There has been no study in which apical root resorption in vivo has been assessed using mathematical reconstruction and digital subtraction radiography. The.
Dentomaxillofacial Radiology (1998) 27, 25 ± 29 1998 Stockton Press All rights reserved 0250 ± 832X/98 $12.00
Assessment of apical root resorption using digital reconstruction E Reukers1, G Sanderink2, A M Kuijpers-Jagtman1 and M van 't Hof3 1
Department of Orthodontics and Oral Biology, University of Nijmegen; 2Department of Dental Radiology, Academic Centre for Dentistry Amsterdam; 3Medical Statistics Department, University of Nijmegen, The Netherlands
Objectives: To assess the in vitro and in vivo accuracy of a mathematical computer-based reconstruction of two images that are not taken with the same recording geometry for the measurement of apical root resorption following orthodontic treatment. Methods: A gold standard for root resorption in vitro was developed from 10 extracted upper central incisors using calipers. Radiographs made with ®ve dierent projection angles were reconstructed mathematically by two observers. The calculated loss of length was compared with the gold standard. Eighty-two upper central incisors from 61 patients were radiographically evaluated for the prevalence and degree of apical root resorption after orthodontic ®xed appliance therapy. The relative amount of reduction was calculated after mathematical reconstruction. Results: The inter-observer error in vitro was 1.8%. The 95% con®dence intervals for the dierence with the gold standard are small. The duplicate measurement in vivo error was 2.2% and the correlation between duplicate measurements was 0.94. The mean loss of tooth length was 7.8% (s.d. 6.9). Conclusions: The prevalence of root resorption corresponds well with that in the literature. Digital reconstruction is a reliable method to correct for dierent projection angles and to monitor the eects of orthodontic movement in serial dental radiographs. Keywords: radiography, dental; radiographic image enhancement; signal processing, computerassisted; tooth diseases
Introduction Root resorption is a common consequence of orthodontic treatment. In most patients it is minimal and of no clinical signi®cance.1 It can be diagnosed radiographically and histologically. Histological studies report a high prevalence whereas clinical radiographic studies are more variable.2,3 One reason may be that radiographic studies reveal only apical root resorption; buccal and lingual resorption is less readily perceptible.4 The diagnosis of apical root resorption is usually based on some measurement of radiographic difference. Measurements can be made in several ways,
Correspondence to: AM Kuijpers-Jagtman, University of Nijmegen, Department of Orthodontics and Oral Biology, School of Dental Science, P.O.Box 9101, 6500 HB Nijmegen, The Netherlands. Based on research submitted by Dr Reukers in partial ful®lment of the requirements for PhD in Medical Sciences, University of Nijmegen, The Netherlands Received 7 March 1997; accepted 8 July 1997
qualitative, quantitative and semi-quantitative. With qualitative measurement the prevalence of root resorption can be assessed (dichotomous assessment) but not the (relative) loss of tooth substance. With semi-quantitative measurement, the radiographs are compared with a predetermined ordinal scale.5 With quantitative measurement the length of a tooth is measured on a radiograph before and after treatment and the actual amount of resorption calculated after correction for any projection errors. If this is not possible, the relative amount of resorption can be expressed as a percentage of the actual tooth or root length: in this case the outcome is a semi-quantitative measurement. Evaluator bias is a factor when interpreting conventional radiographs. Lack of inter observer agreement in assessing root resorption has been reported,6 as well as large variations in intra-observer agreement.7 This variability has been attributed to bias resulting from prior knowledge of the clinical data, variation in ®lm density, equivocal radiographic ®ndings resulting in an increased rate of false negative
Assessment of apical root resorption using digital reconstruction E Reukers et al
26
and false positive diagnoses and observer education, training and experience. The use of computer-aided image analysis may be helpful in increasing the reproducibility of radiographic interpretation. Diagnosis of apical root resorption in orthodontically-treated patients from non-standardised serial radiographs can be misleading. Variation in the projection angle can complicate the interpretation of the apical region. Digital subtraction radiography may be able to solve this problem. Normally, success of the subtraction method is dependent on reproducibility of the image which itself is dependent on the projection angle as well as radiographic density and contrast. Ruttiman et al.8 show that, within certain limits, it is possible to correct for dierences in ®lm density and contrast. Lack of reproducibility in positioning the patient used to be the greatest obstacle in the application of the subtraction technique. Occlusal stents have been used to standardise the position of the source to the ®lm and cephalostats to eliminate the rotations of the patient that are not controllable by the stent. One of the consequences of orthodontic movement is that a mechanical device cannot be used to preserve the imaging geometry. Dunn et al.9 have shown that a mathematical technique can be applied to digital images of radiographs to establish correspondence between pairs of images taken at dierent projection angles. Orthodontic movement causes a change in angulation of one or more single teeth in relation to the radiograph. As a consequence, it should be possible with this technique to establish corresponding images. There has been no study in which apical root resorption in vivo has been assessed using mathematical reconstruction and digital subtraction radiography. The purpose of this study is to assess, ®rst, the reliability of measuring apical root resorption semi-quantitatively in vitro, after mathematical image reconstruction and second, the prevalence and degree of apical root resorption in vivo following orthodontic treatment. Materials and methods Data acquisition and processing In vitro For this part of the study 10 extracted permanent upper central incisors were used. The length of each tooth was measured using vernier calipers (Mitutuyo Dial Calipers 505-633, Tokyo, Japan). Radiographs were taken with a Sidexis intraoral digital CCD system (Siemens, Bensheim, Germany). The long axis of the teeth was placed parallel to the long axis of the sensor. Then, apical root resorption was simulated with a bur in eight of the incisors and the length of the teeth was measured again. Another ®ve radiographs were then taken of each tooth. The ®rst image was taken with the same position as before `resorption'; in the second the incisor was rotated around its long axis through 108; in the third the long axis of the tooth
was angulated 158 to the long axis of the CCD; in the fourth the long axis of the tooth was inclined (tipped) 158 to the horizontal with its incisal edge touching the CCD; while in the ®fth this inclination was 258. Image processing was done on a personal computer (Digital Venturis FX 5100s with a 15 inch PCXCV-Tx SVGA monitor; Digital Equipment Corporation, Maynard, Massachusetts, USA) using the Windowsbased Emago/Advanced v.2.20 software package (SODS, Amsterdam, The Netherlands). This package incorporates the following features: (1) gamma correction to match the density distribution of the image to the reference image to improve subtraction; (2) geometric reconstruction to standardise projection geometry for digital subtraction and (3) linear, logarithmic and colour versions of digital subtraction radiography. For each tooth the images were displayed side by side on the monitor (10246768 pixels). Gamma correction was performed. Two evaluators noted (ER and GS), independently, by pointing with a mouse, four features (e.g. recognisable anatomical landmarks such as the approximo-incisal angle, a typical irregularity in the root canal, or the cementoenamel junction) in the initial image of the tooth which could be readily identi®ed on the second and subsequent images and were also as far from each other as possible. Next, the corresponding locations were identi®ed on the second (`after treatment') image. This procedure was repeated for each of the 50 radiographs. Emago used the reconstruction algorithm described by Dunn et al.9 to identify the co-ordinates of each pixel of the ®rst image in the second image. The second image was reconstructed and subtracted from the initial image using linear subtraction. The resulting subtraction image was evaluated. If (almost) no root and crown structures could now be discerned (Figures 1d and 2d) the reconstruction was considered successful.9 If, on the other hand, root structures and/ or the crown could still be distinguished individually, the subtraction was considered to have failed. The evaluators were allowed one further attempt to get a better ®t between the original and the reconstructed images. After successful reconstruction the tooth length was measured from the geometric centre of the incisal edge to the midpoint of the apex using the measuring device in Emago Advanced. With this device the tooth length is presented as the number of pixels between the two points. The percentage loss of tooth length was calculated as ((L1-L2)/L1)*100 (L1=tooth length before treatment; L2=tooth length following simulated resorption). The same formula was used to calculate the resorption as assessed with the calipers. This result was taken as the gold standard. In vivo This part of the study was based on the intraoral radiographs of maxillary incisors taken before orthodontic treatment and at removal of the ®xed appliance of 61 patients (29 boys and 32 girls) who had participated in a multipractice clinical trial.10 The
Assessment of apical root resorption using digital reconstruction E Reukers et al
a
c
27
b
d
Figure 1 Construction of a reconstructed image of an upper left central incisor. (a) Pre-treatment radiograph of upper central incisors. (b) Post-treatment radiograph of upper central incisors. (c) Reconstructed post-treatment radiograph of tooth 21. (d) Difference image of digitised radiograph A minus digitised radiograph C. Note that almost no root structure of tooth 21 is discernible; this means that the reconstruction of image B into image C has been successful. The length of tooth 21 in a is 318 pixels, and 307 pixels in c. The reduction of 3.5% is less than twice the duplicate error and thus considered not clinically signi®cant
radiographs were obtained using the bisecting-angle technique either with or without ®lmholders and Ekta speed ®lm (Eastman Kodak, Rochester, NY, USA). They were included in the study if on the series of radiographs at least one of the upper central incisors could be evaluated. Teeth were excluded when the apex could not be detected due to cone-cutting, if it was markedly distorted due to positioning errors, if the tooth had been endodontically treated, or if they were failures of exposure or processing. On the basis, 82 upper central incisors were considered suitable for assessment. The radiographs were converted to digital images using a Professional PCD Imaging Workstation and stored on a Photo CD master disk (Eastman Kodak, Rochester, NY, USA). A Photo Cd Lab (Corel Corp, USA) was used to convert the Photo CD images to an 8 bit BMP format with a resolution of 2566384 pixels. Images were processed as described above by one of the observers (ER) and overall tooth length measured. To determine the reliability of the method in vivo, the measurements were repeated on 13 radiographs 1 day later.
a
b
c
d
Figure 2 Example of a patient with apical root resorption, following orthodontic treatment. (a) Pre-treatment radiograph. (b) Post-treatment radiograph. (c) The 21 is reconstructed, correcting for the eects of orthodontic movement on projection angle. (d) Little root structures can be distinguished in the dierence image (AC). The original length of 21 is 301 pixels (a) and after reconstruction (c) 254 pixels resulting in a relative loss of tooth length of 15.6%
Statistical analysis In vitro A paired t-test was used to test for dierences between observers. The inter-observer error was calculated. Dierences between the mean values of all ®ve assessments and the gold standard were calculated for both observers. The deviation from the gold standard was expressed as the 95% con®dence interval (CI) In vivo To check for the absence of any systemic error, a paired t-test was performed using the duplicate assessments. The duplicate error was calculated using H~|d2/2n where d is the dierence between duplicate determinations and n is the number of determinations. The reliability of the measurements was determined by assessing the correlation between both duplicate assessments. The prevalence of root resorption was determined on the basis that root was resorbed if the percentage loss of tooth length was at least twice the duplicate error. The mean loss of tooth length was also calculated. If both central incisors could be evaluated, the mean loss in length of both teeth was used for the calculation.
Assessment of apical root resorption using digital reconstruction E Reukers et al
28 Table 1 In vitro study. The effect of varying the angle of the tooth on its measured length. The results are as expressed as the mean difference (s.d.) between the values obtained digitally by the two observers and the gold standard and as the 95% confidence intervals (in millimetres) Tooth position straight 108 rotation 158 angulation 158 inclination 258 inclination
Observer A Mean diff (s.d.) 70.63 70.37 0.30 71.46 70.43
(0.23) (0.55) (0.62) (0.50) (0.86)
95% Cl 71.2 71.6 71.1 73.6 72.4
to to to to to
70.1 0.9 1.7 70.3 1.5
Observer B Mean diff (s.d.) 70.66 70.48 70.12 71.64 70.83
(0.27) (1.06) (0.72) (0.56) (0.70)
Combined 95% Cl comb.
95%Cl 71.3 72.9 71.7 72.9 72.4
to to to to to
70.1 1.9 1.5 70.4 0.8
71.3 72.2 71.4 73.2 72.4
to to to to to
70.1 1.4 1.6 70.4 1.3
Straight=long axis of the tooth placed parallel to the long axis of the CCD
Results In vitro The paired t-test showed no signi®cant dierence between the two observers (t=0.6). The inter-observer error was 1.8%. The mean gold standard was 10.5% resorption (range 0 ± 26.6%). The con®dence intervals for deviation from the gold standard for the ®ve separate projections is shown in Table 1. Observer A carried out one repeat to get a better ®t between the original and the reconstructed image in 12 of the 50 reconstructions (24%), while observer B repeated it once in nine reconstructions (18%). In vivo No tooth had to be excluded due to less than optimal reconstruction. The results of the duplicate assessments are given in Table 2. There was no signi®cant systematic error (t=1.4). The calculated duplicate error for the relative loss of total tooth length was 2.2%. The correlation between the ®rst and second assessment was 0.94 (P=0.01). The prevalence of root resorption (i.e. loss of tooth length 44.4%) in this group of patients was 66% and in the upper central incisors as a whole 63%. The mean amount of resorption was 7.8% (s.d. 6.9). Figure 2 shows the reconstructed image of the left upper central incisor in a patient with obvious apical root resorption. Discussion In this study apical root resorption of maxillary incisors after ®xed appliance therapy was evaluated radiographically using digital image reconstruction. The quality of the reconstruction was evaluated using subtraction radiography. The 95% CI for the deviation from the gold standard was quite small for all simulated projections. These small values indicate that this reconstruction technique is suciently reliable for clinical applications, provided that the orthodontic movement or variation in beam angulation is limited to 108 rotation, 158 angulation or 258 inclination. The error in duplicate assessment of 2.2% in the in vivo study was relatively small. It must be emphasised, however, that this is a measure of the `technical' error of the method, that is producing reconstructed images digitally, measuring the number of pixels in a row after
Table 2 In vivo study. Values for the assessment of duplicate error, expressed as the percentage loss of overall tooth length Mean Standard deviation
Assessment 1
Assessment 2
Difference
8.2 8.5
7.0 7.3
1.2 3.0
reconstruction and evaluating the radiographs twice. To assess the total duplicate error it would have been necessary to repeat the radiographs, each time. This was rejected on ethical grounds. In a previous study, however, Dunn et al.9 made double exposures from volunteers. They showed that the mathematical technique as implemented in the Emago/Advanced software that we applied, can be used to establish correspondence between pairs of clinical images taken at dierent projection angles and to produce reconstructed images comparable with images taken with occlusal stents. Sampling the analog image on to a larger matrix would bene®t the spatial resolution of the digital image. Theoretically, this would result in more accurate measurements. On the other hand it would lead to a situation where it would no longer be possible to display the pre- and post-treatment images on one computer screen. This would probably have a negative in¯uence on the accuracy of identifying the features needed to reconstruct the images. A major obstacle to the use of digital subtraction in orthodontics has been geometric reproducibility. Although it is possible to preserve the imaging geometry in other dental disciplines using mechanical devices, these fail to give the essential feature of orthodontics; tooth movement. Dunn and van der Stelt11 have shown that invariants on a radiographic image can be used to describe the relationship of pairs of images with angular disparity of up to 168. In a subsequent study Dunn et al.9 showed that this registration procedure can be used to establish correspondence between pairs of clinical images taken at dierent projection angles. In our study we have shown that it is within certain limits well applicable to single teeth following orthodontic movement. Identification of invariants on such teeth is not always simple. In the in vitro part of this study both observers repeated the procedure in 24% and 18% of the cases, respectively. In all these cases the second subtraction was deemed acceptable. The main problem in vivo was to ®nd reliable landmarks that had a sucient distance
Assessment of apical root resorption using digital reconstruction E Reukers et al
29
between each other, especially when the incisal edge of an incisor was not completely depicted on both radiographs, while the apex was. The quality of the reconstruction was evaluated using digital subtraction of the ®rst image and the second (reconstructed) images. In an ideal case with no root resorption, the tooth should be visible as a single unpatterned structure. Setting a criterion for success or failure of the reconstruction after subtraction is arbitrary and subjective. As the main criteria for success we considered a good superior position of the crown structure together with the cervical one-third of the root since these are the structures that should not change shape during orthodontic treatment. If these structures were still visible separately, the reconstruction was considered to have failed. Minor irregularities (e.g. in Figure 2d where the root canal is still slightly visible) were accepted as successful. Visibility of the periodontal ligament space after reconstruction was not considered as a failure since it will be widened due to orthodontic movement. In spite of this arbitrary and subjective element, the correlation between the ®rst and second assessment was very good at 0.94 (P=0.01). The absolute amount of resorption could not be calculated in the in vivo part of the study because we had no means of determining the enlargement. The relative amount of resorption was therefore measured from the overall tooth length and found to be 7.8%. This value is much less than the ®gure of 18% reported by Dermaut and De Munck.4 They calculated the relative resorption from the cemento-enamel junction to the apex. This procedure automatically leads to a
higher relative degree of loss of length when the absolute amount of resorption is equal. If our calculation is corrected for the dierence between root length and overall tooth length (mean over-all length 24 mm; mean root length 12.4 mm12) the relative root resorption would be 15.2%. Converted to absolute ®gures this would be about 1.9 mm loss of root-/toothlength. This is consistent with most other studies on the amount of root resorption on upper central incisors after ®xed appliance therapy.2 The prevalence of apical root resorption found in this study is also consistent with the literature. It has been reported to occur in 39 ± 99% of patients and in 34 ± 93% of maxillary incosors.2 In our study the comparable values were 66% and 63% respectively. We agree with Remington et al.13 in ®nding gross root shortening in only a few cases. Only two out of the 82 incisors in our series showed loss of tooth length of more than 25%, compared with their ®gure of four out of 200 upper central incisors with resorption of more than one-third of the root. In conclusion, this study suggests that the application of digital reconstruction to radiographs of orthodontically-treated upper central incisors can provide a good diagnostic performance in detecting the prevalence and relative degree of apical root resorption. The method is reliable with only a small duplicate measurement error. The results are comparable with those in the literature. Future studies should explore the possibility of correcting for magni®cation so that the absolute amount of apical root resorption can be assessed using this method.
References 1. Kaley J, Phillips C. Factors related to root resorption in edgewise practice. Angle Orthod 1991; 61: 125 ± 131. 2. Brezniak N, Wasserstein A. Root resorption after orthodontic treatment: Part 1. Literature review. Am J Orthod Dentofac Orthop 1993a; 103: 62 ± 66. 3. Brezniak N, Wasserstein A. Root resorption after orthodontic treatment: Part 2. Literature review. Am J Orthod Dentofac Orthop 1993b; 103: 138 ± 146. 4. Dermaut LR, De Munck A. Apical root resorption of upper incisors caused by intrusive tooth movement: A radiographic study. Am J Orthod Dentofac Orthop 1986; 90: 321 ± 326. 5. Levander E, Malmgren O. Evaluation of the risk of root resorption during orthodontic treatment: a study of upper incisors. Eur J Orthod 1988; 10: 30 ± 38. 6. Petrowski CG, ElBadrawy HE, Boehlau EE, Grace MGA. Interobserver variability in radiographic interpretation of pediatric dental diseases: a pilot study. J Can Dent Assoc 1996; 62: 723 ± 730. 7. Goldman M, Pearson AH, Darzenta N. Reliability of radiographic interpretations. Oral Surg 1974; 38: 287 ± 293.
8. Ruttiman UE, Webber RL, Schmidt E. A robust digital method for ®lm contrast correction in subtraction radiography. J Periodont Res 1986; 21: 486 ± 495. 9. Dunn SM, Stelt PF van der, Ponce A, Fenesy K, Shah S. A comparison of two registration techniques for digital subtraction radiography. Dentomaxillofac Radiol 1993; 22: 77 ± 80. 10. Reukers HAJ, Kuijpers-Jagtman AM. Eectiveness of orthodontic treatment: a prospective clinical trial. Eur J Orthod 1996; 18: 424. 11. Dunn SM, Stelt PF van der. Recognizing invariant geometric structure in dental radiographs. Dentomaxillofac Radiol 1992; 21: 142 ± 147. 12. Sicher H, Du Brul EL. Oral anatomy. Fifth edition. St Louis: Mosby Company, 1970, p207. 13. Remington DN, Joondeph DR, AÊrtun J, Riedel RA, Chapko MK. Long term evaluation of root resorption occurring during orthodontic treatment. Am J Orthod Dentofac Orthop 1989; 96: 43 ± 46.