Thanks Rosalia.

Gavin

From: FSL - FMRIB's Software Library <[log in to unmask]> on behalf of Rosalia Dacosta Aguayo <[log in to unmask]>
Reply-To: FSL - FMRIB's Software Library <[log in to unmask]>
Date: Friday, October 20, 2017 at 12:42 AM
To: "[log in to unmask]" <[log in to unmask]>
Subject: Re: [FSL] Registering Stroke Brains to Standard Space

Hi Gavin,

This question was answered by Mark Jenkinson a few days ago. See below:

El 17 oct. 2017 14:05, "Mark Jenkinson" <[log in to unmask]> escribió:

Hi,

Working with lesions can be challenging but this is our recommended pipeline, which normally works:
1) Run fslreorient2std on the image (make sure you give it both an input and an output image name).
2) Use robustfov to remove the neck and crop the image.
3) Create a mask of the lesion (with a value of 1 inside the lesion and 0 outside).  This normally has to be done by hand.
4) Run BET to get a brain extracted image.  Here you might need to try lots of options to get a good result. Try the -R and -c options in particular.  Also, if it is always failing in the area of the lesion then try adding a scaled version of the lesion mask to the brain image in order to make a (temporary) fake image for brain extraction purposes where the lesion is effectively filled in.  The scaling value that you need should be equal to a value that is representative of the grey matter or white matter in your image.  You can do the scaling and addition with fslmaths like this:
        fslmaths lesion_mask -mul scaling_value -add orig_image filled_image
where scaling_value is the GM or WM intensity value and the other names are filenames of your images.  The try running brain extraction on "filled_image" and see if you can get that to work.  If you can then you can make the brain mask from this (use the -m option in bet to get the brain mask) and then apply this brain mask to your original image like this:
        fslmaths orig_image -m brain_mask extracted_image
5) Do a linear registration with FLIRT between the brain extracted image and the MNI152_T1_2mm_brain
6) Do a nonlinear registration with FNIRT between the non-brain-extracted image and the MNI152_T1_2mm, using the FLIRT registration matrix as the input for the --aff option and using the T1_2_MNI152_2mm.cnf for the configuration file.

This procedure should hopefully solve your problems.
If not then let us know.

All the best,
        Mark

El 20 oct. 2017 6:26, "Gavin Elias" <[log in to unmask]> escribió:

Hi,

I'm interested in registering structural T1 scans of stroke patients to MNI space for the purposes of lesion segmentation and exploration of resting state functional networks.  Some of these brains have small focal lesions, while others have rather large lesions (please see attached images for examples).

I'm wondering what - if any - pipeline you might recommend for registering these kinds of brains to standard space.

Many thanks,

Gavin