MEDx Frequently Asked Questions - Registration

Q: Top 8 Things That Can Cause Problems When Using AIR

1. Accurate Mensuration Units.

Make sure the X, Y, and Z voxel sizes are set correctly for both the Standard and Reslice images. AIR uses the voxel sizes to determine the relative scale between the two volumes being registered.

2. Use thresholds that define "brain".

The thresholds determine what is "brain" or more generally what the objects are in each image that should be registered. Everything below the threshold is ignored. Too low of threshold will include too much and one that is too high will not include enough. Use the pixel reporting mode to find a threshold that is appropriate for your data. Also, you can use Thresholding (under the Segmentation menu) and Undo/Redo (under the Image menu) to try different thresholds. The goal is to find thresholds that define "brain" and roughly the same parts in both data sets.

3. Eliminate the scalp.

If one of the images includes scalp and the other doesn't then you will need to remove the scalp. To do this you can use the "Interactive Segmentation" under the Segmentation menu for most images. In general, you should always remove the scalp and other non-brain strucutes (e.g. nose and eyes). This is especially true when using the Warp Algorithm. As Roger Woods notes: "Even if the algorithm does run successfully, you will have invested a lot of computation time in making sure that your subjects' noses are of similar size and shape, even if this means that their cerebellums don't line up so well."

4. Anatomically Orient data acquired in different orientations.

If you are registering images that were acquired in different orientations, such as Axial and Coronal, then you need to properly orient both images in the Orthogonal Viewer (under the Display menu). When the data is anatomically oriented correctly the anatomic labels (A, P, S, I, etc.) are next to the correct part of the anatomy in the image. See Chapter 7 for details on how to perform this operation.

5. Choose the correct Registration Type.

If you are using the Linear Algorithm, it is important to choose the correct Registration Type. In particular, if you are registering MRI's acquired with different pulse sequences then you should choose "Intramodality: MRI T1/T2" rather than simply "Intramodality."

6. Choose the appropriate Cost Function.

If you are using the Linear Algorithm with EPI data then it is recommended that you use the "Least Squares with Intensity Scaling" Cost Function found under the Algorithm Parameters part of the dialog box.

7. Choose the appropriate Initial Model.

If you are using the Warp Algorithm, it is important to choose an initial model that is likely to succeed. Typically, this means choosing "First Order Linear (12 parameter model)."

8. Edit the data such that they cover the same parts of the brain.

This is generally less of a problem with the Linear Algorithm than the Warp Algorithm. Roger notes that with the Warp Algorithm, "Missing data due to a limited field of view can lead to unexpected or even bizarre results with high order warping. If you are dealing with a restricted field of view, you should probably stay with second or third order nonlinear models (or at least carefully inspect the results obtained with higher order models." Also, you can edit the data such that they cover the parts of the brain.


Other FAQ Topics

If you have a question that you would like to see addressed in our list of Frequently Asked Questions, please contact customer support.

Back to FAQ Home