It sounds like what you are doing is fine, though personally I would
probably check to see how well the FA is registered to the T1 rather than
the T1 to the FA.  It sounds like the registration is failing for this
particular subject.  You could try performing it again to see if it still
fails.  Are the brain extractions of both the b0 and the T1 of good quality?
You might need to reduce the freedom that FLIRT has to make angular
transformations.  I'm not sure how to do that through the FDT GUI, but you
can do it on the flirt commandline by specifying:

-searchrx <min_angle> <max_angle>  (angles in degrees: default is -90 90)
-searchry <min_angle> <max_angle>  (angles in degrees: default is -90 90)
-searchrz <min_angle> <max_angle>  (angles in degrees: default is -90 90)

I don't personally use the GUIs, but they can be good if the default
pipelines are working for you and you don't want to fool with the
commandline or scripting.

Peace,

Matt.
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of Lindgren, Kristen, Ann
Sent: Sunday, February 01, 2009 2:31 PM
To: [log in to unmask]
Subject: Re: [FSL] FDT Registration error

I need to generate transforms because my seeding points for probtrackx
are in structural space.  

I'm using the FDT Registration tool to generate str2diff.mat and
diff2str.mat using the input/options mentioned in my previous e-mail.
After I run the registration tool, I check the transform by applying
str2diff.mat to my structural image with ApplyXFM_gui and then comparing
the realigned structural with nodif_brain to make sure it matches.

This has been working fine for all of the people I've put through so
far, but if you look at my original e-mail you'll see the issue I came
across with my latest subject.  

Is this not the way to register DTI and structural images?  Based on the
FDT website I thought this was the recommended way and I thought I'd
seen e-mails saying it's better to use the registration tool within FDT
rather than FLIRT.  

I guess I'm just looking for thoughts on why my subject's
transformations do not seem to be calculating correctly and if this is
the way I should be doing this.  Thanks!

Kristen

-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
Behalf Of Matt Glasser
Sent: Sunday, February 01, 2009 3:22 PM
To: [log in to unmask]
Subject: Re: [FSL] FDT Registration error

Why would you want to do that?  How are you generating str2diff.mat?

Peace,

Matt.

-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
Behalf
Of Lindgren, Kristen, Ann
Sent: Sunday, February 01, 2009 12:44 PM
To: [log in to unmask]
Subject: Re: [FSL] FDT Registration error

Sorry, I mistyped.  I'm applying the transformation to the structural
image
to realign that to DTI space, so that's why I'm using str2diff.mat.  

I've been doing the same steps for all of my subjects (over 60 so far)
and
this is the first one that's given me a problem.  Here's what I'm doing:

1) Select Registration from the FDT pulldown menu
2) Enter subject's bedpostx directory
3) Check "Main structural image" and uncheck "Standard Space"
4) I've been using the defaults - Normal search, 12 DOF, Mutual
Information

Any thoughts?  Thanks for your help!

Kristen


-----Original Message-----
From: FSL - FMRIB's Software Library on behalf of Markus Gschwind
Sent: Sun 2/1/2009 1:20 PM
To: [log in to unmask]
Subject: Re: [FSL] FDT Registration error
 
Hi!

For the DTI_to_T1 registration you best use the dti_FA and not the b0,
it
will give better results (the contrasts are more similar between FA and
T1).

But another thing:
You did DTI_to_structural but the mat file's name is str2diff (= the
inverse). Are you sure that this is correct?
How many deg of freedom did you use?
If it is the same subject's brain, you would take 6. If you do
T1_to_MNItemplate, you would use 12.

All the best,
Markus

2009/2/1 Lindgren, Kristen, Ann <[log in to unmask]>

> I just tried using FDT to register a subject's DTI data to their
structural
> MRI scan (already skull stripped in freesurfer).  I've been checking
my
> alignment using ApplyXFM_gui, and when I applied str2diff.mat to this
> subject's MRI, the output was very odd.  Basically one view looked
like a
> twisted brain (parts next to each other that shouldn't be) and the
other
two
> looked like scans of phantoms.  I tried rerunning registration and
ApplyXFM
> and I got the same results.  I don't see any issues with the
nodif_brain
or
> the MRI scan, so I'm unsure what's going on.  For FDT registration,
does
it
> use anything else besides the nodif_brain?  Do you think my bedpostx
data
> could be corrupted and I need to rerun that?  Thanks for your help!
>
> Kristen
>



-- 
Dr. med. Markus Gschwind, M.D.
Laboratory for Neurology and Imaging of Cognition
Dept of Neurosciences
University Medical Center (CMU)
1 Michel-Servet - 1211 GENEVA - CH

Tel 0041 (0) 22 379 5324
Fax 0041 (0) 22 379 5402
email: [log in to unmask]
http://labnic.unige.ch