Francesca Particelli, Laboratorio di Tecnologia Medica, Istituto Ortopedico Rizzoli, Via di Barbiano 1/10, 40136 Bologna, Italy. Tel.: +39-051-6366858; fax: +39-051-6366863; e-mail: firstname.lastname@example.org
Cortical bone microstructure is an important parameter in the evaluation of bone strength. The aim of this study was to validate the characterization of human cortical bone microarchitecture using microcomputed tomography. In order to do this, microcomputed tomography structural measurements were compared with those obtained through histological examination (the gold standard). Moreover, to calculate structural parameters, microcomputed tomography images have to be binarized with the separation between bone and nonbone structures throughout a global thresholding. As the effect of the surrounding medium on the threshold value is not clear, an easy procedure to find the global uniform threshold for a given acquisition condition is applied. This work also compared the structural parameters of microcomputed tomography cortical sample scan in air or embedded in polymethylmethacrylate; histology was used as a reference. For each acquisition condition, a fixed threshold value was found and was applied on the corresponding microcomputed tomography image for the parameters assessment.
Twenty cortical bone samples were collected from human femur and tibia diaphyses. All samples were microcomputed tomography scanned in air, embedded in polymethylmethacrylate, rescanned by microcomputed tomography, examined by histology and finally compared.
A good correspondence between the microcomputed tomography images and the histological sections was found. Paired comparisons in cortical porosity, Haversian canal diameter and Haversian canal separation between histological sections and microcomputed tomography cross sections, first in air and then embedded in PolyMethylMethAcrylate, were made: no significant differences were found. None of the comparisons showed significant differences for cortical porosity, Haversian canal diameter and Haversian separation over a three-dimensional volume of interest, between microcomputed tomography scans in air and with samples embedded in PolyMethylMethAcrylate.
The very good correlation between bone structural measures obtained from microcomputed tomography datasets and from two-dimensional histological sections confirms that microcomputed tomography may be an efficient tool for the characterization of cortical bone microstructure. Moreover, when the corresponding threshold value for each condition is used, structural parameters determined by microcomputed tomography are not affected by the surrounding medium (PolyMethylMethAcrylate).
In the past few years, there has been an increasing attention to the investigation of cortical bone microstructure, up until now limited by two-dimensional histological methods. It is well known that cortical bone porosity is an important factor affecting cortical bone strength (Currey, 1988; Schaffler & Burr, 1988; Martin & Ishida, 1989; Wachter et al., 2002). For this purpose, microcomputed tomography (micro-CT) imaging represents an excellent alternative compared to histological examination of bone specimens, the “gold standard” until now (Weibel, 1989). The micro-CT scanner has many advantages: specimens can be evaluated three-dimensionally, nondestructively and more rapidly; they do not require a special preparation (histological examination requires embedding in polymethylmethacrylate (PMMA)) and the morphometric parameters can be calculated over the whole sample's volume (Feldkamp et al., 1989).
In order to calculate structural parameters, the key point in the analysis of micro-CT datasets is the binarization procedure, which can differentiate bone from nonbone structures. The global threshold value is normally obtained by identifying the middle point between the peaks (bone tissue and ‘other tissue’) in the grey scale distribution function (Basillais et al., 2007). Usually the threshold value obtained with a binarization method based on a bimodal is very difficult to calculate, especially when samples with low porosity are considered, as in this study. In these cases, the peak of pores resulted often undetectable and unnoticeable from background noise. The importance of using a global threshold is due to the fact that a variation of 0.5% in thresholding can result in a 5% difference in bone volume fraction, as shown in Hara et al. (2002). The threshold value could be obtained using an external method as a reference (histology) (Perilli et al., 2007).
It is therefore important to validate the micro-CT method against histology, the gold standard, also for the cortical tissue. Many studies focused their attention on this area of research, especially for trabecular bone tissue (Muller et al., 1998; Perilli et al., 2007). Conversely, only few studies have tried to validate micro-CT in human cortical bone. Wachter et al. (2001a) studied the comparison between intracortical porosity measured by micro-CT and by histological sections. Although their correlation is good, the resolution of the scanner (30 μm) is too close to the dimensions of the smaller Haversian canals (30–200 μm); in addition, they choose the threshold value in an arbitrary way (assessed density) (Wachter et al., 2001a; Cooper et al., 2004). The importance of validating the micro-CT method against conventional histological analyses is observed also by Britz et al. (2010). They investigated rat cortical bone porosity using both methodologies and found no statistically significant differences with the exception of the canal diameter mean values. However, they did not provide any explanation about their binarization technique. Cooper et al. (2003) showed also a great interest in three dimensional methods to analyze cortical bone microstructure with micro-CT, but they did not validate the method by comparison with histological examinations. Moreover, they did not use a standardized thresholding method. The same author published a different work where a comparison between porosity-related parameters from 2D micro-CT images and matched microradiographic sections was presented (Cooper et al., 2004). The resolution of the microradiography was 4.95 μm and significant differences were found between micro-CT and microradiographic measures, even if the biases between the techniques were relatively small. The micro-CT slices and the microradiographic sections were segmented using a local threshold algorithm. Bousson et al. (2004) studied cortical bone in human femoral neck using micro-CT based on synchrotron radiation, with 10.13 μm resolution. They used the same global threshold for all the samples, taking advantage of the fact that the histogram was clearly bimodal. The most important limitation of this study is that they did not compare the 3D porosity values with those obtained from histological sections, even if they underlined the importance of doing so. The microarchitecture of the human femoral cortical bone by 3D analysis on micro-CT images was also investigated by Basillais et al. (2007). They validated the micro-CT technique comparing structural measurements with ultrasonic techniques and scanning electron microscopy (SEM). The correlation found was strong, but their method lacked a standardized binarization procedure for the micro-CT cross sections. Moreover, the resolution of the SEM images was 5 μm and the comparison was not obtained between the same slice, but comparing a cross section with the whole sample taken from a different region inside the femur. In general, there are no studies that report a standardized procedure in order to find the global threshold value for micro-CT images of cortical tissue.
Moreover, as far as the micro-CT is concerned, the effect of surrounding medium (air, saline solution, PMMA) on the threshold value is not clear. Since the conventional histological examination requires an embedding treatment in PMMA, a study of the effect of this surrounding medium on structural parameters has to be performed for the cortical bone tissue as well. For this reason, the porosity derived from bone samples acquisitions in air and embedded in PMMA was compared.
Therefore, the aim of this study was to compare the porosity of cortical bone specimens obtained by micro-CT scanned in air and then scanned embedded in PMMA, using histological sections as reference. The best threshold value for each different media (air, PMMA) was found by direct comparison with the corresponding histological section. Then, Haversian canal diameter and Haversian canal separation were calculated using the two global thresholds found and compared with the same parameters obtained histologically.
Finally, micro-CT structural parameters (Cortical Porosity, Haversian canal diameter, Haversian canal separation) were compared in the same volume of interest (VOI) for cortical specimens surrounded by different media (air, PMMA).
Material and methods
Bone sample preparation
Twenty cortical bone samples were collected from femurs and tibias of four Caucasian donors (age of the donors 74, 74, 73 and 62). They were obtained from deceased persons without skeletal disorders thanks to a donor program (International Institute for the Advance of Medicine, IIAM, Jessup, PA, USA). Donors had agreed to be part of the program by signing a written informed consent. The biopsies were obtained from the diaphyses of the bones. In order to minimize the risk of transmission of infectious diseases and for proper handling and storage over time, the samples were fixed in a 70% ethanol solution for at least 4 weeks.
From the diaphyses, slices of 20 mm were cut perpendicularly to the longitudinal axis of the bone. The procedure used for the extraction of the samples has been previously described in Tassani et al. (2011). Shortly, a holed diamond-coated milling cutter, with the bone slice immersed in water was used. The final specimen obtained had a cylindrical shape with a nominal diameter of 3 mm and 20 mm height.
All the samples, obtained with the above-mentioned procedure, were micro-CT scanned in air, embedded in PMMA, rescanned by micro-CT, processed for histology and finally compared.
The device used for the samples scanning was Skyscan 1072 X-ray micro-CT (Skyscan, Kontich, Belgium). It consisted of a microfocus X-ray source, a rotating specimen holder and a detector system, with a 1024 × 1024 pixel CCD camera. The scanning geometry was of the cone beam type (Sasov & Van Dyck, 1998). The cortical samples were acquired using a previously published protocol (Tassani et al., 2011): 80 kV, 125 μA, 1 mm aluminium filter, exposure time 5.9 s, image averaged on 2 frames, rotation 180°, rotation step 0.90° and field of view 8 × 8 mm2 with a pixel size of 8 μm. In order to acquire the whole sample with this resolution type, the oversize acquisition procedure was applied as implemented in the acquisition software (TomoNT, Skyscan, Kontich, Belgium). Moreover, all the projection images were automatically smoothed by a median smoothing prior to reconstruction. After the acquisition, the cross-section reconstruction was made using the NRecon v. 1.10.1 software (Skyscan), based on the cone beam reconstruction algorithm described by Feldkamp et al. (1984). In NRecon reconstruction software the Fourier transform-based ring artefact correction is used. For each examination, a stack of 1750 cross-sectional images stored in 8-bit file format (256 grey levels) was produced, with a separation of one pixel (8 μm). The same micro-CT protocol was used both for the acquisitions of samples in air and for the PMMA embedded samples.
In the acquisition in air, samples were exposed for at least 2 h at room temperature before the scan, to avoid the creation of images artefacts due to the partial ethanol evaporation. They were then placed vertically into a polyethylene cylinder and positioned inside the micro-CT.
Once embedded in PMMA (procedure described below), bone samples were acquired again positioning them inside the micro-CT without the polyethylene cylinder.
PMMA embedding of the samples
Each bone specimen was placed in a 13 mm diameter glass: particular attention was used in placing the bone samples into the cylinder, in order to maintain the flat bottom surface of the bone sample attached to the bottom of the glass.
After accurate washing in water, samples were dehydrated and degreased under vacuum with the following series of alcohol washes: ethanol 80% overnight, three steps in ethanol 95% of 1 h each, five steps in methanol of 1 h each at 37°C. Infiltration occurred with four passages in PMMA (Carlo Erba Reagenti SpA, Rodano, Milano, Italy), first destabilized by chromatography with aluminium oxide (Carlo Erba Reagenti SpA) and finally embedded in a mixture of PMMA and di-n-butylphthalate (Carlo Erba Reagenti SpA) 3.5% w/v at 25°C. The reaction mixture catalyst was benzoyl peroxide (Carlo Erba Reagenti SpA) 25% v/v. The glass was then manually broken and the bone embedded in the remaining PMMA cylinder (Fig. 1).
After the micro-CT acquisition, serial 30 ± 5 μm slices were obtained from the embedded blocks with a 300 μm thick wafering blade (Leica 1600, Leica Microsystems, Wetzlar, Germany). One histological slice was produced for each sample from its central part. Each slice was stained with light green and visualized by an optical microscope (Leica DMR-HC, Leica Microsystems). A digital image of each histological slice was taken with Leica DC 300 camera (Leica Microsystems) mounted on the microscope. The final magnification was 50×, with a pixel size of 0.7 μm.
From the histological digital images, a rectangular region of interest (ROI) was extracted (1.5 × 1 mm2) with the ‘Leica Qwin’ software (Leica Microsystems). Over each ROI, the pores area (sum of the pixels marked as pore) was determined manually using Adobe Photoshop 7.0. The pores area was calculated tracing the Haversian canal perimeter and counting the pixels. The final measure was determined as the mean value of three measurements. The tissue area is simply the total pixels of the ROI. The cortical bone porosity (Ct.Po) was calculated as follows (Parfitt et al., 1987):
Over each ROI, Haversian canal diameter (H.Ca.Dm) and Haversian canal separation (H.Ca.Sp) were also determined manually used Image-Pro Plus (version 220.127.116.11). The parameters were determined as follows:
1H.Ca.Dm: the diameter of the largest inscribed circle in the Haversian canal and the diameter of the smallest circumscribed circle were provided. The final value is the average of the two.
2H.Ca.Sp: the distance between the centres of the circles of the Haversian canal considered, after subtracting the radii of the circles.
During the analysis it was established that all the Haversian canal considered ‘in a state of fusion’ should be excluded from the reading since they had noncircular or partially deformed geometry (Beraudi et al., 2010).
Micro-CT: thresholding and structural parameters calculation
The procedure described here was applied both to the bone samples acquired in air and to the embedded samples.
From the stack of micro-CT cross sections, the one corresponding to the histological image was visually chosen for each sample. A rectangular-shaped ROI 1.5 × 1 mm2 (183 × 136 pixels) was selected and then placed in the position that looked like the one of the histological ROI. After the selection of the ROI, a binarization of the micro-CT images was necessary in order to distinguish bone from nonbone tissue and a threshold value had to be found. For both procedures (ROI extraction and binarization) the ‘CTAnalyzer’ software (Skyscan) was used.
The two global threshold values for the acquisitions of bone in air and embedded in PMMA were determined thanks to a procedure described by Perilli et al.: calculation of the porosity of the samples based on an external method (measured on the histological images); determination of a micro-CT optimal threshold for each sample that corresponds to histologically calculated porosity; and calculation of the global threshold for the segmentation of the micro-CT images, as the mean value of the optimal thresholds (Perilli et al., 2007). In this way two global thresholds were calculated: the first one for the micro-CT images of the samples scanned in air and the second one for the acquisitions of the embedded bone samples. Finally, only these global thresholds were applied to the segmentation of the corresponding micro-CT datasets.
After the global thresholds were calculated, structural parameters were determined using the CTAnalyzer software (Skyscan). For the comparison in cortical porosity, Haversian canal diameter and Haversian canal separation between histology and the corresponding cross section of bone sample in air and embedded in PMMA, only one histological slice for each sample was used (morphometric parameters were extracted according to the definition given by Parfitt for bidimensional sections) (Parfitt et al., 1987). Conversely, for the comparison between micro-CT datasets of bone in air and of bone embedded in PMMA, a parallelepiped VOI (Schneider et al., 2007) of 183 slices was chosen (1.5 × 1.5 × 1 mm3) for each sample and the following parameters were calculated:
1Cortical porosity, Ct.Po (%), using Eq. (1) over all the VOI (183 slices). It is the volume of porous canal within the sample tissue volume, and complementary of bone volume fraction (Ct.Po = 1 – BV/TV).
2Haversian canal diameter, H.Ca.Dm (μm): this is the measure of the cortical pore diameter over the VOI and the correspondence of the direct trabecular thickness (Tb.Th*), also called model-independent thickness (Hildebrand & Ruegsegger, 1997). This parameter, based on the estimation of volume-based local thickness, independently of an assumed structure type, was calculated by fitting maximal spheres to every point contained in the 3D structure. The mean thickness of the structure is given by the arithmetic mean value of the local thicknesses (i.e. of the diameters of the maximal spheres) taken over all points of the structure.
3Haversian canal separation, H.Ca.Sp (μm): this is the measure of the cortical pore separation over the VOI and the correspondence of the direct trabecular separation (Tb.Sp*). It is calculated with the same procedure used for the H.Ca.Dm but giving a volume-based estimation of the thickness of the marrow cavities.
Cortical porosity, Haversian canal diameter and Haversian canal separation calculated on histological sections and with micro-CT, for the acquisitions in air and in PMMA, were compared. Moreover, another comparison was made in the parameters Ct.Po., H.Ca.Dm and H.Ca.Sp estimated over the VOIs between micro-CT scans in air and embedded in PMMA. The following comparisons were determined for each sample: the actual differences in the parameters di, the percentage differences di% and then the mean actual difference and the mean percentage difference for each structural parameter (Perilli et al., 2007):
Finally, in order to determine the normality of the distributions of the variables, a Kolmogorov–Smirnov test was used. If the data were compatible with a normal distribution, a Student's t-test for paired samples was used in the comparisons. The differences were deemed to be statistically significant at p < 0.05.
A good correspondence between micro-CT images and histological sections was found as shown in Figure 2. Two different thresholds were found, one for samples acquired in air (grey level value 118) and one for PMMA-embedded samples (grey level value 107).
Table 1 shows the descriptive statistics of cortical porosity, Haversian canal diameter and Haversian canal separation estimated over the 2D sections obtained by histology, micro-CT images acquired in PMMA and micro-CT images acquired in air. The Kolmogorov–Smirnov tests revealed a normal distribution; therefore, parametric statistics was used. The mean actual difference , the mean percentage difference and the paired t-test found in the comparison between histology and micro-CT are shown in Table 2. There were no significant differences in cortical porosity, Haversian canal diameter and Haversian canal separation between histology and micro-CT.
Table 1. Descriptive statistics of the cortical porosity (%), Haversian canal diameter and Haversian canal separation estimated by histology and by micro-CT in air and with the sample embedded over 2D sections of the 20 cortical bone samples (ROI: 1.5 × 1 mm2).
Micro-CT in air
Table 2. Comparison between histology and micro-CT on cortical porosity, Haversian canal diameter and Haversian canal separation calculated over 2D sections. represents the actual mean difference in porosity between histology and micro-CT, is the mean percentage difference and p is the probability for the paired Student's t-test.
Micro-CT embedded- histology
Micro-CT in air-histology
Cortical porosity estimated from micro-CT images embedded in PMMA and in air has been plotted in Figure 3 as a function of cortical porosity assessed from 2D histological sections. Both regressions had a high coefficient of determination (R2= 0.97 and R2= 0.96, respectively) (Figs 3(a) and (b)).
Table 3 shows the micro-CT 3D analysis over the VOI of 1.5 × 1.5 × 1 mm3, with the basic descriptive statistics of the structural parameters for the cortical bone samples scanned in PMMA and in air. The mean actual difference , the mean percentage difference and the paired t-test found in the comparison between micro-CT images acquired in PMMA and acquired in air are shown in Table 4. There were no significant differences in cortical porosity, H.Ca.Dm. and H.Ca.Sp. over the VOIs between micro-CT scans of bone in air and embedded in PMMA.
Table 3. Basic descriptive statistics of the structural parameters obtained by micro-CT over a VOI of 1.5 mm × 1.5 mm × 1 mm, first scanned in air and then scanned after the embedding in PMMA.
Micro-CT in air
Table 4. Comparison in structural parameters determined by micro-CT over a VOI of 1.5 mm × 1.5 mm × 1 mm between samples scanned in air and then scanned after embedding in PMMA. represents the actual mean difference in the parameters obtained by the two scan modalities, is the mean percentage difference and p is the probability for the paired Student's t-test.
Micro-CT embedded- micro-CT air
Figure 4 shows the cortical porosity estimated from micro-CT images in air as a function of cortical porosity assessed from micro-CT images embedded in PMMA over the VOIs. The linear regression fitting these data had a high coefficient of determination (R2= 0.96).
This study confirms that micro-CT may be an efficient tool for characterizing cortical bone microstructure. A very good correlation between bone structural measures obtained from micro-CT datasets and from 2D histological sections for human cortical bone specimens was found. Comparisons were made between samples scanned by micro-CT in air and after embedding in PMMA, using the corresponding histological sections as a reference. Following the recommendations that Perilli et al. (2007) used for the trabecular bone, measurement of porosities based on an accurate external method (histology) was used and a global threshold value was found for each micro-CT acquisition condition. The difference between the two global threshold values was evident (11 grey levels), but this is due to the different surrounding media (air, PMMA). The special attention paid during the process, from the embedding of the samples to the histological examination, and the spatial resolution (8 μm pixel size) ensured good visual correspondences between micro-CT slices and histological sections. Cortical porosity, which is considered to be one of the main parameters of the mechanical strength of cortical bone, showed no statistically significant differences from histology, both for acquisitions in air and PMMA embedded. The same was true for Haversian canal diameter and Haversian canal separation. Moreover, the structural parameters over the VOIs (1.5 × 1.5 × 1 mm3) did not show significant differences either for Ct.Po, H.Ca.Dm. or H.Ca.Sp. between samples scanned in air and embedded in PMMA.
In the literature, only few works have described human cortical bone and its microstructure (Wachter et al., 2001a; Cooper et al., 2003; Wang & Ni, 2003; Bousson et al., 2004; Basillais et al., 2007). However, it has been demonstrated that cortical bone significantly contributes to the mechanical strength of bone (Currey, 1988; Schaffler & Burr, 1988; Martin & Ishida, 1989; Wachter et al., 2002). The importance of validating micro-CT method against the ‘gold standard’ (histology) has been highlighted by numerous authors, whose works focused on trabecular bone (Muller et al., 1998; Perilli et al., 2007). However, up until now little has been written for the validation of the method for the human cortical tissue. Wacther et al. (2001a), for example, found a good correlation between porosity obtained with micro-CT images and with histological sections (r= 0.95), that is smaller than our correlation (R2= 0.97). Unfortunately, they operated at 30 μm pixel size and their positive result is probably facilitated by the pore size of their samples. In fact, as their samples derived from an elderly and pathological population, diameters of the Haversian canals tend to increase and lower spatial resolutions could lead to better results, compared to samples with smaller Haversian canal diameters. The dependence of the parameters on the spatial resolution was studied by Cooper et al. (2007).
In general, a global thresholding was used in order to segment the greyscale images into binary black and white for morphometric analyses. In fact, despite potential limitations due to the artefacts such as beam hardening, this thresholding method is straightforward and the most accessible and reproducible method available also for cortical bone. Many authors did not describe the thresholding procedure that they used (Cooper et al., 2003; Britz et al., 2010). Others used as global threshold value the one corresponding to the middle point between the peak relative to cortical bone and the peak relative to cavities (Bousson et al., 2004; Basillais et al., 2007). This, according to our experience, could lead to some overestimation or underestimation errors. A standardized thresholding procedure like the one described in our work (using histological section as reference) is fundamental for the accuracy of the structural parameters and indeed many studies stress on the importance of the choice of an appropriate segmentation technique to separate bone from nonbone structures (Hara et al., 2002; Waarsing et al., 2004; Parkinson et al., 2008). As shown in the study of Hara et al. on trabecular bone, threshold selection is important for the accurate determination of structural and mechanical properties. In addition, small errors in the selection of the ‘right’ threshold value could affect the correlations between structural parameters and mechanical properties and lead to higher errors in the prediction of fracture risk. In fact it is known that also cortical tissue contributes to the mechanical strength of bone and that the assessment of cortical bone might be relevant for the prediction of fracture risk (Wachter et al., 2001b). If power regression models are used to link structural and mechanical properties, little variations in the threshold value may vastly affect the correlation between structural and mechanical parameters of cortical bone. The validation of the micro-CT measures against the conventional histological analyses was also described by Britz et al. (2010). The matches of their images were often not very accurate; this is probably due to the difference in thickness between the micro-CT slice and the histological section (3 and 100 μm, respectively) and to the fact that the low resolution was insufficient to segment small canals like those present in the small animals’ cortical bone. Basillais et al. (2007) found a significant correlation between the porosity measured by SEM and by micro-CT (r= 0.91); a correlation with ultrasonic parameters was also observed for all microstructural parameters. Despite this, the application of their binarization procedure (based on a bimodal) could be limited only for samples with high porosity, such as those showed in their work. Moreover, the resolution of SEM is 5 μm, whereas the resolution of our microscopic images is 0.7 μm. Nevertheless, Cooper et al. (2004) reported a comparison between micro-CT images and matched microradiographic sections. Significant differences are found between micro-CT and microradiographic measures, probably because the alignment of the bone sample in the scanner and in the diamond saw was done manually and therefore a perfect match is impossible. Some authors investigated cortical bone with micro-CT based on synchrotron radiation (SR micro-CT) (Bousson et al., 2004; Matsumoto et al., 2006). SR micro-CT has many advantages over conventional micro-CT: the availability of monochromatic and high-density X-rays, higher resolutions, no beam-hardening artefacts and a high signal-to-noise ratio. Despite all this, the system has the disadvantage of having a limited access whereas conventional micro-CT offers a more accessible alternative for certain level of imaging (Britz et al., 2010). Matsumoto et al. showed the utility of SR micro-CT to quantify the microstructure of tibial cortical bone in rats. They used a high resolution (5.83 μm pixel size) and a threshold value determined through the comparisons between SR micro-CT images at various threshold values and the light microscopy image of the sliced sample. However, they did not validate the method against histological examination.
Another method to examine cortical tissue is the one described by. Wang & Ni (2003). They used the histological examination in comparison with low field pulsed nuclear magnetic resonance (NMR). In this case, calculations are based on the relaxation time measure of fluid in the pores of cortical bone. The comparison shows a high determination coefficient for bone porosity (R2= 0.72). However, some interference effect from surrounding tissues could be possible with this approach: in fact the NMR signals by soft tissues could interfere with the signals of the pores and overestimate the true cortical porosity of bone.
As the effect of the surrounding medium on the threshold value is not clear, it was necessary to perform a study also for cortical bone. The easiest way of conducting a scan is in air (Goulet et al., 1994) because this provides the highest contrast in the images. Nevertheless, bone specimens are usually put in a cylinder filled with a liquid solution (Muller et al., 1998) or embedded in PMMA (Thomsen et al., 2005) in order to allow sample storage and manipulation. Moreover, PMMA embedding is a necessary treatment for comparison of micro-CT versus histology as it preserves the structure during the cutting procedure. At the energies used in micro-CT (tensions of 80 kV), PMMA and water have a similar attenuation coefficient (at 60 keV: μPMMA= 0.23 cm−1, μwater= 0.21 cm−1) (Hubbell, 1996). In this way, storage in PMMA and storage in water during a micro-CT acquisition are two similar conditions, compared to acquisitions in air. Conversely, scanning samples in air or in PMMA are two different conditions. In this work, different global thresholds for each acquisition condition were found and the surrounding medium did not affect the structural parameters obtained by micro-CT compared to those histologically determined (Perilli et al., 2007).
This study has some limitations. The chosen samples are a group of human femoral and tibial diaphysis from nonpathological patients. Further analyses should be made increasing the number of samples and including pathological bone samples in order to have a more representative group. However, the range of porosity is quite representative of all possible cortical bone (3–30 μm). Moreover, even if 8 μm pixel size is good, better spatial resolution is achievable, in particular with SR micro-CT, in order to succeed in visualizing smaller pores. However, it has been reported that Haversian/Volkmann's canals in the range of 30–200 μm are the biggest portions of the total porosity, whereas the other cavities porosity contributed to a smaller extent (Wang & Ni, 2003). Unfortunately, increasing micro-CT spatial resolution normally means decreasing the field of view; so a compromise has to be found. The visual match between micro-CT image and histological section could introduce small errors in the measures: a registration algorithm that aligns the images could resolve this problem. Another limitation concerns the fact that micro-CT scanner investigates only a very confined ROI and not an entire human femur or tibia. However, even if this would be important to see the homogeneous distribution of the cortical microstructure, it does not invalidate the thresholding procedure. Finally, the threshold value found is optimum for this type of samples and with these dimensions. If different sample sizes are used, another analysis has to be made. This is a common limitation for all histology validations, but it is clear that even for little differences in sample dimensions, when changing the acquisition parameters, the global threshold value found should be very similar.
In conclusion, cortical bone samples were scanned by micro-CT first in air and then embedded in PMMA. Two global threshold values were found using histological sections as a reference. Comparisons with histology sections showed no significant differences in cortical porosity, Haversian canal diameter and Haversian canal separation for both micro-CT acquisitions. No significant differences were observed also for Ct.Po, H.Ca.Dm. and H.Ca.Sp. obtained over the VOIs between samples scanned in air and after embedding in PMMA. Structural parameters determined by micro-CT are not affected by the surrounding medium (PMMA) for cortical bone, if the corresponding global threshold value for each condition is used.
We would like to thank Caroline Ohman for the technical support in the specimens’ extraction, Paolo Erani for the specimen preparation, Stefano Falcioni for the statistical analysis and Luigi Lena for the illustrations. This work was partially supported by the Region-University Research Program 2007–2009 and by the European Community (project number: FP7-223865; acronym: VPHOP).