3D Dental Imaging by Spiral CT

3D Dental Imaging by Spiral CTMichael W. Vannier, Charles F. Hildebolt,Robert H. Knapp, Gary Conover, Naoko Yokoyama-Crothers, Ge WangMallinckrodt Institute of RadiologyWashington University School of Medicine, St. Louis, Missouri 63110ABSTRACTThree-dimensional image acquisition, display, and analysis of dental structures was performed and validated using spiral computed tomography (SCT) with metal artifact suppression. Isolated extracted teeth, a dry mandible, cadaver mandible, and cadaver head were scanned and reconstructed using a spiral CT scanner (Siemens Somatom PLUS-S) with 1 mm detector collimation, 1-mm table feed, and 0.1-1 mm reconstruction interval using specially developed software. Algorithms for metal artifact reduction including extended attenuation range and interpolation of missing projections were applied. V olumetric rendering of voxel sum images was performed to synthesize images comparable to conventional intraoral dental radiographs. Direct comparison of voxel-based synthetic and digitized film images was made. Several isolated, extracted teeth were sectioned with a diamond saw and submitted for histomorphometric analysis to aid in direct comparison with CT slice images obtained by multiplanar reconstruction.Metal artifact reduction was successful in markedly reducing the streaks and star patterns that usually accompany metallic restorations and intraoral appliances. Individual teeth were comparable to CT slice images. V oxel sum images were comparable to dental radiographs; however, for the SCT images, the spatial resolution was higher within the plane of section than it was orthogonal to the plane of section.Serial examinations were obtained by SCT, registered by surface matching, and interval change measured by 3D subtraction. Simulated lesions and restorations were introduced and quantitatively evaluated pre- and post-interventionally to assess imaging method performance. Key words: spiral CT, helical CT, dental imaging, three dimensional1. INTRODUCTIONThe dentomaxillofacial complex is susceptible to diseases that manifest early as subtle changes in bony density and geometry. These changes may be detected by imaging. Our long term goal is to develop, characterize, and apply volumetric 3D x-ray imaging based upon spiral x-ray computed tomography to the dentomaxillofacial complex. We anticipate major gains in diagnostic performance for early disease detection, treatment monitoring and planning, and population studies using this technology.Conventional film-based dental radiography is the de facto standard for clinical and research examination of the oral hard tissues. With this technique, an extraoral x-ray source is aligned with a radiographic film placed either in the patient's mouth or along the side of the patient's head. The x-rays are attenuated based upon the physical characteristics of the structures of interest. The transmitted x-ray beam interacts with the film's emulsion producing a latent image. The film is chemically processed to provide a visible image.[1]The vast majority of dental radiography is performed for evaluation, location, magnitude, and extent of dental caries. Caries diagnosis is most commonly performed with intraoral periapical and bitewing radiographs. The periapical film includes the entire tooth and 3? mm of periapical bone. The bitewing displays the crowns and several/mm誷of the roots and supporting bone of themaxillary and mandibular teeth. Because the bitewing contains views of both maxillary and mandibular teeth, it is the most used radiograph for caries diagnosis.[2] There is, however, agreement that bitewings have only limited success in caries diagnosis,[3,4,5,6,7,8] with some studies indicating that only 60% of the lesions are detected and that often 20% of noncarious tooth surfaces are diagnosed as having cavities.[9,10]Bitewing radiographs and periapicals are also used to detect and monitor alveolar bone loss, which is most commonly attributed to bacteria in dental plaque (periodontal or gum disease), smoking, and diabetes. Treatments to reduce or prevent alveolar bone loss are also evaluated with bitewings and periapicals. For the past seven years, we have been involved with developing and evaluating digital imaging methods for radiograph-based quantification of alveolar bone.[11,12,13,14,15,16,17,18,19] Although we and others have shown that bitewing and periapical radiographs can be effective in quantifying alveolar bone, these methods have one severe limitation: they are based upon two-dimensional representations of complex three-dimensional bony anatomy.Panoramic radiography is used to supplement bitewing and periapical radiographs. This modality, introduced in the 1950s, is employed in over 50% of dental practices. The method produces a tomographic image of the mandible and maxilla on a single 5 x 12-inch film. The panoramic image suffers from the same limitation as bitewings and periapicals due to its two-dimensional representation of complex three-dimensional bony anatomy. Only structures within a narrow curved slice through the maxilla and mandible are in focus. Compared to intraoral radiography, panoramic radiography has poor accuracy for caries diagnosis and alveolar bone quantification.[20,21,22,23,24,25]Within endodontics, periapical radiographs are used to diagnose periapical bone loss, study root canal morphology, detect the presence of lateral canals, determine working lengths for the mechanical enlargement of canals, and perform postoperative evaluations. Restorative dentistry also utilizes bitewings and periapicals to check margins of restorations (precementation), evaluate crown:root ratios, and evaluate cavity preparations. Panoramic and periapical radiographs are used within oral surgery for presurgical evaluations of bony impactions, implant sites, root morphology, mandibular canal and maxillary sinus location. Within orthodontics, cephalometric radiographs are used to assess growth and tooth eruption patterns and to predict treatment outcomes. Cephalometric radiographs are lateral and frontal skull films taken under highly controlled conditions. Cephalograms are also incorporated in orthognathic studies, morphometric studies, and studies of growth and development.All x-ray transmission-based radiographs suffer the limitation in that they are two-dimensional projections of complex intrinsically three-dimensional anatomy. The result is that buccal-lingual (facial-oral) structures are usually indistinguishable. There is currently no traditional method of dental radiography that permits viewing internal dental anatomy without the superimposition of other structures nor is there a traditional method of viewing dental anatomy in a buccal-lingual, cross-sectional perspective (tangential view). In sum, by virtue of the source-detector geometry requirements, only a limited number and types of projections can be obtained through transmission radiography. Soft tissues are not diagnostically imaged and for practical reasons, the exposure settings, dose, and level of detail are relatively fixed and represent a significant compromise. With these methods, there is an acknowledged lack of sensitivity for detecting and quantifying small changes in the hard tissues.The advent of computerized transverse axial scanning (computed tomography, or CT)[26] greatly facilitated access to the internal morphology of soft tissue and skeletal structures. Conventional CT scanning is accomplished by acquiring a series of individual images. Typically, the images represent cross-sections through the body. The image slices are from 1 to 10 millimeters thick and the distances between them are from 1 to 10 or 20 millimeters. Projection data are acquired and reconstructed into images as the patient is moved incrementally through the CT gantry (that is, an image is obtained; the patient is moved to the next scanning position, and the next image is obtained). CT scans possess no magnification errors caused by geometric distortions. Such errors are common in conventional radiographs. A limitation of conventional CT is that although it has a high degree of accuracy within individual slices, it has relatively low between-slice accuracy[27] even with relatively narrow collimation (2 mm) and no interslice gaps. CT scans avoid the superimposition of structures and are, therefore, more desirable than conventional radiography as a morphometric tool. Since its inception, computed tomography has provided quantitative measurements for many different biological systems and has been used in pre- and post-surgical mapping procedures,[28] the evaluation of developmental and regressive dental abnormalities,[29] facial trauma, and temporomandibular joint disorders.[30,31,32,33]Recently, a new CT technique, spiral CT (SCT or volume acquisition CT), has been developed. This method has several advantages over standard CT imaging. By employing simultaneous patient translation through the x-ray source with continuous rotation of the source-detector assembly, SCT acquires raw projection data with a spiral sampling locus in a relatively short period.[34] Without any additional scanning time, these data can be viewed as conventional transaxial images, as multiplanar reconstructions, or as three-dimensional (3D) reconstructions. Such images provide an opportunity to obtain accurate images at any arbitrary location within the volume data set.The unique arrangement of the gantry and rotating x-ray source assembly radically reduces scan times. Partial body scans can be completed during a single breath hold. With standard incremental CT, small objects can be missed or their detection compromised if the patient's degree of inspiration and expiration varies from scan to scan. Moreover, multiplanar and 3D image reconstructions of structures from standard incremental CT data are degraded by motion-induced misregistration artifacts.Because SCT is free from respiratory misregistration, and because with SCT it is possible to reconstruct overlapping structures at arbitrary intervals, the ability to resolve small objects is increased. Several researchers have tested the potential of SCT over standard incremental CT for improved lesion detection.[35,36,37,38] In each study, the authors found that SCT increased the probability for the detection of small lesions. SCT does, however, result in a broadening of the section-sensitivity profile (SSP) with a concomitant increase in volume averaging artifacts.[39,40] Thus, the sharpness of certain structures is decreased even though the spatial resolution is unaffected.The advantages of spiral CT include the rapid acquisition of a volume data set in a relatively short period of time and the reconstruction of images at any plane within the helical data set. In addition, SCT holds the possibility of reducing the amount of radiation exposure to the patient.[41] SCT is being used to evaluate both soft tissue and bony anatomy of several anatomical regions, including the evaluation of lung lesions,[35,36] the liver, the pancreas, and other intra-abdominal structures.[42,43]Although SCT is a relatively new technique, few studies have been focused on the head and neck. These have been limited to qualitative evaluations of images compared with those of conventional CT.[44,45,46,47] While virtually all of these studies show SCT to be at least comparable with conventional CT, none has attempted to define the precision (repeatability) of quantitative measurements from SCT, compare these measurements with those of conventional radiography, or validate the measurements with homologous histological sections.There are several limitations of current SCT methods: (1) Isotropic submillimeter spatial resolution is not possible, (2) Motion and metal artifacts compromise the accuracy of SCT data, (3) Image registration of sequential data sets is suboptimal.In sum, there is currently no validated practical method for detailed 3D study of the dentition and surrounding structures in vivo nor in vitro. Destructive examinations by histomorphometry have limited practicality, even in cadaver specimens. SCT has the potential to overcome the limitations of standard transmission radiography and current CT-based methods. SCT can acquire a contiguous raw projection data set in a relatively short period (15-60 seconds, typically 24-30 seconds). Without any additional scanning time, these data can be viewed as conventional transaxial images, as multiplanar reconstructions, or as three-dimensional (3D) reconstructions. Such images can provide accurate images at any arbitrary location within the volume data set. Such images are essential for the diagnosis of subtle abnormalities, identification of aberrant anatomy, treatment planning, treatment evaluation, and quantification of hard tissue change. If SCT is shown to be practical for oral-facial imaging, this would be a major breakthrough in dental imaging.2. Spiral CT Acquisition MethodThree Spiral CT scanners (Somatom Plus S, Siemens Medical Systems, Inc. Iselin, NJ) are available at the Mallinckrodt Institute of Radiology. Specimens are placed on the examination table as for a normal axial CT examination. Patient registration information is entered at the Diagnostic Main Console (DMC). AP and Lateral topograms (scanned projection radiographs used for localization) are acquired and the appropriate ranges (e.g., set of table locations) for the study are defined. Image acquisition parameters are then chosen from a preset protocol menu. A typical protocol for dental Spiral CT is: head mode, 137 kV, 240 mA, 18 sec., 2 x 2 mm.3. V olume ScanningIn Spiral CT scanning, the selection of detector collimation and table speed influence the effective slice thickness. Typically, spiral table speed is chosen to remain equal to the collimation (Pitch=1).18 tube rotations at a table speed of 2 mm/second will yield a longitudinal table travel of 3.6 centimeters producing an effective slice thickness of approximately 2 mm. Images with slice intervals of less than 1 mm may be generated by further processing of the raw data sets at the remote diagnostic satellite console using specialized sub-millimeter reconstruction programs. This sub-mm software and metal artifact reduction (MAR) software were developed Dr. W. Kalender and associates at Siemens-Erlangen.Once the CT images and raw data are archived to optical disk they are transferred over an Ethernet network to a remote Diagnostic Satellite Console (DSC, from Siemens Medical Systems, Inc.) for further processing and reconstruction.The remote console system consists of a DEC 礦AX 3100 host computer with a Siemens SMI 5 medical imager, 2 each 1 GByte internal magnetic disk drives, ethernet interface, a 9 track 1/2" reel-to-reel tape drive, 12 inch and 5.25 inch WORM optical drives, and two 12 inch digitalmonochrome display monitors. A Kodak XLP laser film recorder is available.The raw data sets are loaded onto the host system's internal disk and reprocessed to create sub-millimeter interval reconstruction images. Sub-millimeter slice interval CT images are created from the raw data by selecting an interpolation algorithm as well as the image centering coordinates, magnification factor, and the desired slice thickness. Initially, 32 each 1 millimeter thick slices are created from each raw data set. The images are reviewed for quality assurance and then may be further processed using three-dimensional (3D) software, Metal Artifact Reduction (MAR) software, and V olumetric Analysis Reconstruction (V AR) software methods. Additionally, the 1 mm image data sets are transferred over the ethernet network to imaging workstations for further processing using ANAL YZE?(Mayo Clinic, Bioimaging Resource) and other image analysis programs.4. CT Metal Artifact Reduction (MAR)Streak artifacts in CT are caused by the attenuation characteristics of metal within the field of view. Because of their higher atomic number, metals attenuate x-rays in the diagnostic energy range much more than soft tissues and bone. The most severe effect of metals is missing data. The x-ray beam is attenuated so strongly that almost no photons reach the detectors. The resultant effects will show up in the images as pronounced dark and bright streaks, non-linear edge gradients, and sampling errors arising from the surface of the implant or restoration (Figure 1).Additionally, the high linear x-ray attenuation coefficients of metals lie outside the range of normal CT numbers. The CT numbers of metallic implants are in the range of 8000 to 50,000 Hounsfield units (HU), with the standard upper limit of most medical scanners being approximately 3000 HU. This causes clipping of the reconstructed image. If low frequency artifacts near the implant have high amplitudes, the result can be a complete blurring and distortion of the true contours of a metal implant or appliance. Simple scaling down of the raw data by an appropriate factor will show that detailed relevant information about the implant (or appliance) is not lost. This is important for a precise and consistent extraction of the geometrical outline of the implant. Scaling alone can permit a more reliable visualization of the implant borders.MAR image processing is used to reduce the artifacts created by metal implants, braces, and dental restorations by extending the range of gray scales used for image display. Streak artifacts are removed by interpolating raw data in the "shadows" of the metal object with adjacent raw data which does not contain the source of the artifact. The "removed" metal object is significantly scaled down in CT density and added back into the image (Figure 2).In certain cases, missing or disturbed data in the "shadow" of an implant may be substituted by interpolating data from adjacent areas. These interpolation methods have met with varying degrees of success and appear to depend on the complexity of the of structures examined. (Figure 3) Highly complex structures may generate additional new artifacts (Figure 1).Among various metal artifact reduction (MAR) algorithms, two types are most important: 1) algorithms with iteration in the spatial and frequency domains, and 2) algorithms with correction in the projection domain.The iterative MAR algorithms were motivated by studies on extrapolation of a band-limited function. It is common in practice to recover or extrapolate a band-limited function f(x) from only its segment(s). The Fourier spectrum of f(x) could be estimated as the transform of the product of f(x) and a window w(x) or based on various a priori assumptions. Generally speaking, iterativealgorithms go back and forth between spatial and frequency domains. Gerchberg[48] and Papoulis[49] studied an iterative algorithm for extrapolation of a band-limited function. Medoff et al.[50] proposed a framework that allows all types of limited data and incorporates a priori information using constraint operators. Sezan and Stark[51] addressed the incomplete data problem using the method of projections onto convex sets (POCS). POCS is a recursive image restoration technique that finds a solution consistent with the measured data and a priori known constraints in both the spatial and frequency domains. The constraints include finite spatial and frequency supports, energy and interval constraints.The corrective MAR algorithms are conceptually straightforward: the corrupted projection data are identified and estimated before reconstruction. Lewitt and Bates[52] developed an algorithm for image reconstruction from hollow projections, in which data gaps are either bridged by polynomial interpolations or filled with data satisfying consistency criteria. Hinderling et al.[53] applied Lewitt and Bates' data gap bridging technique for in vivo evaluation of artificial hip joints. Among various MAR algorithms we favor a corrective MAR scheme, because the iterative MAR methods are sensitive to noise, particularly when the assumption of a band-limited function is violated. In this project we will develop a corrective MAR algorithm with an extended detector dynamic range for spiral CT dental imaging, in which dental implant portions in projection data are so modified that a priori constraints are approximately satisfied. The constraints include the contour and x-ray linear absorption coefficient of a dental implant. The implant contour can be outlined based on direct filtered backprojection, and the implant characteristics determined pre-operatively.5. 3D Reconstruction Methods3D reconstruction images provide a familiar means of viewing anatomy which is more easily interpreted by non-radiologist physicians (Figure 5). 3D images may be produced on the satellite console (Siemens DSC) using software supplied by the vendor. The data set parameters are defined to indicate the range of the study, image size, segmentation threshold, and image orientation. Frontal, rear, top, bottom, right lateral, and left lateral 3D surfaces of the bony anatomy are created using preset programs. User selectable choices permit the adjustment of lighting direction, rotation, and coronal or sagittal slicing.ANAL YZE?is advanced image processing software that adds many enhancements to the standard 3D reconstructions. Sub-millimeter CT images are transferred to a SUN SPARCstation 20 workstation and imported into the ANAL YZE?database at the full 16-bit resolution. By defining multiple objects, individual teeth may be viewed alone or in combination with the semi-transparent mandible. Panoramic views of the dental anatomy may be shown as 3D surfaces or as digital radiographs by displaying a brightest pixel projection of the image data (Figure 6). The advantage of the image processing software lies in its ability to produce images comparable to standard radiographs but with the added ability to display data free of geometric and radiometric distortion. Additionally, the display parameters may be adjusted to improve edge sharpness as well as image contrast and brightness.We acquired sample SCT data sets from an adult cadaver head, a dry mandible, and extracted teeth (Figures 6-8). The same objects were imaged with conventional dental radiography at the SIU-Edwardsville School of Dental Medicine (by Dr. Conover). In Figure 6, the synthetic images are compared with those directly obtained.The images in Figures 3-8 demonstrate the feasibility of the processing methods, includingsuppression of metal artifacts, synthesis of radiographs (bitewing, periapical, and panoramic), and reformatted (MPR) slice views for comparison with histological sections. Figure 8 demonstrates that the 3D nature of SCT data permits the segmentation of dental structures.Serial examinations were obtained by CT, registered by surface matching, and interval change measured by 3D subtraction (Figures 9-10). A simulated lesion was introduced and quantitatively evaluated pre- and post-intervention to assess imaging method performance. The process for image reconstruction, pairwise volume registration, and differencing is outlined in the flowchart of Figure 9 and illustrated in Figure 10.3D dental imaging by spiral CT is feasible. Since the oral hard tissues are complex, multiple displays are needed to complete a comprehensive examination.6. Acknowledgments> We appreciate the technical support in spiral CT from Drs. Willi Kalender and Arec Polacin from Siemens AG in Erlangen, Germany who developed the submillimeter reconstruction software, and their associate, Dr. Ernst Klotz who introduced the extended range and MAR methods. The ANAL YZE? software system was provided by Dr. Richard A. Robb and associates at the Mayo Biomedical Imaging Resource in Rochester, MN. We are most grateful for their assistance in the computer graphics rendering of volumetric dental images.7. References1. Curry TS III, Dowdey JE, and Murry RC Jr. Christensen's Physics of Diagnostic Radiology, 4th ed. Malvern, PA: Lea & Febiger, 1990.2. White SC, Kaffe I, and Gornbein JA. "Prediction of efficacy of bitewing radiographs for caries detection" Oral Surg Oral Med Oral Pathol 60:506-513, 1990.3. Gwinnett AJ. "A comparison of proximal carious lesions as seen by clinical radiography, contact microradiography, and light microscopy" J Am Dent Assoc 83(5):1078-80, 19714. Marthaler TM and Germann M. "Radiographic and visual appearance of small smooth surface caries lesions studied on extracted teeth" Caries Res 4:224-42, 1970.5. Rugg-Gunn AJ. "Approximal caries lesions" Br Dent J 133:481-84, 1972.6. Purdell-Lewis DJ, Groeneveld A, Pot TJ, and Kwant W. "Proximal carious lesions: a comparison of visual, radiographical and microradiographical appearance" Neth Dent J 81:6-15, 1974.7. Buchholz RE. "Histologic-radiographic relation of proximal surface carious lesions" J Prevent Dent 4:23-24, 1977.8. Waggoner WF and Ashton JJ. "Comparison of Kodak D-speed and E-speed xray film in detection of proximal caries" J Dent Child 55:459-462, 1988.9. White SC, Hollender L, and Gratt BM. "Comparison of xeroradiographs and film for detection of proximal surface caries" J Am Dent Assoc 108:755-759, 1984.10. Douglass CW, Valachovic RW, Wijesinha A, Chauncy HH, Dapur KK, and McNeil BJ. "The clinical efficacy of dental radiography in the detection of dental caries and periodontal disease" Oral Surg Oral Med Oral Pathol 62:330-339, 1986.11. Hildebolt CF and Vannier MW. "Automated classification of periodontal diseases using bitewing radiographs" J Periodontol 59(2):87-94, 1988.12. Hildebolt CF, Vannier MW, Shrout MK, Pilgram TK, Province MP, Vahey EP, and Rietz DW. "Periodontal disease morbidity quantification II: Validation of alveolar bone loss measurements and vertical defect diagnosis from digital bitewing images" J Periodontol 61(10):623-632, 1990.13. Hildebolt CF, Vannier MW, Pilgram TK, and Shrout MK. "Quantitative evaluation of digital dental radiograph imaging systems" Oral Surg Oral Med Oral Pathol 70(5):661-668, 1990.14. Hildebolt CF, Vannier MW, Shrout MK, and Pilgram TK. "ROC Analysis of observer-response subjective rating data--application to periodontal radiograph assessment" Am J Phys Anthropol 84(3):351-361, 1991.15. Hildebolt CF, Vannier MW, Gravier MJ, Shrout MK, and Knapp RH. "Digital dental image processing of alveolar bone: Macintosh?II Personal Computer Software" Dentomaxillofacial Radiology 21(3):162-169, 1992.16. Hildebolt CF, Zerbolio DJ Jr., Shrout MK, Ritzi S, and Gravier MJ. "Radiometric-based classification of alveolar bone quality" J Dent Res 71(9):1594-1597, 1992.17. Hildebolt CF, Vannier MW, Gravier MU, Fineberg M, Walkup R, Knapp R, Zerbolio DJ Jr, and Shrout MK. "Semiautomatic classification of alveolar bone quality" Proceedings of Engineering in Medicine and Biology Symposium SPIE, 1652, Medical Imaging VI: Image Processing, pp.678-684, 1992.18. Hildebolt CF, Rupich R, Vannier MW, Zerbolio DJ Jr, Shrout MK, Cohen S, and Pinkus A. "Interrelationships between bone mineral content measures: Dual energy radiography (DER) and bitewing radiographs (BWX)" J Clin Periodontol 20(10):739-745, 1994.19. Hildebolt CF, Bartlett TQ, Brunsden BS, Hente NL, Gravier MJ, Walkup RK, Shrout MK, and Vannier MW. "Bitewing-based alveolar bone densitometry: Resolution requirements for digital imaging" Dentomaxillofacial Radiology 23:129-134, 1994.20. Muhammed AG and Manson-Hing LR. "A comparison of panoramic and intraoral radiographic surveys in evaluating a dental clinic population" Oral Surg Oral Med Oral Pathol 54:108-117, 1982.21. Valachovic RW, Douglass CW, Reiskin AB, Chauncey HH, and McNeil BJ. "The use of panoramic radiography in the evaluation of asymptomatic adult dental patients" Oral Surg Oral Med Oral Pathol 61:289-296, 1986.22. Stephens RG, Kogon SL, Reid JA, and Ruprecht A. "A comparison of panorex and intraoral surveys for routine dental radiography" J Can Dent Assoc 43:281-286, 1977.23. Hurlburt CE and Wuhrmann AH. "Comparison of interproximal carious lesion detection in panoramic and standard intraoral radiography" JADA 93:1154-1158, 1976.24. Mileman P. "Radiographic caries diagnosis and restorative treatment decision making" Groningen: Drukkerij Van Denderen, 1985.25. Espelid I. "Radiographic diagnosis and treatment decision on approximal caries" Community Dent Oral Epidemiol 14:265-270, 1986.26. Hounsfield GN. "Computerized transverse axial scanning (tomography) I. Description of system" Br J Radiology 46:1016-22, 1973.27. Hildebolt CF, Vannier MW, and Knapp RH. "Validation study of skull three-dimensional computerized tomography measurements" Am J Phys Anthropol 82:283-94, 1990.28. Vannier MW, Marsh JL, and Warren JO. "Three-dimensional CT reconstruction images for craniofacial surgical planning and evaluation" Radiology 150:179-184, 1984.29. Ericson S and Kurol J. "CT diagnosis of ectopically erupting maxillary canines--a case report" European J. Orthodontics 10(2):115-121, 1988.30. Helms CA, Katzberg RW, Moorish R, and Dolujick MF. "Computed tomography of the temporomandibular joint meniscus" J Oral Maxillofac Surg 41(8):512-517, 1983.。