Many thanks Christian, that all makes sense!
Regards,
Tim
On 01/12/10 19:55, Christian F. Beckmann wrote:
> Hi
>
> Nope, this is intended behaviour (it's a feature, not a bug ;) - resampling is only necessary for melodic group analysis (Tensor-ICA or Concat), if yous elect single ICA (even with multiple input files) the analysis will be done in native space as this avoids the resampling artefacts to bias the ICA. Once melodic is finished it's easy to get the melodic_IC files into whatever space you like by simply using flirt and/or fnirt to perform the resampling.
> hth
> Christian
>
>
> On 30 Nov 2010, at 23:20, Timothy Rittman wrote:
>
>> Hi,
>>
>> I have run in to a problem where melodic is not resampling my images correctly. I am melodic 3.10 in FSL 4.1.6 to analyse resting state fMRI with original voxel size 3x3x3.75. One slight quirk of my set up is the registration. The brains are from dementia patients and elderly controls, so I have found the best registration results by using fnirt to register structural MPRAGE images to MNI standard space (2mm brain), then in the melodic setup to register the functional scans to the fnirt'd structural scans with 12 DOF, then to standard space with 3 DOF.
>>
>> I am going on to compare the output to 2x2x2mm network templates, so require the output melodic_IC.nii file to be resampled to 2mm. The setting on the registration tab of the melodic GUI works fine when I run multi-tensor ICA, but when I run single tensor ICA the images refuse to be resampled and the melodic_IC.nii file remains with voxel size 3x3x3.75. Below are some outputs from fslinfo.
>>
>> I've run it twice and on different subject groups with the same result. Am I missing something?
>>
>> Kind regards,
>>
>> Tim Rittman
>>
>> Clinical Research Fellow
>> Department of Clinical Neuroscience
>> University of Cambridge
>>
>> fslinfo for output of single tensor ICA:
>> data_type FLOAT32
>> dim1 64
>> dim2 64
>> dim3 32
>> dim4 41
>> datatype 16
>> pixdim1 3.0000000000
>> pixdim2 3.0000000000
>> pixdim3 3.7500000000
>> pixdim4 1.0000000000
>> cal_max 0.0000
>> cal_min 0.0000
>> file_type NIFTI-1+
>>
>> fslinfo for output of multi tensor ICA (identical to the templates I am using):
>> data_type FLOAT32
>> dim1 91
>> dim2 109
>> dim3 91
>> dim4 25
>> datatype 16
>> pixdim1 2.0000000000
>> pixdim2 2.0000000000
>> pixdim3 2.0000000000
>> pixdim4 1.0000000000
>> cal_max 0.0000
>> cal_min 0.0000
>> file_type NIFTI-1+
|