Hi
If everything ran via the GUI (and there are no errors reported in the log or on the command line) then there should be 2 versions of filtered_func_data. One is the output of standard pre-processing in native space (looks like you used these to feed into DR) and then (in the /reg sub folder I think) there's a transformed version of the first. These files are in MNI152 space (though they might not be _sampled_ at 2mm isotropic resolution, e.g. if in the GUI you select 4mm isotropic resampling resolution these will be 4mm iso.
You can check at the command line if you have the right input files: fslsize <filename> gives you x/y/z numbers and dimensions and they need to match between the filtered_func files and melodic_IC.
hth
Christian
On 15 Jan 2013, at 13:55, Helen Sawaya <[log in to unmask]> wrote:
> Dear Christian,
>
> Thank you for your reply.
>
> What we are feeding into dual regression are the melodic_IC file and the filtered_func_data of every participant that is generated by Melodic (which should have been already registered to the structural image and the MNI standard through the Melodic GUI). Those filtered_func_data though are not in MNI space and cannot be overlaid over the standard. The melodic_IC file is in MNI space but also cannot be overlaid onto the standard. Without tinkering with the melodic_IC and registering it to the preprocessed files, dual_regression is not outputting any result. When we did register melodic_IC onto the preprocessed individual subject files, we did have (goodlooking) results.
> Obviously though we are using the wrong inputs to dual_regression or we are missing something when we first run the Melodic GUI because we should get preprocessed files that are already in MNI space. We appreciate your input.
>
> Thanks again,
>
> George Chahine
> Helen Sawaya
> ________________________________________
> From: FSL - FMRIB's Software Library [[log in to unmask]] On Behalf Of Christian F. Beckmann [[log in to unmask]]
> Sent: Tuesday, January 15, 2013 11:14 AM
> To: [log in to unmask]
> Subject: Re: [FSL] dual regression mask
>
> Hi
>
> Interesting side note! If you hack the pipeline you should not be too surprised that things do not work. I guess you used the wrong inputs to dual regression: it look slike you used the filtered_func in native space, rather than the filtered func data in MNI152 space. Rather than use a tranformed melodic_IC to make things work you should fix the inputs to work with the group melodic_IC file.
>
> hth
> Christian
>
>
>
> On 15 Jan 2013, at 09:01, Helen Sawaya <[log in to unmask]> wrote:
>
>> Dear all,
>>
>> I can't seem to be able to figure out why the output mask of my dual regression procedure looks incorrect (I attached the image). All the images were registered to standard MNI.
>> (As a side note, the dual regression would only work when I registered the melodic_IC to one of the functional images, because it would give me an error otherwise stating that the images are not the same size).
>>
>> Any ideas?
>>
>> Thanks a lot!!!
>> Helen<maskall.png>
|