Hello everyone, 2 quick questions regarding MELODIC, command line version.
I am inputting detrended / % signal change converted data into MELODIC using this following command:
melodic -i func_bpass.nii.gz -o ica -m EPI_mask.nii.gz --Oall --report --nl=gauss --dimest=aic --approach=tica -v
Next, I run a small program I wrote to identify unwanted components per run. I then use this command:
fsl_regfilt -i func_bpass.nii.gz -o func_ICA.nii.gz -d ica/melodic_mix -f (components) -v -m EPI_mask.nii.gz --out_data=ICstats.nii.gz
Questions:
1) I know that the first MELODIC call variance normalizes my data, which I can optionally turn off (but would prefer not to). Is FSL_REGFILT smart enough to apply these z-scored components to my % signal change data properly, and output files with % signal change, not z-scored, units? If not, should I z-score my time series before running MELODIC? Eventually, I will be taking an average over the data using a program I've written that is designed to deal with noise. I would prefer to z-score the data after I take the mean rather than before, if possible.
2) What, exactly, is being spit out by --out_data=ICstats.nii.gz? I have not be successful at finding any information about this option.
Thank you so much in advance,
Joseph
|