Dr. Buechel,
What whould the difference be between the bilinear and sinc
interpolations be when reslicing after the normalization step? How do
these two methods change the images in regards to the final
statistical analysis?
thank you,
Jason.
Dear Catherine,
> I understand that reslicing is important for individual data, but
> I was wondering if reslicing only adds additional error if one is
> normalizing. In other words, might it be better not to go through
> reslicing if one does realignment followed by normalization?
>
> When one realigns and reslices, the rI* images lose the pixels on
> the edge due to interpolation. If one then does normalization, one
> gets a second source of error from the interpolation in the
normalization
> process. It seems that the realignment doesn't need the reslicing
because
> the coregistration already appears in the Mat files of the I* files.
You are absolutely right. The only reason for rewriting after
normalisation is (i) you want to analyse non-normalised data or (ii) you
want to "adjust for sampling errors", which isn't terribly useful. An
important note: If you do not reslice immediately (ie reslice after
normalisation), make sure you use sinc interpolation. This is important,
because all volumes are slightly different (because of the realignment
parameters).
-Christian
|