Dear FSL experts,
I have encountered a problem with MELODIC that seems to me like the program is not reading the --mix flag as it should be. I cannot, for the life of me, figure out why. Below I report some command line output that I hope rules out some mundane possibilities and is also helpful in finding the problem. I previously ran MELODIC from the command line on my computer for a group ICA with 124 4D images and automated dimensionality estimation (resulting in 40 components). The data were preprocessed using Nipype to wrap FSL commands. My operating system is Ubuntu 15.04 and the version of FSL is 5.0.9. I have not changed any of the software versions along the way.
First, I show the versions of FSL and MELODIC being called by my computer.
$ which fsl
/usr/share/fsl/5.0/bin/fsl
$ which melodic
/usr/share/fsl/5.0/bin/melodic
Next, I check the read write access of the output directory from the first melodic run. All of the files are original output from MELODIC except for the Classifications*.txt files which were made by melview and edited by hand.
$ls -l Smooth_Auto.ica
total 86404
-rw-rw-r-- 1 929 Jan 15 11:48 Classifications~
-rw-rw-r-- 1 62 Jan 18 12:21 Classifications_list.txt
-rw-rw-r-- 1 991 Jan 18 12:21 Classifications_list.txt~
-rw-rw-r-- 1 930 Jan 18 12:21 Classifications.txt
-rw-rw-r-- 1 991 Jan 15 16:02 Classifications.txt~
-rw-rw-r-- 1 68869 Jan 14 21:06 eigenvalues_percent
-rw-rw-r-- 1 111582 Jan 14 21:22 log.txt
-rw-rw-r-- 1 21473 Jan 14 17:21 mask.nii.gz
-rw-rw-r-- 1 2321762 Jan 14 17:21 mean.nii.gz
-rw-rw-r-- 1 7974916 Jan 14 21:22 melodic_FTmix
-rw-rw-r-- 1 44417756 Jan 14 21:22 melodic_IC.nii.gz
-rw-rw-r-- 1 2223 Jan 14 21:22 melodic_ICstats
-rw-rw-r-- 1 16741816 Jan 14 21:22 melodic_mix
-rw-rw-r-- 1 16613 Jan 14 21:22 melodic_PPCA
-rw-rw-r-- 1 16741816 Jan 14 21:22 melodic_Tmodes
-rw-rw-r-- 1 154 Jan 15 11:49 melview.ini
drwxrwxr-x 2 4096 Jan 14 21:22 stats
Now, I input the command to melodic to remove a number of components that I have classified as noise and as described at the following URL http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/MELODIC/FAQ
$melodic -i Smoothed_Input.txt -o Smooth_Auto_Filtered.ica --mix=Smooth_Auto.ica/melodic_mix --filter="2,6,9,10,11,12,15,16,17,18,19,24,25,26,28,29,32,35,37,38,40"
ERROR:: no mixing matrix for filtering (use --mix='filename')
I have also tried many variations on this theme--putting in the input files as a list to the command line, switching the command orders, using different flags, putting quotations around file names, etc., all to no avail.
I find the error message extremely perplexing since I am using the --mix flag. I have attempted to locate the error in the FSL scripts, but, honestly, I am not well versed in Linux coding and found this to be a rather daunting task. Any help would be greatly appreciated. My ultimate goal is to do dual regression. If for instance, it would be reasonable to include all of the ICs, but to count the unwanted components as nuisance regressors when calculating contrasts, then bypassing this issue would also be an acceptable solution for the time-being.
Cheers,
Benjamin
|