Dear FSL experts,
I want to do source-based morphometry using Melodic in FSL5.0.9. I know it's implemented in GIFT, but I prefer using probabilistic ICA of Melodic. Theoretically I think it's OK, but could you tell me if the procedures are OK?
1. Preprocessing of T1 image (segmentation, normalization, and modulation)
2. Createing 4D file containing preprocessed gray matter data for all subjects (say this as all_gray), and creating a mean image of them (mean_gray)
3. Melodic for preprocessed T1 image using following command:
melodic -i all_gray -o melodic.ica -v --nobet --bgthreshold=10 --report --bgimage=mean_gray --mmthresh=0.5 --Oall
4. Group comparison of the components using the unmixing matrix
The melodic mode should be single session ICA, but I'm not sure which -a option should be assigned.
I really appreciate if you can give me advice.
Best,
Jun
|