Hi Úrsula,
You are right, interpolation as used in registration will normally result in correlations between voxels - the intensities in the the registered image are based on neighbourhoods of voxels in the original image. Depending on what you want to do, you might be fine using nearest neighbour interpolation, which just uses the intensity from the closest voxel instead of combining a neighbourhood of voxels. This does not guarantee that there will be no correlations, though, as the intensity from a single voxel in the original image can appear in multiple voxels in the registered image (unless you resample to a resultion significantly lower than that of the original image).
If your analysis allows for it, it may be easier to perform the steps that are sensitive to the interpolation effects in native space and then register the intermediate results.
Hope that helps!
Cheers,
Eelke
> On 4 Aug 2015, at 13:55, Úrsula Pérez <[log in to unmask]> wrote:
>
> Hi all,
>
> I'm preprocessing functional images, and after normalization to standard template with FLIRT my workmates and I are founding added correlation in the Z axis (not allowed for our study). The register has been: 1) FLIRT functional to ref:anatomical 6dof, 2) FLIRT anatomical to ref:standard 12 dof, 3) convert_xfm to concatenate the transformations leading to func.2std, and 4) FLIRT functional to standard 12 dof with -applyxfm -init func.2std, all with trilinear interpolation).
>
> We are guessing that maybe the added correlation in Z axis it's due to interpolation. Trying to undo that interpolation, I returned the functional image to its original resolution (after normalization, it had the resolution of the template). For that I created a template with fslcreatehd and I used FLIRT with an identity transformation, also trilinear interpolation. With this strategy there is much less added correlation in Z-axis.
>
> Do you have any idea to further reduce the added correlation in Z-axis or avoid it during normalization? My images have an anisotropic relation of 4 and the standard template has 4.15.
>
> Thank you very much,
> Best regards,
>
> Úrsula
>
>
|