 |
 |
Hi Ryan, When applying any linear or non-linear transform the data has to be interpolated from the old voxels to the new voxels. This interpolation inevitably leads to some smoothing of the data. When you then transform the image back to the original space, the data will again be interpolated and hence you would get a smoothed version of the original image. This smoothing is the reason that it is better to combine all the linear and non-linear transforms that you want to apply to an image in a single step, because in that case the data will only have to be interpolated once, which leads to a minimum of smoothing. So, the smoothing that you observed, is just an inevitably problem arising from the resampling of the images and is not a sign of any problem with your transforms. Cheers, Michiel > On 3 Feb 2018, at 17:51, Ryan Hyon <[log in to unmask]> wrote: > > I should add that the FA image transformed into standard space and back-transformed into native space looks like it has been slightly spatially smoothed when compared to the original FA image. Why would spatial smoothing occur when using FLIRT/FNIRT transformations/warpfields and then applying their inversions? > > Thank you!