Hi,

When using flirt for downsampling, it does apply a blurring which can be switched off by the option -noresampblur. Anyway, I guess resampling data requires some kind of interpolation, and when resampling binary data you have to apply a threshold and binarize the mask again. If you just binarize the blurred unthresholded mask, you likely end up with a slightly dilated mask.

For your kind of downsampling you might look into the fslmaths option -subsamp2. Since you apply it on a mask, it does not matter, in which way the data of the old voxel is averaged into the lower resolution voxel.

good luck,
wolf

On 02/23/2013 03:02 AM, Tseng Mark wrote:

Thank you Paul.

My mask_1mm is a T1 structure image and I want to use it to mask a fmri map, which is in 2x2x2 dimension. If I want to transform mask_1 into the space of the fmri map, I need the transformation matrix produced during the preprocessing of my fmri data. Right?

The problem is that the matrix data are not with me now....that's why I tried using "flirt -applyisoxfm" but unsure if the problem I raised is "normal" or not.....

Mark

2013/2/23 Chou Paul <[log in to unmask]>

Dear Mark

Have you ever tried the "ApplyXFM" with identity transformation to resample your "mask_1mm" image to "mask_2mm" image ? I think this tool may do the job for you. 

Best

Paul

> Date: Sat, 23 Feb 2013 00:45:57 +0000
> From: [log in to unmask]
> Subject: [FSL] changing voxel size in a mask
> To: [log in to unmask]

>
> Hi,
>
> I have a mask (mask_1mm) with voxel size of 1x1x1mm. I want to use it to mask a 2x2x2mm map so I use below command to change it into 2x2x2 (mask_2mm):
>
> flirt -in mask_1mm -ref /usr/local/fsl/data/standard/MNI152_T1_2mm_brain -applyisoxfm 2 -out mask_2mm
>
> Then I check the voxel number/volume of the 2 masks:
>
> mask_1mm.nii.gz
> 1226669 / 1226669
>
> mask_2mm.nii.gz
> 181444 / 1451552
>
> I found the voxel size in mask_2mm is not 1/8 that in mask_1mm, and the volume of the two masks are different.
>
> Is it normal?
>
> Thanks.
>
> Mark