Hi Steve,
sorry for the confusion, I may not have been writing out things properly, will do this now:
1.I first use feat to pre-process the subjects data, ie. Motion correction, B0 unwarping, smoothing and high-pass filtering. I also do the linear/nonlinear registration to highres and then to standard space.
2.Then I bring the individuals filtered_func data into standard space:
applywarp --ref=${FSLDIR}/data/standard/MNI152_T1_2mm --in=$filtered_func --warp=$warp --premat=$premat –out=$output
3.I apply command-line melodic in temporal concatenation mode on standard space data. Note I run it on all subjects (patients and controls) without specifying any contrasts:
melodic -i filtered_func_warp .txt -o melodic -a concat --nomask --nobet --mmthresh=0.5 --tr=3 --report -v –Oall
4.Then I run dual_regression, specifying the groups. Note I included a 3rd EV column to account for slightly different acquisition parameters (ie there are a several subjects where the acquisition matrix was slightly different from the majority of subjects, I attach the simple design)
dual_regression melodic_ados/melodic_IC 1 design.mat design.con 5000 melodic_dual_reg `cat filtered_func_warp .txt `
So my questions:
1. Does this procedure look OK to you? Particularly the simple design...
2. If I wanted to include the motion parameters, I'd need to use fsl_glm before? Is this generally recommended to do?
Thank you so much in advance for your patience.
All the best,
Torsten
|