Measuring Bone Volume at Multiple Densities by Micro-computed Tomography

[Abstract] Bone strength is controlled by both bone mass, and the organization and quality of the bone material. The current standard method for measuring bone mass in mouse and rat studies is micro-computed tomography. This method typically uses a single threshold to identify bone material in the cortical and trabecular regions. However, this single threshold method obscures information about the mineral content of the bone material and depends on normal morphology to separately analyze cortical and trabecular structures. To extend this method to identify bone mass at multiple density levels, we have established a protocol for unbiased selection and application of multiple thresholds using a standard laboratory-based micro-computed tomography instrument. This non-invasive method can be applied to longitudinal studies and archived samples and provides additional information about bone structure and strength.


mass, Mineralization
[Background] Micro-computed tomography is a well-established standardized method for measuring bone mass and structure in experimental mice and rats. Typically, the inner network of bone (trabecular) is analyzed separately from the thick outer shell of bone (cortical). Identifying trabecular and cortical regions is usually either manual or uses algorithm-based segmentation on the basis of morphology (Buie et al., 2007;Bouxsein et al., 2010). A single threshold within each region is used to define bone: cortical bone is detected using a high threshold, and trabecular bone at a lower threshold (Ansari et al., 2018;McGregor et al., 2019). Recently however, while analyzing a genetically altered mouse model with severely disrupted bone structure (Dmp1 Cre :Socs3 f/f ), it was not possible to delineate cortical and trabecular structures (Cho et al., 2017;Walker et al., 2020), and standard segregation of bone into trabecular and cortical was not meaningful. Such segregation would also be challenging in other mouse models where cortical bone structure is abnormal, such as during ageing or in chronic kidney disease when it is porous (Piemontese et al., 2017;Metzger et al., 2020), or in conditions with poorly formed cortex such as osteopetrosis (Abe et al., 2000). Using a single threshold also reduces the amount of information that can be gained from micro-computed tomography, in which raw images show multiple levels of gray that are reduced to a single level for quantitation.
To overcome this problem and move away from the dichotomy of "trabecular" and "cortical" bone, we made use of multi-level Otsu thresholding to measure bone mass at multiple density levels. The Otsu Copyright  method makes use of an algorithm that segments pixels from the raw images into different classes based on the gray level intensities within the image (Otsu, 1979). Initially, we applied this method to the metaphyseal region (Walker et al., 2020) because it contains both immature and mature cortical bone, with the most mature bone at greatest distance from the growth plate (Rauch, 2012). Using this new method applied to a slice-by-slice analysis of the metaphyseal region enabled quantification of the process by which normal cortical bone matures over time, specifically, the replacement of a low density structure with a more robust higher-density structure (Walker et al., 2020). It also revealed a higher proportion of low density cortical bone in the mature skeleton of Dmp1 Cre :Socs3 f/f mice (Walker et al., 2020), which we previously reported to have impaired strength, even though the structure was normal when measured by a single-threshold method (Cho et al., 2017). We have also used this method to identify deposition of new bone resulting in a change in bone shape due to pharmacologically-induced muscle hypertrophy (Chan et al., Submitted).
This unbiased method of measuring bone at multiple densities is non-destructive and can be used to measure changes in bone over time, since it can be used in micro-computed tomography scans of live animals (Walker et al., 2020). The method may also be applied to the study of bone development (Bortel et al., 2015), and to the analysis of human bone, particularly during ageing (Zebaze et al., 2010).
The method may also be helpful for providing new methods for age estimation in bioarchaeology and forensic anthropology (Maggiano et al., 2015), and would provide valuable new information in study of other species, such as measuring bone mass at different density levels that have been observed with We describe here the use of this method as it is applied to the metaphyseal region of fixed murine femora which contains both immature and mature bone, since this is likely to be the most common application. The same approach can be used for smaller or different regions, and for scans of bones taken from mice, or other species, including longitudinal studies under anaesthetic. This method can also be applied to archived scans, in which case the protocol need only be followed from Procedure C onwards. This protocol assumes a basic familiarity with micro-computed tomography.

Materials and Reagents
1. Fixed or frozen murine bone samples (e.g., tibiae or femora) for analysis  3. At the same time, and using the same scanning settings, scan both calibration rods.
3. Reconstruct images using your standard protocol. You may find it helpful to refer to method notes from your manufacturer (e.g., For Bruker instruments, "An overview of NRecon: reconstructing the best images from your microCT scan Method note MCT-062", available when you register at https://www.brukersupport.com/). We reconstruct with SkyScan software NRecon using the following settings: 2. Orientate the sample in all three planes so that it is vertical and symmetrically aligned ( Figure   1). Using the coronal plane, manipulate the bone orientation with the mouse while holding 'Ctrl'.
By directing the mouse to the center of the growth plate, the second transverse view can be manipulated so that the bone is orientated with the patella at the top of the image (see Figure   1). Use the sagittal view to align the bone horizontally.  Figure 2B). We define the proximal end of the bone as the slice where the femoral head and the trochanter first meet ( Figure 2C).
In the distal femur, we identify the growth plate by identifying the site where the four quadrants of the femoral growth plate begin to merge. As you scroll from the distal end towards the center of the bone, you will notice the presence of four circular quadrants of the growth plate, that enlarge and meet ( Figure 2D). The medial and lateral growth plate regions meet first. We call this the first meeting. As you continue to scroll, the anterior and posterior growth plate regions meet. We call this convergence of the growth plate "the second meeting" and define it as the "growth plate" slice. Use the measurement values for each landmark to calculate bone length by subtracting the distal measurement from the proximal measurement ( Figure 2E). Image in panel A is taken from http://www.informatics.jax.org/cookbook/figures/figure42.shtml. 6. Calculate the length of the metaphyseal region of interest (ROI) for each sample based on the bone length. We do this so that the same anatomical region is measured in each sample, regardless of whether bone growth is modified. In the original publication using this method, we defined the distal end of the region of interest as the slice at a distance equal to 7.5% of the femur length from the growth plate (see Figure 2A  2. Load all samples from your control group, and run the above custom processing using CTAn's batch manager (BATman). This will generate an output file for each sample that reports 4 Otsu-based thresholds ( Figure 3A).
3. Check the BMP files to ensure they correctly represent the amount of bone visualized. You will notice that the images for the lowest threshold will include much non-bone material. This lowest threshold will not be used for analysis of your samples ( Figure 3B). D. Use the results from scanning the calibration standards to calculate the equivalent calcium hydroxyapatite levels (g/cm 3 ) represented by your thresholds (Figures 3C, 3D) Note: These are required for reporting, so your levels can be reproduced and understood by other researchers who may be using different instruments (reporting the 0-255 gray levels will not achieve this).
2. In CTAn, under the Binary Selection tab, select the "From dataset" tab. Type in the gray level threshold for Level 2 (in our example is 40). Select the "Bone mineral density" tab in this window and scroll down to find the first BMD value that is not greyed out. In our example, the threshold level 40 corresponds to a mineral density of 0.632 g/cm 3 CaHA.  i. Thresholding for mid density bone: Global, and add the Level 3 threshold value calculated in Procedure C. In our example it is 89. Set the maximum (minus 1) at the Level 4 threshold.
In our example it is 147. j. 2D Analysis: Select "All results", deselect "Append summary results to file". This will generate Bone Area measurements for each slice included in the dataset, within the mid density threshold (Level 3 to Level 4) and will automatically place them into the file called 3. Run the above custom processing using BATman batch processing. This will take some hours, depending on the number of samples being assessed.

Data analysis
Procedure E above will generate a summary of the 3D measurements for each sample, at each density threshold. The values for low, mid and high-density Bone Volume will be in their respective separate comma-separated value (CSV) files that you created. Another CSV file will be automatically generated

Statistical analysis
Prior to any statistical analysis, it is important to graph individual values for each sample to ensure there has not been any error made along the way. You should see an increase in the proportion of high-density bone with increasing distance from the growth plate, and a decrease in the proportion of low density bone along the same distance. If major outliers are observed, check back on the BMP files generated during the data processing to visualize the threshold regions analyzed and identify any errors.
1. To determine whether there is an overall difference in the volume of bone at each density level throughout the entire region, we compare overall Bone Volume/Total Volume for the entire region by a one-or two-way ANOVA, depending on the number of factors analyzed.
2. To determine whether there is a significant difference in the amount of bone at a single density threshold between two genotypes at each slice, we apply a two-way ANOVA using "genotype" and "distance from the growth plate" as categories.

Notes
1. The most common cause of variability is when samples have been scanned, or reconstructed, under differing conditions. It is vital to ensure this is the same for comparable results.
2. It is important to check that your scans are of high quality and that your reconstructions are conducted properly -for example, any ring artefact or flaring will interfere with this analysis method.