Hi Claire
> Before conducting preprocessing for the entire fMRI dataset, I preprocessed one subject's data as a testing run and I wanted to make sure that what I did was correct (I am unsure of 'segmentation' and 'normalization').
>
> Here is what I did...
>
> After slice timing and realignment (estimate&reslice), I ran 'segment' to which I added the ORIGINAL T1 image with all of the default options kept. <-- IS THIS RIGHT?
First, you probably don't need to reslice during realignment; the
image headers are updated to be aligned, and not reslicing will save
having to rewrite the data. (In practice it probably won't make that
much difference to your final results though.)
After realignment, you will want to coregister the structural image to
the mean functional image (and all the other functional images).
After you've done so use checkReg to make sure it looks reasonable.
Then, for 'segment', yes, select the T1, and the default options are
generally fine.
> After 'segment', I ran 1) normalise (estimate&write) - for functional images: I put as a source image the output T1 image from segmentation (m~.img)- bias-corrected image. <-- IS THIS CORRECT?
If you've segmented the structural image, want to just do normalise
(write), and use the parameter file generated during segmentation
(*seg_sn.mat) to normalize your functional images. (The idea being
that, if your structural image and functional images are aligned, you
can just apply normalization parameters based on the structural image
to the functional images.)
> 2) normalise (write) - for structural image: I chose as my parameter file the output from the previous normalizaion which looks like m~.sn.mat) and for images to write, I chose again the bias-corrected structual image from segmentation (m~.img) <--- ARE THESE TWO CHOICES CORRECT?
This is fine. Depending on what you are doing with your structural
image, you could write out the bias-corrected or the original; it
probably doesn't really matter.
Hope this helps,
Jonathan
|