resliceToiImage - interpolation type suggestion

[CerenB]

This is a suggestion of interpolation type change for reslicing images.
The suggestion was due to @mohmdrezk warning. I want to ask of this is what he meant + if someone else has an objection to it or not.

[CerenB]

I might be wrong but here is a bit more thought on it.
interpolation == 4 could work well for the zMap, but not for the atlases where the labelling of the voxels are e.g. [1, 2, 3, 4, ...]
interpolation == 1 would perform the following [1.2 1.3 1.4] values which would not be good for the atlas masks.
Looking forward to hearing your ideas/input.

[mohmdrezk]

If you are re-slicing a multi-label atlas to different resolution and you want to preserve the same labels, then the interpolation method that should be used is the nearest neighbor (interpolation == 1).

[CerenB]

ok, if I understood @mohmdrezk right, then we would need a if case here. Please correct me if I'm wrong.

interp = 4 % for neurosynth (having zscores), tmaps, already extracted ROI from an atlas (not carrying mutliple labels)

if multilabelAtlas or binaryMap is used
interp = 1

But I'm guessing in atlas masks, this repo does not do reslicing. I haven't figure it out why/how though. WIP.

Again correct me if I'm wrong/missing a step:

• we take ROIs from an atlas (making 1 label), binarize it
• we reslice it, then due to interpolation I need to binarize it again
• then I need to make 4D.nii in subject space

[mohmdrezk]

interp = 4 % for neurosynth (having zscores), tmaps, already extracted ROI from an atlas (not carrying mutliple labels) For the already extracted ROI from an atlas (not carrying multiple

labels), I think it's better to do nearest neighbour while reslicing instead of doing interpolation and then binarizing.

…

if multilabelAtlas or binaryMap is used interp = 1 Or any ROI that you intend to use as a mask.

[Remi-Gau]

Will merge but this just contains only typo fixes now.