Hi Chris,

Do a search in the list, I think this has been brought up before. I can't remember what was the issue, but it can be fixed easily, without having to run again, with fslcpgeom (make a backup copy of the results before applying).

All the best,

Anderson


--
Anderson M. Winkler
FMRIB / Analysis Group
[ Blog | Twitter ]


On 6 August 2015 at 07:46, C Steward <[log in to unmask]> wrote:
Hello,
I'm using randomise on a 4d skeletonized file which has isotropic 2.5mm pixel dimensions, the voxelwise regressor images (in this case, white matter hyperintensity masks) also have isotropic 2.5mm pixels, but the stats files when running randomise have 1mm pixel dimensions.

I'm wondering if this is a known bug?

Specifically,

fslinfo all_FA_diff_skeletonized.nii
data_type      FLOAT32
dim1           77
dim2           90
dim3           58
dim4           93
datatype       16
pixdim1        2.500000
pixdim2        2.500000
pixdim3        2.500000
pixdim4        1.000000
cal_max        0.0000
cal_min        0.0000
file_type      NIFTI-1+

fslinfo mean_FA_skeleton_mask.nii
data_type      FLOAT32
dim1           77
dim2           90
dim3           58
dim4           1
datatype       16
pixdim1        2.500000
pixdim2        2.500000
pixdim3        2.500000
pixdim4        1.000000
cal_max        0.0000
cal_min        0.0000
file_type      NIFTI-1+

The voxelwise regressor images are all Volumes with the same as the mask above.

But the stats files generated by randomise are, for example,

 fslinfo test_tfce_corrp_tstat2.nii

data_type      FLOAT32
dim1           77
dim2           90
dim3           58
dim4           1
datatype       16
pixdim1        1.000000
pixdim2        1.000000
pixdim3        1.000000
pixdim4        1.000000
cal_max        0.0000
cal_min        0.0000
file_type      NIFTI-1+


Thanks,
Chris