Print

Print
We've also tried running the applywarp command to transform our FIX-cleaned single-subject MELODIC functional data to MNI space. We ran singe-subject MELODIC with nonlinear registration (default settings) to the 1mm MNI standard brain. The command we've run is identical to Greg's except that our input is fitlered_func_data_clean.nii.gz. The command runs successfully but it outputs niftis that are >5 GB in size and we haven't been able to inspect them due to this.  Is this the expected output? 

Thanks,

Riley

On Thu, Oct 8, 2015 at 11:09 AM, Gregory Book <[log in to unmask]> wrote:
Mark,
Thanks for the response. We've been using SPM for years, and are moving toward FSL, so we're learning about all the differences. SPM does all normalization to the template before stats. For now, we wanted to do the stats with SPM since we have all the models already setup, but preprocessing with FSL. This piece of info below really helps us along.

I tried running the applywarp command:
applywarp -i filtered_func_data -o app_mni -r reg/standard -w reg/example_func2standard_warp -v
but got only the following output after about 30 seconds: "Killed", and no output files were written.

The input file is approximately 250MB and 1122 volumes.
-Greg



Hi,

I have a rather basic question about the output from FEAT. When preprocessing data through FEAT, I end up with the file filtered_func_data.nii.gz. I assume this is the 4D output with all preprocessing, registration, and normalization steps applied. And that stats can be calculated on that data. However, when comparing the outputs of multiple subjects, it looks like there is no normalization to MNI space.

We do the registration (or spatial normalization if you prefer) after we run the first level statistics.  So the filtered_func_data is in the native functional space and not transformed to MNI space.  All pre-processing steps have been applied though (i.e. motion correction, unwarping, spatial smoothing, temporal filtering).

If I run stats through FEAT on the output, everything comes out fine. However, if I run stats using SPM on the FEAT preprocessed data, it doesn't turn out well.

Which file contains all of the normalization/registrations that can then be used in other stats programs?

You can transform the filtered_func_data to MNI space if you want (and if SPM requires this) by using applywarp with filtered_func_data as the input, and the "standard" image from the reg sub-directory as the reference and the "example_func2standard_warp" file as the transformation (warp).  This will then give you an output in MNI space.

In FEAT the registrations are applied to the cope and varcope images from the first-level GLM at the start of the second-level analysis.

All the best,
Mark