Rapid Small-Animal Dual-Energy X-Ray Absorptiometry Using Digital Radiography


  • D. W. Holdsworth,

    Corresponding author
    1. Imaging Research Laboratories, John P. Robarts Research Institute, London, Ontario, Canada
    2. Department of Medical Biophysics, University of Western Ontario, London, Ontario, Canada
    • Address reprint requests to: Dr. David W. Holdsworth, Imaging Research Laboratories, John P. Robarts Research Institute, 100 Perth Drive, London, Ontario N6A 5K8, Canada
    Search for more papers by this author
  • M. M. Thornton,

    1. Imaging Research Laboratories, John P. Robarts Research Institute, London, Ontario, Canada
    Search for more papers by this author
  • D. Drost,

    1. Department of Medical Biophysics, University of Western Ontario, London, Ontario, Canada
    2. Lawson Research Institute, St. Joseph's Health Center, London, Ontario, Canada
    Search for more papers by this author
  • P. H. Watson,

    1. Lawson Research Institute, St. Joseph's Health Center, London, Ontario, Canada
    2. Department of Medicine, University of Western Ontario, London, Ontario, Canada
    Search for more papers by this author
  • L. J. Fraher,

    1. Lawson Research Institute, St. Joseph's Health Center, London, Ontario, Canada
    2. Department of Medicine, University of Western Ontario, London, Ontario, Canada
    Search for more papers by this author
  • A. B. Hodsman

    1. Lawson Research Institute, St. Joseph's Health Center, London, Ontario, Canada
    2. Department of Medicine, University of Western Ontario, London, Ontario, Canada
    Search for more papers by this author


Although dual-energy X-ray absorptiometry (DEXA) is an established technique for clinical assessment of areal bone mineral density (BMD), the spatial resolution, signal-to-noise ratio, scan time, and availability of clinical DEXA systems may be limiting factors for small-animal investigations using a large number of specimens. To avoid these limitations, we have implemented a clinical digital radiography system to perform rapid area DEXA analysis on in vitro rat bone specimens. A crossed step-wedge (comprised of epoxy-based materials that mimic the radiographic properties of tissue and bone) was used to calibrate the system. Digital radiographs of bone specimens (pelvis, spine, femur, and tibia from sham-ovariectomized [SHAM] and ovariectomized [OVX] rats) were obtained at 40 kilovolt peak (kVp) and 125 kVp, and the resulting areal BMD values were compared with those obtained with a clinical fan-beam DEXA system (Hologics QDR 4500). Our investigation indicates that the cross-wedge calibrated (CWC) DEXA technique provides high-precision measurements of bone mineral content (BMC; CV = 0.6%) and BMD (CV = 0.8%) within a short acquisition time (<30 s). Areal BMD measurements reported by the CWC-DEXA system are within 8.5% of those reported by a clinical fan-beam scanner, and BMC values are within 5% of the known value of test specimens. In an in vivo application, the CWC-DEXA system is capable of reporting significant differences between study groups (SHAM and OVX) that are not reported by a clinical fan-beam DEXA system, because of the reduced variance and improved object segmentation provided by the CWC-DEXA system.


DUAL-ENERGY X-RAY absorptiometry (DEXA) provides a fast, noninvasive technique to quantify areal bone mineral density (BMD) in humans. The widespread use of the ovariectomized (OVX) rat model (1–3) in investigations of new pharmaceutical therapies has resulted in great interest in applying DEXA techniques to small animal bones. (4–7) Current clinical systems using either pencil- or fan-beam scanners have been adapted to scan small regions of interest (ROI) with higher resolution, (8–11) but scan times remain long (several minutes) and the ultimate spatial resolution of clinical systems is limited by beam collimation. Reduced spatial resolution results in inaccuracy and imprecision when scanning small bones, because of increased uncertainty in edge location.(12)

We have developed a high-resolution DEXA system using an area detector (X-ray image intensifier) that is capable of quantifying areal BMD in small animal bones. An area detector offers advantages over pencil- and fan-beam scanning systems because the image data are acquired much more quickly, with more efficient use of X-ray tube power and with isotropic spatial resolution. A potential disadvantage with area systems is increased sensitivity to scatter, because scatter rejection by collimation is necessarily reduced. When scanning small objects, such as rodent bones or even human phalanges, the artifactual contribution of scatter is expected to be small. In this case, an area X-ray detector produces DEXA measurements with increased precision and accuracy, resulting in several potential advantages: (1) a reduction in the number of animals required to reach statistical significance; (2) the potential of using small, but metabolically active bones (such as the rat pelvis) that may not have been possible with previous DEXA techniques; and (3) the possibility of rapid in vivo DEXA measurements during longitudinal studies, using anesthetized animals.

In this article we describe an area DEXA system that uses standard clinical digital radiographic imaging components and a simple calibration step-wedge to produce BMD measurements from high-resolution (∼0.1 mm pixels) images. The system we describe is compared with a clinical fan-beam DEXA scanner during the investigation of small rodent bones.


X-ray imaging equipment

Dual-energy radiography is a technique that uses the energy dependence of X-ray attenuation coefficients to produce a pair of basis-material thickness images from two digital radiographs, obtained with different mean X-ray energies. The implementation of such a system has been described in detail previously.(13) The system we describe here uses a standard clinical digital radiographic system (Siemens Multistar; Siemens Medical Systems, Erlangen, Germany) capable of acquiring 880 × 880 digital images with a logarithmic amplifier and a 10-bit analogue-to-digital converter. The effective field-of-view (FOV) of each image was about 8.8 cm (using geometric magnification of 1.5), resulting in image pixel spacing of approximately 0.1 mm. Within each image a crossed step-wedge (comprised of bone- and soft-tissue-mimicking material) provided calibration measurements for the dual-energy decomposition; for this reason we will refer to the area DEXA technique as cross-wedge calibrated (CWC)-DEXA. The “bone” wedge consists of bone-mimicking material (SB3; Gammex-RMI, Middleton, WI, U.S.A.) machined into four steps, ranging from 1.5 to 10 mm thick. This calibration material was chosen because it mimics the radiographic properties of cortical bone over the range of diagnostic X-rays.(14) The soft-tissue wedge is fabricated from polymethyl methacrylate (PMMA) with four steps ranging from 2 to 9 mm thick. Note that the specific thickness of the calibrator steps is not important, as long as each is known exactly, and that the range of thicknesses covers the entire measurement range without extrapolation. The four-step wedge used in this study provides 25 unique combinations of bone and soft-tissue thickness. Each calibrator step is approximately 7 mm wide and the entire step-wedge covers a 35-mm square. Bone specimens were arranged around the step-wedge, as shown in Fig. 1A.

Figure FIG. 1.

(A) Bone specimens are arranged near the crossed step-wedge to obtain (B) low-energy and (C) high-energy digital radiographs. (D) Shows the resulting cortical-bone equivalent image.

Optimal dual-energy imaging requires that two images be acquired at widely separated energies, using X-ray spectra that are as narrow as possible. In this case, we chose the following parameters for the low- and high-energy images, respectively: 40 kilovolt peak (kVp; 1.5 mm Al inherent filtration, no added filtration, 300 mA exposure for 180 ms) and 125 kVp (1.5 mm Al inherent filtration, 2 mm Cu filtration added, 17 mA exposure for 170 ms). Total image acquisition time is about 30 s, including the time required to introduce the copper filter between low- and high-energy images.

For each CWC-DEXA acquisition, low- and high-energy images were acquired and corrected for fixed-pattern sensitivity variations as previously described.(13) Five pairs of images were acquired and averaged for each CWC-DEXA measurement, further reducing image noise. Representative corrected images are illustrated in Figs. 1B (low-energy) and 1C (high-energy). These images were processed as described in the following paragraphs to obtain localized BMD measurements within a user-selected ROI.

Image postprocessing and analysis

The high- and low-energy digital images were processed using standard dual-energy decomposition techniques. We have described such a system for in vitro specimens in a previous paper.(13) The crossed step-wedge of bone- and tissue-mimicking materials provides the required calibration, allowing the generation of bone and soft-tissue basis images. In the cortical-bone equivalent image (Fig. 1D), the gray-scale image intensity is directly proportional to the thickness of cortical bone within the ROI. Using the known percentage of hydroxyapatite within normal cortical bone (58%),(15) we can convert cortical-bone mass into bone mineral content (BMC). The area of bone within each ROI is then used to determine object-projected area, and hence BMD. In our CWC-DEXA system, the threshold for the ROI is user-selectable to ensure that background signal is excluded, without excluding thin or weakly mineralized bone. Note that in the standard fan-beam system, the background threshold is determined by a proprietary algorithm that may not be well suited for small-animal bones and is not adjustable by the user.

All CWC-DEXA measures of BMC and BMD for selected bones were recorded for comparison with values obtained from the clinical fan-beam DEXA equipment.

High-resolution fan-beam DEXA imaging

For comparison purposes, all the bone density measurements of test models or bone specimens also were repeated with a clinical DEXA system (Hologic QDR 4500 v1.0; Hologic Inc., Bedford, MA, U.S.A.) in high-resolution small-animal mode. With this fan-beam system, it was possible to scan a 4.6 cm × 8 cm FOV in 7 minutes.

Spatial resolution tests

Tests were performed with a high-contrast lead bar pattern to estimate the square-wave spatial frequency response of both fan-beam and CWC-DEXA systems.(16) The bar pattern included spatial frequencies ranging from 2 to 10 line-pairs/mm. When testing the fan-beam scanner, the bar pattern was positioned in both directions (i.e., parallel and perpendicular to the scan direction). The CWC-DEXA system exhibits isotropic spatial resolution, and it was tested with the bar pattern positioned at a slight angle with respect to the detector matrix. In both cases, images were viewed with optimal contrast and the maximum spatial frequency visible was recorded.

Animal studies

All the animal experiments were performed under the guidelines established by the Canadian Council on Animal Care. Six-month-old virgin female Sprague-Dawley rats were obtained from Charles River Laboratories and surgery was performed to produce either OVX or sham-OVX (SHAM) animals. After surgery, animals were separated into six groups, consisting of SHAM or OVX animals to be killed (euthanized under CO2 gas) at one of three time points: 4, 16, or 24 weeks postsurgery. During the postsurgery period, OVX animals developed osteopenia because of ovarian hormone deficiency. All experimental groups comprised 8 animals. From each animal the following skeletal components were removed, stripped of most soft tissue, and placed in 70% ethanol: right femur, right tibia, right pelvis, and lumbar spine (L1-L5). BMD was then determined for each of the 192 bones, using both the Hologic QDR 4500 fan-beam DEXA scanner and our CWC-DEXA system. Data were analyzed by analysis of variance (ANOVA) followed by Student's t-test to determine significant differences between SHAM and OVX animals at all time points.

Precision and accuracy tests

The reproducibility of the CWC-DEXA system was tested by repeated measurements of area, BMC, and BMD for representative SHAM specimens. Samples of pelvis, spine, femur, and tibia were scanned 10 times (with repositioning between each scan). The images were then processed with identical threshold values, and measurement precision was characterized by the CV of 10 measurements, where CV is defined as the SD divided by the mean value. Precision of BMC and BMD measurements for the fan-beam DEXA system was estimated from 40 measurements of a small (0.94 g) specimen of bone-mimicking material (SB3) that was placed in each scan as a routine component of quality assurance.

Accuracy was assessed for both fan-beam and CWC-DEXA systems by measuring five specimens of bone-mimicking material (SB3) of known BMC, ranging from 0.22 to 0.87 g. Both BMC and BMD were determined from the average of five repeated measurements of each specimen. True BMC was determined from the total mass of each specimen and a knowledge of the SB3 material composition.(14) Because the specimens were imaged with both DEXA systems, it also was possible to compare BMD measurements obtained with fan-beam and CWC-DEXA techniques.


Spatial resolution

Figure 2 illustrates the high spatial resolution obtained with the CWC-DEXA system. It was not possible to save high-quality digital image data from the Hologic QDR 4500, so the spatial resolution of the fan-beam system was evaluated from the display screen. For the CWC-DEXA system, the limiting spatial resolution was determined to be approximately 3.5 line-pairs/mm in both x and y directions. In comparison, the spatial resolution of the fan-beam system generally was lower and significantly anisotropic. Spatial resolution was observed to be highest in the direction of the fan beam (1.6 line-pairs/mm) and lowest in the scan direction (0.7 line-pairs/mm). Note that resolution in the scan direction of the fan-beam scanner is determined largely by the collimator spacing.

Figure FIG. 2.

Digital radiographic image of the bar pattern, obtained with the CWC-DEXA system. Limiting spatial resolution is about 3.5 mm−1.

Threshold thickness selection

A critical aspect of DEXA analysis is the selection of a threshold value of bone thickness, which is used in subsequent calculations to segment bone from background. The CWC-DEXA system provided complete control over choice of background threshold thickness and we found that optimal threshold thickness varied, depending on the bone under analysis. Figure 3 illustrates the effect of varying background threshold in the case of an OVX rat pelvis. In this case, the measured BMD varies from 0.096 g cm−2 (threshold = 0.24 mm) to 0.107 g cm−2 (threshold = 0.5 mm), where the increase is caused by exclusion of thin bone near the periphery. We have found the optimal threshold for pelvic specimens to be about 0.24 mm, because this includes the entire bone.

Figure FIG. 3.

Images of a rat pelvis specimen (OVX), illustrating the (A) cortical-bone equivalent image and images with threshold values of (B) 0.24 mm and (C) 0.5 mm.

In Fig. 4 we demonstrate the effect of varying background threshold in the case of an OVX rat spine. In this case, the measured BMD varies from 0.177 g cm−2 (threshold = 0.24 mm) to 0.226 g cm−2 (threshold = 0.75 mm), where the increase in BMD is due to exclusion of the thinner processes of the spine. We have found that a background threshold of 0.75 is optimal for rat vertebrae, because this excludes processes in all specimens, thereby reducing variability.

Figure FIG. 4.

Images of a rat spine specimen (OVX), illustrating the (A) cortical-bone equivalent image and images with threshold values of (B) 0.24 mm, (C) 0.5 mm, and (D) 0.75 mm.

Note that in all instances it is important not only to select the optimal background threshold level but also to apply it consistently to all bones of the same type within a study. In addition, it must be noted that threshold selection is less critical for bones that are more nearly cylindrical (such as the femur and tibia).

Accuracy and precision

Table 1 summarizes the results of accuracy studies on the fan-beam and CWC-DEXA systems. The average (absolute) error in BMC for the fan-beam and CWC systems was 8.1% and 5.3%, respectively. In the case of areal BMD, there was no absolute standard available, but we can compare the BMD values obtained with fan-beam and CWC systems; the average difference between the two systems was 8.5% for the five samples studied, with the fan-beam DEXA reporting the smaller value.

Table Table 1.. Accuracy Data for Test Samples of Known BMD
original image

The results of our precision study for the CWC-DEXA system are presented in Table 2. For measurements of BMC, projected area, and BMD we have calculated the CV, averaged over all four bone types. The results indicate that the CV for BMD (0.8%) is determined largely by the uncertainty in projected area (0.9%) and less so by uncertainty in BMC (0.6%). The CV of the fan-beam DEXA system, as estimated from repeated studies of a calibration standard, was 0.52% for BMD and 1.01% for BMC.

Table Table 2.. Results of Precision Study Performed on Bone Specimens from SHAM Animals Studied with the CWC-DEXA System
original image

Performance in animal tests: rat pelvis and spine BMD

Figure 5 compares pelvic BMD measurements obtained with the fan-beam (Fig. 5A) and CWC-DEXA (Fig. 5B) systems. Both DEXA systems show the same trend of decreased BMD as a function of time postsurgery, but only the CWC-DEXA system shows a significant (p < 0.05) difference in BMD between SHAM and OVX groups at 4 weeks and 16 weeks postsurgery. Figure 6 illustrates a similar comparison of rat spine BMD. In this case, fan-beam DEXA (Fig. 6A) reports somewhat lower values of BMD (possibly because of a lower background threshold) and no significant differences between groups. CWC-DEXA (Fig. 6B) reports generally higher values of BMD (possibly because of selection of threshold to exclude processes) and shows a significant difference (p < 0.05) between SHAM and OVX BMD at 16 weeks and 24 weeks postsurgery.

Figure FIG. 5.

Plot of BMD versus weeks postovariectomization for rat pelvis specimens, analyzed with either (A) fan-beam DEXA or (B) CWC-DEXA. Error bars indicate SE; asterisks indicate significant differences between SHAM and OVX (p < 0.05).

Figure FIG. 6.

Plot of BMD versus weeks postovariectomization for rat spine specimens, analyzed with either (A) fan-beam DEXA or (B) CWC-DEXA. Error bars indicate SE; asterisks indicate significant differences between SHAM and OVX (p < 0.05).


The CWC-DEXA system that we describe provides several advantages in the analysis of small-animal bones. The improved spatial resolution of the area detector system reduces the blur at bone boundaries, thereby improving the precision of boundary determination. This advantage, combined with the ability to set the background threshold level as desired, may make it possible to study smaller, thinner bones (such as the pelvis) that may be more metabolically active. This potential advantage in discriminatory power is illustrated in the pelvic data shown in Fig. 5. In this case, the fan-beam DEXA values for the OVX group are consistently higher than the CWC-DEXA values in the same bones. This overestimation of BMD by fan-beam DEXA in the OVX pelvis—caused by limitations in segmentation that cause artifactual exclusion of the thin regions near the periphery—removes the significant difference between SHAM and OVX groups that is reported by CWC-DEXA.

Our investigation of system accuracy indicates that BMC may be measured more accurately with the CWC-DEXA system, possibly because of the improved boundary determination. The CWC-DEXA system reports areal BMD values that are within 8.5% of those reported by a clinical fan-beam system in small-animal mode. In the case of BMD measurements, it is difficult to determine which system is correct, because there is no independent standard value for the projected area. Our results indicate that clinical fan-beam DEXA systems may underestimate BMD in very small bones, in agreement with previous investigators.(7) The precision of the CWC-DEXA system is better than 1%, with an average CV in BMD of only 0.8% for the rat bones studied (tibia, femur, spine, and pelvis). This precision is comparable with the short-term reproducibility reported by other investigators using clinical DEXA scanners in a small-animal model.(8, 9, 17) These previous estimates of precision in intact bones range from 0.5% to 1.5%. Note that the best precision reported for fan-beam DEXA systems in previous studies (0.5%) is in agreement with the value reported in the current study for repeated measurements of a bone calibration sample, indicating the best-case performance.

One other important advantage of the CWC-DEXA system is the relatively short acquisition time provided by the area X-ray detector. Total imaging time in the experiments reported here was less than 30 s, with most of that time used to switch filters and change X-ray parameters between high- and low-energy images. The CWC-DEXA images required less than 0.5 s of actual X-ray exposure, indicating that a dedicated small-animal imaging system could achieve acquisitions easily in less than 1 s. Such a short imaging time may make it possible to extend the in vitro system described here to in vivo studies of sedated animals. In this case, a CWC-DEXA system may allow longitudinal animal studies that currently are prohibited by the difficulty of long scans (several minutes) in fan-beam systems.

Although the system we describe makes use of clinical digital X-ray equipment that may not be available to all investigators, the design principles that we describe are applied easily to any small digital radiographic system. This CWC-DEXA approach also has been extended recently to provide peripheral BMD measurements of some human bones in vivo, such as the phalanges of the hand.(18)


We thank Maggie Kisiel for technical assistance in preparing the bone specimens and Misbah Gulam for assistance in analyzing the CWC-DEXA data. Siemens Medical Systems provided the digital X-ray equipment used in this study. This work received partial support from the Medical Research Council of Canada grant MT-13356. D.W.H. is the recipient of a Research Scholarship from the Heart and Stroke Foundation of Canada.