Sunteți pe pagina 1din 6

Dentomaxillofacial Radiology (2003) 32, 8792 q 2003 The British Institute of Radiology http:/ /dmfr.birjournals.



Curvature analysis of the mandibular condyle

M Ueda, K Yonetsu, M Ohki, T Yamada, H Kitamori and T Nakamura*
Department of Radiology and Cancer Biology, Nagasaki University School of Dentistry, Japan

Objective: To determine whether curvature analysis on high resolution CT images can be used as a tool for evaluation of mandibular condyle morphology. Methods: Curvature analysis was performed on reconstructed oblique coronal CT images of 634 normal condyles from 317 patients (144 men and 173 women; age range 4 89 years) with inner or middle ear disease. The condyles were scanned with 1 mm collimation using helical CT. The CT images were analysed manually on a personal computer. Results: The condyle CT images could be categorized into ve curvature prole patterns: (1) bipeak; (2) no peak; (3) tri-peak; (4) bi-peak with an intervening bottom above the base line (bi-peak with a col); and (5) bi-peak with an intervening negative phase. A separate evaluation using computer graphic condyle models indicated that these curvature patterns corresponded to at (bipeak), round (no peak), convex (bi-peak with a col), concave (bi-peak with a negative phase) and angled (tri-peak) contours of the condyles superior surface. The curvature proles were identical between bilateral condyles in 40% (126/317) of the patients. Gender-related differences in the incidence of the curvature proles were also found, the bi-peak with a col prole being more frequently observed in women and the bi-peak with a negative phase prole being observed more frequently in men. Conclusion: Curvature analysis on CT images depicts condyle morphology effectively and may be an adjunctive tool for condyle morphometry. Dentomaxillofacial Radiology (2003) 32, 8792. doi: 10.1259/dmfr/23859709 Keywords: tomography, spiral computed; mandibular condyle; image processing, computerassisted; shape analysis Introduction Several studies have attempted to evaluate the morphology of the human condyle. In the 1960s and 1970s, studies were performed mainly on dry skulls and autopsy materials.1 4 These studies used macroscopic observations, radiological cephalometry and tomography. However, the classications of the condyles were fundamentally subjective. In the late 1980s and the 1990s, CT was established as a reliable tool for morphological assessment of hard tissues, and condylar morphology was evaluated using CT.5 In those studies using CT, assessment was performed on axial CT images. As described by Yale et al,3 the mandibular condyles were well characterized by the shape of the superior condylar surface. However, a single use of axial imaging is not sufcient to depict this surface, mainly owing to the poor resolution of reconstructed images.
*Correspondence to: Dr Takashi Nakamura, Department of Radiology and Cancer Biology, Nagasaki University School of Dentistry, 1-7-1 Sakamoto, Nagasaki 852-8588, Japan; E-mail: Received 3 September 2002; revised 30 December 2002; accepted 27 February 2003

The recent introduction of helical CT into the medical eld has greatly improved the ability of conventional CT to generate thin section images, with resultant improvements in the quality of reconstructed 2D and 3D images from axial data. In fact, use of helical CT increased the accuracy of measurements on reconstructed 3D images of the facial bones.6 Therefore, we reasoned that helical CT could be benecial for the evaluation of coronal images of the condyle, and that thin sliced coronal sections of a condyle might provide serial curves of the condyle. The aim of this study was to develop and test a method of curvature analysis to characterize the superior surface of the condyle. Materials and methods Curvature analysis of digital images In Euclidean geometry, curvature is dened as the rate of change in direction. However, in digital images curvature


Condyle shape analysis M Ueda et al

is dened somewhat differently. A digital curve is dened as a sequence of integer-coordinate points p1, p2, , pi. If pi1 is a neighbour of pi xi ; yi ; we can dene the curvature at point pi by the differences xi21 2 xi and yi1 2 yi . However, the successive slope angles on the digital curve differ by multiples of 458, so small changes in the slope are impossible. To solve this problem we used smoothing procedures and dened the slope at point pi as yik 2 yi =xik 2 xi , where k is the smoothing factor and is greater than 1.7 On digital images, curvature at a particular point pi along the curve is dened by factor k, which in turn denes the distance along the curve from point pi. Thus, the points pi2k and pik are equally and oppositely distant from the point pi by k (Figure 1). The curvature at point pi was then calculated by the following equation: cosQ pi pi2k pi pi pik =jpi2k pi jjpi pik j where pi2k pi and pi pik are vectors. The factor k denes the slopes between points pi2k and pi and between pi and pik ; and, in general, a smaller k is more sensitive to change in the curve than a larger k. To determine the curvature of a digital curve accurately, an appropriate k value should be chosen. In this study we tested two different k values, 20 and 50. Curvature analysis of condyle computer graphics models Computer graphics (CG) was used to create models that were used to test whether curvature analysis could differentiate mandibular condyle morphology effectively. Five CG condylar models were used: (1) convex; (2) angled; (3) at; (4) round; and (5) concave. This categorization was similar to that proposed by Yale et al,3 except for the concave type. These models were created using a 512 512 array and Adobew Photoshop 5.5 software (Adobe Systems,

San Jose, CA) on a personal computer (Macintosh G4; Apple Computer, Cupertino, CA). Curvature analysis of the human mandibular condyle Six hundred and thirty-four helical CT images were analysed from 317 patients (4 89 years of age; 144 men and 173 women) with middle or inner ear disease. The temporomandibular joints of these patients were otherwise normal as judged by axial CT images, medical histories and inquiries. Approval for the study was given by the local Ethics Committee. The patients were scanned using a HighSpeed Advantage SG CT imaging system (General Electric Medical Systems, Milwaukee, WI). Scanning orientation was parallel to the Frankfort horizontal line. Scanning was performed with a collimation of 1, pitch of 1:1, 512 512 matrix and display eld of view of 18 cm. Obtained data were transferred to an Advantage Windows Workstation (General Electric Medical Systems), and planar and 2D reconstructed oblique coronal images of the condyles, the orientation of which was along the long axis of the condylar head, were stored for curvature analysis. The following procedures for curvature analysis were performed on the personal computer. The image extraction protocol is shown in Figure 2. Briey, oblique coronal images of the condyles, which were hard copied on lm, were scanned using a atbed scanner (Epson, Tokyo, Japan) at 8-bit resolution. These images were converted to binary data and skeletized using NIH image 1.62 (National Institute of Health, Bethesda, MD). The curvature proles of the condyles were obtained along the thinned condylar surfaces and were analysed as waveform patterns.

Results Curvature analysis of condyle CG models First, we tested whether the curvature analysis used in the present study depicts condylar morphology effectively. To do this we used ve condyle CG models (Figure 3A E). The at condyle model yielded two distinctive peaks after

Figure 1 Schematic illustration of curvature (QPi) at point Pi on digital images. Points Pi2k and Pik are located on the curve and are separated from Pi by a distance equal to the value of k
Dentomaxillofacial Radiology

Figure 2 Flow chart showing the image extraction protocol for curvature prole analysis of reconstructed CT images of human condyles

Condyle shape analysis M Ueda et al


Curvature analysis of human condyles Given that curvature analysis can characterize condyle morphology effectively, we next asked whether this method could delineate reconstructed CT images of human condyles as effectively as in vitro assessment. We found that the human condyles could be categorized into ve patterns by curvature analysis, namely bi-peak (Figure 4F), no peak (Figure 4G), tri-peak (Figure 4H), bi-peak with a col (Figure 4I) and bi-peak with a negative phase (Figure 4J). As with the CG models, the condyles with the bi-peak prole had at superior surfaces (Figure 4A). The condyles with no peak prole were round (Figure 4B). The condyles with the tri-peak prole had angled surfaces (Figure 4C). The condyles exhibiting the bi-peak prole with a col represented those with convex surfaces (Figure 4D) and the condyles exhibiting the bi-peak prole with a negative phase represented those having concave surfaces (Figure 4E). These results suggest that curvature analysis can effectively delineate coronal CT images of the human condyle. Laterality, and age- and gender-related changes in the curvature pattern Next, we assessed whether laterality exists in the curvature proles between the left and right condyles of individuals. Table 1 shows that the curvature prole patterns coincided between bilateral condyles in only 40% of the 317 patients, but statistical analysis conrmed that the differences between the sides were not signicant (test of marginal homogeneity, P 0:119). Figure 5 shows age- and gender-related changes in the condyle curvature pattern. We found an age-dependent predilection for the condyle curvature prole in women (Kruskal Wallis test, P , 0:001) but not in men. Condyles with bi-peak proles with cols were more frequently observed in women than in men. On the other hand, condyles with bi-peaks and negative phase proles were more frequently observed in men. For example, 21% and 27% of the condyles exhibited bi-peaks with cols and bipeaks with negative phase proles, respectively, in 30 49year-old men, while 49% and 11% of the condyles, respectively, exhibited bi-peaks with cols and bi-peaks with negative phase proles in women of the same age range. In total, 40% and 25% of the condyles in women and men, respectively, had bi-peaks with col proles, and 17% and 26% of the condyles in women and men, respectively, had bi-peaks with negative phase proles. Together, these results were suggestive of asymmetry within an individual, with gender-related differences in the curvature prole pattern of human mandibular condyles. Discussion Through a series of studies on human condyles from a large collection of dry skeletons, Yale et al1 3 categorized condyles into four basic types: attened; convex; angled; and rounded. Curvature analysis on condyle CG models suggested that the round, attened, convex and angled types appear to correspond roughly to the no peak, bi-peak,
Dentomaxillofacial Radiology

Figure 3 Five condyle computer graphic (CG) models (A E) and the curvature proles (F J) corresponding to each CG model. CG condyle models show (A) at, (B) round, (C) angled, (D) convex and (E) concave patterns. The curvature proles represent (F) bi-peak, (G) no peak, (H) tripeak, (I) bi-peak with an intervening col and (J) bi-peak with a negative phase

curvature analysis (Figures 3A and 3F), where the two peaks resulted from the edges on the lateral and medial poles of the condyle. The curvature prole had a slope of 08 between these two peaks, representing a at surface. In contrast, the round condyle model resulted in proles without a peak (Figures 3B and 3G), consistent with a constant rate of change in curvature along the condylar surface. When the condyle model had a central peak (angled type), the curvature analysis yielded three peaks, with two peripheral peaks being higher than a central peak (Figures 3C and 3H). The convex condyle model had two peaks (Figures 3D and 3I), as with the at condyle, but also had an intervening col, which connected the two peaks with a relatively at line above the bottom. The concave condyle model also resulted in proles with two peaks, but in this case the central part of the prole was negative (, 08) (Figures 3E and 3J).


Condyle shape analysis M Ueda et al

Figure 4

Reconstructed oblique coronal CT images of human condyles (A E) and curvature proles corresponding to each of the images (F J)

bi-peak with a col and tri-peak curvature proles, respectively. Yale et al3 found that the convex type was most frequently (58%) observed, then attened (25%) and angled (12%) types. The round type was seen most infrequently (3%). In the present study, condyles exhibiting the bi-peak with a col curvature prole were most frequently (33%) observed, then the tri-peak prole (27%), then bi-peak with a negative phase prole (21%) and nally no peak prole (15%). In particular, condyles exhibiting the bi-peak curvature prole, which may
Dentomaxillofacial Radiology

correspond to those with a attened superior surface in Yales studies, were rarely seen (4%) as judged from the oblique coronal CT images. These results suggest that there is no 1:1 correlation in the categorization systems proposed by Yale et al and by us. In addition, we found that the condyles with the bi-peak with a negative phase prole, which may be called the condyles with the concave surfaces, were commonly seen (21%). This is probably the type that can be categorized only when sectional imaging technology, such as CT, is used.

Condyle shape analysis M Ueda et al


Table 1 Coincidence of curvature proles of the left and right condyles (number of patients with each prole) Left condyle prole BP BP/C TP NP BP/NP Totala BP 3 8 3 1 2 28 Right condyle prole BP/C TP NP 3 38 38 13 10 211 1 30 30 6 10 169 0 9 6 25 2 93 BP/NP 4 24 15 30 30 133

BP, bi-peak; BP/C, bi-peak with a col; TP, tri-peak; NP, no peak; BP/NP, bi-peak with a negative phase a Total number of condyles categorized into the ve curvature prole patterns

In this study we found that a gender-related predisposition may exist in the prevalence of the condyle curvature proles. Some gender-related differences were found in bone mineral density of the condyles, which were thought to be related to differences in mechanical stress from occlusion.8 Therefore, the observed gender-related differences in the condyle curvature proles may also be related to the mechanical stress from occlusion.

In the present study we focused our attention on the assessment of normal variants of condyles. Recent studies have shown that the relative location of the glenoid fossa and the condyle changed after abnormal masticatory function,9 and that the condylar position can be used as a predictor of temporomandibular joint derangement.10 In addition, posterior disk rotation was positively related to the steepness of the articular eminence, implicating the biomechanics relative to the development of disk displacement.11,12 These preceding results collectively suggest that a well balanced position of the condyle relative to the glenoid fossa may be critical to the ordinate function of the temporomandibular joint. When the proper morphological adaptation of the glenoid fossa and the condyle is lost, temporomandibular joint derangement may occur. In this context, curvature analysis of the glenoid fossa in addition to the condyle may be informative for objectively comparing the morphologies of these structures to assess whether the shape of the condyle matches that of the glenoid fossa. Such studies are now being performed. In conclusion, curvature analysis on CT images depicts condyle morphology effectively and may be an adjunctive tool for condyle morphometry.

Figure 5 Age- and gender-related changes in condyle curvature patterns. The patient population was divided into eight decades of life. Each of the tandem bars in each age group represents percentage of subjects. BP, bi-peak; NP, no peak; TP, tri-peak; BP/C, bi-peak with an intervening col; BP/NP, bi-peak with negative phase
Dentomaxillofacial Radiology


Condyle shape analysis M Ueda et al

1. Yale SH, Rosenberg HM, Ceballos M, Hauptfuehrer JD. Laminagraphic cephalometry in the analysis of mandibular condyle morphology. A preliminary report. Oral Surg Oral Med Oral Pathol 1961; 14: 793 805. 2. Yale SH, Ceballos M, Kresnoff CS, Hauptfuehrer JD. Some observations on the classication of mandibular condyle types. Oral Surg Oral Med Oral Pathol 1963; 16: 572 577. 3. Yale SH, Allison BD, Hauptfuehrer JD. An epidemiological assessment of mandibular condyle morphology. Oral Surg Oral Med Oral Pathol 1966; 21: 169 177. berg T. Arthrosis and deviation in form in the 4. Hansson T, O temporomandibular joint. A macroscopic study on human autopsy material. Acta Odontol Scand 1977; 35: 167 174. 5. Eisenburger M, Haubitz B, Schmelzeisen R, Wolter S, Tschernitschek H. The human mandibular intercondylar angle measured by computed tomography. Arch Oral Biol 1999; 44: 947 951. 6. Kitaura H, Yonetsu K, Kitamori H, Kobayashi K, Nakamura T. Standardization of 3-D CT measurements for length and angles by matrix transformation in the 3-D coordinate system. Cleft Palate Craniofac J 2000; 37: 349 356. 7. Rosenfeld A, Johnston E. Angle detection on digital curves. IEEE Trans Comput 1973; C-22: 875 879. 8. Yamada M, Ito M, Hayashi K, Sato H, Nakamura T. Mandibular condyle bone mineral density measurement by quantitative computed tomography: a gender-related difference in correlation to spinal bone mineral density. Bone 1997; 21: 441 445. 9. Poikela A, Pirttiniemi P, Kantomaa T. Location of the glenoid fossa after a period of unilateral masticatory function in young rabbits. Eur J Orthod 2000; 22: 105 112. 10. Bonilla-Aragon H, Tallents RH, Katzberg RW, Kyrkanides S, Moss ME. Condyle position as a predictor of temporomandibular joint internal derangement. J Prosthet Dent 1999; 82: 205 208. 11. Isberg A, Westesson PL. Steepness of articular eminence and movement of the condyle and disk in asymptomatic temporomandibular joints. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1998; 86: 152 157. 12. Major PW, Kinniburgh RD, Nebbe B, Prasad NG, Glover KE. Tomographic assessment of temporomandibular joint osseous articular surface contour and spatial relationships associated with disc displacement and disc length. Am J Orthod Dentofacial Orthop 2002; 121: 152 161.

Dentomaxillofacial Radiology