OK, voxel dimension wasn't it, it was data type. All the error images were those that I used "fslmaths -fillh" on to fill mask holes. Apparently I needed to specify the output data type as using "-odt short" or "-odt input" otherwise fslmaths saves new images as float.
Fixed using:
fslmaths wonkymask -mul 1 lovelymask -odt short
On Thu, 19 Oct 2017 10:41:45 +0100, Paul Wright <[log in to unmask]> wrote:
>I do believe you've got it. Tiny difference in slice thicknesses. I'm not sure how that came about, but I think I can fix it now at least.
>
>% fslinfo data.nii
>data_type INT16
>dim1 128
>dim2 128
>dim3 72
>dim4 66
>datatype 4
>pixdim1 2.000000
>pixdim2 2.000000
>pixdim3 1.999992
>pixdim4 1.000000
>cal_max 0.0000
>cal_min 0.0000
>file_type NIFTI-1+
>
>% fslinfo nodif_brain_mask
>data_type FLOAT32
>dim1 128
>dim2 128
>dim3 72
>dim4 1
>datatype 16
>pixdim1 2.000000
>pixdim2 2.000000
>pixdim3 2.000074
>pixdim4 0.000000
>cal_max 0.0000
>cal_min 0.0000
>file_type NIFTI-1+
>
>
>On Wed, 18 Oct 2017 11:23:49 -0400, Moises Hernandez <[log in to unmask]> wrote:
>
>>Hi Paul.
>>
>>I am not sure, but it seems a problem of compatibility between the
>>diffusion dataset and the mask.
>>Can you run these commands and tell us what do you get, please?
>>fslinfo data.nii.gz
>>fslinfo nodif_brain_mask.nii.gz
>>
>>Moises.
>>
>>On 18 October 2017 at 10:25, Paul Wright <[log in to unmask]> wrote:
>>
>>> Dear FSLers
>>>
>>> I am running bedpostX on a set of 82 images. Of these, seven have thrown
>>> an error. I am attaching the SGE logs. The important lines seem to be:
>>>
>>> Image Exception : #4 :: Incompatible number of mask positions and matrix
>>> columns
>>> terminate called after throwing an instance of 'RBD_COMMON::BaseException'
>>> /software/system/fsl/fsl-5.0.10/bin/bedpostx_postproc_gpu.sh: line 81:
>>> 22601 Aborted (core dumped) $bindir/bin/merge_parts_gpu $all
>>>
>>> I found three other posts citing this error, but none related to bedpostx
>>> or that seemed to have solutions that were applicable. I thought it might
>>> be something to do with the dimensions of he data relative to the diffusion
>>> directions. I checked the dimensions of in-mask voxels for each subject and
>>> the error subjects did not differ. Another possibility is of slices with
>>> very few voxels in. I tried running erode and dilate on my masks but got
>>> the same error.
>>>
>>> I'd appreciate any help getting these subjects to run. I may try running
>>> these without the GPU to see what happens.
>>>
>>> Thanks for your help
>>> Paul
>>>
>>> -----------------------
>>> Paul Wright, PhD
>>> Postdoctoral Researcher, COSMOS lab (Stroke and Cognition)
>>> Department of Basic and Clinical Neurosciences. Institute of Psychiatry,
>>> Psychology & Neuroscience (IoPPN). King's College London.
>>>
>>>
>>>
>>
|