Print

Print
Mark,

Thank you, as always!  Great ideas.

My goal:  I have run *run first* to extract the hippocampus from a non-betted MPRAGE.  I want to extract percent signal change from the segmented hippocampus. 

Featquery requires the mask to be in standard-space, highres-space or lowres-space.  (Note that the segmented hippocampus is in the following space: 160x192x144 voxels with 1.3x1.3x1.3mm resolution.)  According to the featquery website:

"Featquery will automatically detect which space this mask is in (standard-space, highres-space or lowres-space) and will transform it into the native lowres space of example_func; of course this can only work if FEAT registration was setup and carried out."

I understand that only the non-betted whole brain MPRAGE can be registered to standard space -- before the hippocampus is extracted.  

Unfortunately, I do not see the saved registration file (i.e., the *to_std* file) outputted as a result of the run_first command (i.e., the *to_std* image).  In addition, I would prefer not to run the *run_first_all* command, given a few registration issues when I tried to run this command before.

Should I run *run_first* from scratch for the hippocampus after I register each subject's non-betted MPRAGE to the MNI_T1_2mm_brain?  

What is the most efficient way to do achieve my goal?  I would really appreciate your suggestions about the best way to proceed.

Thank you very much - I apologize for the delayed follow-up.

On Mon, Nov 30, 2009 at 2:24 AM, Mark Jenkinson <[log in to unmask]> wrote:
Hello,


1.  To follow-up on "spaces," how does one know the appropriate display range in FSLview for the different images?

Look at the intensities that are contained in the image.  Either click
around different voxels in FSLView and note the values in the
"intensity" box (near the coordinate displays) or use a tool like
"fslstats" with the -r (or -R) option.  When you have a feeling for
the range of intensity values, just set the display range (min and max)
in FSLView to cover this range (and not much more).



2.  To follow up on the registration --

       a.  Is it possible to register a FIRST hippcampus (already segmented & boundary corrected) to standard space?
       e.g.,

       flirt -in [FIRST hippocampus; segmented & boundary corrected] -ref [MNI152_T1_2mm] -out [name] ?

You should not do this.  You should only register images that look
like each other (e.g. the image of a whole brain and another image
of a whole brain, not of one isolated structure).  However, if you have
run FIRST then the registration has already been done and saved
in a file called something like *_to_std_sub.mat
You can *apply* this transformation (the result of a registration) to
any image in the original space using the -applyxfm flag in FLIRT.


       What is the appropriate display range for this image?

For a labeled image it will depend on the structure with the maximum
label number.  Typically if you set the range to be 0 to 40 then you
should see things fine.  But use the strategy I describe above to know
for sure.


       b.  Or must one perform the flirt registration to standard space from the raw structural image with no brain extraction?  And then use FIRST to segment the desired hippocampus?

If you use run_first_all then this is all done for you.  And yes, it is recommended
to use non-brain-extracted images with this.

All the best,
       Mark





Thank you.  Very much.


On Sat, Nov 28, 2009 at 3:35 PM, Mark Jenkinson <[log in to unmask]> wrote:
Hello,

There is no real reference about "spaces" besides the documentation.
I'll try to answer your questions:

1 - Outputs are in different spaces as the originally acquired images
      are in different spaces.  That is, the Field Of View (FOV) and
      resolution are different for the functional, structural and
      standard/template images.  We try to keep things in the most
      appropriate space, partly because moving between spaces
      involves registration and resampling steps which can be inaccurate
      and the interpolation (in the resampling) will degrade the image
      quality by some amount.

2 - As examples:
      Functional images might be 64x64x40 voxels with 3x3x4mm resolution
      Structural images might be 256x256x200 voxels with 1x1x1mm resolution
      Our standard space template images are 91x109x91 voxels with 2x2x2mm resolution

3 - The spaces correspond to your different acquisition images - so just check
      by comparison with your original images.

For your registration I'm not sure what your [FIRST INPUT] image is.  Is it a
structural image with no brain extraction?  If so, then you should not register
to the avg152T1_brain but use the non-brain extracted version.  Also, we
recommend sticking with the MNI152 naming, so MNI152_T1_2mm would
be the appropriate image in this case.  I am not sure what you mean by
"appropriate brain region".  Also, check that the display range is set appropriately
in FSLView, as having this range set badly (which sometimes happens
with the default settings) could make the whole image look white.

All the best,
      Mark



On 28 Nov 2009, at 17:10, ACE . wrote:

Hello.

Could you please point me to a resource about different "brain space" options? (e.g., high resolution space; standard space; native space?)

I have read the registration information on the FSL website, but I have additional questions such as the following:

1.  Why are different FSL outputs in different spaces?

2.  Could you give me an example of a FSL image in each "space?"

3.  What is the best way to determine an image's "space?"  Is there a list of which voxel dimensions signify which "space?"

To register a FIRST brain area (with boundary correction) to standard space -- I tried the following command:

 flirt -in [FIRST INPUT] -ref avg152T1_brain -out [name]

But I do not see the appropriate brain region in FSLview when I open the output.  I see "white" throughout the entire brain.  Any thoughts?

THANK YOU.