It probably doesn't make so much difference - but I may be wrong. I would
suggest either option 2, or a modification of it whereby a mean image is
created by the realignment - purely for the purpose of coregistration.
Best regards,
-John
> in my perfusion analysis I have several images acquired at the same session
> (for every person). These images have to be averaged to get the perfusion
> image. Which of the following procedures is the best in order to get this
> mean image in a preprocessing protocol (while avoiding unnecessary
> reslicing)?
>
> 1) realign (coregister + reslice: mean image only) -- coregister
> (coregister only) -- normalize (estimate T1 + write mean image);
> Disadvantage: the mean image is resliced twice (realignment +
> normalization)
>
> 2) realign (coregister only, creating mean image only for coregistration)
> -- coregister (coregister only) -- normalize (estimate T1 + write
> individual images) -- average individual normalized images;
> Advantage: Reslicing only once. But: does averaging over (non-linear
> transformed) normalized data introduce any errors?
>
> 3) realign (coregister only, creating mean image only for coregistration)
> -- coregister (coregister only) -- average (unresliced!) individual images
> -- normalize (estimate T1 + write mean image);
> I suppose this one isn’t allowed, is it? (Advantage: Reslicing only once.)
>
> 4) ?
|