> After evaluating the resulting GM_mod_merg file from the
> use of
> fslvbm_3_proc script I've noticed that 4 volumes out of
> 47 in this file are
> distorted. How should i cope with this issue? Can I
> continue the analysis
> (randomise script) with these vols included or this can
> affect the validity
> of my results? If I have to exclude these volumes which is
> the best way to
> be done cause I don't think it's a good idea to
> exclude the initial
> 'problematic' files and re-run all the vbm's
> steps from scratch.
You can use fslroi to eliminate these volumes and then fslmerge to re-concatenate the data (see usage). You can directly exclude these subjects in the GM_mod_merge_sX that you will have chosen.
> The intensity range of the GM_mod_merg file is around [0
> 1.4]. Is this a
> normal range of values?
> What kind of values does a
> GM_mod_merg file
GM "volume", i.e. probability for a given voxel of belonging to grey matter tissue, modulated by the amount of local contraction/expansion implied by the NL registration, see Good et al., NI 2001 for this.
> They are not t-values right?
> The values of GM_mod_merg_s*.nii.gz represent the same
> values as GM_mod_merg
> but under smoothing with different kernel sizes, right?
> do we need these
> files(GM_mod_merg_s*.nii.gz) since
> GM_mod_merg_s*_tstat.nii.gz images are of
> our interest?
You need these images because you're going to run your tests on them!
You need *data* to *be analysed*, GM_mod_merg_s*.nii.gz is the data, GM_mod_merg_s*_tstat.nii.gz is one of the output of the analysis.
> Regarding the decision for the most appropriate smoothing,
> a relevant
> discussion in the forum (Subject: RE :[FSL] FSL_VBM
> selecting a mod_merg
> file) was mentioning that it would be better to choose a
> small smoothing
> kernel when we search for small brain changes (like in
> controls group)
Or between a control group and a patient group with minor abnormalities
> and a
> big one when we expect big changes ( like in patients'
Or between a control group and a patient group with major abnormalities
> I'm bit
> confused here... Since all subjects (controls-patients) are
> merged together
> in the GM_mod_merg_s*.nii.gz and will be used for further
> analysis with
> randomise how can we decide for the most appropriate value?
Only you know what you're studying here... If you expect small abnormalities between controls and patients, then go for a smaller smoothing. If not, go for a bigger kernel size.