Longitudinal Micro-Computed Tomography Image Analysis for User-Defined Region of Interest in Critical-Sized Bone Defects

Summary

We present a method for analyzing a user-defined region of interest (ROI) in a longitudinal in vivo rat radial defect model. This method enables comparative analysis between different scaffolds previously limited by variations in microcomputed tomography (µCT) scan field of view, specimen orientation, and baseline presence of scaffold.

Abstract

Micro-computed tomography (µCT) imaging analysis of bone volume is a necessary quantitative tool for investigating bone regeneration potential and outcomes within longitudinal in vivo studies. Established methods for bone segmentation utilize visualization software for whole bone µCT segmentation and alignment of complex anatomical structures. These segmentation protocols provide a robust, high-accuracy method for segmentation, alignment, and analysis but are limited in abilities of user-defined region of interest (ROI) analysis. We present a protocol expanding upon these methods to permit user-defined ROI bone volume analysis surrounding a critical-sized bone defect. The user-defined ROI surrounding the defect can be analyzed over time for in vivo longitudinal studies. Herein, we investigate µCT images of three unique rat specimens each implanted with a polycaprolactone (PCL) control scaffold. Models are analyzed by three users (2 experienced and 1 novice) at time points of 0 and 6 weeks to illustrate the ability to measure an ROI surrounding a critical-sized defect throughout a longitudinal study.

Introduction

Critical-sized bone defects pose significant clinical challenges in orthopedic treatment management. Per ASTM F2721, a critical-sized defect is characterized as a defect with a length 1.5 to 2 times the diameter of the bone of interest¹. Repair of these defects has traditionally been through the use of autologous and allogeneic transplantations limited by the procedural expenses, associated risks of secondary surgeries, and bone graft volume required². Current bone regeneration techniques focus on the use of allogeneic and xenogeneic scaffolds designed to produce both osteoconductive and osteoinductive effects through optimizing their mechanical properties, biocompatibility, bioactivity, angiogenic potential, and degradation profiles^3,4,5. Biomaterials investigated broadly range from bioceramics and biopolymers to metals and other composite materials⁶. Variations of these biomaterials are tested both in vitro and in vivo to interrogate their potential as bone regeneration scaffolds.

µCT is the gold standard for non-invasive, high-accuracy imaging for the assessment of bone morphology, structure, and microstructure in rodent models^7,8,9. This imaging modality has been described to assess the longitudinal, in vivo progression of bone regeneration in fracture healing models¹⁰. Methods have been developed to standardize the quantification of cortical and trabecular bone from µCT scans⁹. Semi-automated segmentation workflows have been developed utilizing commercially available visualization software for whole bone segmentation with complex anatomical structures¹¹. These methods allow for simplified, approachable methods for users across varying experience levels to produce standardized, reproducible results. However, these methods remain limited in abilities to investigate user-defined ROI.

Here, we present a protocol that expands upon current methods to permit user-defined ROI bone volume analysis surrounding a critical-sized bone defect for longitudinal in vivo rat models using visualization software. Establishing a consistent alignment and ROI selection method between weeks of the longitudinal study was essential for the development of a robust protocol. An initial timepoint is used as the baseline for the alignment of subsequent weeks to ensure consistent orientation of solid models. Provided this alignment, corresponding µCT image slices from the overlayed solid models can be selected, encompassing the critical-sized defect. Consistent ROI is verified not only through slice location but also through comparison of the number of slices within the region. The selected ROI from the baseline model can then be replicated on subsequent weeks, allowing for comparative, quantitative analysis.

Protocol

Longitudinal µCT images for this study were collected at weeks 0 and 6 from 3 mm critical radial defects in adult female Charles River SASCO-SD rats treated with a polycaprolactone (PCL)-based scaffold. All animal use was performed in accordance with protocols approved by the University of Rochester's Committee on Animal Resources (UCAR). µCT image collection was performed using Scanco Medical VivaCT 40.

NOTE: The primary steps in this protocol are divided into µCT image segmentation, alignment, ROI selection and cropping, and analysis and visualization (Figure 1). The protocol for µCT image segmentation is adapted from Kenney et al. (2022)¹¹.

Figure 1: Summary workflow diagram. The protocol steps are primarily divided into image segmentation, model alignment, ROI selection, and volume analysis. Please click here to view a larger version of this figure.

1. µCT image segmentation

1. Opening images within the software
   1. Launch Amira software (hereinafter referred to as visualization software) and select Open Data. Within the pop-up window, navigate to the desired folder directory to select the initial time point .dcm files. Press CTRL+A to select all images and click Open Images.
   2. A pop-up window will display information about the opened image data set; click OK to proceed.
   3. A warning on scaling will appear. Click Convert to Float (the default option) to proceed.
   4. The data set icon will appear within the Project View section. To change the name of the data set, single-click on the data set icon and press F2; once the desired name is input, click OK. Establish a naming pattern to identify the scan (for example, 'Bone Scaffold Material' - 'WKX').
2. Filtering and thresholding the µCT Image
   1. With images loaded into the Project View section, click on Ortho Slice default 2D viewer to view 2D orientations and slices. A 2D visualization will appear in the viewing window to the right.
   2. To visualize the images in 3D, right-click on the data set, search for Volume Rendering, and select. Within the Properties section for the Volume Rendering, adjust the Colormap lower threshold to 2500 and press Enter.
      ​NOTE: 2D and 3D can be turned on and off by clicking the blue square next to Ortho Slice or Volume Rendering, respectively. For the viewing window to the right, multiple view options are available along the top of the bar. These include different tools for moving, rotating, or zooming in/out of the image; tools for changing image orientation to XY, XZ, and YZ views, and tools for simple measurements.
   3. As needed, some data sets may require artifacts or data outside the region of interest to be removed using a Volume Edit. To add a Volume Edit, right-click on the data set, search for Volume Edit, and select.
   4. Click on the Volume Edit in the Project View section. Within the Properties section, select TabBox from the first drop-down window to the right of Tool.
   5. When selecting the regions to retain, adjust the TabBox by clicking and dragging the green corners using the Interact cursor (the pointed cursor icon along the top bar of the viewing window). Adjusting the TabBox may require using alternate views (XY, XZ, or YZ) to identify the regions to remove.
   6. Once the TabBox surrounds the region to retain, select Cut Outside to remove the artifact or data outside of the TabBox. A new, modified data set will be generated in the project view.
   7. Click and drag the Volume Rendering string from the original data set to the modified data set. This will change the image displayed in the viewing window to the modified data set. Within the Properties section for the Volume Rendering, adjust the Colormap lower threshold to 2500 and press Enter.
   8. Right-click on the modified data, search for Median Filter, and select. Within the Properties section for the Median Filter, select 3D for Interpretation and click Apply. This will create a new filtered (.filtered) data set in the Project View section.
   9. Right-click on the filtered data, search for Interactive Thresholding, and select. Within the Properties section for the Interactive Thresholding, select 3D for Preview Type, adjust the lower value for the Intensity Range to 2500, and click Apply. This will create a new thresholded (.thresholded) data set in the Project View section.
   10. Click and drag the Volume Rendering string from the modified data set to the thresholded data set. Turn the Interactive Thresholding off and the Volume Rending on (if turned off).
3. Segmenting the µCT Image
   1. Click on the filtered data set and then change to the Segmentation tab by clicking on Segmentation below the menu bar. Within the Segmentation tab, the controls are similar to those in the Project tab.
      ​NOTE: There are some additional tools relevant to the viewing window within the Segmentation tab. There is a +/- zoom button in the top corner of the viewing window that can be used so that the image fills the viewing window. The image can be centered in the viewing window using the left/right scroll bar along the bottom of the viewing window. Above this scroll bar, a second scroll bar is present to move through the µCT data slices.
   2. To ensure that the Segmentation tab is appropriately set up, under the Segmentation Editor section, the Image is set to the filtered data set, and the Label field is set to the thresholded data set. Under the Materials section, double click on the Material named Material and rename this to Bone.
   3. Click on New under the Segmentation Editor to create a new label field. In the pop-up window, rename this using a similar naming convention (for example, 'Bone Scaffold Material' - 'WKX' - Markers) as before and as Markers and click OK.
   4. Adjust the 2D viewing window using alternate views (XY, XZ, or YZ) to display either a sagittal or coronal plane view of the image.
   5. Use brush, lasso, wand, and threshold within the Selection section to identify the areas of interest. For this protocol, select the brush tool. Adjust the size of the brush tool; generally, a mid-size pointer works well.
   6. Adjust the image stack to the first slice where the radius bone begins to be visible. This may require moving through the slices multiple times to determine what volumes belong to the radius and ulna and when they begin.
   7. Once the slice is identified, use the brush cursor to draw by clicking and dragging the cursor around this segment of bone. A red line will appear where the cursor has drawn. Draw within the section of bone, not outside the bone; this will assist the visualization software with differentiating between the two bones.
   8. Continue moving through slices and drawing around the radius. Generally, through the mid-section of the image stack, drawings can be made every 20-30 image slices. Along the end of the image stack or surrounding the critical-sized defect, make drawings every 5-10 image slices to assist the visualization software as bone fades in or out.
   9. Once the full length of the radius has been completed, click on Selection, Fill, and then All Slices in the menu bar drop-down. This will fill all the drawings made with the brush tool. These will now appear as shaded with red rather than outlines.
   10. Click Selection and Interpolate in the menu bar drop-down. This will fill in shaded regions of radius bone across all slices based on the drawings made.
       ​NOTE: This will occasionally miss some sections of bone. Before proceeding scrolling through all slices to ensure it has generally identified the bone. If a section of bone has been missed, used the brush tool to circle around this region on a few slices surrounding the missing section, and repeat the above two steps (1.3.9 and 1.3.10).
   11. Once the radius has fully been identified, under the Materials section, double-click on the Material named Inside and rename this to Radius.
   12. Click the Add (+) symbol located under the Selection section to add this as a material. The radius is now outlined in the viewing window. Click the Lock icon for the Radius material in the Materials section before proceeding; this will ensure that no changes are made to this material.
   13. Complete the same process for the Ulna bone. Under the Materials section, click Add to create a new material. Double-click on the new material and rename it Ulna.
   14. Repeat steps 1.3.6 through 1.3.12 for the Ulna bone.
   15. Once completing the steps for both the Radius and Ulna bones, unlock the Radius by clicking the Lock icon for the Radius material in the Materials section. Change to the Project tab by clicking on Project below the menu bar. There is now a data set named Markers applied to the filtered data set in the Project View section.
   16. Right-click on the filtered data set and, search for Marker Based Watershed Inside Mark (Image Segmentation) and select. Within the Properties section, by clicking the drop-down menus, set the Data as the filtered data set, the Markers as markers data set, and the binary mask as the thresholded data set and click apply.
   17. A .grown file will be generated within the Project View section. Click and drag the Volume Rendering string from the thresholded data set to the .grown data set to visualize the segmentation of the radius and ulna bone within the viewing window.
   18. To convert the .grown file to 8-bit, click on the .grown data set and, search for Convert Image Type and select. The .grown file will be converted from 16-bit to 8-bit.

2. Alignment

1. Axis alignment and extraction of radius and ulna
   NOTE: If the current model is the initial time point, all steps must be completed. Otherwise, begin this section at step 2.1.4. To allow for a transverse slice of the radius, the initial time point model must be aligned perpendicular to the Ortho Slice plane; this model will serve as the baseline for subsequent weeks' alignment.
   1. Within the Project View section, turn on the Ortho Slice (if off) and then click on the .grown data set. In the Properties section, click on the Transform Editor icon. Using the Interact cursor, adjust the angle of the radius and ulna by clicking on the green axis alignment points so that the Ortho Slice creates a transverse slice through the radius.
   2. Once this has been aligned, a new data set needs to be created to save this transformation. Right click on the .grown data set, search for Resample Transformed Image and select.
   3. Click on the Resample Transformed Image and, within the Properties section, set the Data to the .grown data set, Interpolation to Nearest Neighbor, Mode to extended, Preserve to Voxel Size, and Padding value to 0. Click Apply, and a new .transformed data set will be created.
   4. To extract the radius and ulna from the combined segmentation model, right click on the .transformed data set, search for Extract Label and select.
   5. Click on the Extract Label, within the Properties section, set the Labels to the .transformed data set, Label ID to 1, and Export to Binary checked. Click Apply, and a Result data set will be created.
   6. Click and drag the Volume Rendering string from the .grown data set to the Result data set. Click on the Result data set and press F2 to rename the file (for example, 'Bone Scaffold Material' - 'WKX' - Radius).
   7. Repeat steps 2.1.4 to 2.1.6 for the ulna by setting the Label ID to 2 and renaming the Result file for Ulna.
2. File saving and opening longitudinal images
   ​NOTE: If the current model is the initial time point, all steps must be completed. Otherwise, begin this section at step 2.2.2.
   1. To save the initial time point file, click on File, Save Project As, and Set File Directory Location in the menu bar drop-down. Set the save type as an Amira Project and data files (pack &go) (*.hx) file and name the file as the master version (for example, 'Bone Scaffold Material'_ 'WKX' MASTER).
   2. Open the initial time point master file (if not already open) and save a copy by clicking on File, Save Project As, and Set File Directory Location in the menu bar drop-down. Set the save type as an Amira Project and data files (pack &go) (*.hx) file and name the file (for example, 'Bone Scaffold Material'_ 'WK0' and 'WKX'). This file will be used to compare the initial time point with subsequent weeks without overwriting the master file.
   3. Within the new comparison file, open the images for the comparison time point by clicking Open Data in the Project View section. Within the pop-up window, navigate to the desired folder directory to select the initial time point .dcm files. Press CTRL+A to select all images and click Open Images.
   4. A pop-up window will provide information regarding the image data set opened; click OK to proceed.
   5. A warning on scaling will appear. Click Convert to Float (the default option) to proceed.
   6. The data set icon will appear within the Project View section. To change the name of the data set, single-click on the data set icon and press F2; once the desired name is input, click OK. A naming pattern can be established to identify the scan (for example, 'Bone Scaffold Material' - 'WKX').
3. Model alignment
   NOTE: This process will discuss the alignment of the radius bone segmentations. The same process can be applied to the ulna, as needed. Do not align both bones at the same time; this will cause issues with proper alignment.
   1. Turn off any of the initial time point Volume Rendering. There is nothing displayed within the Viewing Window.
   2. Right-click on the comparison (subsequent week) data set, search for Ortho Slice, and select.
   3. Repeat section 1 (starting at step 1.2), section 2(step 2.1) following any additional notes for step numbers provided. Upon completion of these sections, resume at step 2.3.4. Completion of these sections will produce the segmentation of the subsequent week's data set and extract the radius and ulna bones.
   4. Right-click on the comparison (subsequent week) data set for the extracted radius bone, search for Image Registration Wizard, and select. Within the Properties section, set Data as the comparison week data set for the extracted radius bone, Reference as the initial time point data set for the extracted radius bone.
   5. For the Image Registration Wizard Actions section, click Skip for Step 1 of 4. For Step 2 and 3 of 4, using the Interact cursor, adjust the TabBox to the common region between both the initial time point and the comparison week data sets clicking Apply under Action following each step. For Step 4 of 4, set Metric as Correlation, Transformation as Rigid, and Pre-Alignment as Align Principal Axes and click Apply under Action.
   6. Once the data sets are aligned, right-click on the comparison (subsequent week) data set for the extracted radius bone and, search for Resample Transformed Image and select. Within the Properties section, set the Data as the comparison week data set for the extracted radius bone, Interpolation to Nearest Neighbor, Mode to Extended, Preserve to Voxel Size, and Padding Value to 0 and click Apply. A new .transformed data set will be created for the comparison data set.

3. ROI selection and crop

NOTE: Complete the ROI Crop first to determine the slice numbers around the critical-sized fracture. Once these slice numbers are determined, these may be used at the same time point for the ulna, as needed. For this process, cropping cannot be reversed once applied to the model.

1. Region of interest crop
   1. Click on and turn on the Ortho Slice for the initial time point and set the Data to the initial time point data set for the extracted radius. Set the Orientation so that the plane yields a transverse cut through the radius bone.
   2. Using the Slice Number slide bar in the Properties section, change the slice number to determine the proximal and distal slice locations surrounding the critical-sized defect. For both slices, determine the slice most distal and proximal where the fracture meets the diaphysis of the radius bone. Document the slice number; this number will vary for each model.
   3. Click on and turn on the Ortho Slice for the comparison week and set the Data to the initial time point data set for the extracted radius. Set the Orientation so the plane yields a transverse cut through the radius bone.
   4. Using the Slice Number slide bar in the Properties section and with the initial time point data set showing the distal Ortho Slice, change the comparison week slice number so that it aligns with the distal slice of the initial time point. The visualization software allows for a visual check for alignment; when the slices are overlapped, they will appear as a single slice in the viewing window.
   5. Note the slice number for the distal slice of the comparison week data set. Repeat step 3.1.4 for the proximal slice.
      NOTE: As an additional check, determine the difference between the proximal and distal slice for both the initial and comparison week data sets. These values must be equivalent, illustrating a consistently sized ROI.
   6. Click on the initial time point for the extracted radius. In the Properties section, click on the Crop Editor tool.
   7. Within the Crop Editor pop-up, input the minimum and maximum values into the appropriate X, Y, or Z field (the Viewing Window will alter the ROI as these are input). Once the values are input, click OK. The data set will be cropped to the input ROI.
   8. Repeat step 3.1.7 for the comparison week data set.

4. Analysis and visualization

1. Volume analysis
   1. To determine the volume of the initial time point data set, right-click on the transformed initial time point data set for the extracted radius, search for Material Statistics, and select. Within the Properties section, set Data as the initial time point transformed for the extracted radius and, set Select as Materials and click Apply. A new .MaterialStatistics data set will be created for the initial time point data set.
   2. Click on the .MaterialStatistics data set for the initial time point for the extracted radius and within the Properties window, click on Spreadsheet Show. Click on the Tables tab above the Properties Window. Within this tab, the volume for the cropped initial time point data set for the extracted radius is shown.
   3. Repeat steps 4.1.1 and 4.1.2 for the comparison week data set for the extracted radius. Within the Table tab, the initial time point and comparison week will have separate tabs that can be clicked into.
2. Visualization of data sets
   1. To visualize the change in bone volume, right-click on the comparison week transformed data set for the extracted radius, search for Arithmetic, and select. Click on Arithmetic and within the Properties window, set Input A as the comparison week transformed data set for the extracted radius, set Input B as the initial time point data set for the extracted radius, set Input C as NO SOURCE, set Result Type as Input A, leave Option unchecked, set Result Channels as like input A, and set Expression as A-B.
   2. Click on the Result data set and press F2 to rename the file (for example, 'Bone Scaffold Material' - 'WKX' - Bone Change). Right-click on this result data set, search for Generate Surface, and select. Click on Generate Surface and within the Properties window, click Apply, and in the pop-up window, click Continue. A new .surf data set will be created.
   3. Right-click on the .surf data set, search for Surface View, and select. A Surface View of the Arithmetic result will be shown in the Viewing window.
   4. To change the color of the Surface View, click on the Surface view in the Project View window, and then within the Properties window, click on the drop-down for Colors and select Constant. Click on the Colormap and assign the preferred color.
   5. To view the bone volume change on the initial week data set, right-click on the .transformed data set, search for Extract Label, and select. Click on the Extract Label, and within the Properties section, set the Labels to the .transformed data set, Label ID to 1, and Export to Binary checked. Click Apply, and a Result data set will be created. Click on the Result data set and press F2 to rename the file (for example, 'Bone Scaffold Material' - 'WKX' - Full - Radius).
   6. Right-click on this result data set, search for Generate Surface, and select. Click on Generate Surface, and within the Properties window, click Apply and in the pop-up window, click Continue. A new .surf data set will be created.
   7. Right-click on the .surf data set, search for Surface View, and select. A Surface View of the Arithmetic result will be shown in the Viewing window.
   8. To change the color of the Surface View, click on the Surface view in the Project View window, and then within the Properties window, click on the drop-down for Colors and select Constant. Click on the Colormap and assign the preferred color.
   9. To continue with additional weeks, return to step 2.2.2.

Results

µCT images of three unique rat models, each treated with a polycaprolactone (PCL) scaffold, were investigated to illustrate positive results. Analysis of a longitudinal study across time points requires the collected solid models to be aligned prior to selecting and cropping to an ROI. To illustrate this capability over multiple weeks, solid models collected at weeks 0 and 6 were aligned using common regions (Figure 2).

<...

Discussion

Quantifying bone volume change is essential for investigating bone regeneration potential and outcomes in longitudinal in vivo studies. This protocol builds upon established µCT image segmentation methods¹¹, providing a systematic approach for specifying a user-defined region of interest (ROI). This technique has been critical in the analysis of bone scaffold implant effectiveness for in vivo rat studies. By expanding upon established protocols¹¹, thi...

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Amira 3D	FEI SAS, a part of Thermo Fisher Scientific	v2024.1	Program used for segmentation of microCT images.
Graph Pad Prism	GraphPad Software LLC	v10.0.3 (217)	Program used for graph development.
R Statistical Software	The R Foundation for Statistical Computing	v4.4.0 (2024-04-24)	Program used to perform ICC analysis.
Scanco Medical VivaCT 40	Scanco Medical	NA	microCT scanner used for collection of microCT images.