Dear FSL Experts,
I'm sorry for posting these questions again. It's really importmant for me to understand the problems and comtinue the analysis.
I try to combine sMRI, DTI and fMRI together via FLICA to find co-alterations, the inputs for each modality are:
sMRI: GM_mod_merg_s3.nii.gz
DTI: all_FA_skeletonised.nii.gz
all_AD_skeletonised.nii.gz
all_RD_skeletonised.nii.gz
fMRI: dr_stage2_ic0000.nii.gz
dr_stage2_ic0001.nii.gz
...
I followed the guide of FLICA , setting opts.num_components = 10, opts.maxits = 5000, before running [M, weights] = flica_reorder(Morig), everything was OK, but when I ran that command, I got some warning:
Warning: Non-finite values found in assertalmostequal!
> In assertalmostequal at 37
In flica_reorder at 68
And, when I tried to run flica_save_everything(outdir, M, fileinfo), I got the information:
Undefined function or variable 'order'.
Warning: Failed to make Hcorr.png for some reason -- see lasterr above. Keeping calm and carrying on.
> In flica_save_everything at 114
Finally, it show me some warning if I try to run flica_posthoc_correlations(outdir, des):
Subject_Index, component 0... EPS... PNG...convert: Command not found.
Error using flica_posthoc_correlations (line 105)
epstopdf failed
Can you point out my problems?
We have 2 groups for participants, one for patient, the other for normal control, so we want to do post-hoc tests after finding some across-modalities components, is the design the same as GLM for randomise of FSL? I notice that on the homepage of FLICA, it says that "Each one should be a subjects-by-1 column vector.", do we have to load the EVs one vector by one vector separately, like the example?
des.Age = load('age.txt');
des.Group = load('group.txt');
Or can we do the same as flica_template_script.m, loading the full design.txt at once, and then specify the EVs?
design = load('design.txt');
clear des
des.ev1 = design(:,1);
Is the specified step compulsory? Can we only use the full design.txt file for analysis to get the same result?
For group EVs (patient, control), is it the same as GLM, one column for patient, another for control?
EV1 (patient) EV2 (control)
1 0
1 0
1 0
0 1
0 1
0 1
...
Or a little different from it, setting them in one column (vector)?
EV (group)
1
1
1
2
2
2
...
And, are the variables needed to be demeaned like GLM model of FSL or not?
Thank you very much.
Jian
|