Hi
This does not sound right, melodic should not take this long. First, you should only create 4D files per subject and not do the full concatenation yourself. Instead, the input file should be a simple ascii text file, one line per subject...
Did you check that the data looks fine, e.g. by loading into fslview prior to running melodic? Also, you might want to try running everything through the melodic GUI instead of AFNI for pre-processing - this then also saves you creating the inputs.txt file
hth
Christian
On 16 Aug 2011, at 01:04, Shantanu Majumdar wrote:
> Hello FSL experts,
>
> I am running melodic under the multi-session temporal concatenation mode on a resting-state fMRI dataset of 25 subjects with 164 time samples in each subject data (TR=2.5s). I am trying to perform a group ICA. I did the preprocessing steps (each subject, slice time correction, motion correction etc) and then transformed each subject data onto standard space (Talairach space), in Afni. Then, I created a time concatenated dataset using 3dTcat command in Afni. Brain mask was separately generated.
>
> To run melodic, I first used the following command line
>
> melodic -i concat_fmri_data.nii.gz -m mask.nii.gz --report --Oall
>
> Melodic was running for more than 3 days and the log.txt didnt show convergence, so I reran it with an increased eps value
>
> melodic -i concat_fmri_data.nii.gz -m mask.nii.gz --report --Oall --eps=0.0005
>
> Currently, it shows convergence (took 24 hours), but 'Sorting IC maps' step has been running since last 24 hours too. My workstation has good specifications and single subject melodic runs within 15 to 20 mins.
>
> Here are my questions:
>
> 1) Is it normal that temporal concatenation mode takes this long?
> 2) Is the above mentioned procedure for the usage of the melodic command correct?
>
>
> Thanks in advance,
>
> Shantanu
|