Masking issues

[annap...@gmail.com]

Hi Chris et al.

I just re-ran an analysis using a GM mask thresholded at 50% but I'm
still getting voxels in my results images that are outside this mask!
Anybody got any ideas as to What is happening?

I can't attach images to this but Chris I'll send you some snap shots.

thanks
Anna

[annap...@gmail.com]

Ok just tried it again this time with the binary file as the mask
which works a treat but I'm still confused as to why the 50% threshold
on the mean GM image isn't doing what its supposed to be doing ?

thanks
Anna

[Christian Habeck]

Hi Anna,

there is an additional conjunction masking performed. So the non-zero voxels in any images that come out of our package should always be a SUBSET of the voxels that survive the GM-threshold based mask. This is the case for my own results that I have obtained with this software. Any .img files coming out of an analysis stream always have fewer voxels than for instance the GM>0.5 mask.

Sometimes the discrepancy can be quite big. For instance in the latest analysis I did there are 125,575 voxels that survive the GM >0.5 criterion, but the end result only has 70,448 voxels.

Maybe you could check this out - if you feel like it - by including fewer subjects in the analysis. The net number of surviving voxels should grow and approach what's maximally allowed by the GM or binary mask you're using.

Chris

[annap...@gmail.com]

Cheers will have a look at it. I'm just anticipating reviewers
comments as to why my GM maps seem to have blobs in the white matter!

On Dec 10, 3:15 pm, "Christian Habeck" <chab...@gmail.com> wrote:
> Hi Anna,
>
> there is an additional conjunction masking performed. So the non-zero voxels
> in any images that come out of our package should always be a SUBSET of the
> voxels that survive the GM-threshold based mask. This is the case for my own
> results that I have obtained with this software. Any .img files coming out
> of an analysis stream always have fewer voxels than for instance the GM>0.5
> mask.
>
> Sometimes the discrepancy can be quite big. For instance in the latest
> analysis I did there are 125,575 voxels that survive the GM >0.5 criterion,
> but the end result only has 70,448 voxels.
>
> Maybe you could check this out - if you feel like it - by including fewer
> subjects in the analysis. The net number of surviving voxels should grow and
> approach what's maximally allowed by the GM or binary mask you're using.
>
> Chris
>

[Christian Habeck]

Oh I see, that's the problem. The probabilistic mask doesn't correspond to individual subject's segmentation?

That can happen, and we can't do much about that....  I would recommend making a gray matter mask from the T1 scans and not relying on the SPM5-supplied mask.

Chris