Dear FSL team,
I have a question regarding FSL melodic. I have run the following on a group of 116 participants using the default automatic estimation (Laplace approximation) for dimensionality reduction after excluding participants with excessive motion and using AROMA to
correct for motion in the remaining participants.
melodic -i inputlist.txt -o groupICA --tr=2.20 --nobet --bgthreshold=1 -a concat --bgimage=/usr/local/LKeb/FSL/5.0.8/data/standard/MNI152_T1_2mm_brain.nii.gz -m /usr/local/LKeb/FSL/5.0.8/data/standard/MNI152_T1_2mm_brain_mask.nii.gz --report -v –Oall
This approach yielded over 100 ICs which seems high. Do you know what might have caused this and if it is a problem?
I worry that it points to an underlying problem in the data and that reducing the number of dimensions by using a validated template or constrain the dimensions using – d 20 would only mask the problem rather than solve it.
The sample does have a large age range (8 to 60 years) but I tried running the ICA for children and adults separately but still got over 100
ICs in each group.
Any advise would be appreciated.
All the best,
Katharina