Radiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow MPEG Movie
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Huber, M. B.
Right arrow Articles by Link, T. M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Huber, M. B.
Right arrow Articles by Link, T. M.
(Radiology 2008;247:472-481.)
© RSNA, 2008

Musculoskeletal Imaging

Proximal Femur Specimens: Automated 3D Trabecular Bone Mineral Density Analysis at Multidetector CT—Correlation with Biomechanical Strength Measurement1

Markus B. Huber, MSc, Julio Carballido-Gamio, PhD, Jan S. Bauer, MD, Thomas Baum, Felix Eckstein, MD, Eva M. Lochmüller, MD, Sharmila Majumdar, PhD, and Thomas M. Link, MD

1 From Musculoskeletal and Quantitative Imaging Research, Department of Radiology, University of California, San Francisco, 185 Berry St, Suite 350, San Francisco, CA 94107 (M.B.H., J.C., S.M., T.M.L.); Institut für Röntgendiagnostik, Technische Universität München, Munich, Germany (J.S.B., T.B.); Institute of Anatomy and Musculoskeletal Research, Paracelsus Private Medical University Salzburg, Salzburg, Austria (F.E.); and First University Hospital of Gynecology, Ludwig-Maximilians-Universität, Munich, Germany (E.M.L.). Received June 6, 2007; revision requested August 6; revision received September 6; final version accepted October 4. Supported by AO Foundation grant "Fracture Fixation in Osteoporotic Bone," Project X-ray and CT-Analysis, and German Research Society Foundation grant (DFG) LO 730/3-1. Address correspondence to M.B.H. (e-mail: mbh{at}radiology.ucsf.edu).


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCES IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
 References
 
Purpose: To prospectively evaluate an automated volume of interest (VOI)-fitting algorithm for quantitative computed tomography (CT) of proximal femur specimens, correlate bone mineral density (BMD) with biomechanically determined bone strength in vitro, and compare that correlation with those observed at dual-energy x-ray absorptiometry (DXA) measurement of BMD.

Materials and Methods: The study was compliant with institutional and legislative requirements; donors had dedicated their body for education and research before death. Multidetector CT and DXA scans were acquired in 178 proximal femur specimens harvested from human cadavers (91 women, 87 men; mean age at death, 79 years ± 10.2; range, 52–100 years). An automated VOI-fitting algorithm was used to calculate BMD and bone mineral content (BMC) in the head, neck, and trochanter from CT findings and pixel distribution parameters. The femur failure load (FL) was determined by using a mechanical test. Quantitative CT BMD, quantitative CT pixel distribution parameters, DXA BMD, and FL were correlated at multiple regression analysis.

Results: Mean precision errors in quantitative CT BMD measurements at segmentation with repositioning were 0.56%, 2.26%, and 0.61% for the head, neck, and trochanter, respectively. For the head, neck, and trochanter, respectively, r values were 0.77, 0.53, and 0.59 for the correlation between quantitative CT BMD and FL and 0.74, 0.55, and 0.65 for the correlation between quantitative CT BMC and FL (P < .001). Values ranged from 0.77 to 0.80 for correlations between DXA BMD and FL and from 0.73 to 0.82 for correlations between DXA BMC and FL (P < .001). In a multiple regression model that included quantitative CT pixel distributions, adjusted multivariate correlation coefficient values for correlations with FL increased to up to 0.88.

Conclusion: Regional BMD of the proximal femur can be determined in vitro from quantitative CT data with high precision by using an automated VOI-fitting algorithm. The best multiple regression model for predicting FL included DXA BMD and regional quantitative CT BMD measurements.

© RSNA, 2008

Supplemental material: http://radiology.rsnajnls.org/cgi/content/full/2472070982/DC1


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCES IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
 References
 
Predicting the biomechanical strength of the proximal femur in vivo as a risk factor of fracture is an important goal of current research on the diagnosis of osteoporosis. Osteoporotic fractures of the proximal femur are among the worst consequences of osteoporosis: They substantially increase the mortality risk, are a major cause of disability, and substantially reduce the quality of life (1). Given current demographic trends, the prevalence of these fractures will increase substantially worldwide (2,3). Thus, accurate in vivo estimation of the biomechanical strength of the proximal femur to quantify the fracture risk and assess treatment effectiveness has become an important goal in the management of osteoporosis (4–8).

Current diagnostic techniques for predicting proximal femur fracture risk are mainly based on dual-energy x-ray absorptiometry (DXA) measurements (9–11). Quantitative computed tomography (CT), however, has been used successfully in the spine (12–15) and is also available for measuring bone density in the proximal femur (15–18). DXA currently represents the standard technique for evaluating bone status in the femur, is easy to perform, and has been shown to facilitate prediction of the biomechanical strength of the proximal femur with reasonable accuracy (4,5). Compared with quantitative CT, DXA exposes patients to less radiation and may have lower associated costs. However, the limitations of DXA include soft-tissue errors and lack of capability for differentiating trabecular bone from cortical bone (19,20). Moreover, the DXA-derived bone mineral density (BMD) values determined for patients with and those without prevalent and incident femur fractures have been shown to overlap (8,21,22).

Quantitative CT–derived BMD of the proximal femur may have an advantage over DXA-derived BMD because it can be used to measure the trabecular compartment in selected volumes of interest (VOIs). Trabecular bone is known to be metabolically more active and thus to display larger changes during the evolution and treatment of osteoporosis (6). The proximal femur, however, is a complex three-dimensional structure in which performing automatic segmentation and VOI placement in a reproducible manner is challenging. In addition, to our knowledge, it is still not clear which region is best suited for the prediction of biomechanical strength.

The purposes of our study were to prospectively evaluate an automated VOI-fitting algorithm for quantitative CT of proximal femur specimens, to correlate BMD with biomechanically determined bone strength in vitro, and to compare the aforementioned correlation with the correlations observed at DXA measurement of BMD.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 ADVANCES IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
 References
 
Femur Specimens
One hundred ninety-five left femur specimens were harvested from formalin-fixed human cadavers for 4 years. The donors had donated their bodies to the Institute of Anatomy and Musculoskeletal Research, Paracelsus Private Medical University Salzburg, for educational and research purposes prior to death, in compliance with local institutional and legislative requirements. To exclude donors with diffuse metastatic bone disease and hematologic or metabolic bone disorders other than osteoporosis, biopsy samples were obtained from the iliac crest and examined histologically as part of the general research protocol. These histologic assessments were performed by a surgeon who had been a trained full-time pathologist for 3 years, with a focus on bone pathology. In addition, radiographs of all specimens were obtained and assessed for osteolytic changes or other focal abnormalities by two radiologists (T.M.L., J.S.B.; 15 and 5 years experience in musculoskeletal imaging, respectively). Specimens with signs of fracture detected either on the radiographs or during preparation for storage and scanning, as well as specimens that showed a fracture of the femur shaft (rather than of the proximal femur) during mechanical testing, were excluded. After these exclusions, 178 human femur specimens remained for inclusion in the study. The donors (91 women, 87 men) had a mean age of 79.4 years ± 10.2 (standard deviation [SD]) (age range, 52–100 years) at death. Although the mean ages of the female and male donors were not significantly different (P > .05), the male donors were significantly taller and heavier than the female donors (P < .05).

The bones, along with a variable amount of surrounding soft tissue, were removed from the cadavers. To create uniform scanning conditions, the soft tissues around the bones were then removed before imaging and biomechanical testing. The specimens were degassed for at least 24 hours before multidetector CT. During the study, the specimens were stored in fixation solution to prevent storage and air artifacts.

DXA Measurements
A radiologist with 3 years experience performed the DXA measurements by using a GE Lunar Prodigy scanner (GE Medical Systems, Milwaukee, Wis). The femur specimens were positioned in a manner similar to the positioning of femurs in in vivo conditions: They were mildly internally rotated in a container filled with tap water up to 15 cm in height, which was intended to simulate soft tissue. The evaluations were performed by using Lunar Prodigy Encore 2002 software (GE Medical Systems). Area BMD and bone mineral content (BMC) were measured with DXA in four regions of interest (ROIs) in the femur: the neck, the trochanter, the shaft, and the entire proximal femur.

Multidetector CT Measurements
Cross-sectional images of the femora were acquired by using a 16-detector CT scanner (Sensation 16; Siemens Medical Solutions, Erlangen, Germany). The specimens were placed in plastic bags filled with a 4% formalin–water solution. The plastic bags were sealed after air was removed from them by using a vacuum pump. These bags were positioned in the scanner to simulate the conditions in an in vivo examination of the pelvis and proximal femur, with mild internal rotation of the femur. Each specimen was scanned at least once by using a protocol involving a collimation and table feed of 0.75 mm and a reconstruction index of 0.5 mm. A high-spatial-resolution reconstruction algorithm (kernel U70u) was used, with a resulting in-plane resolution of 0.29 x 0.29 mm. Additional scanning parameters were 120 kVp, 100 mA, an image matrix of 512 x 512 pixels, and a field of view of 100 mm. For calibration purposes, a reference phantom (Osteo Phantom, Siemens Medical Solutions) was placed in the scanner below the specimens.

Image Processing and VOI Selection
In a first step, the outer surface of the cortical shell of the femur was segmented by using the bone attenuations of the phantom on each image. The specimens were segmented automatically; however, the shape of the binary mask was manually corrected if errors in segmentation occurred owing to either a thin cortical shell caused by high-grade focal bone loss or adjacent anatomic structures, such as blood vessels, penetrating the cortex. For all specimens, these corrections were performed by one of two radiologists (T.M.L. or J.S.B.).

On the basis of a priori knowledge about the orientation of the specimens on the CT scans, the superior part of the femur head was identified automatically. On the basis of the size and shape of the contours and the center of mass of the contours on consecutive sections, the superior part of the femur head was detected. Since Naish et al found that a sphere could approximate the shape of the femur head (23), a sphere was fitted to the superior surface points of the femur head by using a Gaussian Newton least-squares technique. The fitted sphere was scaled down to 75% of its original size to account for cortical bone and shape irregularities such as fovea capitis and then saved as the femur head VOI.

Because a cylinder can approximate the shape of the femur neck, a cylindric VOI was computed and automatically fitted to the neck region. Starting with an initial estimate for the axis of a cylinder between the center of mass of the fitted sphere and the intersection between the prolonged neck axis and the lateral bone surface, fitting was then performed by using the Gaussian Newton least-squares technique. The axis derived from this fitting was used to orient the neck axis, and the radius of the fitted cylinder was used to establish the neck axis length (65% of the radius of the fitted cylinder). To account for cortical bone and shape irregularities, the radius of the initially fitted cylinder was discarded and another radius was computed on the basis of the distances of the bone surface points in this region to the neck axis. The final radius was the minimum of all distances, minus 50% of the SD of all distances. The resulting cylinder was saved as the femur neck VOI.

To detect the trochanter region, the neck axis was prolonged downward to intersect the bone surface laterally. On the basis of this intersection and the positions of the bone surface points relative to the neck axis, the surface regions corresponding to the trochanter, inferior part of the neck, and superior part of the shaft were detected. The main eigenvector of these regions was used as an initial estimate of the axis of a cone that was fitted to the bone surface points in these regions. Bone surface points in these regions were matched to the fitted cone axis and to the original neck axis. The trochanter bone surface points were then saved as the trochanter VOI. Figure 1 shows all of the VOIs on three-dimensional CT depictions of a representative proximal femur specimen. The sphere in the femur head, the cylinder in the neck, and the trochanter volume are highlighted (movie, http://radiology.rsnajnls.org/cgi/content/full/2472070982/DC1).


Figure 1
View larger version (80K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1: Coronal anterior (left) and coronal posterior (right) quantitative CT depictions of head (sphere), neck (cylinder), and trochanter (irregular) VOIs in femur specimens.

 
Three authors (J.C., M.B.H., T.B.; 10, 6, and 2 years experience, respectively) supervised by two radiologists (T.M.L., J.S.B.) conducted all of the image-processing steps and calculated measurements at Sun workstations (Sun Microsystems, Santa Clara, Calif) with software developed in house by using the MATLAB, version 7.0, program (MathWorks, Natick, Mass).

Quantitative CT BMD Measurements
The mean BMD of each VOI was calculated by converting the pixel attenuations (in Hounsfield units) into BMD values (in milligrams per cubic centimeter) by using the calibration phantom composed of hydroxyapatite: HAW = 0 mg/cm3 and HAB = 200 mg/cm3 for the densities of the water-like and bonelike parts of the phantom, respectively. In addition to these constants, the attenuations of the sections in the phantom were measured on each image: HUW and HUB for the attenuations of the water-like and bonelike parts of the phantom, respectively. By assuming the linear relationship, BMD is proportional to attenuation (HU), one can use the following conversion formula to calculate the BMD: BMD = [HAB/(HUB – HUW)] · (HU – HUW).

Note that by definition, pixel attenuations lower than the attenuation of the water-like part of the phantom will yield negative BMD values, as compliant with the convention for adipose tissue. This technique was previously applied and described by Bauer et al (24). BMD was defined as the mean transformed pixel attenuation value in the VOI. To derive the BMC of the VOIs, each BMD value was multiplied by the volume of the corresponding VOI.

To calculate the short-term precision error in the fitting techniques (25), the segmentation results for six specimens (from three female and three male cadavers, randomly chosen) were analyzed and, if required, manually corrected by two radiologists (T.M.L., J.S.B.). From this subsample, the precision error was calculated as the root mean square of the coefficient of variation of the specimens and expressed as a percentage. In a second test, three specimens were scanned twice with repositioning. We applied our VOI selection algorithm for both acquisitions, and two radiologists (T.M.L., J.S.B.) analyzed the segmentation results.

Biomechanical Tests
The failure load was assessed by using a side-impact test in which a lateral fall on the greater trochanter was simulated, as described previously (26). Briefly, the femur shaft and head were faced downward and could be moved independently of one another while a load was applied to the greater trochanter by using a universal materials testing machine (Zwick 1445; Zwick, Ulm, Germany) with a 10-kN force sensor and dedicated software. Failure load was defined as the peak of the load-deformation curve.

Statistical Analyses
Mean quantitative CT BMD pixel measurements (±SD) in the described VOIs were calculated. Kolmogorov-Smirnov testing was used to determine whether the frequencies followed a normal distribution. Differences between the subgroups were evaluated by using the two-tailed t test.

Pearson correlation coefficients (r) were computed to analyze linear dependencies between parameters. To test whether two correlations were significantly different, we calculated the Fisher z transformation of each correlation coefficient. The difference between the two z transformations was thus an estimate of the probability that the two correlations were statistically equal.

Multiple linear regression analyses were performed to determine the most important BMD parameters for predicting femur failure load for all donors, for the female-cadaver donors separately, and for the male-cadaver donors separately. For these three data sets, we sought to determine the best parameter or combination of parameters in five BMD measurement models: (a) quantitative CT BMD (model 1); (b) DXA BMD (model 2); (c) quantitative CT– and DXA-derived BMD (model 3); (d) quantitative CT BMD and SD of quantitative CT BMD (model 4); and (e) quantitative CT BMD, DXA BMD, and SD of quantitative CT BMD (model 5). A standard F statistic was calculated, and a parameter was included in a stepwise procedure. We reported adjusted multivariate correlation coefficient (Radj) values to account for the complexity of the model. In this analysis, P = .05 indicated significance. One author (M.B.H.), supervised by a biostatistician, performed the statistical analyses by using SPSS, version 13.0 (SPSS, Chicago, Ill), and MATLAB, version 7.0 (MathWorks), software.


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 ADVANCES IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
 References
 
Precision Error and Parameters
In the initial segmentations, mean precision errors of 0.18%, 0.76%, and 0.27% were calculated for the head, neck, and trochanter BMD measurements, respectively (Table 1). For the specimens scanned twice, we calculated mean precision errors of 0.56%, 2.26%, and 0.61% for the head, neck, and trochanter BMD measurements, respectively (Table 1).


View this table:
[in this window]
[in a new window]

  		Table 1. Short-term Precision Errors for Manual Corrections in Initial Segmentation and Segmentation with Repositioning

 
Correlations
There were significant sex-related differences in most of the parameters (P < .001), and all parameters were normally distributed within the sample (P > .05) (Table 2). Mean trabecular bone quantitative CT–derived BMD and BMC values were highest in the femur head VOI (224 mg/cm3 ± 60.4 [SD] and 5.46 g ± 2.21, respectively) and lowest in the femur neck VOI (43.5 mg/cm3 ± 52.4 and 0.161 g ± 0.206, respectively). Quantitative CT measurements were significantly different between the head, neck, and trochanter VOIs (P < .001), and DXA BMD values were significantly different between the neck, entire femur, trochanter, and shaft ROIs (P < .001).


View this table:
[in this window]
[in a new window]

  		Table 2. Sex-based Distribution of Measurements in Proximal Femur Specimens

 
In terms of quantitative CT BMD values measured in the VOIs, the highest correlation with femur failure load was observed in the head (r = 0.77, P < .001), and the lowest such correlation was observed in the neck (r = 0.53, P < .001) (Table 3, Fig 2). In the neck and trochanter, quantitative CT BMC (Table 4) correlated with femur failure load slightly better than did quantitative CT BMD. Sex-based correlations between quantitative CT BMD and failure load and between quantitative CT BMC and failure load were not significantly different. In the trochanter, we observed the highest correlation between CT VOI size and failure load. Coefficients (r) for the correlation between DXA BMD and failure load ranged from 0.77 to 0.80 (Table 3), and those for the correlation between DXA BMC and failure load ranged from 0.73 to 0.82 (P < .001 for all comparisons) (Table 4). None of the correlations between failure load and DXA-derived BMD or BMC was significantly higher than the correlations between failure load and quantitative CT–derived BMD or BMC in the head VOI (P > .05).


View this table:
[in this window]
[in a new window]

  		Table 3. Coefficients for Correlations between Parameters

 

Figure 2A
View larger version (13K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2a: Scatterplots show relationships between failure load (FL) and BMD. Coefficients (r) for correlations with failure load were as follows: (a) 0.80 for correlation with DXA BMD, (b) 0.77 for correlation with quantitative CT (QCT) BMD in femur head, (c) 0.53 for correlation with quantitative CT BMD in femur neck, and (d) 0.59 for correlation with quantitative CT BMD in trochanter (troch.). All correlations were significant (P < .001). Each solid line represents the fit to a linear regression model.

 

Figure 2B
View larger version (14K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2b: Scatterplots show relationships between failure load (FL) and BMD. Coefficients (r) for correlations with failure load were as follows: (a) 0.80 for correlation with DXA BMD, (b) 0.77 for correlation with quantitative CT (QCT) BMD in femur head, (c) 0.53 for correlation with quantitative CT BMD in femur neck, and (d) 0.59 for correlation with quantitative CT BMD in trochanter (troch.). All correlations were significant (P < .001). Each solid line represents the fit to a linear regression model.

 

Figure 2C
View larger version (15K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2c: Scatterplots show relationships between failure load (FL) and BMD. Coefficients (r) for correlations with failure load were as follows: (a) 0.80 for correlation with DXA BMD, (b) 0.77 for correlation with quantitative CT (QCT) BMD in femur head, (c) 0.53 for correlation with quantitative CT BMD in femur neck, and (d) 0.59 for correlation with quantitative CT BMD in trochanter (troch.). All correlations were significant (P < .001). Each solid line represents the fit to a linear regression model.

 

Figure 2D
View larger version (15K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2d: Scatterplots show relationships between failure load (FL) and BMD. Coefficients (r) for correlations with failure load were as follows: (a) 0.80 for correlation with DXA BMD, (b) 0.77 for correlation with quantitative CT (QCT) BMD in femur head, (c) 0.53 for correlation with quantitative CT BMD in femur neck, and (d) 0.59 for correlation with quantitative CT BMD in trochanter (troch.). All correlations were significant (P < .001). Each solid line represents the fit to a linear regression model.

 

View this table:
[in this window]
[in a new window]

  		Table 4. Coefficients for Correlations between Quantitative CT– and DXA-derived BMC Measurements

 
The highest correlations between the quantitative CT and DXA measurements were observed in corresponding anatomic regions (Table 3). The correlation between BMD measurements in the femur neck (r = 0.72, P < .001) was significantly lower (P < .001) than that in the trochanter (r = 0.85, P < .001). Correlations between quantitative CT– and DXA-derived BMC were similar, with significantly lower (P = .002) correlations in the neck than in the trochanter (Table 4).

All quantitative CT measurements of BMD and VOI size correlated significantly with the weight and height of the cadaveric donors (Table 3). Neither quantitative CT BMD in the trochanter nor any of the quantitative CT VOI size measurements correlated with donor age. All DXA BMD values correlated significantly with donor age, weight, and height. The strongest correlation between DXA BMD and donor weight was measured in the trochanter (r = 0.59, P < .001), whereas correlations between DXA BMD and donor age and height were lower, and all were in a similar range.

The mean SDs of quantitative CT BMD values were 170 mg/cm3 ± 27.8 (SD), 110 mg/cm3 ± 31.5, and 108 mg/cm3 ± 18.8 in the head, neck, and trochanter VOIs, respectively. The mean SD of quantitative CT BMD in the head was significantly (P < .001) different from those in the neck and trochanter. The coefficients (r) for correlations between femur failure load and SD of quantitative CT BMD were 0.42, 0.43, and 0.53 for the head, neck, and trochanter VOIs, respectively (P < .001 for all correlations).

At multiple linear regression analysis, Radj values for the model consisting of quantitative CT BMD VOIs (model 1) were 0.77, 0.82, and 0.74 for all femur specimens, the femur specimens from female cadavers only, and the femur specimens from male cadavers only, respectively (Table 5). For the entire specimen cohort, use of model 1 in the head VOI alone yielded the highest Radj; quantitative CT BMD values in the neck and trochanter VOIs did not contribute to the prediction of femur failure load. However, for the female-cadaver femur specimens, the highest Radj was calculated for a combination of head, neck, and trochanter VOIs. With use of the DXA BMD parameters only (model 2), we calculated Radj values of 0.82, 0.83, and 0.74 for all specimens, the female-donor specimens, and the male-donor specimens, respectively. In all DXA models, the neck ROI was represented.


View this table:
[in this window]
[in a new window]

  		Table 5. Multiple Linear Regression BMD Measurement Models for Predicting Femur Failure Load

 
When quantitative CT– and DXA-derived BMD measurements were combined (model 3), Radj values increased significantly (P < .05): to 0.84, 0.86, and 0.77 for all specimens, the female-donor specimens, and the male-donor specimens, respectively. For all specimens and for the female-donor specimens, DXA BMD measurements in the neck and trochanter ROIs and quantitative CT BMD measurements in the head and trochanter VOIs contributed to the prediction of failure load.

For all specimens and for the female-donor specimens, adding the SD of quantitative CT BMD values (model 4), compared with using the quantitative CT BMD variables alone, significantly improved (P < .05) the correlation. Radj values reached 0.80 (all specimens) and 0.83 (female-donor specimens).

Results for model 5—involving quantitative CT BMD, DXA BMD, and the SD of quantitative CT BMD variables—showed that using SD of quantitative CT BMD variables could significantly improve (P < .05) the correlations of both DXA- and quantitative CT–derived BMD measurements for all specimens (Radj = 0.85) and for female-donor specimens (Radj = 0.88). For all models and data sets, donor age was not associated with significantly improved correlations.


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
ADVANCES IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
 References
 
Our study data show that trabecular bone BMD and BMC measured with quantitative CT can be used to predict biomechanical strength. Strong relationships between these parameters and femur failure load were observed—especially in the femur head. Our study findings show that the variation in BMD pixel values within the VOIs also facilitated the prediction of failure load. Although the trabecular bone quantitative CT BMD measurements were slightly inferior to the DXA BMD measurements, they yielded substantial information in addition to DXA BMD for the prediction of failure load.

Quantitative CT can be used to measure bone compartments individually, whereas DXA enables a purely integral, projectional bone measurement, including that of cortical and trabecular bone. The cortical shell is known to make a substantial contribution to the biomechanical strength of the proximal femur (27,28). In the femur neck, where calculated correlations between quantitative CT and DXA measurements were significantly lower than those in the trochanter region, cortical bone had the greatest effect on failure load. However, trabecular bone is substantially more active metabolically. A study to investigate the influence of various medications for the treatment of osteoporosis revealed that the time required to evaluate the efficiency of an intervention is reduced with use of compartment-specific quantitative CT measurements (6).

Our quantitative CT–derived BMD and BMC measurements in the head, neck, and trochanter compartments, similar to those in other studies, were significantly different (29,30). These findings are compatible with the finite element modeling results in a previous investigation (31), which demonstrated a shift in the distribution of load from mainly the trabecular bone near the femur head (70%) to predominantly the cortical bone at the base of the femur neck (96%) during both gaits and falls. The investigators concluded that the trabecular bone in the trochanter mainly provides mechanical stability for muscle attachments rather than increases the structural integrity of the bone itself.

Quantitative CT–derived BMD and BMC measurements in the femur head were highest and correlated best with failure load, confirming the results of a study that involved assessment of femur head VOIs (29) and supporting the importance of trabecular bone in this compartment. Interestingly, the neck VOI had a high fat content. Kuiper et al (32) estimated the influence of the marrow fat in the femur neck on quantitative CT and DXA measurements. With both techniques, the amount of bone mineral was underestimated by 0.7% per 1% of fat content. This effect may reduce quantitative CT correlations with failure load: In our trabecular bone quantitative CT analysis, the fraction of marrow fat in the neck VOI was higher than that in DXA ROIs, where cortical bone is included.

The performance of quantitative CT BMD in the prediction of failure load has been investigated in several studies. Most of the techniques used in these studies, however, have involved manual or semiautomatic VOI placement. Lang et al (16) obtained quantitative CT BMD scans of 26 femur specimens (mean donor age at death, 73 years ± 11). They manually defined a femur neck axis in the CT volumes, which was used in an algorithm to determine VOIs. The calculated trabecular neck and trochanter VOIs resulted in correlations with failure load (r = 0.77 and r = 0.89, respectively) that were higher than those observed in our study. This may be explained by the smaller sample size and younger donor age in their study (16).

By using a manual two-dimensional ROI placement technique, Buitrago-Tellez et al (29) analyzed head, neck, and trochanter ROIs on the quantitative CT images of 41 femur specimens. The head ROI was found to have the highest correlation (r = 0.76) with failure load; this finding is supported by our results. Cheng et al (33) used a manually segmented two-dimensional ROI and calculated a correlation (r = 0.83) between trabecular bone quantitative CT BMD and failure load in the trochanter for a sample of 64 specimens (mean donor age at death, 69 years ± 15). Our multivariate regression results support the findings of Cheng et al, who concluded that an optimal correlation with DXA parameters is only slightly better than an optimal correlation with quantitative CT parameters.

Our study did not yield the high correlation coefficients calculated in a number of previously published studies. However, specimens from younger donors were used in all of these other studies. In a recent study, Bousson et al (18) created neck fractures in 28 femur specimens from donors of ages similar to the ages of the donors in our study (mean age at death, 84 years ± 12) and reported r values for the correlation between failure load and quantitative CT BMD in the neck of up to 0.60. Our results confirm these findings, and we also confirmed that including age in the analysis did not improve the multivariate regressions. It appears that the variability in failure load in older patients is based less on the mineralization of trabecular bone.

The prediction of failure load with DXA was superior in the neck and trochanter ROIs compared with corresponding predictions based on trabecular bone quantitative CT–derived BMD and BMC. A potential reason that CT analysis did not perform significantly better than DXA in the prediction of mechanical failure load is that the CT analysis was confined to trabecular bone, whereas DXA is also performed in cortical bone, which also has been shown to be important for assessing whole-bone strength (27). This may indicate that the cortical shell has a more important role in the biomechanical strength of the proximal femur in older subjects and during the later stages of osteoporosis. Results of a study with finite element modeling support this hypothesis (31).

With our method, as compared with other algorithms (14,34) used to generate VOIs in the proximal femur for BMD measurement, the image itself is not altered by interpolating pixel attenuations, which can increase precision errors. Instead, our algorithm fits the VOIs to the binary masks of the images. Other potential applications of this approach are trabecular bone structure analysis (10,35,36) and finite element modeling (37,38).

Short-term precision errors in quantitative CT BMD measurements in the head, neck, and trochanter VOIs were smaller than those previously encountered with manual VOI placement techniques (14,24). In addition, precision errors were highest in the neck compartment in these previous studies (14,24). Our precision errors in quantitative CT–derived BMD and VOI size measurements were comparable to those in another automated VOI-finding algorithm: errors of less than 3% for BMD measurements and less than 1.5% for VOI size measurements (39). In that previous work, an anatomic coordinate system was found automatically and VOIs were placed relative to this reference system. Investigators in another study (14) used registration techniques and reported precision errors of 4.53% for the neck VOI and 0.60% for the trochanter VOI.

A limiting factor of our study was the manual correction of the segmentation results, which potentially induces operator-dependent error and in our study resulted in the described quantitative CT BMD precision errors. We performed these corrections in a research setting under well-controlled quality conditions. In a general clinical setting, however, such an environment may not be possible, so the reported precision errors may increase.

In conclusion, we developed an automated algorithm for fitting VOIs to determine the regional BMD of the proximal femur in vitro at quantitative CT. Regional quantitative CT BMD correlated less strongly with failure load than did DXA BMD, but the best multiple regression model for predicting failure load included both DXA- and quantitative CT–derived BMD measurements.


   ADVANCES IN KNOWLEDGE
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCES IN KNOWLEDGE
IMPLICATION FOR PATIENT CARE
 References
 


   IMPLICATION FOR PATIENT CARE
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCES IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
References
 


   ACKNOWLEDGMENTS
 
We thank Matthias H. Priemel, MD, Department of Trauma Surgery, Hamburg University School of Medicine, Hamburg, Germany, for performing the histologic assessments. We thank Holger Boehm, MD, Klinikum Grosshadern, Ludwig-Maximilians-Universität, for performing the DXA measurements. We thank Ying Lu, PhD, Department of Radiology, University of California, San Francisco, for the statistical consultations.


   FOOTNOTES
 

Abbreviations: BMC = bone mineral content • BMD = bone mineral density • DXA = dual-energy x-ray absorptiometry • Radj = adjusted multivariate correlation coefficient • ROI = region of interest • SD = standard deviation • VOI = volume of interest

Author contributions: Guarantors of integrity of entire study, M.B.H., J.C., F.E., S.M., T.M.L.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, M.B.H., J.C., J.S.B., F.E., S.M., T.M.L.; experimental studies, all authors; statistical analysis, M.B.H., J.S.B., T.B.; and manuscript editing, M.B.H., J.C., J.S.B., F.E., S.M., T.M.L.

Authors stated no financial relationship to disclose.


   References
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCES IN KNOWLEDGE
 IMPLICATION FOR PATIENT CARE
 References
 

  1. Tromp AM, Ooms ME, Popp-Snijders C, Roos JC, Lips P. Predictors of fractures in elderly women. Osteoporos Int 2000;11:134–140.[CrossRef][Medline]
  2. Cooper C, Campion G, Melton LJ 3rd. Hip fractures in the elderly: a world-wide projection. Osteoporos Int 1992;2:285–289.[CrossRef][Medline]
  3. Burge R, Dawson-Hughes B, Solomon DH, Wong JB, King A, Tosteson A. Incidence and economic burden of osteoporosis-related fractures in the United States, 2005–2025. J Bone Miner Res 2007;22:465–475.[CrossRef][Medline]
  4. Johnell O, Kanis JA, Oden A, et al. Predictive value of BMD for hip and other fractures. J Bone Miner Res 2005;20:1185–1194.[CrossRef][Medline]
  5. Abrahamsen B, Vestergaard P, Rud B, et al. Ten-year absolute risk of osteoporotic fractures according to BMD T score at menopause: the Danish Osteoporosis Prevention Study. J Bone Miner Res 2006;21:796–800.[CrossRef][Medline]
  6. Black DM, Greenspan SL, Ensrud KE, et al. The effects of parathyroid hormone and alendronate alone or in combination in postmenopausal osteoporosis. N Engl J Med 2003;349:1207–1215.[Abstract/Free Full Text]
  7. Black DM, Steinbuch M, Palermo L, et al. An assessment tool for predicting fracture risk in postmenopausal women. Osteoporos Int 2001;12:519–528.[CrossRef][Medline]
  8. Kanis JA, Borgstrom F, De Laet C, et al. Assessment of fracture risk. Osteoporos Int 2005;16:581–589.[CrossRef][Medline]
  9. Mazess R, Collick B, Trempe J, Barden H, Hanson J. Performance evaluation of a dual-energy x-ray bone densitometer. Calcif Tissue Int 1989;44:228–232.[Medline]
  10. Boehm HF, Eckstein F, Wunderer C, et al. Improved performance of hip DXA using a novel region of interest in the upper part of the femoral neck: in vitro study using bone strength as a standard of reference. J Clin Densitom 2005;8:488–494.[CrossRef][Medline]
  11. Eckstein F, Lochmuller EM, Lill CA, et al. Bone strength at clinically relevant sites displays substantial heterogeneity and is best predicted from site-specific bone densitometry. J Bone Miner Res 2002;17:162–171.[CrossRef][Medline]
  12. Lang T, LeBlanc A, Evans H, Lu Y, Genant H, Yu A. Cortical and trabecular bone mineral loss from the spine and hip in long-duration spaceflight. J Bone Miner Res 2004;19:1006–1012.[CrossRef][Medline]
  13. Bouxsein ML, Melton LJ 3rd, Riggs BL, et al. Age- and sex-specific differences in the factor of risk for vertebral fracture: a population-based study using quantitative CT. J Bone Miner Res 2006;21:1475–1482.[CrossRef][Medline]
  14. Li W, Sode M, Saeed I, Lang T. Automated registration of hip and spine for longitudinal quantitative CT studies: integration with 3D densitometric and structural analysis. Bone 2006;38:273–279.[Medline]
  15. Lang TF, Guglielmi G, van Kuijk C, De Serio A, Cammisa M, Genant HK. Measurement of bone mineral density at the spine and proximal femur by volumetric quantitative computed tomography and dual-energy x-ray absorptiometry in elderly women with and without vertebral fractures. Bone 2002;30:247–250.[Medline]
  16. Lang TF, Keyak JH, Heitz MW, et al. Volumetric quantitative computed tomography of the proximal femur: precision and relation to bone strength. Bone 1997;21:101–108.[Medline]
  17. Riggs BL, Melton LJ 3rd, Robb RA, et al. Population-based analysis of the relationship of whole bone strength indices and fall-related loads to age- and sex-specific patterns of hip and wrist fractures. J Bone Miner Res 2006;21:315–323.[CrossRef][Medline]
  18. Bousson V, Le Bras A, Roqueplan F, et al. Volumetric quantitative computed tomography of the proximal femur: relationships linking geometric and densitometric variables to bone strength—role for compact bone. Osteoporos Int 2006;17:855–864.[CrossRef][Medline]
  19. Bolotin HH, Sievanen H, Grashuis JL. Patient-specific DXA bone mineral density inaccuracies: quantitative effects of nonuniform extraosseous fat distributions. J Bone Miner Res 2003;18:1020–1027.[CrossRef][Medline]
  20. Lochmuller EM, Krefting N, Burklein D, Eckstein F. Effect of fixation, soft-tissues, and scan projection on bone mineral measurements with dual energy x-ray absorptiometry (DXA). Calcif Tissue Int 2001;68:140–145.[CrossRef][Medline]
  21. Taylor BC, Schreiner PJ, Stone KL, et al. Long-term prediction of incident hip fracture risk in elderly white women: study of osteoporotic fractures. J Am Geriatr Soc 2004;52:1479–1486.[CrossRef][Medline]
  22. Cummings SR, Nevitt MC, Browner WS, et al. Risk factors for hip fracture in white women: Study of Osteoporotic Fractures Research Group. N Engl J Med 1995;332:767–773.[Abstract/Free Full Text]
  23. Naish JH, Xanthopoulos E, Hutchinson CE, Waterton JC, Taylor CJ. MR measurement of articular cartilage thickness distribution in the hip. Osteoarthritis Cartilage 2006;14:967–973.[CrossRef][Medline]
  24. Bauer JS, Kohlmann S, Eckstein F, Mueller D, Lochmuller EM, Link TM. Structural analysis of trabecular bone of the proximal femur using multislice computed tomography: a comparison with dual x-ray absorptiometry for predicting biomechanical strength in vitro. Calcif Tissue Int 2006;78:78–89.[CrossRef][Medline]
  25. Gluer CC, Blake G, Lu Y, Blunt BA, Jergas M, Genant HK. Accurate assessment of precision errors: how to measure the reproducibility of bone densitometry techniques. Osteoporos Int 1995;5:262–270.[CrossRef][Medline]
  26. Eckstein F, Wunderer C, Boehm H, et al. Reproducibility and side differences of mechanical tests for determining the structural strength of the proximal femur. J Bone Miner Res 2004;19:379–385.[CrossRef][Medline]
  27. Augat P, Reeb H, Claes LE. Prediction of fracture load at different skeletal sites by geometric properties of the cortical shell. J Bone Miner Res 1996;11:1356–1363.[Medline]
  28. Lochmuller EM, Burklein D, Kuhn V, et al. Mechanical strength of the thoracolumbar spine in the elderly: prediction from in situ dual-energy x-ray absorptiometry, quantitative computed tomography (quantitative CT), upper and lower limb peripheral quantitative CT, and quantitative ultrasound. Bone 2002;31:77–84.[Medline]
  29. Buitrago-Tellez CH, Bonnaire F, Schulze C, et al. Quantitative CT assessment of the proximal femur: experimental studies on its correlation with breaking load in femoral neck fractures [in German]. Rofo 1997;167:627–632.[Medline]
  30. Issever AS, Vieth V, Lotter A, et al. Local differences in the trabecular bone structure of the proximal femur depicted with high-spatial-resolution MR imaging and multisection CT. Acad Radiol 2002;9:1395–1406.[CrossRef][Medline]
  31. Lotz JC, Cheal EJ, Hayes WC. Stress distributions within the proximal femur during gait and falls: implications for osteoporotic fracture. Osteoporos Int 1995;5:252–261.[CrossRef][Medline]
  32. Kuiper JW, van Kuijk C, Grashuis JL, Ederveen AG, Schutte HE. Accuracy and the influence of marrow fat on quantitative CT and dual-energy x-ray absorptiometry measurements of the femoral neck in vitro. Osteoporos Int 1996;6:25–30.[CrossRef][Medline]
  33. Cheng XG, Lowet G, Boonen S, et al. Assessment of the strength of proximal femur in vitro: relationship to femoral bone mineral density and femoral geometry. Bone 1997;20:213–218.[Medline]
  34. Ehrhardt J, Handels H, Plotz W, Poppl SJ. Atlas-based recognition of anatomical structures and landmarks and the automatic computation of orthopedic parameters. Methods Inf Med 2004;43:391–397.[Medline]
  35. Mueller D, Link TM, Monetti R, et al. The 3D-based scaling index algorithm: a new structure measure to analyze trabecular bone architecture in high-resolution MR images in vivo. Osteoporos Int 2006;17:1483–1493.[CrossRef][Medline]
  36. Carballido-Gamio J, Phan C, Link TM, Majumdar S. Characterization of trabecular bone structure from high-resolution magnetic resonance images using fuzzy logic. Magn Reson Imaging 2006;24:1023–1029.[CrossRef][Medline]
  37. Cody DD, Gross GJ, Hou FJ, Spencer HJ, Goldstein SA, Fyhrie DP. Femoral strength is better predicted by finite element models than quantitative CT and DXA. J Biomech 1999;32:1013–1020.[CrossRef][Medline]
  38. Keyak JH. Nonlinear finite element modeling to evaluate the failure load of the proximal femur. J Orthop Res 2000;18:337.[CrossRef][Medline]
  39. Kang Y, Engelke K, Fuchs C, Kalender WA. An anatomic coordinate system of the femoral neck for highly reproducible BMD measurements using 3D quantitative CT. Comput Med Imaging Graph 2005;29:533–541.[CrossRef][Medline]




This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow MPEG Movie
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Huber, M. B.
Right arrow Articles by Link, T. M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Huber, M. B.
Right arrow Articles by Link, T. M.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
RADIOLOGY RADIOGRAPHICS RSNA JOURNALS ONLINE