fMRIprep-to-FEAT: how to get unsmoothed background images?

[bphilip]

Summary of what happened:

I have been using the recommendations from mumfordbrainstats to adapt fMRIprep-processed data into my FEAT analysis pipeline.

Everything works fine EXCEPT that my higher-level background images, and thus all the .png files, are on a smoothed background. The functional data look fine: voxels sizes are correct, the data passes Mumford’s data checks, and activity map looks very similar to a FSL-only analysis. Therefore I think it’s not affecting my results, just my visualizations; i.e. bg_image is from smoothed data.

Anyone know if there’s a way to avoid this? Or must I just avoid & ignore the automatically-generated backgrounds?

Here’s the rendered_thresh_zstat1 from a FSL-only pipeline, looking as it should:

Here’s the rendered_thresh_zstat1.png from the fMRIprep-to-FSL pipeline. Activity map is very similar (despite some registration differences), but the anatomical background image is smoothed.

[paulmccarthy]

Hi @bphilip, I think this might have something to do with step 6 in Jeanette Mumford’s instructions - instead of mean_func.nii.gz, you could copy whatever image you wish to use as the background (making sure that it is in the same space as your data).