Hi Armin, It seems you would like to run the preprocessing just once, then take the preprocessed images and run Melodic in different ways, is this correct? If yes, then here is a suggestion: 1) Open Feat (or Feat_gui on the Mac). On the top of the window, select "1st level analysis" and "Pre-stats only". 2) Select the images of your subjects, the preprocessing options that you would like to use and everything. Make sure you use non-linear registration. When ready, save as a *.fsf file and click "Go". Feat will take care of almost all the preprocessing for you. 3) After Feat finishes, the last preprocessing step is to apply the warps to the timeseries of each subject. This is easy to do: just make a for-loop calling applywarp, something like this (modify it according to your needs): for s in ${subj_list} ; do applywarp \ --ref=${SUBJDIR}/${s}.feat/reg/standard \ --in=${SUBJDIR}/${s}.feat/filtered_func_data \ --out=${SUBJDIR}/${s}.feat/filtered_func2standard \ --warp=${SUBJDIR}/${s}.feat/reg/highres2standard_warp \ --premat=${SUBJDIR}/${s}.feat/reg/example_func2highres.mat done where ${SUBJDIR} is the directory that contains all your subjects and ${subj_list} is a text file with one subject ID per line. Again, modify the filenames to match what you have there. 4) When you finish, inspect the results to make sure you have good quality results. Once you are satisfied, run melodic via command line. Just type 'melodic' to see the options. An example could be like this: melodic -i listfiles.txt -a concat --nobet --tr=2.5 -d 0 -m mymask.nii.gz --Oall -o mymelodic01 --report Of course, you set the parameters and add/remove options according to what you need, and you can run multiple times with different settings. In this example, the "listfiles.txt" should contain as many lines as there are subjects, each line containing the path to the "filtered_func2standard", which is the warped preprocessed image for each subject. Hope it helps! All the best, Anderson On 07/01/12 19:40, Armin wrote: > To whom it may concern, > > First of all I should thank Benjamin Kay for his previous answers. I would be thankful if somebody who is expert with melodic or work in Analysis Group, FMRIB could answers my question. > > I want to do group-ICA analysis for my data more than one time. Therefore, I decided to use images resulted from pre-processing and registration of the first running of the concat-ICA as input for the next analyses. > > So I did the following method for the next analyses: > > > Choosing the images from " reg_standard" folder of each subject (from my first completed analysis) and disabling all of "Pre-Stats" icons. However, since Melodic doesn't let us skip from "Registration state" I just chose 3DOF (for increasing speed) in the registration of Atlas . > > In other words, I tried different situations in which Set A is initial Data and Set B is Data after pre-processing and registration. > > First situation: > Set A ->(Pre_Stats and Registration) -> Set B -> dimensionality estimation. and ICA analysis > > Second situation: > Set B -> dimensionality estimation and ICA analysis. > > > But the results are strange. > > Although it is reasonable to have a difference between each time you run ICA analysis, the difference between when I do the whole process (I have done that twice) and when I use the registered_filtered_images and skip the pre-processing (I have done that 3 times) is too much. For instance, in the first situation I have 48 components in both tries but for the second situation there are 68 component for all of 3 times. > > Would you please explain what my problem is. > > > Thank you in advance, > > Best regards, > > Armin > >