Published online before print May 9, 2002, 10.1148/radiol.2241010948
(Radiology 2002;224:247-252.)
© RSNA, 2002
Bone Mineral Density Measurement with Dental Quantitative CT Prior to Dental Implant Placement in Cadaver Mandibles: Pilot Study1
Peter Homolka, PhD,
Andreas Beer, MD, DDS,
Wolfgang Birkfellner, PhD,
Robert Nowotny, PhD,
André Gahleitner, MD,
Manfred Tschabitscher, MD and
Helmar Bergmann, PhD
1 From the Departments of Biomedical Engineering and Physics (P.H., W.B., R.N., H.B.) and Diagnostic Radiology (A.G.), University of Vienna General Hospital, Währinger Gürtel 18-20, A-1090 Vienna, Austria; the Department of Oral Surgery (A.B., A.G.), Dental School of the University of Vienna, Austria; the Department of Anatomy (M.T.), University of Vienna, Austria; and the Department of Maxillofacial Surgery, Evangelisches Krankenhaus, Vienna, Austria (A.B.). Received May 21, 2001; revision requested July 9; final revision received January 2, 2002; accepted February 1. Address correspondence to P.H. (e-mail: peter.homolka@univie.ac.at).
 |
ABSTRACT
|
|---|
PURPOSE: To correlate torque forces during insertion of screw-type dental implants with bone mineral density (BMD) values determined preoperatively.
MATERIALS AND METHODS: Dental quantitative computed tomography (CT) was performed with simultaneous imaging of five postmortem mandibles and a calibration standard containing defined concentrations of calcium hydroxyapatite. CT numbers were converted to local BMD values by assuming a linear relationship (BMD = a x HU + b), where a and b are calibration coefficients. The a, b, P, and t values, correlation coefficients, and standard errors were calculated. Dental implants (n = 25) were set, and insertion torques were recorded. BMD was determined at the implantation site and correlated with torque forces recorded during implant insertion. Calibration coefficients derived for specimens were compared with those derived for actual patients.
RESULTS: Calibration coefficients (at 120 kV) for the postmortem specimens were a = 0.760 ± 0.03 (mean ± SD) and b = 2.8 ± 3.7 and for the patients were a = 0.804 ± 0.06 and b = 5.2 ± 4.2. Calibrated BMD values at the location of dental implants exhibit a significant correlation (R2 = 0.83, P < .001) with insertion torques on the basis of a second-order model, which yields torque = (0.0055 x BMD + 0.73)2 for the implants used and the surgical technique applied.
CONCLUSION: Correlation exists between BMD measured with dental quantitative CT and the insertion torque of dental implants.
© RSNA, 2002
Index terms: Bones, mineralization, 243.81 • Computed tomography (CT), three-dimensional, 243.12117 • Experimental study • Jaws, CT, 243.12111, 243.12117, 243.92 • Teeth, 25.456
|
INTRODUCTION
|
|---|
Dental implants surgically inserted into the jaw bone (like small titanium posts) are used to support a superstructure that can be either a single tooth, a bridge, or a complete prosthesis. The volume of the bone available and the quality of the bone are two factors that determine the type of surgical procedure and the type of implant. Both of these factors contribute to the success of dental implant surgery (1,2).
Numerous radiologic procedures exist to measure bone volume (3–5). Measures that quantify bone quality (2,6) with rater crude grading methods (3) may be used but will hardly reflect the exact bone situation at a planned implantation site for an individual patient.
Determination of local bone mineral density (BMD) with cortical thickness and bone height may offer a comprehensive description of the bone the surgeon will encounter when he or she actually sets the implant. Quantitative computed tomography (CT) (ie, quantitative interpretation of values derived from Hounsfield units with a suitable calibration procedure) is the modality of choice to determine BMD. Common quantitative CT procedures developed for BMD measurement are used in the assessment of osteoporosis to determine a general BMD status in the patient by averaging BMD values over those obtained from, for example, lumbar vertebral bodies. These procedures cannot be applied directly to the jaws, because in dental quantitative CT, the prediction of implant anchorage at a well-defined position in the jaw bone requires exact local information about the BMD. Thus, BMD measurements in the jaws prior to implant surgery will have to extend to small regions of interest that are comparable in size to the diameter of the implants. Because of the small field of view in dental CT, a typical voxel size of 0.3 mm in the imaging plane allows accurate measurements within these regions.
Several methods for the assessment of implant stability during and after insertion of the implant have been used to evaluate the immediate loading capacity and expected long-term results (7–9). Methods that are available immediately after implantation include the periotest method (10) and the resonance-frequency analysis (8,9,11–13). Measurement of the cutting resistance of the jaw bone—the insertion torque (measured in newton-centimeters), where 1 N · cm is the torque generated by a force of 1 N acting on a lever of 1 cm in length)—is performed intraoperatively (8,12–15). These measurements, since they are available during or after implantation, cannot be used for surgical planning.
Quantitative CT to measure BMD by using simultaneous scanning for calibration has been extended to the jaws (16–19). Different calibration standards containing CaCO3 (19) or calcium hydroxyapatite in either polyurethane (17,18) or water-equivalent base materials have been used. Use of the actual bone mineral (calcium hydroxyapatite) embedded into a water-equivalent substance will minimize dependence of BMD on CT scanning conditions. Therefore, it is a common consensus to use these standards for quantitative CT of other body regions.
Quantitative CT offers the possibility to measure BMD values of cortical and cancellous bone separately (16), although the outcome depends critically on the method used for discriminating these two compartments (20). Furthermore, measurement of average BMD values for both segments does not contain the information sought for assessing implant positions, since BMD values vary locally to a high extent (13,21). Thus, evaluation of BMD locally or averaging over small regions of interest comparable in size with the implants are likely to reflect local bone properties more appropriately.
If a correlation was established between BMD measured with dental quantitative CT and parameters that measure implant stability, dental CT information could be increased without the need for additional examinations in the patients.
We hypothesized that BMD values can be used to predict mechanical properties of the jaw bone. Measurement of the insertion torque (ie, cutting resistance) reflects mechanical properties and is a well-accepted and readily used indicator of primary implant stability (8,14,15). Thus, the purpose of our study was to correlate torque forces during implant insertion with BMD values determined preoperatively.
|
MATERIALS AND METHODS
|
|---|
Materials
Three human mandibles (mandibles A, B, and C) and two specimen heads (mandibles D and E) obtained after death and according to our university policies were used. Four mandibles were edentulous; two teeth were extracted from mandible A prior to scanning and implantation to eliminate possible beam hardening in the vicinity of the roots. Selection criteria included minimum alveolar height for implantation and exclusion of highly resorbed mandibles. Mean cross-sectional bone height measured on the mandibles with a vernier caliper in the incisor region was 25.2 mm (range, 20–29 mm; SD, 3.9) and in the molar region, 17.6 mm (range, 11–21 mm; SD, 4.0). Mean bone height at implant positions was 21.8 mm (range, 17–29 mm; SD, 3.3). All specimens were unfixed, and soft tissue was not removed from the mandibles to best represent the radiologic appearance in a patient.
Five scans obtained in patients undergoing regular dental CT examinations, whose approval and informed consent were obtained, were used to compare calibration coefficients with the respective values calculated for the mandibles obtained after death. Our institutional review board does not require its approval for this type of study. Patient-identifying information was removed from the scans. The data were also used to determine whether individual calibration of Hounsfield units to BMD values is necessary or if dental quantitative CT was stable enough to allow the use of calibration parameters that are patient independent. Because the length of scanning is kept to the minimum necessary to minimize patient dose, the average number of scans (n = 26.9) was lower in patients (n = 20) than in specimens (mandibles A–E). Thus, as a selection criterion, a maximum number of sections was used, which resulted in five patients with 29 or more sections without metal artifacts.
Titanium screw-type implants (Brånemark MK III; Nobel Biocare, Göteborg, Sweden; 11.5-mm nominal length, 3.75-mm diameter, regular platform) were used. Insertion was performed (A.B.) by using an Osseo Care surgical drilling equipment (Brånemark; Nobel Biocare) that is capable of measuring torque while screwing the implants into the predrilled implant canal (3.15-mm diameter in this study). With the surgical drilling equipment, the applied torque is recorded by measuring motor load (14,15) as a function of the rotation angle from which insertion depth is derived. In this study, four torque data points per rotation (360°) were recorded. Average torque was calculated for each implant as a mean value of the torque recorded during insertion.
CT examinations (Tomoscan SR-6000; Philips, Best, the Netherlands) were performed. A standard dental CT protocol was used, with overlapping contiguous scans with 1.5-mm section thickness and 1-mm table feed. Scanning time was set to 2 seconds, with 120 kVp and 125–150-mA tube current (the higher value was used for the heads, mandibles D and E). To ensure defined scanning conditions, the mandibles were scanned in an 18-cm diameter water-filled phantom. Inside the cylindric phantom, the mandibles were mounted on a water-equivalent plate in the usual anatomic position. In the cases in which a whole head was scanned, use of the water-filled phantom was unnecessary. A calibration standard (Dental Phantom; Image Analysis, Columbia, Ky) containing three compartments with 0, 75, and 150 mg of calcium hydroxyapatite per cubic centimeter in a water-equivalent matrix was mounted on the outside of the cylindric phantom or fixed to the cheek when imaging of the whole-head specimen or in a patient, respectively, was performed. Scanning was performed before the insertion of dental implants (preoperative) and afterward (postoperative). In the latter case, the phantom was not filled with water, since these scans were not used to deduce quantitative information but only to locate the exact implant positions. Reconstruction was performed by using a neutral reconstruction rather than a sharp (edge-enhancing) kernel, as is normally used for dental CT to preserve quantitative information in the vicinity of edges.
The scanning protocol for patients was identical to the one used for specimens in regard to tube voltage, section thickness, table feed, and scanning time. Tube current (75–100 mA) and field of view were adapted to the individual patient.
CT Image Analyses
All CT image analyses were performed (P.H.) by using a medical image-processing software system (Analyze AVW; Biomedical Image Resource, Mayo Foundation, Rochester, Minn) and a workstation (Ultra Sparc 10; Sun Microsystems, Palo Alto, Calif) (22–25).
Calibration
To determine the relation between Hounsfield units and BMD values, Hounsfield units were measured in three compartments of the dental phantom. Rectangular regions of interest extending to more than 20 x 20 voxels (36 mm3) per section per compartment were used. Three-dimensional analysis included 32–45 (mean, 36.6; SD, 5.2) sections of the five original preoperative data sets. The different number of sections available for calibration resulted from different scanning lengths adapted to the anatomic situation. Calibration was performed individually for every mandible and was based on a linear relationship: BMD = a x HU + b, where a and b are calibration coefficients, which were calculated from the phantom data along with the corresponding statistical descriptors. Voxel values of the preoperative scans were converted from Hounsfield units to BMD values by using this equation.
Calibration data (for specimens and patients) were analyzed individually and pooled for each of the two ensembles. By using the first approach, calibration coefficients for each case, their standard error, P and t values, correlation coefficient, and standard error of the estimate were calculated, as well as mean values and SDs for the different cases. In the second approach, all calibration data were pooled by using all data from each ensemble with equal weight. Average Hounsfield units, their SDs (SD of mean voxel values measured in the appropriate compartment of the dental phantom in one CT section), the resulting calibration coefficients, and the corresponding statistical values were determined.
Registration and Measurement of BMD at Implant Positions
To assess BMD at the exact implant site, implants were identified on the postoperative scans and aligned with the coordinate system of the preoperative scans, which were coded as BMD values. Prior to this registration process, both data sets were interpolated to cubic voxel size (0.3 mm in every spatial dimension) by using a cubic-spline algorithm. The transformation matrix for the postoperative scan was determined with the normalized mutual information method (26). Thresholds were used to limit normalized mutual information to the mandibular bone. The lower thresholds were set to be markedly higher than the soft tissue, that is, 230 mg/cm3 for preoperative scans and 300 HU for postoperative scans. The upper threshold (used only for postoperative scans) was set to 1,000 HU to be lower than the CT numbers of the titanium implants. Thus, their contribution to normalized mutual information was removed. This was necessary, because the titanium signal was not present on the preoperative scans.
Circular regions of interest of 13 voxels (3.9 mm in diameter) were used to define the implants in each section and measure BMD on the calibrated preoperative images. Stacking these regions in three dimensions yielded a cylinder that circumscribed each implant. BMD values corresponding to those of the implants were calculated by averaging BMD values for each cylinder.
|
RESULTS
|
|---|
Calibration
CT numbers measured in different compartments of the calibration standard (dental phantom), SDs, and resulting calibration coefficients (first equation) are shown in the Table, with t and P values, correlation coefficients, and standard errors of the estimates. The parameter a (slope) varied between 0.724 and 0.793 (mean value, 0.760 ± 0.03 SD) for individual calibration of the scans of mandibles A–E. The intercept b values ranged from -2.3 to 8.2 (mean, 2.8 ± 3.7 SD). For all specimens, correlation coefficients of r > 0.999 were obtained. Pooled analysis of the calibration data of all five mandibles yielded the following results: a = 0.758 ± 0.002, P < .001 and b = 3.1 ± 0.3, P < .001.
These calibration values were compared with values derived by using the five scans of actual patients. Individual calibration yielded a values between 0.750 and 0.868 (mean, 0.804 ± 0.06) and b values between 1.0 and 10.9 (mean, 5.2 ± 4.2; Table). Pooled analysis resulted in an a value of 0.796, standard error of 0.004, and P value of less than .001; and a b value of 5.9, standard error of 0.4, and P value of less than .001. The Table includes details of the statistical analysis. Figure 1 shows data and regression lines for mandibles A–E and actual patients’ scans.
Image Analysis
Registration of pre- and postoperative images was performed with a normalized mutual information algorithm and implemented with the image-processing software system (22–26) without user intervention, except for the setting of thresholds and masking of the cervical spine on scans of mandibles D and E. To illustrate the accuracy of the registration of pre- and postoperative images, Figure 2 shows corresponding sections in registered volumes (pre- and postoperative) together with a fused image. Figures 2–4 show mandible A. Figure 3 illustrates the transfer of regions of interest, which indicate implant positions on the postoperative data set, to the preoperative scan series. In Figure 4, the same mandible is shown three dimensionally. Part A, shows a rendered preoperative image and part B, a transparent rendering of the postoperative data to visualize the implants. Part C, is a three-dimensional rendering of the preoperative scan that shows cylindric regions of interest, which correspond to the implant positions, used to measure BMD.
View larger version (106K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2A. Registration of scans of mandible A. Transverse (A) preoperative (window settings: 200 mg/cm3 center and 2,000 mg/cm3 width) scan, (B) postoperative (window settings: 400 HU center and 3,000 HU width) scan, and (C) both scans merged with 50% weighting to demonstrate registration accuracy. With this registration, implant positions were identified on the postoperative scan and used as regions of interest to measure BMD on the preoperative scan at the corresponding location (see Fig 3). The crosshair marks the center of the implant (6 in Fig 4, B).
|
|
View larger version (115K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2B. Registration of scans of mandible A. Transverse (A) preoperative (window settings: 200 mg/cm3 center and 2,000 mg/cm3 width) scan, (B) postoperative (window settings: 400 HU center and 3,000 HU width) scan, and (C) both scans merged with 50% weighting to demonstrate registration accuracy. With this registration, implant positions were identified on the postoperative scan and used as regions of interest to measure BMD on the preoperative scan at the corresponding location (see Fig 3). The crosshair marks the center of the implant (6 in Fig 4, B).
|
|
View larger version (110K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2C. Registration of scans of mandible A. Transverse (A) preoperative (window settings: 200 mg/cm3 center and 2,000 mg/cm3 width) scan, (B) postoperative (window settings: 400 HU center and 3,000 HU width) scan, and (C) both scans merged with 50% weighting to demonstrate registration accuracy. With this registration, implant positions were identified on the postoperative scan and used as regions of interest to measure BMD on the preoperative scan at the corresponding location (see Fig 3). The crosshair marks the center of the implant (6 in Fig 4, B).
|
|
View larger version (56K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3. Transverse (left) preoperative and (right) postoperative scans of mandible A, with regions of interest indicating implant positions. Because of blurring, implants appear larger on the postoperative scan than they actually are (region of interest diameter: 3.9 mm, 13 voxels). Window settings are the same as in Figure 1.
|
|
View larger version (65K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4A. Three-dimensional scans of mandible A obtained from (A) preoperative (rendered with gradient shading) and (B) postoperative data sets. B, Implant positions (1-8) were depicted with summed-voxel rendering. C, Transparent gradient-shaded rendering of the preoperative data set, with superimposed implant positions from B.
|
|
View larger version (62K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4B. Three-dimensional scans of mandible A obtained from (A) preoperative (rendered with gradient shading) and (B) postoperative data sets. B, Implant positions (1-8) were depicted with summed-voxel rendering. C, Transparent gradient-shaded rendering of the preoperative data set, with superimposed implant positions from B.
|
|
View larger version (72K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4C. Three-dimensional scans of mandible A obtained from (A) preoperative (rendered with gradient shading) and (B) postoperative data sets. B, Implant positions (1-8) were depicted with summed-voxel rendering. C, Transparent gradient-shaded rendering of the preoperative data set, with superimposed implant positions from B.
|
|
Correlation of Torque and BMD
Torque values measured at the implant location for the 25 implants are plotted against the corresponding BMD values (Fig 5). The relation between a mean BMD value and the insertion torque was obtained by using a second-order regression model: y = (kx + d)2 with two parameters (k, d). The resulting regression curve for the type of implant used is given by torque = (0.0055 x BMD + 0.73)2 and plotted as a line in Figure 5.
An analysis of variance showed that the relation between an average BMD at implant position and an average insertion torque is significant (P < .001). Standard error of the estimate was 0.40, P values for the parameters k and d were less than .001 and less than .005, respectively, and R2 was 0.83.
|
DISCUSSION
|
|---|
As a result of this study, prediction of achievable insertion torques for defined insertion positions with dental quantitative CT seems feasible, although more patient data are needed to obtain reliable values for the parameters that link BMD to torque. The correlation of BMD and insertion torque of dental implants, as independently measured parameters, suggests that dental quantitative CT can be used to draw quantitative information about the supportive capacity of jaw bone for dental implants.
For mandibles A–E, calibration coefficients used for calculating BMD from Hounsfield units were determined individually. In a routine setting, it would be desirable to determine calibration coefficients once, thus rendering individual calibrations unnecessary. Calibration coefficients calculated by using pooled data from a sufficiently large group of patients could then be used to calculate BMD from Hounsfield units as part of the clinical routine. To judge whether this method represents a reasonable approach, individual calibration data obtained from the actual patients’ scans were compared with the calibration coefficients that resulted from a pooled analysis. The slope a determined individually (mean, 0.804; SD, 0.06; range, 0.750–0.868) varied by approximately 15% among the patients. Thus, the maximum difference, if calibration coefficients determined with the pooled data (from the same patients) were used as opposed to an individual calibration to calculate the expected torque in bone with the torque equation, with an average BMD of 500 mg/cm3, amounts to 1.6 N · cm (11.6%).
Thus, more patient data are needed to decide conclusively whether predetermined calibration factors will deliver the accuracy needed for applications in implant dentistry or whether additional correction procedures are necessary. However, if a patient-independent calibration is used, checks of constancy should be included in a regular quality-control procedure. Caution is also necessary when CT settings are changed. For example, a change in the kilovolt peak from 120 to 100 kV on the CT unit would imply a change in the slope a of approximately 10%, thus requiring a separate calibration.
In this study, a nonlinear dependence of torque on BMD was used. By using the quadratic model with two parameters, R2 of 0.83 was obtained; whereas, a linear model yields a slightly lower R2 of 0.81. In the literature, a nonlinear dependence of cutting energy on radiographically determined aluminum-referred density of bone slabs (15) was reported, which suggests an addition of a quadratic term. On the basis of these results, a quadratic model was favored, although data in Figure 5 alone would not suffice to decide this issue.
Statistical analysis performed with BMD and torque data did not account for the clustering of data points, with the assumption that all 25 BMD-torque data pairs were independent. Some degree of clustering can be assumed, especially in mandibles, where advanced atrophy restricts the choice of a possible implant site. However, the number of data points did not permit a more detailed analysis.
Practical application: In this in vitro pilot study, a strong correlation between BMD and insertion torque was shown with a retrospective determination of BMD at implant positions, yet the relationship demonstrated in this study will have to be verified in a clinical study in which a larger number of implants is involved. By knowing the relation between BMD and insertion torque for the particular implant and surgical procedure, predictions of achievable insertion torque could be drawn preoperatively from a dental quantitative CT scan, which would enable the surgeon to predict implant properties and immediate loading capacity before surgery. Individual preoperative measurements of local BMD from dental CT could be used to determine the best-suited implant sites and types and optimize the prosthetic plan accordingly.
Two planning strategies can be envisaged. In completely or partly edentulous jaws, there may be the possibility to choose implantation sites to optimize immediate loading capacity and long-term success. The other strategy would be to allow preoperative optimization of implant anchorage at a predetermined implant site by adjusting the preparation technique to local BMD, that is, by recommending suitable implant diameters and ratios of pilot drill to implant diameter to achieve an optimum result.
|
ACKNOWLEDGMENTS
|
|---|
The authors acknowledge many useful discussions with Anders Holm, DDS, PhD (Nobel Biocare, St Pölten, Austria), and Håkan Lindström (Nobel Biocare, Göteborg, Sweden) and thank the radiographers, Teresa Keindl and Helge Schöchtner, at the Department of Oral Surgery. Last, the authors owe a note of thanks and respect to all people who, following a long tradition in Vienna, donate their bodies to the Institute of Anatomy for teaching and medical research.
|
FOOTNOTES
|
|---|
Abbreviation: BMD = bone mineral density
Author contributions: Guarantors of integrity of entire study, P.H., A.B., H.B., R.N.; study concepts and design, P.H., A.B.; literature research, P.H., A.B.; experimental studies, P.H., A.B., M.T.; data acquisition, A.G., P.H., A.B.; data analysis/interpretation, P.H., A.B.; statistical analysis, P.H.; manuscript preparation and definition of intellectual content, P.H., A.B., A.G., W.B.; manuscript editing, P.H., A.B.; manuscript revision/review, P.H., A.B., H.B.; manuscript final version approval, all authors.
|
REFERENCES
|
|---|
- van Steenberghe D, Lekholm U, Bolender C, et al. The applicability of osseointegrated oral implants in the rehabilitation of partial edentulism: a prospective multicenter study on 558 fixtures. Int J Oral Maxillofac Implants 1990; 5:272-281.[Medline]
- Jaffin R, Bermann C. The excessive loss of Branemark fixtures in type IV bone: a 5-year analysis. J Periodontol 1991; 62:2-4.[Medline]
- Jacobs R, van Steenberghe D. Radiographic planning and assessment of endosseous oral implants Berlin, Germany: Springer, 1998.
- Quirynen M, Lamoral Y, Deheyser C, et al. The CT scan standard reconstruction technique for reliable jaw bone volume determination. Int J Oral Maxillofac Implants 1990; 5:384-389.[Medline]
- Rothman SLG, Chaftez N, Rhodes ML, Schwarz MS. CT in the preoperative assessment of mandible and maxilla for endosseous implant surgery. Radiology 1988; 168:171-175.[Abstract/Free Full Text]
- Lekholm U, Zarb G. Patient selection and preparation. In: Branemark PI, Zarb G, Albrektsson T, eds. Tissue integrated prostheses: osseointegration in clinical dentistry. Chicago, Ill: Quintessence, 1985; 199-209.
- Branemark PI, Engstrand P, Ohrnell LO, et al. Branemark novum: a new treatment concept for rehabilitation of the edentulous mandible—preliminary results from a prospective clinical follow-up study. Clin Implant Dent Relat Res 1999; 1:2-16.[Medline]
- Friberg B, Sennerby L, Meredith N, Lekholm U. A comparsion between cutting torque and resonance frequency measurements of maxillary implants. Oral Maxillofac Surg 1999; 28:297-303.[CrossRef]
- Friberg B, Sennerby L, Linden B, Grondahl K, Lekholm U. Stability measurements of one-stage Branemark implants during healing in mandibles: a clinical resonance frequency analysis study. Int J Oral Maxillofac Surg 1999; 28:266-272.[CrossRef][Medline]
- Olive J, Aparicio C. Periotest method as a measure of osseointegrated oral implant stability. Int J Oral Maxillofac Implants 1990; 5:390-400.[Medline]
- Meredith N, Rasmusson L, Sennerby L, Alleyne D. Mapping implant stability by resonance frequency analysis. Med Sci Res 1996; 24:191-193.
- Meredith N. Assessment of implant stability as a prognostic determinant. Int J Prosthodont 1998; 11:491-501.[Medline]
- Friberg B, Sennerby L, Roos J, Lekholm U. Identification of bone quality in conjunction with insertion of titanium implants: a pilot study in jaw autopsy specimens. Clin Oral Implants Res 1995; 6:213-219.[CrossRef][Medline]
- Friberg B, Sennerby L, Roos J, Johansson P, Strid CG, Lekholm U. Evaluation of bone density using cutting resistance measurements and microradiography: an in vitro study in pig ribs. Clin Oral Implants Res 1995; 6:164-171.[CrossRef][Medline]
- Johansson P, Strid KG. Assessment of bone quality from cutting resistance during implant surgery. Int J Oral Maxillofac Implants 1994; 9:279-288.
- Lindh C, Nilsson M, Klinge B, Petersson A. Quantitative computed tomography of trabecular bone in the mandible. Dentomaxillofac Radiol 1996; 25:146-150.[Abstract]
- Maki K, Okano T, Morohashi T, Yamada S, Shibaski Y. The application of three-dimensional quantitative computed tomography to the maxillofacial skeleton. Dentomaxillofac Radiol 1997; 26:39-44.[Abstract]
- Maki K, Miller A, Okano T, Shibasaki Y. Changes in cortical bone mineralization in the developing mandible: a three-dimensional quantitative computed tomography study. J Bone Miner Res 2000; 15:700-709.[CrossRef][Medline]
- Iwashita Y. Basic Study of the measurement of bone mineral content of cortical and cancellous bone of the mandible by computed tomography. Dentomaxillofac Radiol 2000; 29:209-215.[Abstract]
- Beer A. Methoden der knochendichtemessung des unterkiefers. Thesis Austria: Department of Biomedical Engineering and Physics, University of Vienna, 2000.
- Ulm C, Blahout R, Matejka M, Gruber H. Reduction of compact and cancellous bone substances of the edentulous mandible caused by resorption. Oral Surg Oral Med Oral Pathol 1992; 74:131-136.[CrossRef][Medline]
- Robb RA, Barillot C. Interactive display and analysis of 3-D medical images. IEEE Trans Med Imaging 1989; 8:217-226.[Medline]
- Robb RA, Hanson DP, Karwoski RA, Larson AG, Workman EL, Stacy MC. ANALYZE: a comprehensive, operator-interactive software package for multidimensional medical image display and analysis. Comput Med Imaging Graph 1989; 13:433-454.[CrossRef][Medline]
- Robb RA, Hanson DP, Camp JJ. Computer-aided surgery planning and rehearsal at Mayo Clinic. IEEE Comput 1996; 29:39-47.
- Robb RA. Biomedical imaging, visualization and analysis New York, NY: Wiley, 1999.
- Studholme C, Hawkes DJ, Hill DLG. A normalized entropy measure for multimodality image alignment. Proc SPIE Med Imaging 1998; 3338:132-143.
TOP
ABSTRACT
INTRODUCTION
, dashed line) and actual patients (
, solid line). Regression parameters are the same as those in the 
= mandible B, 
= mandible C, 
= mandible D, and x = mandible E.