Hi Stephan,
Sorry if my question was a bit unclear. By normalization of the time series, I meant if we should be scaling the time series by dividing each timepoint by the mean of the timeseries as is often times done with task-based FMRI. Although this is done more for percent signal change comparisons, I wanted to know if with ICA, seed-based or other approaches you see this step as necessary or not as I am not as familiar with FSL's analysis approach just yet. I have seen it sometimes done in resting state papers which I suspect is a carry over from task based experiments but I do not think this step is necessary. Since I am not as familiar with ICA preprocessing,I thought it would be good to see your thoughts on this matter. If it is necessary, I would think it would be performed after regression of physio signals.
Thanks,
Ajay
|