In general you do NOT want to manually adjust *single volumes* of a EPI session at all. The thread started with alignment of several EPI *sessions*. As the range in which the algorithms operate are limited (as true motion is limited), there might be some instances in which the automatic realignment fails, for example if subjects were repositioned between the two sessions and the displacement is large. You would try to reorient one session so that it matches the orientation of the other session before starting with the automatic preprocessing. For the realignment plot this manual adjustment is not taken into account because you want to see the true extent of motion (e.g. does the subject move from the left to the right and if so, to which extent) and not motion minus your manual adjustments.
I'm still not sure whether you really have realignment problems. If you enter your raw data into realign and then checkreg the volumes afterwards, do they agree well? In case they do, everything's fine. Of note, when opening the realigned files with other programs like MRIcro then the volumes might look identical to the raw files (no shifts/orientations visible) because they don't load the information stored in the header part. But if you go on with preprocessing with SPM the reorientation parameters should be taken into account during Normalise: Write.
The second aspect is the automatic coregistration. As previously stated, it does not matter whether you coregister structure onto EPI mean or EPI mean onto structure (although in the latter case you would have to add the EPI session as "other images", otherwise only the mean is altered but not the EPI session in which you're interested in). The coregistration might fail if the displacement between mean and structure is too large, and animal data is often problematic because there are large signal drop-outs or deformations.
The third aspect is the segmentation. If the structure differs from the tissue probability maps to a larger extent (very different orientation, drop-outs, ...), segmentation might fail.
Without screenshots it is unfortunately difficult to guess at which point your problems arise.
Best,
Helmut
|