Dear Alessandro,
> I would like to better understand how you perform resampling after normalization by means of fnirt.
> You provide options for trilinear, sinc and bspline interpolation. However, after reaching maxima, your points/voxels in native space become a non structured grid, since non linear displacement occurs. How do you then implement interpolation and extrapolation? I am familiar with interpolation methods on standard grids, but with unstructured ones this is not the case. I am asking because I need to carefully check how new intensities are estimated.
>
> Any clarification or hint on these aspects will be well accepted!
the new coordinates x’ (after non-linear transformation) are on an irregular grid. In order to “figure out” what intensity to assign to that voxel x’ is “mapped into” the regular grid of the original image. So effectively what you have is a non-integer coordinate x’ in a regular grid so it is straightforward to use and “traditional” interpolation method such as tri-linear, spline etc.
Jesper
>
> Best Regards,
>
> Alessandro
|