Hi Stefano,
Probtrackx takes all of the standard space data (ROIs, surfaces, etc.) and
warps them linearly or nonlinearly to diffusion space. Then the streamline
counts are warpped back to standard space. Therefore the diffusion data
never moves. Things like ROIs are undoubtedly interpolated with nearest
neighbor.
You could register the FA nonlinearly to the TBSS nonlinear FA template that
is also in MNI152 space.
Peace,
Matt.
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of Marenco, Stefano (NIH/NIMH) [E]
Sent: Thursday, August 11, 2011 2:26 PM
To: [log in to unmask]
Subject: Re: [FSL] probtrackx registration
Hi Saad, thanks for the first suggestion, stupid of me not to think about
this.
I am not sure I understand the second point, though. What does probtrackx
actually do? The probabilities are saved for each voxel in the diffusion
space of each individual by bedpost. By normalizing to a template, don't we
run the same problem as with the normalization of raw acquisitions, where
the bmatrix would have to be adjusted in order to take into account
rotation, shear and interpolation effects on the directionality of the
vectors?
A related query is the following. All our DTI scans have been corrected for
epi distortion with fieldmaps. However, for some we are missing high
resolution structurals. What would be the best way to normalize to the MNI
template? Use the undistorted b0, the FA image or the magnitude images from
the fieldmaps we acquired (they have the same 2 mm isotropic resolution than
the DTI scans)? And what template combination would be recommended? For
example, do you know whether a T2MNI to b0 registration works better than a
T1MNI to FA or field-map magnitude image?
Thanks for your help, Stefano Marenco
-----Original Message-----
From: Saad Jbabdi [mailto:[log in to unmask]]
Sent: Wednesday, August 10, 2011 4:16 AM
To: [log in to unmask]
Subject: Re: [FSL] probtrackx registration
Hi
> When running probtrackx using classification targets in standard (MNI)
space, is it possible to get the results in diffusion space? In previous
runs, I've chosen the 'seed space is not diffusion' option and supplied the
'standard2diff' linear transformation matrix, and have gotten results in
standard space (the original space of the ROIs). However, I would like to
visually verify the results by examining the fdt paths for each subject in
fslview. This would require the probtrackx results to be in diffusion
space. If this is not an option, given masks in standard space, how can I
be confident that the registrations from standard to diffusion space worked
adequately for every subject?
You can check the registration using a single image (e.g. FA versus T1
versus MNI152). If you are confident that registration is accurate for all
your subjects, then you can provide the transformation to probtrackx and run
everything in standard space, which we recommend for multi-subjects
comparisons.
> A second issue relates to the following. When registering ROIs from
standard to diffusion space, there are interpolation effects that make the
ROIs not binary any longer and adjacent ROIs can overlap. I assume that
these effects are ignored by protrackx, so if one wanted to make sure that
there was no overlap across regions, it would always be better to transform
the ROIs to individual space and then do some editing before running the
tractography. Could someone please give their experience on this?
You don't need to get your ROIs into diffusion space, just stick to standard
space and make sure registration is as accurate.
Saad.
>
>
> Thank you for your help.
>
> Katie Gambale
>
> Catherine Gambale, B.S.
> Post-Baccalaureate IRTA
> Genes, Cognition, and Psychosis Program
> Clinical Brain Disorders Branch, NIMH
> 10 Center Dr., Bldg. 10, Room 3C205
> Bethesda, MD 20852
> Phone: (301)-496-9672
> Email: [log in to unmask]
>
--
Saad Jbabdi
University of Oxford, FMRIB Centre
JR Hospital, Headington, OX3 9DU, UK
(+44)1865-222466 (fax 717)
www.fmrib.ox.ac.uk/~saad
|