Micro-computed Tomography Assessment of Human Femoral Trabecular Bone for Two Disease Groups (fragility Fracture and Coxarthrosis): Age and Gender Related Effects on the Microstructure

The aim of this study was to identify three-dimensional microstructural changes of trabecular bone with age and gender, using micro-computed tomography. Human trabecular bone from two disease groups, osteoporosis and osteoarthritis was analyzed. A prior analysis of the effects of some procedure variables on the micro-CT results was performed. Preliminary micro-CT scans were performed with three voxel resolutions and two acquisition conditions. On the reconstruction step, the image segmentation was performed with three different threshold values. Samples were collected from patients, with coxarthrosis (os-teoarthritis) or fragility fracture (osteoporosis). The specimens of the coxarthrosis group include twenty females and fifteen males, while the fragility fracture group was composed by twenty three females and seven males. The mean age of the population was 69 ± 11 (females) and 67 ± 10 years (males), in the cox-arthrosis group, while in the fragility fracture group was 81 ± 6 (females) and 78 ± 6 (males) years. The 30 μm voxel size provided lower percentage difference for the microarchitecture parameters. Acquisition conditions with 160 µA and 60 kV permit the evaluation of all the volume's sample, with low average values of the coefficients of variation of the microstruc-tural parameters. No statistically significant differences were found between the two diseases groups, neither between genders. However, with aging, there is a decrease of bone volume fraction, trabecular number and fractal dimension, and an increase of structural model index and trabecular separation, for both disease groups and genders. The parameters bone specific surface, trabecular thickness and degree of anisotropy have different behaviors with age, depending on the type of disease. While in coxarthrosis patients, trabecular thickness increases with age, in the fragility fracture group, there is a decrease of trabecular thickness with increasing age. Our findings indicate that disease, age and gender do not provide significant differences in trabecular microstruc-ture. With aging, some parameters exhibit different trends which are possibly related to different mechanisms for different diseases.


INTRODUCTION
As for other materials, the mechanical properties of bone depend on its structural characteristics.Trabecular bone is formed by an interconnected network of rods and plates and can be found at the epiphysis of long bones and in the vertebral body.Almost all fragility fractures occur at regions with trabecular bone, for which the trabecular microarchitecture was affected by a disease mechanism.In this sense, the bone structural characterization, in particular of trabecular bone, is fundamental to assess the risk of fracture and to help in the prevention of bone failure.
Osteoarthritis (OA) and osteoporosis (OP) are pathologies that affect the quality of life of patients.Osteoporosis occurs due to the discrepancy between bone formation and bone resorption, which may lead to bone loss, associated to an increase on the risk of bone failure.In this sense, the trabecular microarchitecture is an important determinant of osteoporosis.Osteoarthritis is a chronic inflammatory disease which mainly affects the cartilage of the joint compromising the bone properties, whose influence on the microstructure of bone is not very well known.
Micro-CT is widely used to evaluate the microstructure of trabecular bone and is regarded to be a gold standard technique.Depicted the number of papers on this subject, some aspects may be missing.Although some works refer to the comparison between diseases as osteoporosis and osteoarthritis [30,35], while others mention the age and gender effect on the structure [20,21] a survey of the literature did not allow to find any combination of the age, gender and disease.
Although, there are studies on the micro-computed tomography applied to trabecular bone, which deal with the impact of the procedure variables on the accuracy of the measurements, each case is a different case with its own particular characteristics depending on the apparatus properties, process variables and sample properties [2,43,44].Prior to the application of the micro-CT technique to a large number of samples, a study on the effects of some procedure variables is advantageous.In this work, an optimization of the micro-CT parameters with emphasis to the voxel size, voltage, intensity and threshold values was performed.
The aim of this study was to identify three-dimensional (3D) microarchitectural changes of trabecular bone simultaneously with age, gender, and two different pathologies, coxarthrosis (CA) and fragility fracture (F) using micro-computed tomography (micro-CT).We hypothesized that, in the diseased samples, there would be age-related changes in the trabecular bone microarchitecture, in similitude to age effects in healthy bone.

Materials
Femoral epiphyses were collected from patients submitted to hip replacement surgery in the Orthopedic Department of Hospital de Santa Maria, Lisbon, Portugal.Two pathologies were evaluated, namely coxarthrosis which is a particular form of osteoarthritis, and fragility fractures, which probably occurred due to osteoporosis.Trabecular bone cylinders were drilled to extract cylinder-shaped samples of 5 mm in diameter and approximately 15 -20 mm in length.This study was approved by local Ethics Committee and followed the International Guidelines stated by Declaration of Helsinki (Seoul, 2008).Patient's agreement to these experiments was obtained by written informed consent.
Bone cylinders were prepared following a procedure which included five steps: fixation, dehydration, clearing, impregnation and inclusion [45].During the fixation phase, the sample is placed in alcohol 70% for a minimum of 72 hours at 5˚C, after which they were dehydrated in ethanol 96% to 100%, over a period of 24 hours, at 5˚C.On the third step, samples were cleared in order to replace the alcohol by an intermediate solution of xylene for 24 hours at 5˚C.Then, the specimens were embedded in methylmethacrylate (MMA) for a minimum of 72 hours at −20˚C, after which they were included in the polymer at a constant temperature between 5˚C and 10˚C, to polymerize [46].
Initially, three trabecular bone samples from patients with CA were studied, which belong to two male patients with mean age of 64 ± 7 years, and one female with 61 years to establish imaging parameters to be used in this study.Then, a total of sixty-five samples were analyzed and two bone diseases were compared, using thirty-five coxarthrosis specimens (20 females with mean age of 69 years, 5 males with mean age of 67 years) and thirty fragility fracture specimens (23 females with mean age of 81 years, 7 males with mean age of 78 years).

Procedure Variables in the Micro-CT Evaluation
The goal of the first part of the work was to establish an optimized methodology to measure 3D trabecular bone structural characteristics, in order to apply it to a wider sample population.For this purpose a set of three sam-ples was used.Measurements were determined three times for each of the samples, under the same testing conditions.
Micro-CT analysis involves three main steps: acquisition, reconstruction and image analysis [3].In the acquisition stage, the variables to be chosen are the X-ray tube potential, voxel size, and location of the volume of interest [3].On the reconstruction phase, the binarization (or segmentation) is carried out, separating bone from no bone regions.Despite the number of methods found on the literature, no completely reliable method of bone segmentation has been established and an optimal thresholding protocol should be proposed for each case studied.In general, for bone samples, local threshold are preferred, especially for comparison studies with histomorphometric evaluation [16].On the third phase, the reconstructed image data can be interpreted with a 3D analysis software that enables the quantification of the bone microstructure parameters.Although several parameters may be used to characterize the 3D structure of trabecular bone, the most widely used are [3,47,48] The micro-CT study was performed on a SkyScan 1172 device (SkyScan, Kontich, Belgium) [49].The apparatus is controlled by a computer programmed with the SkyScan software package, which is based on Feldkamp algorithm [1]. Figure 1 shows an example of the shadow image (or projection image), the reconstructed slice, the region of interest (ROI), the binarized region, and the 3D rendering (i.e., manipulation in a virtual space) of a human trabecular bone sample.
The acquisition protocol involves several scan parameters, some of which were changed to verify their influence in the results.The X-ray projection images were collected under two conditions of X-ray voltage and Xray current denoted respectively by A) and B), where A) corresponds to 160 µA and 60 kV and B) to 100 µA and 100 kV.Images were acquired over an angular range of 180˚ with an angular step of 0.45˚, with scanning times around 30 min.The tested values of the image pixel size were 10, 15 and 30 μm.During the micro-CT scan, each sample was entirely contained in the field of view (FOV) and a stack of 1000, or more, cross-section images was obtained with a slice to slice increment of 10 μm.The projection images were stored in TIFF file format, as 16-bit shadow images, in a size of matrix around 640 × 512 pixels.
The size and position of the volumes of interest were chosen by the operator to obtain the maximum possible volume.Samples were only repositioned between each image acquisition and the analysis was conducted by the same operator.
Following scanning, the projection images were reconstructed, in about 500 slices along the ZZ axis, by using the cone-beam reconstruction software NRecon (SkyScan, Kontich, Belgium) [49].The segmentation method applied is based on the local minimum of the grayscale bimodal histogram function of the stack of slices images corresponding to the volume of interest.Three different threshold values on the histogram function were tested which correspond to 1) the grey value of the local minimum (g min ); 2) the minimum value minus 5 units of the grey level scale (g min − 5); and 3) the minimum value plus 5 units (g min + 5).Each sample has a unique g min and therefore the threshold values were different for distinct samples.An example is given in Figure 2.
On the third step, the reconstructed slice images were processed, quantified and interpreted by means of 3D image analysis software (CTAn and ANT software, Sky-Scan, Kontich, Belgium).This 3D rendering enabled the determination of the previously mentioned structural parameters for the analysis of trabecular structure.
The reproducibility of measurements of the morphological parameters was assessed by determining these parameters, making three scans for each sample, under the same acquisition conditions, and calculating the coefficients of variation, CV(%) given by Eq. 1.To compare the results of the two scanning conditions (p.A, p.B), the percent difference, ΔDiff (%), was also evaluated with the Eq. 2.

 
A comparison between the two conditions A) and B) for the three voxel sizes (10, 15 and 30 μm), is evaluated by the ΔDiff (%) values, which are illustrated on Figure 3.
In general, the percentage difference for the microarchitectural parameters is lower for the results obtained with 30 µm voxel size, with the exceptions of BV/TV and DA.Some works report that the scanning of large specimens may require the use of special resolutions correspondent to voxel sizes greater than 100 μm [2].energy (50 to 90 kV) [3].
We determined the threshold influence on a set of three samples, using three values of the global grayscale histogram, g min − 5, g min and g min + 5. Table 2 quantifies the threshold effect on the binarized BVI for one sample, where the microarchitectural parameters determined under different values of the grayscale histogram threshold, are shown.The average of the three results is very similar to the results obtained with the g min value.The CVs associated with the measurements are in the interval of 0.5% to 8%.The higher CV, around 8%, was also obtained for BV/TV, while the coefficient of variation for BS/BV and Tb.Th varied between 6% to 7%.For SMI, the CV value is 4.36% and the coefficients for BS/TV, Tb.N, Tb.Sp, DA and FD are equal or lower than 2%.Our values are in the range of a similar work performed by Beaupied et al. [44].We decided to use values of g min which provides results similar to the average of the results obtained with these three threshold values.However, for voxel sizes larger than 100 μm, the microarchitectural parameters are strongly dependent on the voxel size [2].As, when the voxel size decreases, the resolution increases, it is convenient to choose a voxel size lower than 100 μm, which allows scanning the entire sample's volume.The voxel size of 30 µm was the ideal value for our samples.

Micro-CT Imaging
Table 1 presents the microarchitectural parameters, for the three samples, determined under scan conditions A and B, with a voxel size of 30 µm.The average values of the coefficients of variation are almost the same either for settings A or B conditions.
Taking into account the preliminary results obtained in the study, the optimal parameters for scanning and reconstruction of trabecular bone were chosen as: acquisition conditions of voltage equal to 60 kV, intensity of 160 μA, 30 µm of voxel size and the threshold value adjusted at the minimum of the global grayscale histogram from each specimen evaluated.
As there is not much influence of the acquisition voltage and current on the results, A conditions were chosen to evaluate the rest of the samples, as it allows to evaluate all the volume's sample and reduces the artifacts.This is in accordance with authors that recommend the use of micro-CT works at the medium range of X-ray

Statistical Analysis
First, a Shapiro Wilk test was conducted to evaluate the normality of the distributions.The test indicated that all bone microarchitectural parameters had non-normal distributions.Therefore, the Mann-Whitney (Wilcoxon) test was performed to assess comparisons between female and male population, and also between the two bone diseases groups.
In addition, the univariate correlation given by the Spearman's correlation coefficient of age as a continuous variable for each bone microarchitectural parameter was tested.Finally, for each bone disease group, Spearman's correlation coefficients were obtained for correlation between each microarchitectural parameter.
Statistical analysis was performed using a statistical software SAS (version 9.2, Institute Inc., Cary, NC, USA) and differences were considered statistically significant between groups for two-side p-value lower than 0.05, and for p-value lower than 0.0001, it was considered that those differences were highly statistically significant.

RESULTS
Table 3 presents the global results obtained from the extended trabecular microarchitectural study, where a comparative microarchitectural evaluation between two bone diseases and both genders was made.The Mann-Whitney (Wilcoxon) test revealed no significant differences between disease and gender.For both diseases, specimens from male and female donors presented nearly identical BV/TV values, and for these reasons the 3D reconstruction of male and female samples did not present great microarchitectural differences.Figure 4 shows the 3D reconstruction for coxarthrosis samples and Figure 5 presents the fragility fracture samples.In order to evaluate the effect of age on microarchitectural properties, the Spearman correlation coefficients were determined for all the samples from both groups of coxarthrosis and fragility fracture (Table 4).Negative coefficients were obtained for both CA and F groups, meaning that the bone volume fraction, trabecular number and fractal dimension decrease with increasing age.
Trabecular separation and the Structure Model Index increase with age, also for both disease groups.The parameters BS/BV, Tb.Th and DA show different Spear-   man coefficient signs, meaning that there are differences between the coxarthrosis specimens and fragility fracture groups.While BS/BV and DA decrease with age on the CA group, an increase with age was found in the F group.
The trabecular thickness has a tendency to increase with age on the coxarthrosis group, while it decreases in fragility fracture group.Additionally, to evaluate disease-related effects on the relationship between each microarchitectural parameter, the Spearman coefficients were determined for each disease group, and the results are presented in Tables 5 and  6.For the majority of the results, the Spearman statistics revealed similar trends for both disease groups.In general, the microarchitectural parameters are well correlated with BV/TV, showing high values of the coefficients.The worst correlation values were obtained for DA.Trabecular network complexity (FD) and trabecular Structure Model Index are strongly correlated with BV/TV and, consequently, with bone relative density.

DISCUSSION
A better understanding of the trabecular bone microstructure might be relevant for the evaluation of the fracture risk of hip femoral bone.Several factors may affect the structure of bone, as gender, age and disease.The most important age-related change in trabecular structure is bone loss leading to an enhanced risk of fracture.In our work, micro-CT evaluation revealed no significant differences in microarchitectural parameters between diseases groups and gender.
No significant differences were found from the Spearman coefficient determination for age effect, but both diseases presented similar microstructural relationship with age for the same parameters.This is the case of  BV/TV, Tb.N, FD, SMI and Tb.Sp.In healthy human trabecular bone, BV/TV, Tb.Th and Tb.N, decrease, while Tb.Sp and SMI increased with age [21].With the exception of Tb.Th, our findings are in accordance with previous studies [20,21,50].However, the parameters BS/BV, Tb.Th and DA show different Spearman coefficient signs, for the two disease groups, meaning that aging has different effects on some microarchitectural parameters.This indicates different aging mechanisms for different diseases.With aging, trabecular thickness increases in the CA group, but in the F group, it decreases.The increase in trabecular thickness can be attributed to a compensatory mechanism that tries to maintain bone strength, even during bone loss [20].
An increase in the trabecular separation with age can be achieved by an increase in the intertrabecular distance or by the appearance of large areas with no trabeculae [20].This potentially happens in both CA and F groups.
Structure Model Index indicates the relative proportion of rods and plates in a 3D structure such as trabecular bone.An ideal plate and an ideal cylinder have SMI values of 0 and 3 respectively.Our SMI mean values revealed that both groups present a mixture of plate-and rod-like model microstructure.It is reported [21] that, human healthy trabecular bone structure changes from a plate-like to a more rod-like structure with age.The increase of SMI with age is demonstrated on Table 4, for our two disease groups.
The Degree of Anisotropy mean values are very close to 1 affirming a rather homogeneous trabecular structure in both groups.With aging, DA tends to decrease in the CA group, while it tends to increase in the F group.It is mentioned that in healthy bone, trabeculae with aging seem to align to the direction of principal loading, which means a decrease on DA.Probably the patients with CA have a tendency to follow this mechanism, while the F patients will maintain the bone anisotropy.
The fractal dimension, FD, which is an indicator of surface complexity of an object that quantifies how the object's surface fills the space, tends to decrease with age, which is explained by a decrease in the bone volume fraction.
The correlation between microarchitectural parameters, determined by the Spearman statistics, shows that the majority of the parameters are well correlated with BV/TV.Moreover, our results are consistent with previous studies for which Tb.Th and Tb.N presented positive significant coefficients with BV/TV, while Tb.Sp showed negative significant coefficients [35].These findings confirm the very important correlation between BV/TV and trabecular morphometric parameters (Tb.N, Tb.Th and Tb.Sp) that influence porosity and, consequently trabecular bone density.

CONCLUSIONS
We emphasize the need for a preliminary study on mi-cro-CT variables prior to extending it to a large number of samples.In fact, comparison with literature data is not straightforward, even if the same device is used.Changes in the parameters may bias the reconstructed images, giving rise to a high variability of the 3D microarchitectural parameters.
As on other works, bone volume fraction BV/TV mean values and its good correlation with the other microarchitectural parameters demonstrated the great importance of this parameter in trabecular structural prediction.
Gender, age and disease variations showed no signifycant effects on the microarchitectural parameters of trabecular bone.
As for healthy bone, there seems to be effects of aging in the microarchitecture of bone.Age-related changes in the microstructure of trabecular bone are not the same in different pathologies, such as, coxarthrosis and fragility fracture.With aging, bone specific surface, and the degree of anisotropy decrease in the CA and increase in the F group, while trabecular thickness increases in the CA group and decreases in the F group.This means that, depending on the disease, the age effects in the microbarchitecture of bone are different.

Figure 2 .
Figure 2. Effect of the grayscale histogram thresholding value on the binarized ROI: (a) g min − 5; (b) g min ; and (c) g min + 5, where g min is the grey level correspondent to the minimum (A scan conditions: 160 µA and 60 kV, 30 μm pixel size).

Figure 4 .
Figure 4. Three-dimensional reconstruction of coxarthrosis samples from a female with 59 years (a) and 75 years (b), and from a male with 60 years (c) and 74 years (d).

Figure 5 .
Figure 5. Three-dimensional reconstruction of fragility fracture samples from a female with 76 years (a) and 84 years (b), and from a male with 70 years (c) and 84 years (d).

Table 1 .
Effect of the scan acquisition conditions A and B (A: 160 µA/60 kV, B: 100 µA/100 kV) on the microarchitectural parameters for three different samples (Mean ± SD values and CV), for a voxel size of 30 μm.

Table 2 .
Microarchitectural parameters determined under the same acquisition conditions (15 μm voxel size, scan acquisition condition A and with different values of the grayscale histogram threshold, namely, g min − 5, g min and g min + 5, where g min is the grayscale minimum (Mean, standard deviation, SD, and coefficient of variation, CV).

Table 3 .
Descriptive statistics of the micro-CT measurements for the two bone disease groups, CA (coxarthrosis) and F (osteoporosis or fragility fracture).

Table 4 .
Spearman correlation coefficients between trabecular microarchitectural parameters and age for coxarthrosis (CA) and fragility fracture (F) groups.

Table 6 .
Spearman correlation coefficients between each trabecular microarchitectural parameter for fragility fracture (F) group.