Projects:RegistrationLibrary:RegLib C28

Back to ARRA main page
Back to Registration main page
Back to Registration Use-case Inventory

v3.6.1 Slicer Registration Library Case 28:
In Vitro cerebral body CT (intra subject)

Input

baseline image		follow-up

Modules

• Slicer 3.6.1 recommended modules: BRAINSFit

Objective / Background

This dataset contains in vitro CT of two human vertibral body segments (L4) before and after regional damage. Goal is to align the two images to study in detail the local changes that occurred.

Keywords

CT, spine cancer, vertebrae

Issues Challenges

• the baseline container contains the vertebra upside down (as apparent from orientation of the pedicle/roots of the vertibral arches coming out of the body, visible in sagittal or coronal view). Hence this image needs to have its z-axis (IS) reversed before registration
• the containers walls and lid contain dominant image content that would prevent a successful automated registration; cropping is required
• the two vertebrae have strong rotational misalignment; manual initial alignment is recommended

Input Data

• fixed: CT , 0.342 x 0.342 x 0.20 mm voxel size, axial; 512 x 512 x 193 unsigned short image
• moving: CT , 0.342 x 0.342 x 0.20 mm voxel size, axial; 512 x 512 x 250 unsigned short image

Procedure / Pipeline

1. Open case scene file or import image data: RegLib_C28_SlicerScene1.mrml
2. Place L4_intact and L4_defect in back- and foreground respectively.
3. Scroll through sagittal view and note the inverted axis orientation
4. Flip axis
   1. Go to the Data module
      1. right-click on the Scene node and select "Insert Transform Node" from the menu
      2. select the newly created transform node, then in the MRML Node Inspector below, rename it to "Xf0_AxialFlip"
      3. select the "L4_intact" node and drag it inside/on top of the "Xf0_AxialFlip" node
      4. ensure you have L4_intact selected for background and the fade slider set on background (B) so you can see changes happening.
   2. Go to the Transforms module
      1. Select "Xf0_AxialFlip" from the Transform Node menu
      2. In the 4x4 matrix displayed, click in the field containing the "1.0000" in row 3 and column 3.
      3. hit Return key to activate editing
      4. change the number to -1. Hit Enter.
      5. you should see the image flip in the sagittal and coronal slice views
5. Crop Volumes
   1. Go to the Crop Volume module
   2. select "L4_intact" as Input Volume; select "Create new MRML ROI Node" in the ROI menu, select "create new volume" in the Output Volume" menu
   3. you should see a small blue rectangle in all 3 views. Click in the views to expand the region to contain only the L4 vertebra (body and processes)
   4. when finished, click on Do ROI resample
   5. go to back to the Data module and change the name of the new "SubVolume" to "L4_intact_crop"
   6. delete the ROI nodes (right click: delete node)
   7. repeat the same for the "L4_defect" image
6. Manual Initial Alignment
   1. in the Data module again select "Insert new Transform node" as done above
   2. select the newly created transform node, then in the MRML Node Inspector below, rename it to "Xf1_ManualInit" or similar
   3. select the "L4_defect_crop" node and drag it inside/on top of the "Xf1_ManualInit" node
   4. ensure you have L4_intact_crop selected for background and L4_defect_crop for background
   5. set the fade slider to halfway between background (B) and foreground (F) so you can see changes happening.
   6. Go to the Transforms module
   7. Select "Xf1_ManualInit" from the Transform Node menu
   8. adjust the sliders to roughly align the two images
   9. for more details on manual registration, see the "Manual Registration" tutorial here: [1]
7. Run final registration
   1. Go to the BRAINSfit module
      1. 'Presets: select the "BRAINSFit_Affine" from the parameter set menu. See here for a guide on how to Load/Save Registration Parameter Presets]]
      2. otherwise select the following parameters:
      3. Fixed Image Volume: :L4_intact_crop ; Moving: L4_defect_crop
      4. Initialize with previously generated transform: select "Xf1_ManualInit" generated above
      5. check boxes for Include Rigd registr. phase , Include ScaleVersor3D, include Affine
      6. Slicer Linear Transform: select "create new transform", rename to "Xf1_Affine" or similar
      7. leave rest at defaults. Click Apply
      8. registration should take ~ 10 secs.
      9. use fade slider to verify alignment; compare with result snapshots shown below.

1. 1. Input Surface: S02_femur
   2. Target Surface: S01_femur
   3. Output Transform: create new transform, rename to "Xf1_surf_S2-S1"
   4. Select/check Mean distance mode: RMS
   5. Leave defaults: iterations: 50, landmarks: 200
   6. Select/check Start by matching centroids
   7. Select/check Check mean distance
   8. Click: Apply; process duration: ~ seconds
2. Go to the Data module
   1. find the newly created transform node "Xf1_surf_S2-S1" and drag the "S02_femur" model node (and all other S02 entities) inside the transform.
   2. for the 3D viewer turn on the visibility of the surface models (should be on already)
   3. the surfaces for the two femur models should now align in basic position and orientation
3. for additional details see tutorial downloads below

Registration Results

Download

• Data
  • RegLib_C05_Data: Example Dataset (MRI images, surface models,Presets, Solution, zip file 9.5 MB)
• Presets
  • RegLib_C05_Presets.mrml: registration parameter presets (.mrml text file 12 kB)
• Link to User Guide: How to Load/Save Registration Parameter Presets
• Documentation
  • RegLib_C05_Tutorial: Powerpoint file (.ppt file 1 MB)
  • RegLib_C05_Tutorial: PDF file (.pdf file 0.5 MB)