 |
 |
On Fri, 16 Jan 2009 22:01:12 +0000, <John> <Gelburg> <[log in to unmask]> wrote: >Hi, > >Attached two spm2.ps files with different head movements patterns: slow >drift in the session and abrupt head movement between sessions. > >A couple of questions: >1. In general, is movement between sessions preferable than during the >session and it can be better corrected in realignment procedure? >Consequently, 2 mm drift during the session is much more harmfully than >between sessions? I think displacement (translational and rotational) is more harmful within a session than between sessions, for a reason having to do with the statistical analysis stage. In most fMRI experiments, all conditions occur within all sessions. So we can view a contrast as within-session, with the results then _averaged_ across sessions. Contrasts involve subtraction (within a session), and subtraction interacts with motion in a "bad" way. But averaging uses addition, which is in a way a "smoothing" operation. So, for example, if you take two images, one of which is slightly displaced from the other, and subtract, you'll pick up things like edges. On the other hand, if you add, you'll kind of blur the image. _Apart_ from this statistical point, and only looking at the problem of motion correction itself, it's not clear to me which is worse. One difficulty with EPI is image distortion: the distortion (as far as I know) depends on position in the field. So large movements (perhaps even purely rotational) will also modify the shape of the brain in the image, which can only be dealt with by "unwarp" procedures. In that sense, more movement is probably worse regardless of whether it's between- or within-session. So I don't think you can say that movement between sessions is easier to correct. Certainly there's nothing in the algorithm in SPM that would make it so. You can only argue as I do above that motion between sessions is less harmful to the statistical design. >2. As far as I understand, what I see in spm2.ps is what was really >corrected during the realignment. Is the any tool to make visual inspection >to assess whether realignment was good? Something similar to AnalyzeMovie of >SPM99? I don't know of any; maybe someone wrote a script. >Thanks, >John >