Visualization of woven bone structure through analysis of biopsy specimens using synchrotron radiation and conventional X-ray microcomputed tomography

This study explores the application of synchrotron radiation and conventional microcomputed tomography (SR-µCT and C-µCT, respectively) in evaluating bone-biopsy specimens. Bone-biopsy specimens were obtained using a trephine bur during bone-graft removal for implant placement six months after performing a maxillary sinus bone-graft procedure. Image data of specimens were obtained using SR-µCT and C-µCT. SR-µCT was performed using the 6C biomedical imaging beamline at the Pohang Accelerator Laboratory with a monochromatic X-ray beam of 23 keV, and C-µCT was performed using a table-top CT scanner (Skyscan 1272). Reconstruction images obtained using the two methods were qualitatively compared with 2D images evaluated under 3D visualization. The SR-µCT images, especially of the new-bone–graft–woven-bone formation, were less noisy and sharper than the C-µCT images. To evaluate the new-bone–graft–woven-bone formation, only the SR-µCT images showed areas of new bone (NB) formation with bone substitute (BS; Bio-Oss) and woven bone (WB) contact, and correctly visualized true 3D structures of bone formation. Hence, µCT techniques are non-destructive and can provide detailed images of bone biopsy. In particular, SR-µCT can be used to obtain improved image quality with contrast of NB, BS and WB, demonstrating a level of detail comparable with bone formation. SR-µCT could be an unbiased 3D alternative for imaging WB formation and for high-throughput analysis.

1. Introduction

The purpose of bone regeneration in medicine is to restore and maintain the lost function of bones. In the dental field, alveolar bone lost due to tooth loss or periodontal disease is regenerated through bone grafting (Finkemeier, 2002). The clinical success of bone regeneration is achieved through tissue analysis and measurement, and the accurate evaluation of the bone healing is important (Kulak & Dempster, 2010). Microanalysis of biomaterials and biologically regenerated tissues is essential for the accurate assessment of bone healing.

The key aspects of microanalytic tools are image quality (high-quality histological sections), resolution (labeling fluoro­chromes) and accuracy. Bone-biopsy samples contain information on bone regeneration and remodeling along with their complex structure (Kulak & Dempster, 2010). The amount of biopsy tissue [including new bone (NB) with graft materials and soft tissue] after a surgical treatment of maxillary sinus floor elevation (MSFE) has been observed and estimated (Yildirim et al., 2001). Among the current methods to analyze biopsy tissue in general, the hematoxylin-eosin staining method is the most widely used histological technique (Froum et al., 2008; Jensen & Sennerby, 1998). Histological techniques are exceptionally useful in highlighting biological structures but they involve long pre-processing and sectioning of the sample, which may create deformation. On the other hand, the analysis of information for bone dynamics is only for the longitudinal direction view and uses a limited number of sections.

Avoiding the demerits of conventional histomorphometry, microcomputed tomography is a non-destructive imaging technique. It is used in the bone analyses of animal models, such as bone growth, remodeling and regeneration (Irie et al., 2018). Microcomputed tomography (µCT) has been previously used in bone biopsy to visualize and analyze NB formation with bone substitute (BS) (Stelzle & Benner, 2010; Klein et al., 2013; Trisi et al., 2006). However, the demerits of µCT images are noise, artifacts and low-contrast resolution (Bouxsein et al., 2010). This poses a hurdle for its ability to visualize soft tissues of low density, which makes segment separation of hard tissue from soft tissue difficult without using histological sections as references.

Despite the continued development of analytical equipment, analysis using tissue staining is still in use. It is impossible to make measurements in three dimensions using the method of tissue staining, but it is accurate, based on analysis and evaluation at the cellular and molecular levels.

The use of synchrotron radiation X-ray µCT (SR-µCT) can potentially solve these shortcomings and yield a high-resolution 2D/3D image. The advantage of this approach compared with conventional µCT (C-µCT) is that using it with a highly collimated and parallel monochromatic beam eliminates beam-hardening artifacts. The photon flux is several orders of magnitude higher than the conventional X-ray tube sources. This collimated beam gives an increased spatial resolution for the accurate measurement of tissue-mineral density. The additional advantage of using this method is that beams pass through dense structures effectively, making internal structures visible. SR-µCT has been used for a precise volume analysis of NB, BS (Seo et al., 2015) and trabecular bone architecture (Martín-Badosa et al., 2003) in the bone biopsy of humans and, recently, to investigate genetic variations in the mineral density and microstructural properties of murine cortical bone, including the vascular canals and osteocyte lacunae (Raum et al., 2007; Schneider et al., 2007).

Unlike conventional absorption-based imaging techniques, phase-contrast imaging has been widely used to improve soft-tissue contrast as the phase shift varies with materials of different X-ray refractive indices (Raven et al., 1996). This phase shift is observed at the edges of structures, enhancing the contrast when the X-ray absorption is mainly of low contrast. High-contrast images make it easier to apply threshold methods such as segmentation and analysis. However, X-ray CT images still exhibit motion artifacts and are noisy. Possible solutions include the correction of beam hardening, the removal of motion artifacts, mitigation by high-resolution CT and the implementation of a density-based threshold method. Pre-processing steps improve the output image when included before segmentation (Bouxsein et al., 2010). Therefore, to obtain a higher-quality output, pre-processing is necessary for X-ray CT imaging.

This study aimed to provide an enhanced image contrast and a threshold approach to quantify a bone-biopsy sample using high-resolution µCT. The specific aims were (i) to investigate the effects of image contrast and the pre-processing filtering method on segmentation measurements of SR-µCT and C-µCT (Fig. 1); and (ii) to demonstrate the segmentation that can be used for high-resolution microstructure and high-throughput analysis of NB, BS and woven bone (WB).

Figure 1 Image-processing diagrams.

2. Materials and methods

2.1. Sample preparation

Ethical approval for this study was obtained from the Ethics Committee, Kyungpook National University Hospital Institutional Review Board. Informed consent was obtained from the patient, and the biopsy sample was from a partially edentulous patient with at least the posterior upper tooth missing. With the surgeon using the lateral approach, after flap elevation an osteotomy on the lateral wall of the maxilla was performed to create a window. The space created between the maxillary alveolar process and the elevated Schneiderian membrane was filled using BS (Bio-Oss, Geistlich Pharma AG, Wolhusen, Switzerland). To avoid infection, the patient received Augmentin (625 mg), Etodin (200 mg) and Stilen (60 mg) daily over a seven day period. After six months of uneventful healing, the implant sites were prepared with trephine burs (3.0 mm outer diameter and 2.0 mm inner diameter) to obtain bone cores (vertical biopsies). The bone specimen was fixed in 4% neutral-buffered paraformaldehyde for use in SR-µCT and C-µCT investigations. In the analysis, the dimensions of the bone specimen were ∼3 mm (height) × 2 mm (diameter) for SR-µCT and 6 mm (height) × 2 mm (diameter) for C-µCT in the grafted area.

2.2. Conventional µCT acquisition

Conventional µCT scanning was performed using a table-top µCT scanner (Skyscan1272, Bruker, Belgium). This scanner consists of an X-ray tube with a current of 200 µA, a source voltage of 50 kVp and no filter. We used an exposure time of 200 ms and a rotation step of 0.3° over 180°. This resulted in an image pixel size of 2 µm.

2.3. SR-µCT acquisition

The SR-µCT scans were performed at the 6C Biomedical Imaging (BMI) beamline of the Pohang Light Source, Pohang City, South Korea (PLS-II), which has an electron energy of 3.0 GeV and a stored current of 350 mA in top-up mode. This beamline provided a monochromatic beam of relatively large spectral bandwidth (ΔE/E) of 1–2% using multilayered crystals. The multilayered crystals were coated with tungsten and boron carbide bilayers (Incoatec, Germany). The default beam size of the 6C beamline is 100 µm and it delivers a photon flux of 10 × 10¹¹ photons s⁻¹ mm⁻². The sample was scanned using a monochromatic X-ray beam energy of 23 keV. The images were acquired using a charge-coupled-device (CCD) camera (Zyla, Andor, UK) system, which consisted of a fluorescent screen (YAG:Ce, thickness 50 µm, Crytur, Czech) that converts monochromatic X-rays into visible light on the object lens (×4). The field of view was 4.2 mm × 3.0 mm with an effective pixel size of 1.8 µm. The sample-to-detector distance was set to 20 cm. For the tomographic scan, 360 projection images were taken, with an exposure time of 200 ms, each at 0.5° during the rotation of the sample from 0 to 180°. The beamline layout is shown in Fig. 2.

Figure 2 Layout of the SR-µCT experiment performed using the 6C BMI beamline at PLS-II. A 23 keV monochromatic beam from a double multilayer monochromator (DMM) attenuated by a graphite attenuator was used. The beam propagated 20 cm from the bone-biopsy sample to a scintillation crystal (YAG:Ce, thickness 50 µm) that converted the monochromatic X-rays into visible light on an object lens (×4).

2.4. Image reconstruction

SR-µCT reconstruction was performed using the commercial software Octopus (version 8.9; Inside Matters, Aalst, Belgium). The process was as follows: removal of white spots, normalization, ring filtration, phase filtering (modified Bronnikov algorithm), sinogram creation. It implemented the ordinarily filtered back-projection algorithm. The resulting images mapped the local X-ray attenuation coefficient and the boundaries enhanced by the interference of the transmitting coherent X-rays. C-µCT reconstruction was performed using the NRecon software provided by the scanner manufacturer (version 1.6.9; Skyscan, Kontich, Belgium).

2.5. Measurement of mean contrast value

After normalized images of 0 to 255 gray level intensities were exported to reconstruction images, intensity profiles of bone formation (BF) were qualitatively evaluated via the plotting of a line profile. Changes along the line between the NB, BS and WB corresponded to variations in the intensity. The mean contrast value of BF was quantitatively assessed by calculating the contrast-to-noise ratio (CNR) using the following formula (Bushberg et al., 2003),

where IBF is the maximum intensity value for the BF and IBG is the minimum intensity value from an equal size area (5 pixels × 5 pixels) enclosing the entire background area (BG). σBG is the standard deviation of the intensity values in the background image. The selections of equal sizes used for calculating IBF, IBG and σBG were outlined in slices clearly separating the NB, BS and WB measurements of 12 slices.

2.6. Image segmentation and analysis

For 3D image visualization as well as processing and analysis, another commercial software, Amira (version 6.2; FEI Co., Hillsboro, OR, USA), was used. The noise reduction was median filtered and smoothened with the software's non-local means noise (NLMF) filter function on the reconstruction images. The volume analysis (see Table 1) was preceded by image segmentation. The individual pixels of the 3D images were classified into the categories NB, BS and WB. The gray levels in a single segment fluctuated, often exceeding the threshold values set to discriminate between the different categories. In order to suppress the noise, we applied the NLMF filtration algorithm. The same procedure was successfully applied for the C-µCT 3D image sets. 2D area analysis of each bone material was performed using the following formula (Seo et al., 2015): Area percentage per slice (%) = N × Pixels (width × height) × 1/πr² × 100, and the ratio and distribution of NB and BS were measured every 8 µm from the specimen. The volume analysis was conducted by counting the number of pixels belonging to each category.

Table 1
Results obtained from the volume analysis of SR-µCT and C-µCT

	NB (mm^3)	BS (mm^3)	WB (mm^3)	Total (mm^3)
SR-µCT	3.54	1.61	1.6	6.75
C-µCT	3.02	1.29	—	4.31

3. Results

3.1. Improvement of image quality and contrast

Fig. 3 presents the sliced images of the bone-biopsy specimen harvested before implant placement in the graft area. The images obtained by SR-µCT and C-µCT were compared using the reconstruction program Octopus, version 8.9 (Inside Matters, Aalst, Belgium). As shown in Figs. 3(a) and 3(d), some of the main characteristics of the bone-graft fractions appear lighter and the NB appears denser, and soft tissues are invisible, including the WB region. The sliced images differentiated the two major constituents of contrast quite well, i.e. the NB and BS material Bio-Oss. However, a number of lower-contrast issues were identified in the C-µCT image compared with the SR-µCT image due to the beam-hardening effect and the ring noise.

Figure 3 Improved bone-biopsy cross-sectional view in SR-µCT images compared with C-µCT images. (a) and (d) The reconstruction original images, (b) and (e) after NLMF filtering, (c) and (f) intensity line profiles corresponding to the yellow lines shown through the images. The C-µCT images and line profile show the beam hardening effect (red arrow).

To distinguish between the main characteristics of the bone graft fractions possibilities, noise reduction was carried out. Figs. 3(b) and 3(e) show high-contrast images with the main characteristics of NB, bone graft and WB. The WB was observed by both SR-µCT and C-µCT. The image quality was quantitatively evaluated by plotting the intensity profiles along a horizontal line, i.e. the yellow dashed line in Figs. 3(b) and 3(d). The results in Fig. 3(c) of the SR-µCT image at the peak point are robust and sharp. The C-µCT image intensity was of high value (same as the graft material) at the beginning and at the end point due to the beam-hardening effect, and the peak point was not as sharp compared with the SR-µCT image intensity [Fig. 3(f)].

3.2. Mean contrast values

The mean image-contrast values measured from the SR-µCT images for the NB, bone graft and WB were 105.1 ± 3.78, 164.8 ± 3.19 and 78.5 ± 2.54, respectively, whereas the corresponding values from the C-µCT images were 85.5 ± 2.71, 105.1 ± 4.04 and 42.5 ± 2.01, respectively. Fig. 4 shows the mean values of the contrast difference between background values. A comparison of the results reveals that the contrast values in the SR-µCT images were different in each region while the contrast values in the C-µCT images were larger in the NB and BS regions and smaller in the WB region compared with the threshold limit.

Figure 4 Mean contrast comparison between the SR-µCT and C-µCT images. The limit for applying threshold segmentation is indicated by the red dashed line.

3.3. 3D visualization and analysis for C-µCT

The 3D datasets generated by C-µCT consisted of about 3236 × 3236 × 3114 voxels. The full field of view and spatial resolution of the C-µCT images allowed a whole-sample segmentation of the NB and BS, as shown in Fig. 5(a). The sagittal section shows the inner structure and the NB formation with BS [Fig. 5(b)]. The 2D analysis shows a graph of the NB and BS superimposed onto a 3D image background, which was correlated with slice position [Fig. 5(c)]. Our results demonstrated that the bone dynamics between NB and BS along the inferior–superior direction after MSFE could be analyzed using 400 slices. The specimen was analyzed in the cross-sectional position and the analysis readily extended over the entire volume. The average volume of the NB and BS measured is summarized in Fig. 5(d). The average values were 29.44% and 30.45% for NB formation and 13.39% and 13.07% for BS, giving total values (NB + BS) of 42.83% and 43.52% for SR-µCT and C-µCT, respectively.

Figure 5 Visualization of the 3D image and analysis of the bone-biopsy specimen acquired by C-µCT. (a) A whole bone specimen of a complete segmentation of the NB (red) and BS (white), and (b) a sagittal section view. (c) Visualization of the 2D analysis of NB and BS in graft sites versus slice number, which was correlated with slice position; the background image shows the 3D image of the bone biopsy. (d) Comparison of volume analysis between SR-µCT and C-µCT (without WB).

3.4. 3D visualization and analysis for SR-µCT

The 3D datasets generated by SR-µCT consisted of about 2379 voxels × 2379 voxels × 2058 voxels. The high-contrast and spatial resolution of the SR-µCT images allowed a robust and sharp segmentation of the NB, BS and WB, as shown in Figs. 6(a) and 6(b). The 2D analysis shows a graph of the NB, BS and WB superimposed onto a 3D image background, which was correlated with slice position [Fig. 6(c)]. Our results demonstrated that the bone dynamics between NB, BS and WB along the inferior–superior direction after MSFE could be analyzed using 400 slices. The specimen was analyzed in the cross-sectional position and the analysis readily extended over the entire volume. The positional variation of the NB and WB fractions clearly indicated the dependence on BS. The average volumes of the NB, BS and WB and the total volume (TV) measured are summarized in Fig. 6(d). The average values measured from the SR-µCT data for NB, BS, WB and TV were 29.44, 13.39, 13.29 and 56.12%, respectively. The results suggest that the WB is constantly remodeled by activity via bone regeneration. Previous CT studies revealed the analysis of NB and BS. However, we showed the 3D visualization and analysis of WB. SR-µCT delivers significantly better results than C-µCT because of phase contrast.

Figure 6 Visualization of the 3D image and analysis of the bone-biopsy specimen acquired by SR-µCT. (a) A whole bone specimen of a complete segmentation of the NB (red), BS (white) and WB (purple), and (b) a sagittal section view. (c) Visualization of 2D analysis of NB, BS and WB in graft sites versus slice number, which was correlated with slice position; the background image shows the 3D image of bone biopsy. (d) Comparison of volume analysis between SR-µCT and C-µCT.

3.5. Visualization of WB formation

The high contrast of the SR-µCT images allowed a robust and sharp segmentation of the NB formation, graft material and WB formation using the threshold function in the Amira software. Fig. 7(a) illustrates a sample of the segmentation results and the corresponding 3D reconstructed model. A clear separation between NB and BS can be seen. The region enclosing the WB is illustrated in the high-quality 3D view in Figs. 7(b) and 7(c).

Figure 7 3D visualization image of the bone-biopsy specimen acquired by SR-µCT. (a) A whole bone specimen of a complete segmentation of the NB (red), BS (white) and WB (purple), and (b) the corresponding high-quality 3D image along with a magnified view (c).

4. Discussion

This study is the first work to report on the visualization of a biopsy sample with a WB structure using SR-µCT. In addition, we presented a 3D image obtained by threshold segmentation of the complete NB, BS and WB, and an area or volume analysis. The threshold-segment method presented here can be used to develop the bone analysis of SR-µCT. Results suggest that SR-µCT provides a superior visualization of bone structures over C-µCT.

WB is usually studied via histology, and pre-processing is necessary to create such histological slides. The sample may be damaged in the demineralization process and the results may be distorted in the limited two-dimensional cross-sectional analysis. Previous studies have shown that histomorphometric analysis easily distinguishes between graft particles and surrounding NB, based on staining and morphology (Froum et al., 2008; Jensen & Sennerby, 1998; Yildirim et al., 2001; Moy et al., 1993; John & Wenz, 2004). The analysis in this paper was a quantitative evaluation of 2D images taken from a number of sections for specimens, and the analyzed parameters were for NB, BS and WB regions. We showed that the improved high-contrast SR-µCT images allow for an effective threshold segmentation. The clear and accurate segmentations resulted in a realistic and detailed 3D image of WB structures such as the NB and BS (Fig. 7), and we only assume that these images represented all the elements on the grafted sites.

Recently, for three-dimensional evaluation using µCT, most studies have focused on bone regeneration, and studies of hard tissues or the condition (optics tool, beam condition and image pre-processing) of the WB have not been found. This has been previously assessed only to a limited extent, because weak contrast, such as an unclear boundary due to low absorption coefficient, makes WB invisible on conventional µCT images. Several reports have evaluated NB formation for bone biopsy using C-µCT (Trisi et al., 2006; Klein et al., 2013; Soardi et al., 2014). However, most studies have focused on bone regeneration and analysis of the hard tissues. Soft tissues are more difficult to depict in conventional absorption-based µCT imaging because of low-energy X-ray blocking by attenuation filters, and because WB, with a low absorption coefficient, could not be depicted in the image. In contrast, here we tried not to use a C-µCT attenuation filter in order to utilize the low-energy X-rays. As a result, we observed WB in a 2D image through the image-filter processing and showed the potential to visualize WB. However, it is more difficult to carry out 3D segmentation for WB by C-µCT imaging due to the presence of fine low-absorption connective tissue and the beam-hardening effect.

WB is either immature or pathologic bone, with similar physical properties to soft tissues, unlike the lamellar bone. Therefore, for weak contrast, such as an unclear boundary due to low absorption coefficient, it is difficult to make an accurate standard in segmentation. Understanding the presence of WB in bone regeneration elucidates the BF potential. The approach to evaluating human-bone-biopsy specimens by high-energy X-rays from SR-µCT and the edge enhancement achieved by phase-contrast imaging resulted in an improved visualization and contrast of soft tissues such as WB. The edge enhancement provided by the phase-contrast method (Langer et al., 2009; Betz et al., 2007) and the depth of penetration provided by the SR-µCT images were sharper than the corresponding C-µCT images that had the same spatial resolution. Seo et al. (2015) previously investigated the ability of SR-µCT to evaluate NB formation in a human-bone-biopsy sample. Another study reported that a resolution of at least 4 µm is required to evaluate the bone structure (Palacio-Mancheno et al., 2014; Bouxsein et al., 2010). Indeed, the SR-µCT and C-µCT used in this work had similar spatial resolutions of ∼2 µm, and the number of projection images was 300 and 600, respectively. The edge-sharpening capability of SR-µCT allows it to depict the boundaries of NB, BS and WB. These improvements in visualization and sharpness are directly translated to the increased signal intensity contrast and the 1.5-fold increase illustrated in Fig. 4. Another possible explanation for this might be that higher-quality images can be obtained when images are acquired with the same projection number and pixel size as those in C-µCT.

The area and volume data were compared with the results of the C-µCT and SR-µCT analysis of the biopsy, using a similar region of interest (ROI). By using specific threshold settings (limit value = 50), SR-µCT was able to distinguish between NB, WB and grafted ROIs, and to gain insight into the 3D structure of the bone grown into the grafted maxillary sinus floor. Contrary to expectations, the C-µCT results did not exhibit a significant difference between the background and WB contrast values. The contrast value is the key parameter for the threshold segmentation. In this case, some pixels have intensity values below the threshold value, and some background pixels lie above the threshold value due to image inhomogeneities and additive noise. With C-µCT images, these pixels cannot be classified correctly because the contrast value adjacent to the background value is limited to the threshold setting. However, in phase-contrast imaging, the distance between the sample and the CCD camera can be adjusted, and there is a clear boundary with the structure due to the phase effect. It also provides improved contrast through image-filter processing. Thus, it is easy to set a threshold value using a clear boundary between the hard and soft tissues.

These results demonstrate two things. First, the observation region was limited to a field of view of 3 mm along the upper direction using SR-µCT; additional studies will be needed to extend to a large field of view with collimation crystal optics. Second, it is necessary to study the NB formation process in the early stage of bone regeneration through an in situ experiment. Therefore, further studies will be needed to determine the low-dose exposure, fast scan and full field of view with high-resolution CT imaging.

5. Conclusions

This study demonstrated that for all the elements represented on grafted sites, clear and accurate segmentations resulted in realistic and detailed 3D images of WB structures such as NB and BS. SR-µCT can be used to improve the quality of visualization and analysis (NB, BS and WB) without the need for pre-processing of the sample. It also provides a remarkably high depiction of structural details and soft-tissue contrasts as well as higher CNRs than C-µCT. Threshold segmentations of NB, BS and WB structures are possible using these images and yield accurate reconstructions of 3D images. This SR-µCT tool and these threshold segmentations represent a promising development in high-quality-based segmentation methods for bone regeneration and remodeling.