Hi
On 16 Feb 2012, at 17:40, SURIL GOHEL wrote:
> Hi Experts,
> While running melodic ICA from command line
>
> i am using following option
>
> melodic -i group_1_sublist -o group_1_melodic_concat --bgimage=/usr/local/fsl/data/standard/MNI152_T1_2mm.nii.gz -a concat --report --Oall.
>
> I am getting 29 ICA components and following are the .nifti files present in the ICA directory. I wanted to know what the data in those files means.
>
>
> melodic_IC.nii.gz -> is it the un-thresholded un-normalised Z score maps?
>
Yep
> melodic_oIC.nii.gz -> What does this file represent?
>
These are the original components maps before transofrming to voxel-wise Z-stats (i.e. before sclaing by the residual noise std deviation)
> melodic_pca.nii.gz -> i understand that this is PCA calculated at first step but is there a one to one correspondence between 1st ICA component and 1st PCA component?
no, see the melodic paper or see the technical report http://www.fmrib.ox.ac.uk/analysis/techrep/tr02cb1/tr02cb1.pdf
> if is there a way to derive corresponding component from both ?
>
Well, yes, this is what ICA does, it takes PCA components and transforms these to give ICs... in the case of melodic this means that there is a rotation involved, see the 2004 TMI paper
> I am interested in thresholding each of the groups Z score maps separately. which input files should I use in the easythresh command?
>
Either use the z-maps melodic_IC or (given that you did run melodic with the --report --Oall options) use the probmap files in the stats subfolder. This then thresholds on the basis of the Gaussian/Gamma mixture model.
hth
Christian
|