Hi Steve,

Thanks. When we register a mask into the standard space. What threshold do we always use? 0.5? Or I just need to keep the similar volume? For example, the number of voxels is n of the 1*1*1 structure image. After registration into the 2*2*2 standard space, the registered mask should be n/8 voxels?

Mingxia

On Mon, Apr 15, 2013 at 12:27 AM, Stephen Smith <[log in to unmask]> wrote:

Hi - the general interpolation/thresholding issues would be the same for linear and nonlinear registration.

Steve.

On 15 Apr 2013, at 08:23, zhang mingxia <[log in to unmask]> wrote:

Thanks for all you nice answers! I was thinking about it again. Because the structure image would be register to the standard space. I think I should use nonlieaner registration. Because I had the highres2standard_warp for every subject from the preprocessing. I think I should use applywarp:

applywarp -r $FSLDIR/data/standard/MNI152_T1_2mm_brain -i binary_mask.nii.gz -o std_mask.nii.gz -w highres2standard_warp

There was the same issue. In this situation, should I use --interp=nn or the default one? If I use the default one, and then what threshold should I use? My understand is the threshold of .5 will come the similar result as using the nearest neighbour.

Thanks!

Mingxia

On Fri, Apr 12, 2013 at 4:02 PM, Mark Jenkinson <[log in to unmask]> wrote:

Hi,

Alternatively, we prefer to recommend the use of trilinear interpolation and then thresholding the result.  This gives more flexibility in terms of deciding whether you need the result to be conservative/exclusive/inclusive/size-preserving/etc.

See the registration part of the FSL course for more information on this.

All the best,
        Mark

On 11 Apr 2013, at 18:45, John Herrington <[log in to unmask]> wrote:

> Hi Mingxia - the default interpolation method for flirt will combine signal from adjacent volumes, including those with zero values.  Use nearest neighbour interpolation to avoid this, like so:
>
>
> flirt -in binary_mask.nii.gz -ref $FSLDIR/data/standard/MNI152_T1_2mm_brain -output std_mask.nii.gz -interp nearestneighbour
>
> Cheers - John
>
> ----- Original Message -----
> From: "zhang mingxia" <[log in to unmask]>
> To: [log in to unmask]
> Sent: Thursday, April 11, 2013 1:30:58 PM
> Subject: [FSL] binary mask is "flirt" into a non-binary image?
>
>
> Dear FSL experts,
>
>
> I am analyzing the resting state data. As previous studies did, I wanted to get the timecourse of WM and CSF as the nuisance signal. I used fast to segment 1*1*1 structure image and got the binary mask of WM and CSF (also some questions I mention in last letter, hoping to get your reply).
>
>
> Because my resting state data is 2*2*2, I used flirt for the registration:
> flirt -in binary_mask.nii.gz -ref $FSLDIR/data/standard/MNI152_T1_2mm_brain -output std_mask.nii.gz.
>
>
> But I found the std_mask.nii.gz was not binary. So I can not use it as the mask to extract timecourse. I could not understand it and checked. Yes, binary_mask.nii.gz is a binary image and std_mask.nii.gz is not. What should I do?
>
>
> I have an extra question: should I do the registration on the structure images firstly and then use fast to segment them?
>
>
>
> Thanks!
>
>
> Mingxia Zhang

---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director,  Oxford University FMRIB Centre

FMRIB, JR Hospital, Headington, Oxford  OX3 9DU, UK
+44 (0) 1865 222726  (fax 222717)  2176
[log in to unmask]    http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------