Hi Marios,
> I always run things from the command line but never managed to get it to go
> this far. Which version of fSL are you using? I have the latest version
> installed on my machine.
>
This is the version I am using, too.
> Perhaps I misinterpreted what the alternative mask option does, but it would
> seem from the log info you sent that it merely masks the filtered data, then
> uses regular thresholding (consistent with BET being turned off) to generate
> a mask anyway.
My understanding of the alternative mask is exactly, that it replaces
bet. The thresholding might be turned off in the GUI (or design.fsf
file) by setting the threshold level to 0 %. In the preprocessing step
it is used to remove non-brain signal from the image to speed up later
processing, since you don't want to run the analysis on voxels covering
e.g. muscle or eyes. However, sometimes the bet results in feat are
suboptimal, hence it is useful to define a mask selecting brain matter
in advance to feat and use this as alternative mask. Of course this mask
affects latter on the main stats, but here you have an extra option in
the GUI in the Post-stats tab (Pre-threshold masking). Here you could
enter any mask for selecting voxels used for the stat images. It is
possible to use different Pre-threshold masks when running Feat without
the Pre-Stats, but you would not recover any voxels that are set to zero
in the Pre-Stats by using the alternative mask or bet option.
My recommendation would be to use more generous mask for the Pre-Stats
(it is exactly the idea by using bet in the liberal way from the Feat
GUI) and to constrain voxel selection in the Post-Stats by using a
threshold mask. Usually, Feat gives good results with human data, but it
might be worth to create masks manually and ensure a sufficient good
quality. I guess hat this is even a must-do when using non-human data,
or data acquired with surface coils.
all the best,
wolf
> I thought that what one enters as an alternative mask is the
> actual mask used for the main stats but the info you sent does not seem to
> support that.
>
> Thanks
> Marios
>
|