Cortical Break Detection Module
The Cortical Break Detection module segments the cortical breaks on a greyscale scan. The bone structures are binarized first. The algorithm then outputs a cortical break mask and a list of seed points inside the cortical breaks. The user will manually select the erosions from those cortical breaks. The periosteal bone mask is required and can be created in the AutomaticContour Module if it is not already available.
Load the greyscale scan, and the contour/mask if available, into Slicer. Either drag-and-drop the files to the application window, or hit the Add Data button
and select the files. The contour/mask need to be loaded as LabelMapVolumes. Check the Show Options check box at the top right corner of the Add Data prompt window, and check the LabelMap option.
It is recommended to enhance the cone beam CT image with a sharpening image filter due to the lower resolution. XtremeCT images need not be enhanced.
Go to the Simple Filters built-in module.
Apply the SimpleITK LaplacianSharpeningImageFilter to the cone beam CT image, preferably twice (by selecting the correct input/output volume and hitting the Apply button twice).
In the screenshot above, the input and output volumes are the same, so the greyscale image will be overwritten with the enhanced image.
A "_SEG" bone segmentation will be generated with the bones separated from the background. Cortical breaks will be identified based on the greyscale values, the periosteal mask and the bone segmentation.
Parameters:
- Input Volume: the greyscale scan.
- Output Volume: the output "_SEG" bone segmentation.
- Lower/Upper Threshold: global thresholds, in Scanco intensity units.
- Gaussian Sigma: the standard deviation in the Gaussian smoothing filter
Once all the parameters are set, hit the Preprocess button. The bones will be segmented using a Gaussian filter and thresholds.
The thresholds should be set between 686 and 4000 for the XtremeCT images and above 600 for the enhanced CBCT images. The Gaussian sigma should be 0.8 [1] for XtremeCT images and 0.0001 (minimum) for CBCT images. The thresholds and the Gaussian sigma should result in the bone separated from the background and should give a representation of the bone structure.
Parameters Cont'd:
- Input Volume: the greyscale scan.
- Preprocessed Volume: the "_SEG" bone segmentation.
- Contour: the periosteal bone mask.
- Output Cortical Breaks: Output binary cortical break mask.
- Output Seed Points: a list of seed points, one seed point in each cortical break, generated based on the output cortical break mask
- Cortical Thickness: thickness of the cortex in voxels. For simplicity, the algorithm assumes the cortex has a constant thickness, and it selects cortical breaks penetrating the cortex.
- Dilate/Erode Distance: dilate/erode distance in voxels used in the internal morphological operations in order to remove undesired voids in the cortex.
- Voxel Size: voxel size of the greyscale image in mm.
- XtremeCT I/XtremeCT II/CBCT: select the CT type. XtremeCT images will undergo a slightly different algorithm than CBCT images.
For XtremeCT I, the default cortical thickness is 4 voxels [1] and the default dilate/erode distance is 1 voxel [1]. For XtremeCT II, default cortical thickness is 5 voxels and default dilate/erode distance is 2 voxels.
Once all the parameters are set, click the Get Cortical Breaks button.
Parameter:
- Seed Points: the output seed points from the previous step
View the individual seed points. Select the true erosions by removing the extra seed points. In general, there will be non-erosion seed points, but no true erosion seeds should be missing.
[1] Peters M, Scharmga A, de Jong J, van Tubergen A, Geusens P, Arts JJ, Loeffen D, Weijers R, van Reitbergen B, van den Bergh J. An automated algorithm for detection of cortical interruptions on high resolution peripheral quantitative computed tomography images of finger joints. PLOS ONE (2017) 12:e0179138. doi:10.1371/journal.pone.0179138
