Hi all:
We are looking into using FSL tools (e.g., MELODIC) for analysis of rs-fMRI time series.
One problem we have is that in some acquisitions, certain "frames" (i.e., 3D volumes at one time point) are bad in the sense that, due to motion and interleaved acquisition, they do not represent a spatially faithful head image at all. Since these frames are essentially a composite of the head in two or perhaps even three different poses, we cannot use frame-to-frame motion estimation to correct the problem.
Consequently, we'd like to exclude them from the modeling procedure, but without pretending that they weren't there in the first place. That is, we'd like these to be treated as actual missing data that nonetheless still occupy "space" along the time axis.
Is this type of exclusion something that can be done in MELODIC? Or does anyone have a better idea how to address the problem?
Thanks!
Torsten
|