Hi Yi-Fen,

That is very interesting. I would have thought the interpolation only re-sampled as well. I did have a chance to look at the source code, and the interpolation does change the vox_to_ras transformation in the .nii header which may cause it to not overlay properly in a program which uses this information. Right now there is not an option to change this, although by default we should keep this information the same.

I checked an example dataset that I have and found I could overlay the images correctly in mricro but I have not checked in FSL. 

Sorry for the confusion.

Best,

Russell