 |
 |
Hi Anderson, Thank you for your answer. Well, to be more precise, I have no hypothesis, I compare the resting-state of the whole brain beetween 13 young control people and 13 young people with autism syndrome disorder without any a priori ROIs. Otherwise, the two acquisitions were to do some average, because is two similar acquisitions (7 min of resting-state with eyes opened, with the same scanner ...). So that's why I asked if, instead of doing an average between the 2 sessions per subject, I consider the 2nd acquisition as new subjects (that increases the size of my samples : 26 control / 26 autism) was correct. For the merging of the component, I first split my dr_stage2_IC#subcomponent1 and dr_stage2_IC#subcomponent2, then I merged for each subject the 2 subcomponents (with fslmerge), then I did the average of that (still for each subject) with fslmaths -Tmean, and finally I merged all these new files to have a new 4d image where each time point is the new IC map (mean(subIC1+subIC2)) of 1 subject, i.e, I have a new 'dr_stage2_ic#[(subIC1+subIC2)/2]'. Is it correct ? and is it not better to do that before the dual regression, i.e, create a new 'melodic_IC' where some component are merged ? finally, you are rigth even if for some RSNs I have no significant difference, It's still a significant result (in the sens there is no difference between the 2 groups significantly). Moreover, even if I have only 3 or 4 RSNs showing a difference (asd<control) in my statistical contrast images, it seems to be in the same networks (Default Mode, executive control, Salience network) I could find in previous literature. Best regards, Antoine