Hi,
I am having some frustrations while trying to conduct a meta-analysis of
resting-state FMRI data. Hopefully someone can help.
So far, I have processed the raw data using pre-stats in FEAT and run
MELODIC on each subject in their native space.
Following the history of posts on this issue, I have tried both of the
following using a test case of 4 subjects:
1) Transformed each subject's filtered_func_data (feat output) file to
standard space; concatenated the data together across time using
avwmerge++ -z; run melodic on the new 5D file.
Problem: melodic finishes in < 5 sec w/ no results.
2) Transformed each subject's melodic_IC file (melodic output - 30
components) to standard space; smoothed the data using a 5 mm Gaussian
kernel (avwmaths++ -fmean -kernel gauss 5); concatenated the data together
across time (avwmerge++ -z); run melodic on the new 5D file.
Problem: melodic hangs in the command window while eating up most
available CPU speed and RAM (after 30 minutes of 0 progress I killed the
process).
It certainly seems possible that method (2) may eventually work, but if it
takes all available CPU and most RAM over an extended period of time for
only 4 subjects, I'd hate to see what happens when I run it on 60+.
I'm not sure what's going on with method (1) since I have no problem
running melodic on an individual subject's filtered_func_data.
Any suggestions on what I'm doing wrong or alternative methods to attempt
would be much appreciated.
Cheers,
Peter Fried
Center for Magnetic Resonance Research
University of Minnesota
|