Image Registration Module
The Image Registration module performs a longitudinal registration of a baseline and a follow-up image. The algorithm will determine a transform that minimilizes differences between the two images, so they are aligned by features on the bone. The visualization tool can then be used to create a subtraction of the images, showing changes in the bone over the time period.
Parameters:
- Baseline: Baseline (initial) image
- Follow-up: Follow-up image
- Similarity Metric: Method used to calculating difference between images
- Metric Sampling Percentage: Percentage of image used for similarity calculation
- Output: Output volume
Upload the baseline and follow-up image into slicer and select them in the appropriate dropdown. Select a similarity metric to use in the registration algorithm. Each metric uses a different method to determine the difference between images at each iteration. The chosen metric can significantly affect the results of registration, so choose carefully.
-
Mean Squares: Computes the mean squared difference between pixel values. Requires intensity values to be within the same thresholds for images.
-
Correlation: Computes the normal correlation between pixel values. Requires images in the same modality, but can be in any intensity range.
-
Mattes Mutual Information: Computes the mutual information (ability to determine intensity of the second image based on the first) of the images. Can be used with multiple modalities (i.e. CT scan and MRI).
-
ANTS Neighborhood: Computes the correlation of a small neighbourhood for each pixel. Ideal for images that are very close already since this method is slower.
The Metric Sampling Percentage will determine what percentage of the image is analyzed per iteration. Points on the image will be randomly selected for calculations. Increasing the sampling percentage will improve the final results, but will significantly increase the time required.
Recommended percentages:
- Small Image (<50 MB): 0.1
- Medium Image (50-200 MB): 0.01
- Large Image (200 MB - 1 GB): 0.001
- X-Large Image (>1 GB): 0.0001
Click Apply to begin registration. This will take a few minutes.
Parameters:
- Baseline: Baseline (initial) image
- Registered Follow-up: Registered follow-up image
- Output: Output volume
3D Subtraction Only:
- Lower Threshold: Lower threshold of bone
- Upper Threshold: Upper threshold of bone
- Gaussian Sigma: Sigma for Gaussian filter
Select the baseline image and the registred image from step 1 to create a subtraction of the images. Subtractions can only be created for unimodal images
3D Subtraction This tool creates a color-coded subtraction mask that can be rendered as a 3D volume. This image shows the regions which are different between the baseline and follow-up, which can be used as a measure of registration quality or as a visual tool for detecting developing erosions.
Color code:
- White: Present in both images
- Peach: In baseline only
- Blue: In follow-up only
Grayscale Subtraction This tool creates a direct subtraction image of the baseline and follow-up, which is used to evaluate scan-rescan consistency.
Paramaters:
- Baseline: Baseline (initial) image
- Registered Follow-up: Registered follow-up image
- Output: Output volume
- Number of Cells: Number of cells in each dimension
This tool will create a checkerboard view of the baseline and follow-up image. Elements from each image will be placed into a grid layout, which can be used as a measure of registration quality. Can be used for multi-modal images.
Click Show Checkerboard Grid to create a mask with an overlay of the generated grid. Recommended when comparing unimodal images with this tool.
