Hi
Running de-noising of individual data sets using group spatial maps is quite sub-optimal as the spatial maps relate to the group on average and are not optimized for any one specific data set. As such I'd expect to see less of a benefit compared to running de-noising on the original data sets first, before running group ICA. In any case if you do want to de-noise based on group results you will first need to genereate subject-wise mixing matrices using fsl_glm:
fsl_glm -i filtered_func_data_subject_i -d group_ICA_maps.nii.gz -m mask -o mixing_subject_i
Then you can use the group maps together with the mixing matrix to denoise as described for single subject ICAs
Wrt you second question this depends on how you run randomise and what output you're looking at. The XX_corr_p files are corrected, the ones without the corr are not.
hth
Christian
On 24 Oct 2010, at 07:09, Dan wrote:
> Dear list,
>
> I got two questions that may appear simple, but I can not find a solution:
>
> 1. How is it possible to filter out unwanted IC components from a group-ICA-file. It is described for single ICAs, where it works pretty well using the prefiltered_func_data first. Should I have to concatenate all prefiltered input-files into one 4D-File first to use the same procedure for my gica-file?
>
> 2. After calculating group-statitsics of a certain IC-compontent using randomise -c command I got t-values (I concatenate the specific and individual probmap IC-results). Are these results corrected for multiple comparisons?
>
> Hopefully someone can help me and thanks in forward...
>
> Dan
>
|