Dentomaxillofacial Radiology (2003) 32, 80–86 q 2003 The British Institute of Radiology http://dmfr.birjournals.org
RESEARCH
Accuracy of panoramic radiography in assessing the dimensions of radiolucent jaw lesions with distinct or indistinct borders V Chuenchompoonut, M Ida, E Honda, T Kurabayashi and T Sasaki* Department of Oral and Maxillofacial Radiology, Graduate School, Tokyo Medical and Dental University, Tokyo, Japan
Objectives: To evaluate the accuracy of panoramic radiography (PR) in assessing the dimensions of mandibular lesions. Methods: One hundred and fifteen cases exhibiting radiolucent lesions in the mandibular premolar, molar or ramus area were selected retrospectively. They were divided into four types: cyst with sclerotic margin (Type I); cyst without sclerotic margin (Type II); ameloblastoma (Type III); or squamous cell carcinoma (Type IV). Maximum mesiodistal length and superoinferior height were measured on PR (Lpmax and Hpmax, respectively) and on CT (Lcmax and Hcmax, respectively) and the results were compared. Results: Correlation coefficients between Lpmax and Lcmax and between Hpmax and Hcmax were high for Type I, II and III lesions but were significantly lower for Type IV lesions. Regression coefficients between Lpmax and Lcmax and between Hpmax and Hcmax were indistinguishable from 1.0 for all types of lesions. The mean relative difference between Lpmax and Lcmax varied from 1.2% to 8.2%. The difference was only 2 0.3% for larger lesions of combined Types I and II. The mean relative difference between Hpmax and Hcmax varied from 2 3.5% to 1.1% depending on the type of lesions. Conclusion: PR is accurate for assessing the dimensions of radiolucent lesions in the posterior mandible when the margins are well defined. Dentomaxillofacial Radiology (2003) 32, 80–86. doi: 10.1259/dmfr/29360754 Keywords: radiography, panoramic; mandible; tomography, X-ray computed Introduction Panoramic radiography (PR) has been widely used for obtaining a comprehensive overview of the maxillofacial complex. One of the limiting factors in the clinical use of PR is the uncertainty regarding the actual dimensions of structures given their radiographic appearance. Although a number of studies have shown that measured dimensions on PR are reliable,1 – 6 there is no clinical comparison between measured dimensions on PR and those obtained from three-dimensional structures. CT is one of the most useful modalities for assessing jawbone lesions. It clearly depicts soft tissues and hard tissues without superimposition of anatomical structures.7 – 9 Moreover, CT scans exhibit no magnification and no geometric distortion.10 Within the limits of its spatial resolution, CT is considered more reliable than *Correspondence to: Takehito Sasaki, Department of Oral and Maxillofacial Radiology, Graduate School, Tokyo Medical and Dental University, 5-45, Yushima 1 chome, Bunkyo-Ku, Tokyo 113-8549, Japan; E-mail:
[email protected] Received 20 August 2002; revised 9 January 2003; accepted 20 February 2003
conventional projection radiography as a morphometric tool.11 – 14 However, the disadvantages of CT, including high cost, limited availability and high patient dose, make this imaging modality unsuitable for routine dental applications. The purpose of this study was to evaluate the accuracy of PR in assessing the mesiodistal length and the superoinferior height of mandibular lesions compared with those measured on CT images. Materials and methods Patients All patients who had been examined with CT from September 1996 to September 2001 at the Department of Oral and Maxillofacial Radiology, Tokyo Medical and Dental University Dental Hospital, were retrospectively reviewed. Approval by the Ethics Committee, Tokyo Medical and Dental University, was granted for this study.
Dimension of jaw lesions V Chuenchompoonut et al
Patients who had radiolucent lesions in the premolar, molar or ramus region of the mandible were selected. Patients who had no PR, no histopathological diagnosis and who had previous surgical treatment for the lesion were excluded. CT images of poor quality resulting from metal restorations were also excluded. One hundred and fifteen cases were selected, comprising 66 males and 49 females with a mean age of 44.5 years (range 6 – 83 years). Cases were classified into four types according to the histopathological diagnosis and the radiographic appearance of the lesion border (Table 1). Cysts were defined as having a sclerotic margin if at least 75% of the margin appeared sclerotic. Otherwise they were defined as cysts without sclerotic margin. Imaging CT examinations had been performed using a Somatom Plus S Scanner (Siemens Medical Systems, Erlangen, Germany) operated at 120 – 137 kVp, 75 – 125 mA using a spiral scanning mode. A slice thickness of 1 mm was used for 101 cases and a slice thickness of 2 mm was used for 14 cases. Axial CT images were reconstructed parallel to the occlusal plane with a reconstruction increment of 1 mm using a bone contrast algorithm. Only CT images without contrast enhancement were used for the present investigation. They were printed on KODAK DryView DVB laser imaging film with a Laser Imager Plus DryView 8700 printer (Kodak Co. Ltd., Tokyo, Japan). Digital panoramic radiographs were taken using a Super Veraview Epocs (J. Morita Corporation, Kyoto, Japan) operated at 60 – 80 kVp and 5 – 10 mA using photostimulable phosphor plates (ST III; Fuji Film Co. Ltd., Kanagawa, Japan). The plates were processed with a FCR 7000 system (Fuji Film Co. Ltd., Kanagawa, Japan). PR images were printed on Fuji medical dry imaging film DI-AL with a Laser Imager CR-DP L printer (Fuji Film Co. Ltd., Tokyo, Japan). Radiographic measurement Measurement on CT images All CT measurements were performed on the printed images with a window width of 2500 and a centre of 500. An axial CT image showing the maximum mesiodistal length (Lcmax) of the lesion was selected. The length was
81
first defined on the monitor by drawing a line connecting the most mesial and distal points of the lesions. A CT density profile in Hounsfield units (HU) was plotted for every unit of 3 pixels along this line. The border of the lesion was defined as a point half the maximum value of the CT density profile, as shown in Figure 1. The distance between the two points of the half maximum values was measured on the printed image using calipers with a precision of 0.05 mm. The measured value was calibrated with the magnification factor of the CT density profile scale. Maximum superoinferior height on CT (Hcmax) was obtained as the distance between two table positions (TPs) of the CT unit providing images that showed the superior border and the inferior border of the lesions. Measurement on PR (Figure 2) All measurements on PR were done on the laser printed film. The measurement level for the maximum mesiodistal length on PR (Lpmax) was selected such that it matched the plane of the CT image used for measurement of Lcmax. Since the axial CT image was taken parallel to the occlusal
Figure 1 Example of CT measurement procedure. The locations of point A and point B are defined by points at half the maximum density value shown in the CT density line profile. The distance between these points represents the maximum mesiodistal length (Lcmax) of the lesion
Table 1 Classification of lesion types based on histopathological diagnosis and radiographic appearance Lesion
Histopathological diagnosis
Type I
Cyst with sclerotic margin Radicular cyst Odontogenic keratocyst Dentigerous cyst Cyst without sclerotic margin Radicular cyst Odontogenic keratocyst Dentigerous cyst Aneurysmal bone cyst Ameloblastoma Squamous cell carcinoma
Type II
Type III Type IV Total
Number of cases 45 4 13 28 25 6 5 13 1 20 25 115
Figure 2 Example of panoramic radiography (PR) measurement procedure. Dp, distance between the measurement level of the axial plane and the occlusal plane determined from the corresponding table position on the CT image; Lpmax, maximum mesiodistal length of the lesion on PR; Hpmax, maximum superoinferior height of the lesion on PR Dentomaxillofacial Radiology
Dimension of jaw lesions V Chuenchompoonut et al
82
plane, the distance (Dp) between the measurement level of the axial plane of the axial images and the occlusal plane was determined from the difference in TP. Dp was adjusted for PR using a site-specific magnification factor provided by the manufacturer. The magnification factor varied from a maximum of 1.30 in the incisor region to a minimum of 1.25 at the condyle. These magnification factors were confirmed by our own experiment using a phantom with steel balls. The value corresponding to the central portion of the lesion was applied as the magnification correction factor. The borders of the lesion were determined by the observer. Lpmax was measured using a pair of calipers and corrected for magnification. The superior and inferior borders of the lesion were determined by the observer and the maximum superoinferior height on PR (Hpmax) was measured as the distance between these borders as measured with calipers. The measured value was corrected for the same magnification factor as used for Lpmax. All primary radiographic measurements were performed by one observer (VC). Intraobserver and interobserver variability In a test – re-test analysis designed to assess intraobserver variation, 37 of 45 Type I cases were arbitrarily selected for measurement of mesiodistal length. Only 37 cases scanned with a slice thickness of 1 mm were selected for measurement of superoinferior height. All 25 cases of Type IV lesions were re-measured within 1 month by one observer (VC) on both PR and CT. To determine interobserver variability, three observers (VC, MI and EH) measured the maximum mesiodistal length and superoinferior height in each of 10 cases of Type I and Type IV lesions on PR and CT. All the selected cases were CT scanned with a slice thickness of 1 mm. Statistical analyses Correlation of Lpmax with Lcmax and of Hpmax with Hcmax were analysed by determining the correlation coefficient (R) and the regression coefficient with a 95% confidence interval (95% CI). Mean relative differences between observed and expected values on PR from the regression line were also analysed for every type of lesion. Differences between PR and CT measurements and intraobserver variability were assessed using a paired t-test with P , 0:05 implying statistical significance. The coefficient of variation (CV%) was calculated as CVð%Þ ¼ ðSi =Xmean Þ £ 100; where S i is the standard deviation of the difference of the first (X1i) and the second (X2i) measurement expressed as vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi uX un Si ¼ t ðX1i 2 X2i Þ2 =2n
Statistical analysis for interobserver variation was performed by ANOVA with P , 0:05 implying statistical significance. Results The results of the various measurements, including their standard deviations, are shown in Table 2. Figure 3 shows scatter plots of Lpmax and Lcmax with regression lines analysed for every type of lesion. All correlations in Figure 3 are statistically significant (Table 3), with the correlation coefficient for Type IV lesions being somewhat smaller than those for the other types of lesions. Table 3 also shows the regression coefficients with 95% CI between Lpmax and Lcmax for each type of lesion. Mean relative differences between observed and expected Lpmax values calculated from the regression line ðY ¼ XÞ are also shown in Table 3. The regression coefficients for all four types of lesion ranged from 0.999 to 1.031 and were not significantly different from 1.0. Paired t-test analysis indicated that Lpmax was not significantly different from Lcmax for each type of lesion. Mean relative differences ranged from 1.2% to 8.2%, with a relatively wide 95% CI. The observed Lpmax was generally higher than the expected Lpmax for Type I and II lesions with a length smaller than the mean length. Since all statistical values shown in Table 3 are similar for Type I and Type II lesions, the combined Type I and II lesions were re-analysed comparing the smaller lesions with the larger lesions. The results are shown in Table 4, which shows that the correlation coefficient was significantly lower for smaller lesions than for larger lesions. Paired t-test also showed that the difference between Lpmax and Lcmax is significant only for smaller lesions. The mean relative difference between observed and expected Lpmax values was þ 11.0% for smaller lesions, with a wide CI, and only 2 0.3% for larger lesions, with a narrow CI. Figure 4 shows the scatter plots and regression lines for Hpmax and Hcmax for each type of lesion. All correlations were statistically significant, with the correlation coefficient being higher for Type I, II and III lesions and significantly lower for Type IV lesions (Table 5). The regression coefficients for all four types of lesions were not significantly different from 1.0. Paired t-test analysis indicated that there was no significant difference between Table 2 Mesiodistal length and superoinferior height measured on CT and on panoramic radiography (PR). Values are mean ^ SD (range in parenthesis) Lesion Type I
i¼1
with n being the number of cases. Xmean is the mean of all values of X1i and X2i expressed as n X Xmean ¼ ðX1i þ X2i Þ=2n i¼1
Dentomaxillofacial Radiology
Type II Type III Type IV
Mesiodistal length (mm) CT (Lcmax) PR (Lpmax)
Superoinferior height (mm) CT (Hcmax) PR (Hpmax)
24.3 ^ 12.2 (7.4 –58.4) 22.5 ^ 10.6 (7.6 –45.7) 31.8 ^ 14.0 (6.7 –64.6) 22.8 ^ 9.8 (4.8 –43.3)
27.6 ^ 12.4 (8.0 – 80.0) 22.9 ^ 9.3 (9.0 – 55.0) 35.1 ^ 14.0 (10.0– 64.0) 15.8 ^ 7.3 (7.0 – 34.0)
24.9 ^ 11.2 (8.2 –55.1) 23.2 ^ 10.1 (9.7 –47.1) 31.9 ^ 15.2 (8.2 –68.9) 23.8 ^ 9.3 (5.4 –39.1)
27.2 ^ 11.8 (10.9 – 78.7) 22.7 ^ 8.8 (9.0 –50.4) 33.5 ^ 12.9 (11.1 – 59.5) 16.4 ^ 7.3 (4.4 –29.9)
Dimension of jaw lesions V Chuenchompoonut et al
83
Figure 3 Scatter plot with regression line ðY ¼ aXÞ for mesiodistal lengths measured on panoramic radiography (PR) (Lpmax) and on CT (Lcmax) for Types I, II, III and IV lesions (see also Table 3)
Table 3 Correlation between mesiodistal length on panoramic radiography (Lpmax) and on CT (Lcmax), with mean relative difference Lesion
R
Type I
0.958p
Type II
0.950
Type III
0.950
Type IV
0.885p
Regression coefficient (a) (95% CI)
Mean % difference (95% CI)
0.999 (0.963 – 1.034) 1.011 (0.958 – 1.063) 1.009 (0.945 – 1.074) 1.031 (0.960 – 1.102)
6.3 (1.8 –10.8) 7.5 (20.1 – 15.1) 1.2 (26.7 –9.1) 8.2 (0.1 –16.3)
R, correlation coefficient; 95% CI, 95% confidence interval Regression equation: Y ¼ aX, where Y ¼ Lpmax and X ¼ Lcmax Mean % difference is the difference between observed and expected Lpmax calculated from the regression line p Significant difference between Type I and Type IV lesions at P , 0.05
Hpmax and Hcmax for Type I, II and IV lesions, but a significant difference for Type III lesions. Mean relative differences ranged from 2 3.5% to þ 1.1% depending on the type of lesion. These differences were smaller than those observed for the mesiodistal lengths. The combined Type I and II lesions were again reanalysed to compare the smaller with the larger lesions, as shown in Table 4. In this analysis, cases scanned with more than 1 mm slice thickness were excluded and only the cases scanned with a slice thickness of 1 mm were used. The correlation coefficients for these two groups were quite similar and the regression coefficients were indistinguishable from 1.0. Paired t-test also showed that there was no
significant difference between Hpmax and Hcmax. Mean relative difference between observed and expected Hpmax values was þ 2.1% for smaller lesions and 2 1.9% for larger lesions. There was no significant difference between the values from smaller and larger lesions. Intraobserver variability for the maximum mesiodistal length and the superoinferior height was analysed by paired t-test for Type I and Type IV lesions. These lesions are representative of lesions with distinct borders and indistinct borders, respectively. There were no significant intraobserver differences between the first and the second measurements for all dimensions on PR and CT. The results are shown in Table 6 as CV%. All values of CV% for intraobserver variability were less than 4% for Type I lesions. The CV% values for Type IV lesions were larger than those for Type I lesions but were less than 5%, which corresponds to a maximum of 1.1 mm. The CV% was smaller for Lcmax than for Lpmax for both Type I and Type IV lesions. CV% was similar for Hcmax and Hpmax. Interobserver variability results are shown in Table 7. There was no significant difference in the mesiodistal lengths of Lpmax and Lcmax among the three observers for Type I lesions. The mean differences between the minimum and the maximum values among the three observers were 1.1 mm and 2.1 mm for Lpmax and Lcmax of Type I lesions, respectively. Differences between observers in measuring Lpmax for Type IV lesions were statistically significant. The mean difference between the minimum and maximum values was approximately 10 mm. The mean difference in Lcmax measurements for Type IV lesions was also large, but was not statistically significant. Differences Dentomaxillofacial Radiology
Dimension of jaw lesions V Chuenchompoonut et al
84 Table 4
Correlation for Type I and Type II lesions combined
Correlation Lpmax and Lcmax
Hpmax and Hcmax
p
Size of lesions
No. of cases
R
smaller
35
0.820p
larger
35
0.928
smaller
31
0.951
larger
31
0.962
Significant difference between smaller and larger lesions at P , 0:05;
pp
Regression coefficient (a) (95% CI)
Mean % difference (95% CI)
1.058 (0.998 –1.118) 0.986 (0.950 –1.021) 1.018 (0.991 –1.044) 0.969 (0.944 –0.994)
11.0pp (4.3 – 17.7) 20.3 (24.0 – 3.4) 2.1 (21.4 – 5.6) 21.9 (24.5 – 0.7)
significant difference between Lpmax and Lcmax at P , 0:02
Figure 4 Scatter plots with regression line (Y ¼ aX) for superoinferior heights measured on panoramic radiography (PR) (Hpmax) and on CT (Hcmax) for Types I, II, III and IV lesions (see also Table 5)
between observers in measuring the superoinferior height were all statistically significant except for the Hcmax measurements of Type IV lesions. Mean differences for the superoinferior heights of Type IV lesions were again larger
than those for Type I lesions for both Hpmax and Hcmax, but was significant only for Hpmax. Discussion
Table 5 Correlation between superoinferior heights on panoramic radiography (Hpmax) and on CT (Hcmax), with mean relative difference Lesion
R
Type I
0.986
Type II
0.982
Type III
0.990
Type IV
0.940p
Regression coefficient (a) (95% CI)
Mean % difference (95% CI)
0.977 (0.958 – 0.997) 0.983 (0.955 – 1.011) 0.949 (0.927 – 0.972) 1.000 (0.942 – 1.059)
20.4 (23.1 –2.3) 20.3 (23.1 –2.5) 2 3.5pp (26.9 –0.1) 1.1 (26.2 –8.4)
R, correlation coefficient; 95% CI, 95% confidence interval Regression equation: Y ¼ aX, where Y ¼ Hpmax and X ¼ Hcmax p Significant difference between Type IV lesion and any other lesion at P , 0:05; pp significant difference between Hpmax and Hcmax at P , 0:05 Dentomaxillofacial Radiology
PR has been widely used for the detection and evaluation of jaw lesions in the primary dental care setting. PR would Table 6 Intraobserver variability expressed as coefficient of variation (CV%) for mesiodistal length and superoinferior height (P values from paired t-tests in parenthesis) Lesion
Mesiodistal length PR (Lpmax) CT (Lcmax)
Superoinferior height PR (Hpmax) CT (Hcmax)
Type I (37 cases) Mean ^ SDa Type IV (25 cases) Mean ^ SDa
3.7 (0.704) 25.0 ^ 11.9 4.5 (0.100) 23.8 ^ 9.3
3.0 (0.721) 26.3 ^ 11.0 4.8 (0.971) 16.4 ^ 7.3
a
1.7 (0.681) 24.8 ^ 13.1 3.2 (0.196) 22.8 ^ 9.8
3.7 (0.579) 26.6 ^ 11.8 4.9 (0.600) 15.8 ^ 7.3
Mean and standard deviation (mm) of the first measurement
Dimension of jaw lesions V Chuenchompoonut et al
Table 7
Interobserver variability analysed by ANOVA with mean difference between observers Mesiodistal length
Lesion Type I (10 cases) Mean ^ SDa Mean difference ^ SDb Type IV (10 cases) Mean ^ SDa Mean difference ^ SDb a b p
85
Superoinferior height PR (Hpmax) CT (Hcmax)
PR (Lpmax)
CT (Lcmax)
NS
NS
P , 0.05p
P , 0.05p
26.5 ^ 13.0 1.1 ^ 0.4 (0.4 – 1.9)
27.4 ^ 14.6 2.1 ^ 2.7 (0.2 –8.7)
25.3 ^ 9.9 2.0 ^ 1.9 (0.2 – 6.5)
25.0 ^ 10.7 1.4 ^ 0.7 (0.0 – 2.0)
P , 0.05p
NS
P , 0.01p
NS
26.1 ^ 8.9 10.8 ^ 9.9 (0.3 –26.9)
25.2 ^ 7.9 9.0 ^ 10.2 (2.1 – 31.9)
18.6 ^ 7.8 6.1 ^ 5.2 (0.9 – 16.8)
18.5 ^ 8.7 4.7 ^ 3.3 (0.0 –10.0)
Mean and standard deviation (mm) from one observer (VC) Mean difference (mm) between the minimum and maximum values among three observers ^ standard deviation (range in parenthesis) Significant difference at P , 0:05; NS, not significant
clinically be most valuable for surgical planning or follow-up observation of jaw bone lesions,7,9,15,16 dental implants, 8,12 – 14 bone transplants17,18 or craniofacial deformity19 – 21 if it has the potential for assessing the true dimension of bone lesions. Reliability of PR measurements has been evaluated by comparing measured values with those from phantoms,1 – 3 dry mandibles4 – 6 or autopsy specimens.8,14 Although direct measurements of the lesion dimensions are generally possible in these experiments, it is difficult to precisely match them with measurements on PR images. This is mainly because the object for imaging is three-dimensional and the PR image is a two-dimensional projected tomographic image. Thus, any dimensions measured on PR images vary depending on the projection geometry of the panoramic machine and on patient positioning. In clinical experiments, measurements obtained during surgery12 or on post-surgical specimens16 could be used as a gold standard. However, it is not always possible to obtain the real dimensions of large cysts or ameloblastomas, because a number of patients with such lesions are treated by enucleation, marsupialization or bone morphogenic therapy. CT, a non-invasive imaging method allowing threedimensional measurements, depicts both soft tissue and hard tissue changes without superimposition of anatomical structures.7 – 9 As a morphometric tool, CT has been considered more reliable than projection radiography because CT scans exhibit no magnification or geometric distortion.10 – 14 Although the spatial resolution of CT is lower than the spatial resolution of PR, it is reasonably high in the scan plane with modern equipment. A detection limit of 0.35 mm has been claimed by the manufacturer of the CT scanner used in the present study. In this study, the superoinferior height measurements on CT images were estimated from differences in TP of those axial images showing the superior and inferior margins of the lesions. The measurement error using this method is at least equivalent to the axial slice thickness. In the present study, both mesiodistal length and superoinferior height were measured on PR. The measurements were made at a level matching the measurements on the corresponding CT images, as shown in Figure 2. CT images were reconstructed with an axial slice thickness of
1 mm using a bone contrast algorithm. This enabled the observers to identify corresponding features on PR images. The borders of the lesion were defined as points where the CT density was half the maximum. With correction for magnification, this method allowed a more direct comparison between PR measurements and CT measurements, which is otherwise impossible under clinical conditions. It may still be argued that the horizontal dimensions of the image recorded with PR will be magnified or minified depending on the object depth in relation to the central plane of the focal trough and the object angulation relative to the horizontal plane. Several investigators have reported that the horizontal distance measured on PR is unreliable owing to non-linear variation in the magnification at different object depths,1 – 3 whereas vertical distance is relatively reliable providing that the patient is properly positioned.1 – 5 However, horizontal length measurements in the mandible on PR have been reported to correspond with dry mandible measurements as long as the measurements do not cross the mid-sagittal plane.6 Therefore, in this study we selected lesions located in the premolar region, molar region or the ramus of the mandible. Correlations between mesiodistal lengths measured on PR and CT images were highly significant for all types of lesions. The regression coefficients were also indistinguishable from 1.0. The mean relative difference between Lpmax and Lcmax varied from 1.2% to 8.2%, which is roughly comparable with the reported difference (3.3 – 7.0%) between PR measurements and dry mandible measurements.6 The correlation coefficient was higher for Type I, II and III lesions, but significantly lower for Type IV lesions. This may be the result of poor visualization of the margins of the Type IV lesions as a result of bone invasion.15,16 This result was supported by a significant interobserver variability in PR measurements of Type IV lesions. The mean difference between the minimum and maximum values shown in Table 7 was 1.1 – 2.1 mm for Type I lesions and 9.0 – 10.8 mm for Type IV lesions, the latter also having a wider range. Since the borders of Type IV lesions have no clear sclerotic margin, measurements are more subjective and less reliable. Dentomaxillofacial Radiology
Dimension of jaw lesions V Chuenchompoonut et al
86
Intraobserver variability was not significant for either mesiodistal or superoinferior measurements. All CV% were less than 5% for both imaging modalities. However, CV% was always smaller for Type I lesions than for Type IV lesions. Thus, measurement of a single observer is relatively reproducible with either method. Since correlation results were quite similar for Type I and Type II lesions, these two types were combined. The combined group was divided into two groups using a lesion size criterion: one-half with smaller lesions and one-half with larger lesions (Table 4). The correlation coefficient between Lpmax and Lcmax was significantly higher for larger lesions than for smaller lesions. A possible explanation might be the relative effect of spatial resolution in Lcmax measurements. A spatial resolution limit may yield a 5% error for a 10 mm lesion, while causing a 1% error for a 50 mm lesion. Uncertainty in determining the maximum mesiodistal length on CT images may also be larger as a result of the subjective evaluation in using a monitor. Quantitative evaluation of the effect of uncertainty in determining Lcmax was not possible in this study. Thus, a low correlation coefficient for smaller lesions does not necessarily imply that the dimensions measured on PR are less accurate than on CT.
Re-analysis of correlation for superoinferior heights showed no significant difference between smaller and larger lesions. The correlations were higher than those for mesiodistal length, as shown in Table 4. This is probably owing to a different definition of maximum superoinferior height compared with the definition of mesiodistal length. The overall error may be caused largely by the slice thickness and the reconstruction increment, both of which were 1 mm in this study. In fact, exclusion of images with a slice thickness of 2 mm showed that the correlation coefficients for superoinferior height are increased. The regression coefficients were again indistinguishable from 1.0 for all types of lesions. The paired t-test between Hpmax and Hcmax had previously shown a significant difference for Type III lesions (Table 5), but exclusion of images taken with a 2 mm slice thickness resulted in no significant difference for all types of lesions (data not shown). In conclusion, measurements of the horizontal and vertical dimensions of bone lesions in the posterior mandible on panoramic radiographs are accurate when the margins of the lesion are well defined.
References 1. Tronje G, Welander U, McDavid WD, Morris CR. Horizontal and vertical magnification. In: McDavid WD (ed). Imaging characteristics of seven panoramic x-ray units. Dentomaxillofac Radiol (Suppl) 1985; 8: 29 – 34. 2. Tronje G, Welander U, McDavid WD, Morris CR. Image distortion in rotational panoramic radiography. I: general considerations. Acta Radiol Diagn (Stockh) 1981; 22: 295 – 299. 3. Tronje G, Eliasson S, Julin P, Welander U. Image distortion in rotational panoramic radiography. II: vertical distances. Acta Radiol Diagn (Stockh) 1981; 22: 449 – 455. 4. Batenburg RHK, Stellingsma K, Raghoebar GM, Vissink A. Bone height measurements on panoramic radiographs: the effect of shape and position of edentulous mandibles. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1997; 84: 430– 435. 5. Larheim TA, Svanaes DB. Reproducibility of rotational panoramic radiography: mandibular linear dimensions and angles. Am J Orthod Dentofacial Orthop 1986; 90: 45 – 51. 6. Catic A, Celebic A, Valentic-Peruzovic M, Catovic A, Jerolimov V, Ivana M. Evaluation of the precision of dimensional measurements of the mandible on panoramic radiographs. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1998; 86: 242– 248. 7. Cavalcanti MGP, Ruprecht A, Quets J. Progression of maxillofacial squamous cell carcinoma evaluated using computer graphics with spiral computed tomography. Dentomaxillofac Radiol 1999; 28: 238 –244. 8. Lindh C, Petersson A, Klinge B. Measurements of distances related to the mandibular canal in radiographs. Clin Oral Implants Res 1995; 6: 96 – 103. 9. Close LG, Merkel M, Burns DK, Schaefer SD. Computed tomography in the assessment of mandibular invasion by intraoral carcinoma. Ann Otol Rhinol Laryngol 1986; 95: 383 –388. 10. Vannier MW, Hildebolt CF, Conover G, Knapp RH, YokoyamaCrothers N, Wang G. Three-dimensional dental imaging by spiral CT. A progress report. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1997; 84: 561 – 570. 11. Kalender WA. Image quality. In: Kalender WA (ed). Computed tomography. Munich, Germany: Publicis MCD Verlag, 2000, pp 82 –118.
Dentomaxillofacial Radiology
12. Wovern N. In vivo measurement of bone mineral content of mandibles by dual-photon absorptiometry. Scand J Dent Res 1985; 93: 162 –168. 13. Tol H, Moses O. A comparison of panoramic radiography with computed tomography in the planning of implant surgery. Dentomaxillofac Radiol 1991; 20: 40 – 42. 14. Bjorn K, Arne P, Pavel M. Location of the mandibular canal: comparison of macroscopic findings, conventional radiograph, and computed tomography. Int J Oral Maxillofac Implants 1989; 4: 327 – 332. 15. Brown JS, Griffith JF, Phelps PD, Browne RM. A comparison of different imaging modalities and direct inspection after periosteal stripping in predicting the invasion of the mandible by oral squamous cell carcinoma. Br J Oral Maxillofac Surg 1994; 32: 347 – 359. 16. Store G, Larheim TA. Mandibular osteoradionecrosis: a comparison of computed tomography with panoramic radiography. Dentomaxillofac Radiol 1999; 28: 295 –300. 17. Lim KC, Nabil S, Antonio CK, Henk T. Mandibular reconstruction with the Cadron urethane tray: a radiographic assessment of bone remodeling. J Oral Maxillofac Surg 1994; 52: 373 – 380. 18. Soderholm A, Hallikainen D, Lindqvist C. Radiographic follow-up of bone transplants to bridge mandibular continuity defects. Oral Surg Oral Med Oral Pathol 1992; 73: 253 – 261. 19. Sasaki T, Honda E, Umino K, Takahashi Y. A simulation system with three-dimensional CT images for maxillofacial surgery and prediction of post-operative facial appearance. In: Farman AG, et al (eds). IADMFR/CMI’97 Advances in Maxillofacial Imaging. Amsterdam, The Netherlands: Elsevier, 1997, pp 197 –203. 20. Huang CS, Ko WC, Lin WY, Liou EJ, Hong KF, Chen YR. Mandibular lengthening by distraction osteogenesis in children—a one-year follow-up study. Cleft Palate Craniofac J 1999; 36: 269 – 274. 21. Tharanon W, Sinn DP. Mandibular distraction osteogenesis with multidirectional extraoral distraction device in hemifacial microsomia patients: three-dimensional treatment planning, prediction tracings, and case outcomes. J Craniofac Surg 1999; 10: 202 – 213.
