compactness, and moment) between measurements from photographs and radiographs. A further comparison showed that measurements from the radiographs correlated more closely with those from photographs when the mastoid processes were used as a baseline, rather than latero-orbitale. Repeatability of mandibular outline digitization proved satisfactory. Although digitization from standardized photographs is the preferred approach, the results indicated that posteroanterior cephalometric radiographs can be used similarly. Unlike other cephalometric analyses for mandibular asymmetry, this method avoids problems of landmark identification, thus presenting a clinically useful method of quantifying asymmetry, eg, in auditing the surgical-orthodontic correction of asymmetry or monitoring change over time. (Am J Orthod Dentofacial Orthop 2003;123:167-74)
T
he relative conformity with which orthodontists measure anteroposterior and vertical skeletal discrepancies is not matched in the transverse plane, particularly in the assessment of asymmetries. Posteroanterior (PA) cephalometry has provided the most commonly used means of measuring asymmetry, and many methods have been described, including the use of angular1 and linear measurements.2-5 Additionally, combinations of linear and angular measurements have been presented as indexes6 and as part of a comprehensive analysis.7-9 Other methods have involved area calculations, such as those by Mongini and Schmid10 and Schmid et al,11 or triangulation.12 The existence of so many approaches perhaps indicates that, for everyday clinical purposes, none is ideal. Certainly, there can be problems of landmark identification13,14 because of bony superimposition, and cautionary coma
Consultant orthodontist, Norman Rowe Maxillofacial Unit, Queen Mary’s Hospital, Roehampton, London, United Kingdom. b Reader, School of Computing and Information Systems, Kingston University, Kingston-upon-Thames, Surrey. c Senior lecturer, School of Computing and Information Systems, Kingston University, Kingston-upon-Thames, Surrey. Reprint requests to: R. Edler, Orthodontic Departement, Norman Rowe Maxillofacial Unit, Queen Mary’s Hospital, Roehampton Lane, Roehampton, London SW15 5PN United Kingdom; e-mail, [email protected]. co.uk. Submitted, January 2002; revised and accepted, May 2002. Copyright © 2003 by the American Association of Orthodontists. 0889-5406/2003/$30.00 ⫹ 0 doi:10.1067/mod.2003.16
ments about the reliability of measurements taken from PA cephalometric radiographs have been made in established texts.15,16 Alternatives, such as submentovertex (SMV) film,17-20 have also received some criticism,2 and the elegantly described combination of PA radiographs with SMV film21 does not appear to be universally used, possibly because it is rather timeconsuming for everyday purposes. Three-dimensional imaging techniques such as photogrammetry or optical scanning22 involve costly equipment so that these techniques are currently available only to a few clinicians working in specialized centers; yet mandibular asymmetry is a relatively common finding.23 Measurements taken from standardized photographs24,25 have proved useful, eg, in monitoring the quality of nostril asymmetry in correcting the cleft-lip nose.26 As with 3-dimensional imaging, this approach is noninvasive and also, by definition, considers soft tissue appearance, which is very relevant for mandibular asymmetry. Measurement of this type of asymmetry from photographs, with digitized mandibular outlines, has been described.27,28 The measurements have been shown to be clinically relevant for quantification of asymmetry, rather than as an aid to diagnosis or treatment planning, and used specifically for mandibular asymmetry, rather than for generalized facial asymmetry. It is hoped that this technique will prove useful in recording the quality of surgical correction of asymmetry, in monitoring patients before or after treatment, and in identifying 167
168 Edler, Wertheim, and Greenhill
American Journal of Orthodontics and Dentofacial Orthopedics Febuary 2003
Fig 1. A, Photograph of patient; B, digitization of mandibular outline from facial photograph. Ratios for this patient were: area, 1.093; perimeter, 1.018; compactness, 0.949; moment, ⫺1.265. Deviation ratios were: area, 0.093; perimeter, 0.018; compactness, 0.051. Difference indexes were: area, 0.044; perimeter, 0.009; compactness, ⫺0.026.
mild cases.29 Athanasiou et al13 identified several points on the PA radiograph that were less vulnerable to landmark identification error than others; these included latero-orbitale and the tips of the mastoid processes. The bony mandibular outline is also readily identified, and thus it might well be possible to measure mandibular asymmetry from the PA radiograph by digitizing the mandibular outline. Accordingly, the aims of this study were to assess asymmetry analysis of PA radiographs as a method for mandibular asymmetry measurement and to compare it with the digitization of mandibular outlines from facial photographs. MATERIAL AND METHODS
Standardized full-face photographs and PA cephalometric radiographs of 29 patients were used. The patients had all been identified as having mandibular asymmetry, from mild to quite severe, and had been referred for orthodontic treatment, including orthognathic surgery in some cases. The sample was selected to provide an adequate range of mandibular asymmetry. One patient was subsequently excluded because the ear position could not be clearly discerned. Thus, 28 patients were included in this study (12 males, 16
females; median age, 17.0 years; range, 11.2-41.7 years). Photographs were taken under standardized conditions30,31 with visual axes horizontal. The patients were asked to close lightly on their back teeth and avoid expressive activity. Earrings and eyeglasses were removed and long hair tied back. The lighting arrangements were designed to avoid shadows, particularly around the mandibular outline. The technique, described previously,28 involves the use of background lights with diffuse reflectors, situated behind and before the patient to avoid shadow. The lamps were situated at a 45° angle to the patient and were 6 feet high. The photographs were taken with 1:8 magnification and a 100-mm macrolens, and then scanned and stored on a compact disk. The photos were digitized in random order, the photographic and radiographic images having been coded independently. The use of an on-screen digitizing program has been described previously.27,28 The software was developed with Visual C⫹⫹ version 6 (Microsoft Corporation, Redmond, Wash), run on a Pentium-based PC. The mandibular outline was digitized with a series of mouse clicks, after the inferior insertions of the ears (otobasion inferius) had been
Edler, Wertheim, and Greenhill 169
American Journal of Orthodontics and Dentofacial Orthopedics Volume 123, Number 2
Fig 2. Digitization of mandibular outline from radiograph of same patient. A, With line joining latero-orbitale points as baseline, ratios were: area, 1.217; perimeter, 1.081; compactness, 0.960; moment, ⫺1.046. Deviation ratios were: area, 0.217; perimeter, 0.081; compactness, 0.04. Difference indexes were: area, 0.098; perimeter, 0.039; compactness, ⫺0.021. B, With line joining tips of mastoid processes, ratios were: area, 1.127; perimeter, 1.039; compactness, 0.958; moment, ⫺1.019. Deviation ratios were: area, 0.127; perimeter, 0.039; compactness, 0.042. Difference indexes were: area, 0.060; perimeter, 0.019; compactness, ⫺0.022.
Mandibular outline digitization, with Spearman rank correlation of ratios measured from photographs and PA cephalograms
Table I.
Ratios Photo against PA ceph (mastoid baseline) Photo against PA ceph (latero-orbitale baseline) Difference indexes Photo against PA ceph (mastoid baseline) Photo against PA ceph (latero-orbitale baseline) Deviation ratios Photo against PA ceph (mastoid baseline) Photo against PA ceph (latero-orbitale baseline)
Area
Perimeter
Compactness
Moment
P ⫽ .002 r ⫽ 0.566 P ⫽ .002 r ⫽ 0.553
P ⫽ .040 r ⫽ 0.390 P ⫽ .074 r ⫽ 0.343
P ⬍ .001 r ⫽ 0.682 P ⫽ .001 r ⫽ 0.578
P ⫽ .004 r ⫽ 0.530 P ⫽ .007 r ⫽ 0.500
P ⫽ .004 r ⫽ 0.528 P ⫽ .001 r ⫽ 0.585
P ⫽ .071 r ⫽ 0.347 P ⫽ .671 r ⫽ 0.084
P ⬍ .001 r ⫽ 0.636 P ⫽ .001 r ⫽ 0.581
P ⫽ .003 r ⫽ 0.539 P ⫽ .001 r ⫽ 0.591
P ⫽ .078 r ⫽ 0.338 P ⫽ .646 r ⫽ 0.091
P ⬍ .001 r ⫽ 0.645 P ⫽ .001 r ⫽ 0.583
identified; previous work29 indicated that that baseline provided the best repeatability (Fig 1). The baseline was then automatically bisected, so that, following the mandibular outline digitization, the lower part of the face was divided into right and left segments (Fig 1).
The segments were then compared according to 4 ratios: area (size) of right segment compared with left; perimeter (length of mandibular outlines) of right segment compared with left; compactness (shape) calculated as the square of the perimeter divided by the
170 Edler, Wertheim, and Greenhill
American Journal of Orthodontics and Dentofacial Orthopedics Febuary 2003
Fig 3. Correlation plots: A, compactness ratio, photograph vs radiograph/latero-orbitale; B, compactness ratio, photograph vs radiograph/mastoid; C, perimeter ratio, photograph vs radiograph/latero-orbitale; D, perimeter ratio, photograph vs radiograph/mastoid.
area,32 comparing right and left segments; and moment ratio (center of area), the horizontal distance from the center of both areas combined to the baseline perpendicular, expressed as a percentage of total baseline width. For area, perimeter, and compactness, ideal symmetry is represented by a ratio of 1.000. For the moment ratio, it is 0.000. The system has a display resolution of 0.001. Each mandibular outline was digitized twice, with an interval of 1 to 2 weeks, and the mean measurement calculated. All the PA cephalograms were taken on a Proline 2002 CC (Planmeca, Helsinki, Finland). The film distance to x-ray tube was fixed at 160 cm. Films were exposed at 68-70 kV and 12 mA and were 24 ⫻ 30 cm in size. The head was positioned with the Frankfort plane horizontal. Magnification was 10%. The bisected baseline was used to identify the appropriate vertical boundary separating the right and
left segments. For this study, 2 baselines were selected for comparative purposes—a line joining the right and left latero-orbitale points and a line joining the lowest points of the right and left mastoid processes (Fig 2). Athanasiou et al13 have shown that both of these points are readily identifiable as landmarks with minimal landmark identification error. Then the baseline was automatically bisected (Fig 2), and the bony mandibular outline was digitized twice; the mean measurement was calculated. Digitization of photographs and radiographs was undertaken with the same software and by 1 operator (R.E.). For a symmetrical outline, area ratio, perimeter ratio, and compactness ratio would be expected to give a value of 1.000. Thus, the data were expressed in 2 ways to evaluate how far they deviated from a symmetrical outline: (1) absolute deviation from 1, termed the deviation ratio, so that a perfectly symmetrical outline would have a deviation ratio of 0.00; and (2)
Edler, Wertheim, and Greenhill 171
American Journal of Orthodontics and Dentofacial Orthopedics Volume 123, Number 2
the variables that avoids such bias. Both can be assessed from the right/left ratio. Repeatability of soft tissue mandibular outline digitization has been analyzed previously.29 Repeatability of bony mandibular outline digitization with lateroorbitale and the mastoid processes as baseline points was compared according to the method of Bland and Altman.33 This was assessed by comparing the difference in the 2 measurements against the mean for each set of baseline points. Additionally, the photographs and the radiographs of 5 patients were each digitized on 5 separate occasions, and the error of the method was assessed. The distributions for each set of data were tested for consistency with normal with the Ryan-Joiner method using Minitab Release 13.31 (Minitab Inc, State College, Pa). Nonparametric correlation was performed with the Spearman rank correlation; this involved ranking the data and then performing a Pearson correlation on the ranked data. RESULTS
Fig 4. Agreement between measurements of area ratios. A, between photograph and radiograph/lateroorbitale, and B, between photograph and radiograph/ mastoid.
absolute (right segment metric ⫺ left segment metric)/ (right segment metric ⫹ left segment metric) multiplied by 100, termed the difference index, so that a perfectly symmetrical outline would have a difference index of 0.0. The first method is an intuitive assessment of asymmetry, although there might be some bias between left and right for very asymmetric patients. The bias could occur because the deviation ratio is calculated from the deviation from 1 of the ratio of right to left measurements; this might differ when compared with the ratio of left to right measurements. The effect is likely to be more noticeable for very asymmetric outlines. The difference index method is a means of assessing the difference between left and right sides for
Table II.
Many data sets were found to be inconsistent with a normal distribution, and nonparametric statistical analysis was used. With the Spearman rank correlation, a significant relationship between the difference index for the 3 measurement sets of area, compactness, and moment was found, summarized in Table I. For perimeter, the relationship was more variable. When comparing the measurements of area, compactness, and moment ratio calculated in the 3 positions, there is a very clear relationship (except for perimeter measurements) as shown in Table I; some comparisons are shown in Figure 3. Thus there is close agreement among not only the measurements made in the 3 positions, but also in the 3 ways of expressing the results: ratios, deviation ratios, and difference indexes. This is also confirmed by plotting the difference between the 2 measurements against the mean, as in Figure 4. As can be seen, there is better agreement between the mastoid measurements and the photograph measurements than between the latero-orbitale measurements and the photographs. Similarly, the variability of the measurements at the mastoid was smaller than that of the measurements at
Comparative repeatability of digitization for 4 ratios (all patients)
Maximum difference (SD) Area Perimeter Compactness Moment
Photos
Mastoid
Latero-orbitale
0.073 (0.021) 0.022 (0.008) 0.040 (0.011) 0.91 (0.37)
0.026 (0.013) 0.020 (0.008) 0.050 (0.013) 1.44 (0.50)
0.099 (0.034) 0.041 (0.013) 0.040 (0.013) 1.75 (0.55)
172 Edler, Wertheim, and Greenhill
American Journal of Orthodontics and Dentofacial Orthopedics Febuary 2003
DISCUSSION
Fig 5. Repeatability of digitization of area ratio: A, radiograph with latero-orbitale as baseline; B, photo; and C, radiograph with mastoid tip as baseline.
the latero-orbitale, as shown in Table II and Figure 5. Results of the error method assessment, having repeated the digitization of photographs and radiographs of 5 patients, are presented in Table III.
The relationship of the otobasion inferius baseline to the mandibular outline on the photograph is closer to that of the mastoid tip baseline and the mandibular outline on the radiograph, than to a latero-orbitale baseline and mandibular outline on the radiograph. A latero-orbitale baseline might well relate more closely to a photographic baseline involving the outer canthus of the eyes, although this was not investigated in this study. It has previously been found that mandibular outline digitization from photographs can be used to quantify this type of asymmetry and to distinguish between groups of patients with different degrees of asymmetry. The correlation coefficients shown in Table I for 3 ratios—area, compactness, and moment—are high enough to suggest a close relationship between measurements taken from photographs and digitized cephalometric radiographs. Four relevant correlation plots are shown in Figure 3; A and B show that for the compactness ratio, the relationship is strong. Although similar findings apply to the use of both area and moment ratios, the correlations for perimeter ratio were low (Figure 3, C and D). Previous work has shown that perimeter ratios do not seem to relate as well to clinical assessment as the others, perhaps because this type of measurement is more sensitive to minor errors in outline digitization.28 The correlation coefficients in Table I also indicate that the measurements between photographs and PA cephalometric radiographs are closer if the mastoid baseline is used, rather than latero-orbitale, although this does not apply to area deviation ratios or difference indexes. It is too early to identify which ratio or index is most appropriate in quantifying asymmetry, but it appears that the technique can be applied to measurements taken from the PA cephalometric radiograph, and when this radiograph is used, the mastoid baseline would give results closer to those taken from the photographs than latero-orbitale, although this baseline could be used if necessary. The plots of repeatability assessment for area ratio shown in Figure 5, A-C, indicate that repeatability of PA cephalometric radiograph digitization is generally satisfactory, as shown in Table II. Previous work29 on repeatability of digitization of photographs identified a standard deviation of differences for area ratio of 0.03. In general, if the radiograph is used in this way, the baseline involving the tips of the mastoid processes generally gives better repeatability than latero-orbitale. The many asymmetry measurement techniques,
Edler, Wertheim, and Greenhill 173
American Journal of Orthodontics and Dentofacial Orthopedics Volume 123, Number 2
Table III. Intra-observer repeatability, with photographs and radiographs (latero-orbitale and mastoid baselines) of 5 patients, digitized on 5 occasions
Photos Patient number 1 2 3 4 5 PA ceph mastoid Patient number 1 2 3 4 5 PA ceph latero-obitale Patient number 1 2 3 4 5
Mean
SD
Minimum
Maximum
Difference (max-min)
0.881 0.963 1.035 0.947 1.013
0.005 0.008 0.006 0.012 0.007
0.874 0.953 1.030 0.934 1.003
0.888 0.975 1.042 0.965 1.022
0.014 0.022 0.012 0.031 0.019
0.871 0.992 1.021 0.945 1.000
0.012 0.036 0.014 0.022 0.010
0.856 0.931 0.996 0.915 0.988
0.885 1.026 1.030 0.976 1.010
0.029 0.095 0.034 0.061 0.022
0.900 0.839 0.970 0.914 0.887
0.022 0.021 0.027 0.020 0.020
0.865 0.812 0.932 0.899 0.860
0.924 0.866 1.000 0.937 0.913
0.059 0.054 0.068 0.048 0.053
based on the PA cephalometric radiograph, have already been noted. In suggesting yet another method based on this film, we point out that asymmetry measurement by digitizing the mandibular outline avoids the landmark identification problems previously mentioned. The 2 suggested baseline points—mastoid tips and latero-orbitale— have been shown to be less error prone than many others. The technique appears to have potential as a measurement tool in monitoring the quality of surgical orthodontic correction, in long-term monitoring of asymmetry patients, and in identifying mild cases. In general, measurement of asymmetry from a photograph seems appropriate because it is noninvasive and also soft tissue appearance is taken into account. However, if necessary, the PA cephalogram can be used as an alternative. It would be helpful to compare both the validity and the repeatability of the digitization of the mandibular outline from PA cephalometric radiographs with other techniques involving the same film. CONCLUSIONS
1. A significant relationship was found for area, compactness, and moment ratios measured from photographs and PA cephalometric radiographs, but not for perimeter ratio. 2. Repeatability studies generally showed less varia-
tion from the mastoid baseline measurements than measurements involving a latero-orbitale baseline. 3. Repeatability levels appeared satisfactory for all 3 techniques. 4. Digitization of the mandibular outline from PA radiographs, and the different ratios obtained, would seem a potentially useful way to assess mandibular asymmetry. This approach should be explored further.
REFERENCES 1. Letzer GM, Kronman JH. A posteroanterior cephalometric evaluation of craniofacial asymmetry. Angle Orthod 1967;37:205-11. 2. Peck S, Peck L, Kataja M. Skeletal asymmetry in esthetically pleasing faces. Angle Orthod 1991;6l:43-7. 3. Skolnick J, Iranpour B, Westesson P-L, Adair S. Prepubertal trauma and mandibular asymmetry in orthognathic surgery and orthodontic patients. Am J Orthod Dentofacial Orthop 1994;105:73-7. 4. Severt TR, Proffit WR. Postsurgical stability following correction of severe facial asymmetry. Int J Adult Orthod Orthognath Surg 1997;12:251-61. 5. Basyouni AA, Nanda SK. An atlas of the transverse dimensions of the face. Ann Arbor: University of Michigan; 2000. 6. Chebib FS, Chamma AM. Indices of craniofacial asymmetry. Angle Orthod 1981;51:214-26. 7. Grummons DC, Kappeyne Van De Coppello MA. A frontal asymmetry analysis. J Clin Orthod 1987;21:448-65. 8. Ricketts RM, Bench RW, Hilgers JJ, Schulhof R. An overview of computerized cephalometrics. Am J Orthod 1972;61:1-28. 9. Svanholt P, Solow B. Assessment of midline discrepancies on the posteroanterior cephalometric radiograph. Trans Eur Orthod Soc 1977;25:261-8.
174 Edler, Wertheim, and Greenhill
10. Mongini F, Schmid W. Treatment of mandibular asymmetries during growth: a longitudinal study. Eur J Orthod 1987;9:51-67. 11. Schmid W, Mongini F, Felisio A. A computer-based assessment of structural and displacement asymmetries of the mandible. Am J Orthod Dentofacial Orthop 1991;100:19-34. 12. Hewitt AB. A radiographic study of facial asymmetry. Br J Orthod 1975;2:37-40. 13. Athanasiou AE, Miethke RR, Van der Meij AJW. Random errors in localisation of landmarks in postero-anterior cephalograms. Br J Orthod 1999;26:273-83. 14. Cook JT. Asymmetry of the cranio-facial skeleton. Br J Orthod 1980;7:33-8. 15. Proffit WR, White RP Jr. Surgical-orthodontic treatment. St Louis: Mosby Year Books; 1991. 16. Athanasiou A. Orthodontic cephalometry. London: MosbyWolfe; 1995. 17. Rose JM, Sadowsky C, BeGole EA, Moles R. Mandibular skeletal and dental asymmetry in Class II subdivision malocclusions. Am J Orthod Dentofacial Orthop 1994;l05:489-95. 18. Arnold TG, Anderson GC, Liljemark WF. Cephalometric norms for craniofacial asymmetry using submental-vertical radiographs. Am J Orthod Dentofacial Orthop 1994;l06:250-6. 19. O’Byrn BL, Sadowsky C, Schneider B, BeGole EA. An evaluation of mandibular asymmetry in adults with unilateral posterior crossbite. Am J Orthod Dentofacial Orthop 1995;107:394-400. 20. Forsberg CM, Burstone CJ, Hanley KJ. Diagnosis and treatment planning of skeletal asymmetry with the submental-vertical radiograph. Am J Orthod 1984;85:224-37. 21. Grayson BH, McCarthy JG, Bookstein F. Analysis of craniofacial asymmetry by multiplane cephalometry. Am J Orthod 1983;84:217-24. 22. Moss JP, Coombes AM, Linney AD, Campos J. Methods of three-dimensional analysis of patients with asymmetry of the face. Proc Finn Dent Soc 1991;87:139-49.
American Journal of Orthodontics and Dentofacial Orthopedics Febuary 2003
23. Severt TS, Proffit WR. The prevalence of facial asymmetry in the dentofacial deformities population at the University of North Carolina. Int J Adult Orthod Orthognath Surg 1997;12:171-6. 24. Coghlan BA, Laitung JKG, Pigott RW. A computer-aided method of measuring nasal symmetry in the cleft lip nose. Br J Plast Surg 1993;46:13-7. 25. Coghlan BA, Matthews B, Pigott RW. A computer-based method of measuring facial asymmetry. Results from an assessment of the repair of cleft lip deformities. Br J Plast Surg 1987;40:371-6. 26. Timoney N, Smith G, Pigott RW. A 20-year audit of nose-tip symmetry in patients with unilateral cleft lip and palate. Br J Plas Surg 2001;54:294-8. 27. Greenhill D, Edler R, Wertheim D, Belsham A, Jones GA. A system for assessing mandibular outlines. International conference on advances in medical signal and information processing. London: Institution of Electrical Engineers, ISBN— 0-85296728-4; 2000. p. 192-9. 28. Edler R, Wertheim D, Greenhill D. Clinical and computerised assessment of mandibular asymmetry. Eur J Orthod 2001;23: 485-94. 29. Edler R, Wertheim D, Greenhill D. Mandibular outline assessment in three groups of orthodontic patients. Eur J Orthod 2002; 24:605-14. 30. Bengel W. Standardisation in dental photography. Int Dent J 1985:35:210-7. 31. Claman L, Patton D, Rashid R. Standardized portrait photography for dental patients. Am J Orthod Dentofacial Orthop 1990; 98:197-205. 32. Gonzalez RC, Woods RE. Digital image processing. Reading, United Kingdom: Addison Wesley Publishing Co: 1993. 33. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:306-10.
Comparison of radiographic and photographic measurement of mandibular asymmetry Raymond Edler, BDS, FDS, M.ORTH.RCS,a David Wertheim, MA, PhD, CEng, MIEE, CPhys MinstP,b and Darrel Greenhill, BSc, PhD, MIEEc London and Surrey, United Kingdom This study compared measurement of mandibular asymmetry by digitization of mandibular outlines from standardized facial photographs and posteroanterior cephalometric radiographs. Four ratios were used in calculating asymmetry: area (relative size of right and left mandibular segments), perimeter or length of outlines, compactness (shape), and moment. The records of 28 patients with varying degrees of asymmetry were used. A significant relationship was found for 3 of the ratios (area, compactness, and moment) between measurements from photographs and radiographs. A further comparison showed that measurements from the radiographs correlated more closely with those from photographs when the mastoid processes were used as a baseline, rather than latero-orbitale. Repeatability of mandibular outline digitization proved satisfactory. Although digitization from standardized photographs is the preferred approach, the results indicated that posteroanterior cephalometric radiographs can be used similarly. Unlike other cephalometric analyses for mandibular asymmetry, this method avoids problems of landmark identification, thus presenting a clinically useful method of quantifying asymmetry, eg, in auditing the surgical-orthodontic correction of asymmetry or monitoring change over time. (Am J Orthod Dentofacial Orthop 2003;123:167-74)
T
he relative conformity with which orthodontists measure anteroposterior and vertical skeletal discrepancies is not matched in the transverse plane, particularly in the assessment of asymmetries. Posteroanterior (PA) cephalometry has provided the most commonly used means of measuring asymmetry, and many methods have been described, including the use of angular1 and linear measurements.2-5 Additionally, combinations of linear and angular measurements have been presented as indexes6 and as part of a comprehensive analysis.7-9 Other methods have involved area calculations, such as those by Mongini and Schmid10 and Schmid et al,11 or triangulation.12 The existence of so many approaches perhaps indicates that, for everyday clinical purposes, none is ideal. Certainly, there can be problems of landmark identification13,14 because of bony superimposition, and cautionary coma
Consultant orthodontist, Norman Rowe Maxillofacial Unit, Queen Mary’s Hospital, Roehampton, London, United Kingdom. b Reader, School of Computing and Information Systems, Kingston University, Kingston-upon-Thames, Surrey. c Senior lecturer, School of Computing and Information Systems, Kingston University, Kingston-upon-Thames, Surrey. Reprint requests to: R. Edler, Orthodontic Departement, Norman Rowe Maxillofacial Unit, Queen Mary’s Hospital, Roehampton Lane, Roehampton, London SW15 5PN United Kingdom; e-mail, [email protected]. co.uk. Submitted, January 2002; revised and accepted, May 2002. Copyright © 2003 by the American Association of Orthodontists. 0889-5406/2003/$30.00 ⫹ 0 doi:10.1067/mod.2003.16
ments about the reliability of measurements taken from PA cephalometric radiographs have been made in established texts.15,16 Alternatives, such as submentovertex (SMV) film,17-20 have also received some criticism,2 and the elegantly described combination of PA radiographs with SMV film21 does not appear to be universally used, possibly because it is rather timeconsuming for everyday purposes. Three-dimensional imaging techniques such as photogrammetry or optical scanning22 involve costly equipment so that these techniques are currently available only to a few clinicians working in specialized centers; yet mandibular asymmetry is a relatively common finding.23 Measurements taken from standardized photographs24,25 have proved useful, eg, in monitoring the quality of nostril asymmetry in correcting the cleft-lip nose.26 As with 3-dimensional imaging, this approach is noninvasive and also, by definition, considers soft tissue appearance, which is very relevant for mandibular asymmetry. Measurement of this type of asymmetry from photographs, with digitized mandibular outlines, has been described.27,28 The measurements have been shown to be clinically relevant for quantification of asymmetry, rather than as an aid to diagnosis or treatment planning, and used specifically for mandibular asymmetry, rather than for generalized facial asymmetry. It is hoped that this technique will prove useful in recording the quality of surgical correction of asymmetry, in monitoring patients before or after treatment, and in identifying 167
168 Edler, Wertheim, and Greenhill
American Journal of Orthodontics and Dentofacial Orthopedics Febuary 2003
Fig 1. A, Photograph of patient; B, digitization of mandibular outline from facial photograph. Ratios for this patient were: area, 1.093; perimeter, 1.018; compactness, 0.949; moment, ⫺1.265. Deviation ratios were: area, 0.093; perimeter, 0.018; compactness, 0.051. Difference indexes were: area, 0.044; perimeter, 0.009; compactness, ⫺0.026.
mild cases.29 Athanasiou et al13 identified several points on the PA radiograph that were less vulnerable to landmark identification error than others; these included latero-orbitale and the tips of the mastoid processes. The bony mandibular outline is also readily identified, and thus it might well be possible to measure mandibular asymmetry from the PA radiograph by digitizing the mandibular outline. Accordingly, the aims of this study were to assess asymmetry analysis of PA radiographs as a method for mandibular asymmetry measurement and to compare it with the digitization of mandibular outlines from facial photographs. MATERIAL AND METHODS
Standardized full-face photographs and PA cephalometric radiographs of 29 patients were used. The patients had all been identified as having mandibular asymmetry, from mild to quite severe, and had been referred for orthodontic treatment, including orthognathic surgery in some cases. The sample was selected to provide an adequate range of mandibular asymmetry. One patient was subsequently excluded because the ear position could not be clearly discerned. Thus, 28 patients were included in this study (12 males, 16
females; median age, 17.0 years; range, 11.2-41.7 years). Photographs were taken under standardized conditions30,31 with visual axes horizontal. The patients were asked to close lightly on their back teeth and avoid expressive activity. Earrings and eyeglasses were removed and long hair tied back. The lighting arrangements were designed to avoid shadows, particularly around the mandibular outline. The technique, described previously,28 involves the use of background lights with diffuse reflectors, situated behind and before the patient to avoid shadow. The lamps were situated at a 45° angle to the patient and were 6 feet high. The photographs were taken with 1:8 magnification and a 100-mm macrolens, and then scanned and stored on a compact disk. The photos were digitized in random order, the photographic and radiographic images having been coded independently. The use of an on-screen digitizing program has been described previously.27,28 The software was developed with Visual C⫹⫹ version 6 (Microsoft Corporation, Redmond, Wash), run on a Pentium-based PC. The mandibular outline was digitized with a series of mouse clicks, after the inferior insertions of the ears (otobasion inferius) had been
Edler, Wertheim, and Greenhill 169
American Journal of Orthodontics and Dentofacial Orthopedics Volume 123, Number 2
Fig 2. Digitization of mandibular outline from radiograph of same patient. A, With line joining latero-orbitale points as baseline, ratios were: area, 1.217; perimeter, 1.081; compactness, 0.960; moment, ⫺1.046. Deviation ratios were: area, 0.217; perimeter, 0.081; compactness, 0.04. Difference indexes were: area, 0.098; perimeter, 0.039; compactness, ⫺0.021. B, With line joining tips of mastoid processes, ratios were: area, 1.127; perimeter, 1.039; compactness, 0.958; moment, ⫺1.019. Deviation ratios were: area, 0.127; perimeter, 0.039; compactness, 0.042. Difference indexes were: area, 0.060; perimeter, 0.019; compactness, ⫺0.022.
Mandibular outline digitization, with Spearman rank correlation of ratios measured from photographs and PA cephalograms
Table I.
Ratios Photo against PA ceph (mastoid baseline) Photo against PA ceph (latero-orbitale baseline) Difference indexes Photo against PA ceph (mastoid baseline) Photo against PA ceph (latero-orbitale baseline) Deviation ratios Photo against PA ceph (mastoid baseline) Photo against PA ceph (latero-orbitale baseline)
Area
Perimeter
Compactness
Moment
P ⫽ .002 r ⫽ 0.566 P ⫽ .002 r ⫽ 0.553
P ⫽ .040 r ⫽ 0.390 P ⫽ .074 r ⫽ 0.343
P ⬍ .001 r ⫽ 0.682 P ⫽ .001 r ⫽ 0.578
P ⫽ .004 r ⫽ 0.530 P ⫽ .007 r ⫽ 0.500
P ⫽ .004 r ⫽ 0.528 P ⫽ .001 r ⫽ 0.585
P ⫽ .071 r ⫽ 0.347 P ⫽ .671 r ⫽ 0.084
P ⬍ .001 r ⫽ 0.636 P ⫽ .001 r ⫽ 0.581
P ⫽ .003 r ⫽ 0.539 P ⫽ .001 r ⫽ 0.591
P ⫽ .078 r ⫽ 0.338 P ⫽ .646 r ⫽ 0.091
P ⬍ .001 r ⫽ 0.645 P ⫽ .001 r ⫽ 0.583
identified; previous work29 indicated that that baseline provided the best repeatability (Fig 1). The baseline was then automatically bisected, so that, following the mandibular outline digitization, the lower part of the face was divided into right and left segments (Fig 1).
The segments were then compared according to 4 ratios: area (size) of right segment compared with left; perimeter (length of mandibular outlines) of right segment compared with left; compactness (shape) calculated as the square of the perimeter divided by the
170 Edler, Wertheim, and Greenhill
American Journal of Orthodontics and Dentofacial Orthopedics Febuary 2003
Fig 3. Correlation plots: A, compactness ratio, photograph vs radiograph/latero-orbitale; B, compactness ratio, photograph vs radiograph/mastoid; C, perimeter ratio, photograph vs radiograph/latero-orbitale; D, perimeter ratio, photograph vs radiograph/mastoid.
area,32 comparing right and left segments; and moment ratio (center of area), the horizontal distance from the center of both areas combined to the baseline perpendicular, expressed as a percentage of total baseline width. For area, perimeter, and compactness, ideal symmetry is represented by a ratio of 1.000. For the moment ratio, it is 0.000. The system has a display resolution of 0.001. Each mandibular outline was digitized twice, with an interval of 1 to 2 weeks, and the mean measurement calculated. All the PA cephalograms were taken on a Proline 2002 CC (Planmeca, Helsinki, Finland). The film distance to x-ray tube was fixed at 160 cm. Films were exposed at 68-70 kV and 12 mA and were 24 ⫻ 30 cm in size. The head was positioned with the Frankfort plane horizontal. Magnification was 10%. The bisected baseline was used to identify the appropriate vertical boundary separating the right and
left segments. For this study, 2 baselines were selected for comparative purposes—a line joining the right and left latero-orbitale points and a line joining the lowest points of the right and left mastoid processes (Fig 2). Athanasiou et al13 have shown that both of these points are readily identifiable as landmarks with minimal landmark identification error. Then the baseline was automatically bisected (Fig 2), and the bony mandibular outline was digitized twice; the mean measurement was calculated. Digitization of photographs and radiographs was undertaken with the same software and by 1 operator (R.E.). For a symmetrical outline, area ratio, perimeter ratio, and compactness ratio would be expected to give a value of 1.000. Thus, the data were expressed in 2 ways to evaluate how far they deviated from a symmetrical outline: (1) absolute deviation from 1, termed the deviation ratio, so that a perfectly symmetrical outline would have a deviation ratio of 0.00; and (2)
Edler, Wertheim, and Greenhill 171
American Journal of Orthodontics and Dentofacial Orthopedics Volume 123, Number 2
the variables that avoids such bias. Both can be assessed from the right/left ratio. Repeatability of soft tissue mandibular outline digitization has been analyzed previously.29 Repeatability of bony mandibular outline digitization with lateroorbitale and the mastoid processes as baseline points was compared according to the method of Bland and Altman.33 This was assessed by comparing the difference in the 2 measurements against the mean for each set of baseline points. Additionally, the photographs and the radiographs of 5 patients were each digitized on 5 separate occasions, and the error of the method was assessed. The distributions for each set of data were tested for consistency with normal with the Ryan-Joiner method using Minitab Release 13.31 (Minitab Inc, State College, Pa). Nonparametric correlation was performed with the Spearman rank correlation; this involved ranking the data and then performing a Pearson correlation on the ranked data. RESULTS
Fig 4. Agreement between measurements of area ratios. A, between photograph and radiograph/lateroorbitale, and B, between photograph and radiograph/ mastoid.
absolute (right segment metric ⫺ left segment metric)/ (right segment metric ⫹ left segment metric) multiplied by 100, termed the difference index, so that a perfectly symmetrical outline would have a difference index of 0.0. The first method is an intuitive assessment of asymmetry, although there might be some bias between left and right for very asymmetric patients. The bias could occur because the deviation ratio is calculated from the deviation from 1 of the ratio of right to left measurements; this might differ when compared with the ratio of left to right measurements. The effect is likely to be more noticeable for very asymmetric outlines. The difference index method is a means of assessing the difference between left and right sides for
Table II.
Many data sets were found to be inconsistent with a normal distribution, and nonparametric statistical analysis was used. With the Spearman rank correlation, a significant relationship between the difference index for the 3 measurement sets of area, compactness, and moment was found, summarized in Table I. For perimeter, the relationship was more variable. When comparing the measurements of area, compactness, and moment ratio calculated in the 3 positions, there is a very clear relationship (except for perimeter measurements) as shown in Table I; some comparisons are shown in Figure 3. Thus there is close agreement among not only the measurements made in the 3 positions, but also in the 3 ways of expressing the results: ratios, deviation ratios, and difference indexes. This is also confirmed by plotting the difference between the 2 measurements against the mean, as in Figure 4. As can be seen, there is better agreement between the mastoid measurements and the photograph measurements than between the latero-orbitale measurements and the photographs. Similarly, the variability of the measurements at the mastoid was smaller than that of the measurements at
Comparative repeatability of digitization for 4 ratios (all patients)
Maximum difference (SD) Area Perimeter Compactness Moment
Photos
Mastoid
Latero-orbitale
0.073 (0.021) 0.022 (0.008) 0.040 (0.011) 0.91 (0.37)
0.026 (0.013) 0.020 (0.008) 0.050 (0.013) 1.44 (0.50)
0.099 (0.034) 0.041 (0.013) 0.040 (0.013) 1.75 (0.55)
172 Edler, Wertheim, and Greenhill
American Journal of Orthodontics and Dentofacial Orthopedics Febuary 2003
DISCUSSION
Fig 5. Repeatability of digitization of area ratio: A, radiograph with latero-orbitale as baseline; B, photo; and C, radiograph with mastoid tip as baseline.
the latero-orbitale, as shown in Table II and Figure 5. Results of the error method assessment, having repeated the digitization of photographs and radiographs of 5 patients, are presented in Table III.
The relationship of the otobasion inferius baseline to the mandibular outline on the photograph is closer to that of the mastoid tip baseline and the mandibular outline on the radiograph, than to a latero-orbitale baseline and mandibular outline on the radiograph. A latero-orbitale baseline might well relate more closely to a photographic baseline involving the outer canthus of the eyes, although this was not investigated in this study. It has previously been found that mandibular outline digitization from photographs can be used to quantify this type of asymmetry and to distinguish between groups of patients with different degrees of asymmetry. The correlation coefficients shown in Table I for 3 ratios—area, compactness, and moment—are high enough to suggest a close relationship between measurements taken from photographs and digitized cephalometric radiographs. Four relevant correlation plots are shown in Figure 3; A and B show that for the compactness ratio, the relationship is strong. Although similar findings apply to the use of both area and moment ratios, the correlations for perimeter ratio were low (Figure 3, C and D). Previous work has shown that perimeter ratios do not seem to relate as well to clinical assessment as the others, perhaps because this type of measurement is more sensitive to minor errors in outline digitization.28 The correlation coefficients in Table I also indicate that the measurements between photographs and PA cephalometric radiographs are closer if the mastoid baseline is used, rather than latero-orbitale, although this does not apply to area deviation ratios or difference indexes. It is too early to identify which ratio or index is most appropriate in quantifying asymmetry, but it appears that the technique can be applied to measurements taken from the PA cephalometric radiograph, and when this radiograph is used, the mastoid baseline would give results closer to those taken from the photographs than latero-orbitale, although this baseline could be used if necessary. The plots of repeatability assessment for area ratio shown in Figure 5, A-C, indicate that repeatability of PA cephalometric radiograph digitization is generally satisfactory, as shown in Table II. Previous work29 on repeatability of digitization of photographs identified a standard deviation of differences for area ratio of 0.03. In general, if the radiograph is used in this way, the baseline involving the tips of the mastoid processes generally gives better repeatability than latero-orbitale. The many asymmetry measurement techniques,
Edler, Wertheim, and Greenhill 173
American Journal of Orthodontics and Dentofacial Orthopedics Volume 123, Number 2
Table III. Intra-observer repeatability, with photographs and radiographs (latero-orbitale and mastoid baselines) of 5 patients, digitized on 5 occasions
Photos Patient number 1 2 3 4 5 PA ceph mastoid Patient number 1 2 3 4 5 PA ceph latero-obitale Patient number 1 2 3 4 5
Mean
SD
Minimum
Maximum
Difference (max-min)
0.881 0.963 1.035 0.947 1.013
0.005 0.008 0.006 0.012 0.007
0.874 0.953 1.030 0.934 1.003
0.888 0.975 1.042 0.965 1.022
0.014 0.022 0.012 0.031 0.019
0.871 0.992 1.021 0.945 1.000
0.012 0.036 0.014 0.022 0.010
0.856 0.931 0.996 0.915 0.988
0.885 1.026 1.030 0.976 1.010
0.029 0.095 0.034 0.061 0.022
0.900 0.839 0.970 0.914 0.887
0.022 0.021 0.027 0.020 0.020
0.865 0.812 0.932 0.899 0.860
0.924 0.866 1.000 0.937 0.913
0.059 0.054 0.068 0.048 0.053
based on the PA cephalometric radiograph, have already been noted. In suggesting yet another method based on this film, we point out that asymmetry measurement by digitizing the mandibular outline avoids the landmark identification problems previously mentioned. The 2 suggested baseline points—mastoid tips and latero-orbitale— have been shown to be less error prone than many others. The technique appears to have potential as a measurement tool in monitoring the quality of surgical orthodontic correction, in long-term monitoring of asymmetry patients, and in identifying mild cases. In general, measurement of asymmetry from a photograph seems appropriate because it is noninvasive and also soft tissue appearance is taken into account. However, if necessary, the PA cephalogram can be used as an alternative. It would be helpful to compare both the validity and the repeatability of the digitization of the mandibular outline from PA cephalometric radiographs with other techniques involving the same film. CONCLUSIONS
1. A significant relationship was found for area, compactness, and moment ratios measured from photographs and PA cephalometric radiographs, but not for perimeter ratio. 2. Repeatability studies generally showed less varia-
tion from the mastoid baseline measurements than measurements involving a latero-orbitale baseline. 3. Repeatability levels appeared satisfactory for all 3 techniques. 4. Digitization of the mandibular outline from PA radiographs, and the different ratios obtained, would seem a potentially useful way to assess mandibular asymmetry. This approach should be explored further.
REFERENCES 1. Letzer GM, Kronman JH. A posteroanterior cephalometric evaluation of craniofacial asymmetry. Angle Orthod 1967;37:205-11. 2. Peck S, Peck L, Kataja M. Skeletal asymmetry in esthetically pleasing faces. Angle Orthod 1991;6l:43-7. 3. Skolnick J, Iranpour B, Westesson P-L, Adair S. Prepubertal trauma and mandibular asymmetry in orthognathic surgery and orthodontic patients. Am J Orthod Dentofacial Orthop 1994;105:73-7. 4. Severt TR, Proffit WR. Postsurgical stability following correction of severe facial asymmetry. Int J Adult Orthod Orthognath Surg 1997;12:251-61. 5. Basyouni AA, Nanda SK. An atlas of the transverse dimensions of the face. Ann Arbor: University of Michigan; 2000. 6. Chebib FS, Chamma AM. Indices of craniofacial asymmetry. Angle Orthod 1981;51:214-26. 7. Grummons DC, Kappeyne Van De Coppello MA. A frontal asymmetry analysis. J Clin Orthod 1987;21:448-65. 8. Ricketts RM, Bench RW, Hilgers JJ, Schulhof R. An overview of computerized cephalometrics. Am J Orthod 1972;61:1-28. 9. Svanholt P, Solow B. Assessment of midline discrepancies on the posteroanterior cephalometric radiograph. Trans Eur Orthod Soc 1977;25:261-8.
174 Edler, Wertheim, and Greenhill
10. Mongini F, Schmid W. Treatment of mandibular asymmetries during growth: a longitudinal study. Eur J Orthod 1987;9:51-67. 11. Schmid W, Mongini F, Felisio A. A computer-based assessment of structural and displacement asymmetries of the mandible. Am J Orthod Dentofacial Orthop 1991;100:19-34. 12. Hewitt AB. A radiographic study of facial asymmetry. Br J Orthod 1975;2:37-40. 13. Athanasiou AE, Miethke RR, Van der Meij AJW. Random errors in localisation of landmarks in postero-anterior cephalograms. Br J Orthod 1999;26:273-83. 14. Cook JT. Asymmetry of the cranio-facial skeleton. Br J Orthod 1980;7:33-8. 15. Proffit WR, White RP Jr. Surgical-orthodontic treatment. St Louis: Mosby Year Books; 1991. 16. Athanasiou A. Orthodontic cephalometry. London: MosbyWolfe; 1995. 17. Rose JM, Sadowsky C, BeGole EA, Moles R. Mandibular skeletal and dental asymmetry in Class II subdivision malocclusions. Am J Orthod Dentofacial Orthop 1994;l05:489-95. 18. Arnold TG, Anderson GC, Liljemark WF. Cephalometric norms for craniofacial asymmetry using submental-vertical radiographs. Am J Orthod Dentofacial Orthop 1994;l06:250-6. 19. O’Byrn BL, Sadowsky C, Schneider B, BeGole EA. An evaluation of mandibular asymmetry in adults with unilateral posterior crossbite. Am J Orthod Dentofacial Orthop 1995;107:394-400. 20. Forsberg CM, Burstone CJ, Hanley KJ. Diagnosis and treatment planning of skeletal asymmetry with the submental-vertical radiograph. Am J Orthod 1984;85:224-37. 21. Grayson BH, McCarthy JG, Bookstein F. Analysis of craniofacial asymmetry by multiplane cephalometry. Am J Orthod 1983;84:217-24. 22. Moss JP, Coombes AM, Linney AD, Campos J. Methods of three-dimensional analysis of patients with asymmetry of the face. Proc Finn Dent Soc 1991;87:139-49.
American Journal of Orthodontics and Dentofacial Orthopedics Febuary 2003
23. Severt TS, Proffit WR. The prevalence of facial asymmetry in the dentofacial deformities population at the University of North Carolina. Int J Adult Orthod Orthognath Surg 1997;12:171-6. 24. Coghlan BA, Laitung JKG, Pigott RW. A computer-aided method of measuring nasal symmetry in the cleft lip nose. Br J Plast Surg 1993;46:13-7. 25. Coghlan BA, Matthews B, Pigott RW. A computer-based method of measuring facial asymmetry. Results from an assessment of the repair of cleft lip deformities. Br J Plast Surg 1987;40:371-6. 26. Timoney N, Smith G, Pigott RW. A 20-year audit of nose-tip symmetry in patients with unilateral cleft lip and palate. Br J Plas Surg 2001;54:294-8. 27. Greenhill D, Edler R, Wertheim D, Belsham A, Jones GA. A system for assessing mandibular outlines. International conference on advances in medical signal and information processing. London: Institution of Electrical Engineers, ISBN— 0-85296728-4; 2000. p. 192-9. 28. Edler R, Wertheim D, Greenhill D. Clinical and computerised assessment of mandibular asymmetry. Eur J Orthod 2001;23: 485-94. 29. Edler R, Wertheim D, Greenhill D. Mandibular outline assessment in three groups of orthodontic patients. Eur J Orthod 2002; 24:605-14. 30. Bengel W. Standardisation in dental photography. Int Dent J 1985:35:210-7. 31. Claman L, Patton D, Rashid R. Standardized portrait photography for dental patients. Am J Orthod Dentofacial Orthop 1990; 98:197-205. 32. Gonzalez RC, Woods RE. Digital image processing. Reading, United Kingdom: Addison Wesley Publishing Co: 1993. 33. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:306-10.