Spatial Normalization

(1) Open up a new MEDx session, and then open up a New Folder.

(2) In the New Folder, select Image --> Open Image. In the Open Image dialog box, in the Filter entry box, type in the tutorial directory, followed by Ss*.hdr, e.g.

and then click on the Filter button. This will cause a list of the 84 spatially smoothed, temporally filtered, motion-corrected image volumes to be made in the Files: pane. Click on the first one, then (leaving the first one selected) scroll down to the last one and press the Shift key. While pressing on the Shift key, click on the last SsTfMc image volume, which should be #84. This should cause all 84 SsTfMc image volumes to be selected.

(3) In the Open Image dialog box, click on the Apply button. This will load the 84 spatially smoothed, temporally filtered, motion-corrected scans into MEDx. Note the name of the Group page; it should be New Group.

(4) Also load the Mean image that you created in step 14 of Make a Mean Image Volume and Mask. In the Filter entry box, type in the tutorial directory, followed by M*.hdr, e.g.

and then click on the Filter button. Select the Mean.hdr file, and then click on the Apply button.

(5) Finally, load in the MNI EPI template. In the Filter entry box, type in

and then click on the Filter button. Select the EPI.16SI.tif file, and then click on the Open button. Note the name of the image volume shown along the top of the folder.

(6) Select Toolbox --> Registration --> AIR. In the Automated Image Registration (v3.08) dialog box, set Standard Image to the MNI EPI template, and Reslice Group or Image to the Mean.hdr image. We will compute the spatial transform from the Mean image volume, then turn around and apply it to the group of 84 spatially smoothed, temporally filtered, motion-corrected scans (which are themselves aligned with and therefore in the same space as the Mean image volume).

(7) OK, I'm putting you to the test here. Find the maximum value of the MNI EPI template, and of the Mean.hdr image. Calculate what 15% of these values are, rounding to the nearest integer if necessary. Under Thresholds, use the 15% max value computed from the MNI EPI template for the Standard Image value, and 15% max value computed from the Mean image volume for the Reslice Image value.

(8) Under Algorithm and Model, set Algorithm to Warp. Leave Initial Model, Final Model, and Interpolation set to their default values. Click on the OK button. This will compute the spatial transformation needed to warp the Mean image volume into the MNI EPI template space.

(9) Let's save the spatial transformation into a file, so that we can then turn around and apply it to the 84 scans, so that they will be effectively "shadowing" or following along with the Mean image volume. Select Toolbox --> Transformation --> Shadow Transform --> Save.... In the Save Shadow Transform dialog box, in the Filter: entry box, type in the tutorial directory, e.g.

Click on the Filter button (to make sure that MEDx is now "looking at" the tutorial directory), and under Selection type tal.IR1.xform. Then click on the Save button. The spatial transform has now been saved into the tutorial directory.

(10) Now apply the spatial transform to the 84 scans. Use the Page Manager to go to the group page for the 84 scans (it should be named New Group). Then select Toolbox --> Transformation --> Shadow Transform --> Apply....

In the Apply Shadow Transform dialog box, set Transform to AlignWarpReslice. Don't select Use For Pixel Reporting.

In the Filter: entry box, type in the tutorial directory, followed by *.xform e.g.

Under Files:, the tal.IR1.xform created in the previous step should appear. Click on this entry. The full path to this file should now appear under Selection:. Click on the Ok button. The spatial transform will now be applied to the 84 image volumes. A new group of 84 image volumes named Results: ShadowTransform of "New Group" should now appear.

(11) The MEDx folder is getting crowded again. Use the SaveNSetAVW.tcl script to write out the spatially smoothed data to the tutorial directory. Set the Root Name of these images to

to indicate that the images have now been spatially normalized, spatially smoothed, temporally filtered, and motion corrected. This time set X Scale, Y Scale, and Z Scale all to 2, which is the same as the MNI EPI template. Watch the images as they are displayed during the disk write, to make sure that they look "okay".

(12) Make an average volume and mask of the spatially normalized images. (If you don't remember how to do this, go back to Make a Mean Image Volume and Mask.) Save these two image volumes out to disk in AVW format as files named SnMean.hdr and SnMask.hdr. Then exit MEDx.

The next step is to do an actual statistical analysis. I'll show you how to do the analysis in two ways. The easiest way is a simple t-test. A more sophisticated way is to use the Multiple Regression module in MEDx, which takes into account spatial autocorrelations (if you don't take these into account, your p-values may be artificially reduced, resulting in false positives).

Your choice, Neo:

Analysis 1: Simple t-test (blue pill)

Analysis 2: Multiple Regression (red pill)

For more information on Spatial Normalization using AIR, see section 19.5 of Chapter 19 of the MEDx User's Guide; see section 19.7 in the same chapter for more information on Shadow Transform. This tutorial demonstrates the use of Shadow Transform with AIR, although in the context of cross-modality within-subject image alignment, rather than spatial normalization. (You might want to try this to display single-subject functional activations on top of that subject's structural MRI.)

For an alternative method of spatial normalization using SPM96, see section 32.4 of Chapter 32, as well as this tutorial.

Return to tutorial main page