Hi Simon,
-- Original Message -----
From: Simon AU YOUNG <[log in to unmask]>
To: <[log in to unmask]>
Sent: Thursday, May 11, 2000 12:58 PM
Subject: Coregistration problem of multisubject's normalized images
>
> Hi all:
>
> I have questions regarding the image pre-processing steps in spm99. I
> set origin at AC and y-axis passing thru the AC-PC line for each
> subject. Second, I ran each subject's functional data independently
> through realignment and normalization (with the EPI template). Then, I
> checked for coregistration of the norm mean images between subjects.
>
> The origins (which are all located at the same voxel, 40 57 26) do not
> agree very well anatomically. Some of them stay near AC in the images,
> but some were quite a distance away.
>
> What went wrong and how should I fix it? Here're the possible
> options I can think of. Please give me some suggestions.
>
There are a number of reasons why normalization might not produce the best
results. One stage that may really help is to re-orient the data prior to
spatial normalization (use the display button and adjust your images so that
they look as close (in orientation) to the EPI template as possible). This
has been discussed previously on the list and in our experience it helps out
quite a bit. Also, it would be important to know what parameters you are
using for spatial normalization. Depending upon the nature of your data,
one may need to adjust the default options of the spatial normalization. In
particular, you may get better results if you turn off the 'brain mask'
function (according to spm authors this was originally intended for PET
data). A number of groups have found that, if enabled, this option can
distort the images. In particular, brains that are larger than normal show
poor normalization if this procedure is enabled.
Best regards,
Kent
> OPTION 1. The disagreement is due to the misplacement of AC in some
> subjects and I should redo normalization for these subjects.
>
> OPTION 2. I should move the origin to AC in each subject's norm
> mean image and apply to its corresponding norm images before smoothing and
> statistics. (For the statistics, are the *.mat files used or do I need to
> reslice after fixing the origins?)
>
> OPTION 3. Nothing's wrong with normalization. The origins should all be
> set at "40 57 26" now, and I should not move them at this stage.
>
> Thanks in advance.
>
> Sincerely,
> Simon Au Young
> M.Sc. Program, Experimental Medicine
> University of British Columbia
>
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|