Dear FSL experts,
I have a few questions about doing a multi-site ICA analysis using FSL-MELODIC.
What would be the correct approach to combine data from different sites which had 180, 240, and 320 resting state fMRI time-points [the other scanning parameters are comparable]? Does the data need to be truncated to the same timecourse to avoid over-representing subjects with longer time-courses in the multi-session temporal-concatetation ICA? Can MELODIC handle subjects with different timecourses in any other way - for example by somehow weighing down the individual scans within the group analysis? Would you recommend using GIFT instead [https://sourceforge.net/p/icatb/mailman/message/34626652/]?
I have a somewhat related question pertaining to the number of controls and patients and/or their respective time-points. Should the ICA be balanced for the different subject groups in order to avoid generating group components biased for particular subject group (for example if there are more longitudinal scanning sessions for patients and/or more patients are recruited to the study than controls)? In the applied clinical fMRI literature, I have not seen this question properly addressed in studies with unbalanced subject groups. Using the multi-session temporal concatenation, this would bring up the same issue as having subjects with different number of timepoints.
Thank you very much for your advice.
Sincerely,
mj
|