Dear Marta,
yes indeed, you probably have to exclude the subjects completely or go with the first run only. This is quite bad, as in your case with two runs (?) it means that you only have half of the data for some subjects.
The realignment, as described by Donald, corrects for between-session movement. A displacement of 1 or 2 cm along the z-axis should be no problem for preprocessing in general (apart from that you still loose the topmost part of the brain). In some instances it might be problematic though. With head-only scanners, the magnetic field becomes inhomogenous rather fast with increasing distance from the centre. Thus the occipital lobe might be deformed to a much larger extent (bent outward) in the 2nd run compared to the 1st run. In these cases rigid-body transformations (rotations, translations) are not sufficient to properly align the two runs on each other. Preprocessing the 2 runs seperately might be an option then, although it is far from perfect of course!
Having said that, go with the two sessions within one realignment step as Donald said, and then check the data whether there are any large differences between the two runs. I guess this is rather unlikely in your case, but nonetheless, better check :-)
Best,
Helmut
|