JOURNAL OF ENDODONTICS Copyright 2003 by The American Association of Endodontists

Printed in U.S.A. VOL. 29, NO. 4, APRIL 2003

Three-Dimensional Quantitation of Periradicular Bone Destruction by Micro-Computed Tomography

Dietrich von Stechow, MD, Khaled Balto, DDS, Philip Stashenko, DMD, PhD, and Ralph Mu ller, PhD

We have previously shown that two-dimensional, high-resolution, micro-computed tomography is a rapid, reproducible, and noninvasive method for measuring periradicular bone resorption in mice, giving results virtually identical to histology. In this study, we determined whether a three-dimensional volumetric quantitation of bone resorption could be achieved and whether this correlates with twodimensional measurements. Periradicular lesions were induced in the lower first molars of mice by pulp exposure and infection; unexposed teeth served as controls. Mandibles were harvested on day 21 and subjected to: (a) three-dimensional micro-computed tomography imaging; and (b) conventional histology. Using a three-dimensional model and a semiautomatic contouring algorithm, we determined three-dimensional void volume, void surface, void thickness, and the standard deviation of the thickness distribution. The results showed a significant correlation between lesion void volume and two-dimensional lesion area by histology (r2 0.73), as well as high correlations between void volume and void thickness (r2 0.86) and standard deviation of the void thickness (r2 0.87), but no relationship with void surface. These results show that three-dimensional analysis of micro-computed tomography images is highly correlated with two-dimensional cross-sectional measures of periradicular lesions. Nevertheless, microcomputed tomography allows assessment of additional microstructural features as well as subregional analysis of lesion development.

challenging (2). Furthermore, precise anatomical analyses of teeth are only possible after destruction of the specimen (3). In endodontics, changes in the bony structure around the tooth apex are studied by means of periradicular radiographs or experimentally using histologic sections. These two-dimensional projections of threedimensional structures often do not provide an adequate representation of the region of interest and therefore produce a loss of quantification accuracy. Also radiographs are usually interpreted visually, a process that is subjective in nature (4). Conversely, trabecular bone morphometry has traditionally been assessed in two dimensions, where the structural parameters are either inspected visually or measured from sections, and the third dimension is added on the basis of stereology. This conventional approach to morphologic measurements, however, entails substantial preparation of the specimen. Although the method offers high spatial resolution and high image contrast, it is a tedious and time-consuming technique that results in destruction of the specimen. Three-dimensional (3-D) objects are typically reconstituted from planar sections. These can be physical sections, optical sections (confocal microscopy), or computed tomography (CT) reconstructions. Tachibana and Matsumoto (5) investigated the application of X-ray CT in endodontics and found that 3-D reconstruction of teeth was feasible but that the limited resolution of 0.6 mm produced by a clinical CT scanner was inadequate for detailed analysis. High-resolution CT has been used successfully to measure enamel thickness in teeth (6) and surface areas and volumes of dental materials (7). In most CT scanners, objects are reconstructed as a stack of two-dimensional (2-D) sections through the object, with in-plane resolution being better than between-plane resolution. In a previous study (8), we have shown that micro-CT imaging is a rapid, reproducible, and noninvasive method that gives results highly comparable with those obtained by histology. The micro-computed tomography (CT) system uses a stacked fan-beam geometry and produces true 3-D reconstruction of the object with cubic voxels and isotropic resolution. In this study, we determined whether a 3-D volumetric quantitation of periradicular bone resorption could be achieved and how this would correlate with 2-D area measurements.

In many areas of medicine, the ability to monitor biological changes in disease processes over time has gained increasing importance. This applies especially to the field of dentistry, where inflammatory, osteolytic, and neoplastic processes can occur without detectable changes in the applied imaging methods (1). Even after extraction of teeth the detection of anatomical details can be
252

MATERIALS AND METHODS Animals and Periradicular Lesion Induction Eleven-week-old, C57BL6, male mice (n 7) were obtained from Charles River Breeding Laboratory, Wilmington, MA. Ani-

Vol. 29, No. 4, April 2003

3-D Micro-CT in Periradicular Bone Resorption

253

mals were maintained in a conventional environment in the Forsyth Institute Animal Facility according to the guidelines of the Institutional Animal Care and Use Committee. For lesion induction, anesthetized mice (n 5) were mounted on a jaw retraction board and the lower first molar pulps were bilaterally exposed to the oral environment by means of a round bur as described (8). Animals without exposures served as controls (n 2). All animals were killed by CO2 asphyxiation on day 21 after pulp exposure. For histology purposes the mandibles were dissected free of soft tissue, fixed in 10% phosphate-buffered formalin, and stored at 4C overnight.

Micro-CT Imaging Fixed mandibular samples were analyzed by means of a compact fan-beam-type tomograph (CT 20; Scanco Medical, Bassersdorf, Switzerland), also referred to as desktop micro-CT (9). The mandibles were placed in an airtight cylindrical sample holder filled with formalin, with no further sample preparation or operator intervention needed. Approximately 150 microtomographic slices (1024 1024 pixels) with an increment of 17 m were acquired, covering the entire medial-lateral width of each hemimandible. Measurements were stored in 3-D image arrays with an isotropic voxel size of 17 m. Void space was segmented from bone and dentin using a global thresholding procedure (9). An extended marching cubes algorithm was used for 3-D visualization (10). The total examination time for each specimen was approximately 6 h. After microtomographic imaging, mandibles were decalcified and prepared for histologic examination by cutting paraffin sections of 6-m thickness and were subsequently stained with H & E (8).

FIG 1. 2-D micro-CT image through sagittal extent of the first molar in experimental mice with root canal exposure. Both burr hole in the crown of the tooth and destruction around the root apex are clearly visible using a high isotropic image resolution of 17 m.

Statistical Analysis Area and volume measurements determined by histology and CT were evaluated between control and experimental groups using paired t tests. Unpaired t tests were used to compare differences in microtomographic indices between control and experimental groups. Statistical significance was defined by two-tailed p 0.05. The Pearson correlation coefficient (r) was calculated to determine the association between (a) 3-D VV obtained from CT and 2-D area by histology; and (b) 3-D VV and 2-D area both obtained by CT. Analysis of the data was performed by using Excel 2000 (Microsoft Corporation, Redmont, WA). RESULTS The entire medial-lateral width of the jaw was encompassed in approximately 150 radiographic sections taken at a 17-m interval. The mandibular first molar was visible in approximately 100 radiographic sections. Of these, a median of 53 slices was chosen for 3-D quantification of periradicular bone destruction defined by the endpoints for contouring as described above. As described previously (8), teeth without pulp exposure had a narrow periodontal ligament space and normal periradicular bony topography. In contrast, teeth with surgical exposures exhibited readily discernible radiolucencies surrounding the mesial and distal root apices. This is better illustrated in the reconstituted 3-D images for both control and experimental samples (Fig. 3, black arrow). In the experimental image, the periradicular lesion is defined by a larger VV and a higher average lesion thickness (V.Th). Nevertheless, regional differences can be seen as well, with the largest effects seen at the

Image Analysis For the 2-D analysis, both histological and microtomographic sections corresponding to the central portion of the lesion (Fig. 1) were analyzed by 2-D morphometry. In this approach, the crosssectional area of the region of interest from both preparation techniques was analyzed. The region of evaluation included the void area of periradicular destruction and/or the periodontal ligament space surrounding the distal root of the first molar. For the 3-D analysis of the micro-CT data, a volumetric model was used in which lesion and control void spaces were defined using a semiautomatic contouring algorithm providing an accuracy of one voxel (11). Moving in an apico-coronal direction the analysis for every 3-D image started with the first transverse section showing the apex dentis and ended with the last section in which the tooth was completely surrounded by crestal bone. Similar to the 2-D approach, contouring of the periradicular destruction and the periodontal ligament space surrounding the distal root tooth was performed (Fig. 2). In the contoured volume, basic structural metrics were measured in three dimensions using direct 3-D morphometry including the volume of the measured void (void volume; VV), the surface area of the void (void surface; VS), the ratio of VV to VS (surface-to-volume ratio; VS/VV), the average 3-D void thickness (V.Th), as well as the standard deviation of the void thickness (V.Th.SD), which is a measure of the variability of V.Th. For a more complete description of methods used for direct 3-D morphometry, see reference (12).

254

von Stechow et al.

Journal of Endodontics

FIG 3. 3-D visualization of void space for control and experimental distal root. Experimental lesions are defined by a larger VV and a higher average lesion thickness. Nevertheless, regional differences can easily be seen and effects are largest at the root canal exit (black arrow), resulting in a significantly larger distribution of thicknesses as can be measured in different regions

DISCUSSION The aim of this study was to assess the 3-D resorption patterns of periradicular lesions in mice by CT. In addition, morphometric methods were evaluated to allow estimation of 3-D microstructural indices underlying bone resorption. Using these approaches, a high correlation between 2-D and 3-D methods could be demonstrated. Nevertheless, several advantages of the CT methodology can be pointed out. First, data collected from intact teeth are readily available for further evaluation and analysis in different planes. Previously, high-resolution morphological studies required complete tooth destruction to evaluate pulpal volume, surface area, and anatomy (13). The aspect of rapidity regarding CT measurements has been addressed by several authors (2, 8, 14). Depending on the number of samples and their size, CT data acquisition and evaluation can be accomplished in a matter of a few days. Thus avoidance of sample destruction and speed of data acquisition are obviously great advantages compared with conventional histology, the current gold standard. Our results demonstrate that 2-D measurements of periradicular lesion area and 3-D measurements of lesion volume are highly correlated. This suggests that 2-D sections might be sufficient to reflect the volumetric changes associated with bone destruction. Nevertheless, CT allows the computation of additional 3-D indices, such as VS and thickness, as well as the performance of subregional analysis of lesion development. This indices allow novel insight into the course of the destructive process over time. In conventional stereology, the primary parameters VV and VS are usually used to derive indices, such as V.Th, based on an underlying model assumption. In this study, we proposed a direct thickness measurement approach, where mean V.Th is calculated directly and without an underlying model assumption by determining a local thickness at each voxel representing void (15). Such an approach does not only offer direct measurement of mean thickness but also allows computation of the lesion thickness distribution (V.Th.SD). Using this new approach, we were able to demonstrate that bone destruction or increase in VV was highly associated with an

FIG 2. Contouring of resorption or void space from 3-D stacks of high-resolution micro-CT images.

root canal foramen (black arrow). Three experimental samples presented with minimal periradicular bone destruction. A closer examination of the corresponding histological sections revealed that the pulp exposures in these teeth were very small, resulting in delayed pulpal infection and necrosis and slowed periradicular bone destruction (not shown). Basic statistics for the 2-D and 3-D analysis of periradicular bone destruction are shown in Tables 1 and 2. In the 2-D study, the mean difference between control and experimental groups in area of the lesion was 158% for histology (p 0.01) and 165% for CT (p 0.01). In the 3-D analysis, the experimental group exhibited significant increases of 28% for VV (0.17 mm3), 10% for VS (5.5 mm2), 26% for V.Th (0.089 mm), and 118% for V.Th.SD (0.03 mm) compared with the control group. VS/VV (33.4 1/mm) displayed a decrease of 11% and was not significant. Additionally, the results showed a highly significant correlation between 3-D VV and 2-D area by histology (r 0.84, p 0.001) and CT (r 0.85, p 0.001), respectively (Fig. 4). When comparing VV with other 3-D structural indices, we found high correlations between VV and V.Th (r 0.92, p 0.001), and V.Th.SD (r 0.93, p 0.001), but no significant relationship between VV and VS (r 0.56, NS). It was noted that in the experimental group, the standard deviation of the measured V.Th was much larger than in the control group where the thickness distribution was very uniform (Table 2).

Vol. 29, No. 4, April 2003

3-D Micro-CT in Periradicular Bone Resorption TABLE 1. 2-D area measurements obtained by histology and micro-CT

255

Group Control (n 4) Experimental (n 10) Difference p value

Histology 0.08 0.02 0.21 0.08 0.13 0.05

Micro-CT 0.08 0.02 0.22 0.09 0.14 0.06 0.01*

Difference 0.00 0.02 0.01 0.08 0.01*

p value 0.63 0.15

Data are means SD. The two-sample t test was used to evaluate differences between the control and experimental groups. Paired t tests were used to assess differences between histology and micro-CT measurements within each group. *Statistically significant.

TABLE 2. Group differences for 2-D and 3-D microtomographic indices Index 2-D 3-D 3-D 3-D 3-D 3-D Area (mm2) VV (mm3) VS (mm2) VS/VV (1/mm) V.Th (mm) V.Th.SD (mm) Control 0.08 0.2 0.13 0.01 5 0.5 37.3 4.6 0.07 0.01 0.01 0.001 Experimental 0.22 0.09 0.17 0.04 5.5 0.6 33.4 8 0.09 0.02 0.03 0.02 Difference 158% 28% 10% 11% 26% 118% p value 0.01* 0.01* 0.05* NS 0.01* 0.01*

Data are means SD. The two-sample Students-t test was used to evaluate differences between the control and experimental groups. *Statistically significant. NS Not Significant.

FIG 4. 2-D measure of resorption area versus 3-D measure of VV as assessed from the same micro-CT images.

increase in lesion thickness. The data from the experimental group also revealed that the thickness variation in the experimental teeth was much larger than in the controls, indicating that resorptive lesions are not uniform in nature, with most of the destruction taking place right under the root canal foramen and in the midsagittal plane, the standard analysis region for the 2-D analysis (8). Biologically, this reflects the site of egress of bacteria and their components from the infected root canal, which stimulate host cells to express bone resorptive cytokines, particularly interleukin-1 (16). In clinical practice Tachibana and Matsumoto proposed X-ray tomographic imaging to be the method of choice (5, 7). Nevertheless, current clinical tomographic units provide resolutions of approximately 1 mm, resulting in inadequate representation of small objects and substantial losses in quantification accuracy (14, 17). Micro-CT allows remarkably improved resolution over conven-

tional techniques. Although it is technically possible to produce resolutions on the order of 1 m (18), our 17-m thick tomographic sections compared well to histology (6 m nominal resolution). Other studies in endodontics chose slice thicknesses of 34 m and 68 m, respectively, and considered the results to be of acceptable quality (2). Currently, CT can only be used for in vitro measurements of small samples due to the high radiation dose of 900 mSv, which is not compatible with living organisms. However, the in vivo application of CT on a regular basis in hospitals is currently under development. Current experimental patient systems allow resolutions of up to 160 m at low radiation doses (19). In the near future resolutions on the order of 100 m will be achievable in humans. In small animals it is already possible to achieve resolutions of approximately 20 m for in vivo follow-up studies (20), yielding remarkable results on the transient biology of bone adaptive processes. In conclusion, we believe that CT provides a new and innovative tool for endodontic research. Webber suggested that one of the most promising opportunities for technological development would be the generalization of the principles underlying CT to facilitate task-specific 3-D dental radiography. The present report introduces a methodology capable of producing novel 3-D morphometrical data on the structure of periradicular lesions, unavailable from other techniques. Such data are likely to be of considerable importance for the diagnosis and study of mechanisms related to inflammation-induced bone resorption.

This study was partly supported by grants DE-09018, DE-I1664, and AG-13333 from the N.I.H. and the M. E. Mu ller Professorship in Bioengineering at Harvard Medical School. Drs. von Stechow and Mu ller are affiliated with the Orthopedic Biomechanics Laboratory, Beth Israel Deaconess Medical Center and Harvard Medical School; Drs. Balto and Stashenko are affiliated with the Department of Cytokine Biology, The Forsyth Institute; and Dr. Balto is also affiliated with the Department of Endodontics, Harvard School of Dental Medicine, Boston, MA. Address requests for reprints to Ralph Mu ller, PhD, Orthopedic Biomechanics Laboratory, Beth Israel Deaconess Medical Center and Harvard Medical School, 330 Brookline Avenue, RN 115, Boston, MA 02215.

256

von Stechow et al.

References

Journal of Endodontics
new method to analyse and display the three-dimensional structure of trabecular bone. Phys Med Biol 1994;39:145 64. 11. Ru egsegger P, Mu nch B, Felder M. Early detection of osteoarthritis by 3-D computed tomography. Technol Health Care 1993;1:53 66. 12. Hildebrand T, Laib A, Mu ller R, et al. Direct three-dimensional morphometric analysis of human cancellous bone: microstructural data from spine, femur, iliac crest, and calcaneus. J Bone Miner Res 1999;14:116774. 13. Lyroudia K, Samakovitis G, Pitas I, et al. 3-D reconstruction of two C-shape mandibular molars. J Endodon 1997;23:101 4. 14. Mu ller R, Ru egsegger P. Micro-tomographic imaging for the nondestructive evaluation of trabecular bone architecture. Stud Health Technol Inform 1997;40:6179. 15. Hildebrand T, Ru egsegger P. A new method for the model-independent assessment of thickness in three-dimensional images. J Microscopy 1997;185:6775. 16. Stashenko P, Teles R, DSouza R. Periradicular inflammatory responses and their modulation. Crit Rev Oral Biol Med 1998;9:498 521. 17. Gullickson DC, Montgomery S. The study of root canal morphology using a digital image processing technique. J Endodon 1987;13:158 63. 18. Peyrin F, Salome M, Cloetens P, et al. Micro-CT examinations of trabecular bone samples at different resolutions: 14, 7 and 2 micron level. Technol Health Care 1998;6:391 401. 19. Mu ller R, Koller B, Hildebrand T, et al. Resolution dependency of microstructural properties of cancellous bone based on three- dimensional mu-tomography. Technol Health Care 1996;4:1139. 20. Kinney J, Lane N, Haupt D. In vivo, three-dimensional microscopy of trabecular bone. J Bone Miner Res 1995;10:264 70.

1. Cardoso AS, Mitchell DF. Progression of pulpitis to necrosis and periradicular disease in deciduous and permanent teeth of monkeys. J Dent Res 1971;50:934 8. 2. Peters OA, Laib A, Ru egsegger P, et al. Three-dimensional analysis of root canal geometry by high-resolution computed tomography. J Dent Res 2000;79:14059. 3. Green D. Morphology of the pulp cavity of the permanent teeth. Oral Surg 1955;8:74359. 4. Gelfand M, Sunderman EJ, Goldman M. Reliability of radiographical interpretations. J Endodon 1983;9:715. 5. Tachibana H, Matsumoto K. Applicability of X-ray computerized tomography in endodontics. Endod Dent Traumatol 1990;6:16 20. 6. Spoor CF, Zonneveld FW, Macho GA. Linear measurements of cortical bone and dental enamel by computed tomography: applications and problems. Am J Phys Anthropol 1993;91:469 84. 7. Nielsen RB, Alyassin AM, Peters DD, et al. Microcomputed tomography: an advanced system for detailed endodontic research. J Endodon 1995; 21:561 8. 8. Balto K, Mu ller R, Carrington DC, et al. Quantification of periradicular bone destruction in mice by micro-computed tomography. J Dent Res 2000; 79:35 40. 9. Mu ller R, Van Campenhout H, Van Damme B, et al. Morphometric analysis of human bone biopsies: a quantitative structural comparison of histological sections and micro-computed tomography. Bone 1998;23:59 66. 10. Mu ller R, Hildebrand T, Ru egsegger P. Non-invasive bone biopsy: a