Other useful function: "check_epi_template_space(epi_img, discrepency_threshold=30)" that would:

1. compute_epi_mask on the data
2. resample to the MNI template (obtained with nilearn.datasets.load_mni152_template)
3. compute a brain mask out of the template (by thresholding it)
4. Compute a xor between the two masks
5. (optionaly) erode the result
6. Threshold on the number of voxels in the resulting array.

This could be displayed in an example showing that ADHD is in template space, but not Haxby.

Other useful function: "check_epi_template_space(epi_img, discrepency_threshold=30)"

If nobody is already working on this, I would be happy to have a look.

@aabadie : I thought that you were busy lately :).

@dmadou is on it actually.

I thought that you were busy lately :)

I am but I can always find some time to work on awesome features ;)

@aabadie I can take care of that feature ^^

@dmadou sure, go ahead !

Hi everyone,
I implemented the function and when doing so, some questions emerged:

• Why computing the epi mask before resampling ? That may be in order to save time.
• In step 3." compute a brain mask out of the template (by thresholding it)", using "compute_epi_mask" or thresholding the template leads to the same brain mask as the template is a zero background image but thresholding is a pretty fast and straighforward process.
• What does the discrepancy threshold really mean ? Is it a number of voxels or a percentage ?
  Indeed, if it is a number of voxels, it will not have the same meaning when the template itself or its resolution changes.
• The last one, not that important, is should we allow the user to set the number of iterations when eroding ?
  Thanks for your answers !