The four most likely things are:
1) The images were originally acquired using a limited FOV, so that
part of the data is missing.
2) The GM threshold used for masking in the GLM is too high, so it
excludes too much of the brain.
3) There was some weird motion effect that caused much of the data to
be masked out because there was no data for one or more time points).
4) Something else.
You can check the (1) by looking at the brain coverage in the original
data. You can check (2) by comparing the mask.img (generated at the
GLM fitting stage) against the data you fit the GLM to. You can check
(3) by looking at the estimated motion parameters. For (4) you may
need to look at all your images to find the source of the problem.
Best regards,
-John
On 11 January 2012 09:38, Luke Wang <[log in to unmask]> wrote:
> Dear All,
>
>
>
> I’m using spm8 to analyze my fMRI data and encounter a problem. When
> conducting 1st level analysis, I get weird results in model estimation for
> one certain subject. The beta images look like this:
>
>
>
> http://i1104.photobucket.com/albums/h327/madaye/beta.png
>
>
>
> As you can see, the bottom half of the brain becomes NaNs. However, the raw
> and pre-processed images look normal and such problem only occurs for this
> particular subject. I wonder what are the possible causes of this and any
> ideas will be greatly appreciated.
>
>
>
> Best,
>
> Luke
>
> 2012-01-11
|