Region of Interest (ROI) Analyses
The main purpose of a mask is to only show you blobs (potentially) in areas that a priori you expect to see. The second great thing masks do is increase the likelihood that you are going to get results in your area of interest. The third great thing masks do is speed processing time, by radically reducing the number of voxels that must be analyzed. EG, its like doing FSL on squirrel brains. At the statistical level a mask reduces the number of independent comparisons you are correcting for. EG, if the brain has 100,000 voxels, a Bonferoni correction requires you to essentially divide your signal by 100,000. This reduces signal intensity. But if you have a mask of a region with 10,000 voxels, that exact same signal will become more powerful and you will ‘see’ it. This is the nature of zooming in using a mask.
FSL does not currently (as of 12.2006) offer a library of ROI masks at the group level, although the email list should be checked periodically about this topic as they are planning to produce one.
In contrast, SPM offers a library of masks that can be accessed by opening matlab and typing wfu_pickatlas [enter]. Many masks are available here, in axial slices only.
Individual versus Group Masks
Individual versus Group Masks
A deeper problem is that not everyone’s brain is the same size, and some are inflated and warped during standardization, and as a result a person’s amygdala may not fall completely within the group amygdala mask that you draw on the MNI brain. The solution is time consuming: draw an individual mask for each subject! At the group level the solution is to draw somewhat generous mask, but this is a poor solution as you will end up with activations that LOOK like they aren’t on the amygdala even if they are.
For obvious reasons, individual masks are better than group masks; but they take longer, and so it may be in your interest to do a quick mask of your group results before engaging in the 2h/subject work it takes to create individual masks.
Creating a Mask
1. Open FSL View
2. Click on file… Create Mask or just open a blank layer (eg, add a blank mask layer from your ‘masks’ folder’. It is good to keep a blank one always handy, and to rename it first thing before doing any drawing. Note that the name will not change in FSL view, so you may want to close it and reopen.)
3. Make sure it is unlocked. (When you load even the blank mask, it will be locked at first.) this is done by clicking the little symbol that looks like a house, or a lock (depending on your view of things) at the middle bottom of the FSL View GUI, next to the eye and i symbols.
4. Click on the Pencil tool in an otherwise blank layer. This will open up a variety of options. From left to right, the pencil (for drawing on pixels) the eraser (for erasing them) the filler (to fill within a shape you’ve drawn with the line tool) the back button, and finally the value of the mask (set this to 0 to erase your mask on eraser, and to be black on pen) and then the number of voxels of your pen width.
5. Change the color to yellow or some other very obvious color by clicking on the color box and using the i symbol in the bottom left of the FSL view It is probably best not to draw your mask in greyscale.
6. Depending on your anatomical structure it will be easiest to draw in axial, saggital or coronal view. For example the ACC is best drawn in sagital; the amgdala in coronal, the brainstem in coronal.
7. Beginning at the bottom layer and working up one layer at a time (using the x,y,z coordinates in the lower left hand corner to advance) create a mask over the region of interest.
8. It is often easiest to draw a single-voxel ring around the structure with the pen tool and then fill it with the can tool.
9. Overdraw slightly in your initial pass. Then review your work from another perspective (eg coronal and then axial if you drew in sagittal) until you are happy with the result.
10. SAVE the resultant mask in the ‘Masks’ Folder in your directory
Applying a Mask in FSL-FEAT
1. Open a FEAT design.fsf that you applied to the whole-brain, or if one is not yet done, review that chapter. The mask is an add-on.
2. Open post-stats tab
3. Select ‘Pre-Threshold masking’ and navigate to the mask.hdr.gz that you wish to apply
4. Select voxel thresholding or uncorrected. There are strong arguments to be made that small, coherent brain regions (eg the amygdala) are resels and therefore do not need to be voxel corrected. There are also arguments for resel-correction, which isn’t covered here. Voxel-correction thresholding is conservative.
5. Select the p threshold (.05 or more rigorous)
6. Rename the file to indicate that a mask is being used
BOLD Response data for the mask region
1. Open the Mask
2. To get individual BOLD response data for the voxels in the mask area, click on Utils -> get featstat from ROI mask or coordinates
By clicking all the buttons here, a ‘feats-stats query’ html… will show up in the feat folder… [show what to do with it]