Dear Melodic experts,
As all of you know, head motion has turn out to be a series challenge for resting state fMRI analysis and several suggestions have been turned out on how one should optimally minimize the effect of motion. However, to our experience and also several other groups motion do not produce a similar effect on ICA components as have been reported on seed approaches (e.g. localize increased in functional connectivity and distal decreased). My first question is if anyone has already published anything addressing this issue using ICA?
The second question is whether it is better to apply motion regression before ICA approach or on the component's time course. To my knowledhge, it would be better to apply motion correction on ICA time courses, or even during dual regression, since regressing out motion before carrying out ICA may reduce the ability of the algorithm to optimally tease apart independent components. Is there any publication suggesting that motion correction should be done on component's timecourse rather than prior ICA?
Thanks in advance.
Kami
|