3D registration - longitudinal and across subjects
3D registration of craniomaxillofacial images can be performed with voxel based, surface based and landmark based procedures for registration relative to any anatomic region of interest : cranial base or regional maxillary or mandibular superimpositions.
The surface and landmark based procedures are work in progress to be added to the CMFreg extension for the Slicer software by November 2013.
The steps for voxel based registration are described below:
Step by step procedures for Voxel Based Registration using CMFreg for Slicer 4.3 nightly build
All procedures for voxel based registration have now been incorporated as a modules in the CMFreg extension for Slicer software.
Step 1 – Construct surface models of the cranial base for T1 and T2 images on ITK-snap
CBCT T1 and cranial base segmentation constructed on ITK-snap
CBCT T2 and cranial base segmentation constructed on ITK-snap
Step 2 - Perform cranial base registration for T1 and T2 images using the growing or non growing modules of CMF reg in Slicer depending on whether the patient is growing (child) or non-growing (adult).
Step 2a - Open Slicer software and load the data (CBCTs and Segmentations)
Open Slicer and click on 'DATA' on the upper left.
A new window appears, click in 'Choose File(s) to Add'. Another new window appears allowing you to search for your files. After finding them, select them and click in 'Open'.
A new window appears showing the files that you loaded. Double check if they are the right files and make sure to center them by clicking at the centered checkbox option, this will help to reduce computing time. To finish click in 'OK'.
Step 2b - Select the loaded files into the CMF registration - Growing or NonGrowing module.
After you download the CMFreg extension for Slicer, you can find the CMF Registration module by clicking at the toolbar which is localized at the right side of the loupe. And inside the CMF Registration you can choose Growing or NonGrowing Registration depending of your cases.
The first options to be set are the 'Input Files' where you can see the 'Baseline Scan (fixed)', 'Baseline Segmentation (fixed)', 'Follow up Scan (moving)' and 'Follow up Segmentation (moving)'.
Baseline Scan (fixed) should be your CBCT Time 1
Baseline Segmentation (fixed) should be the single label skull-base segmentation that you created on ITK-snap.
Follow up Scan (moving) should be your CBCT Time 2
Follow up Segmentation (moving) should be the single label skull-base segmentation that you created on ITK-snap.
Step 2c - Select the 'Voxel Base Registration Options':
The software allows 3 possible options: 'Rigid+Scale (7DOF)', 'Affine (12DOF)' or 'Rigid+Scale+Skew (10DOF)'.
These 3 options are the different ways that the software can register the CBCT's taking into account differences in shape and size. For most of the longitudinal scans with 1-3 years follow up the 'Rigid+Scale (7DOF)' with 1000 iterations (default number) works properly. If the registration is not adequate as verified by overlay in Slicer or ITK-SNAP don't, you can try to select one of the other two options. If the patient has had remarkable growth and/or treatment changes (in regional registration areas), the Affine (12DOF) or Rigid +Scale+Skew may help improve the quality of he voxel base fit. For any of these 3 options, even though differences in scale and shape may be taken into account to compute the registration parameters, after computing the voxel best fit parameters, only the registration matrix equivalent to a Rigid registration ( x,y,z parameters of translation and rotation) will be applied.
Step 2d - Select the 'Output Registration Matrix':
The Registration Matrix will be the output text file (.txt), which contains the x,y,z translation and rotation coordinates to register the CBCT and Segmentation files of the 2 time points.
In this tab, click in 'Create new LinearTransform'
Step 2e - Select the 'Apply Registration Matrix':
In this tab, we will select the CBCT and segmentation to apply the registration matrix:
'Input Seg to be Registered' is any segmentation (single or multiple color-labels), which the Registration Matrix will be applied to. It means that it should be a segmentation constructed from the CBCT Time 2.
'Output Registered Seg' is the same segmentation selected as the 'Input Seg to be Registered', that now will be renamed and have the Registration Matrix applied onto. For this option select 'Create new Volume" and rename file with -reg suffix.
'Output Registered Scan' is the CBCT Time 2 that will now have the Registration Matrix applied to. For this option select 'Create new Volume' and rename file with -reg suffix.
Step 2f -After selecting all options in steps 2a to 2e, click in 'Apply' and wait for the status to be Completed. This step will compute the Registration MAtrix and apply it to to the CBCT and segmentation at Time 2.
Double check if the Registration worked properly: compare and scroll through all semitransparent slices of 'CBCT T1'(selected at the 'Baseline Scan (fixed') and 'Output Registered Scan' (selected as 'Output Registered Scan').
This is a example of two CBCT's being compared before and after a Growing Registration using CMF module.
Observation: Slicer do not save files automatically. To have the files saved on the computer click in 'File' on the upper left corner and then in 'Save'. A new window will appear allowing the user to modify the file name, format and directory.
For more information about ITK-snap.
For more information about 3D Slicer
For more information about the CMFreg documentation.
For access the CMFreg source code.