> This message is both for SPM and MEDx users. > > We have a study for which we used MEDx implementation of AIR for > realignment (rather than SPM'96 realign module), and now we are trying to > replicate the 2-step realignment in MEDx. Here are my questions: > > 1) In SPM'96, when the 1st realignment is done, the images normally do not > get resliced, and only the mean image is produced. I presume that the mean > image must reflect realignment of Scan 2 to Scan 1 even though rScan2 is > not produced. Am I correct? Yes. > > 2) Because the 1st realignment will create a Scan2.mat file, the 2nd > realignment will see it and use it when realigning Scan 1 and Scan 2 to > the mean. But the mean is already very close to Scan 1 and "realigned > Scan 2". What would change if I delete Scan2.mat file and realign original > Scan 1 and Scan 2 to the mean? The starting estimates for the second realignment will not be as close. In SPM99, this would make very little difference because the realignment runs until a stopping criterion is reached, and should only result in the second realignment taking slightly longer. In SPM96 it makes a slight difference bacause the realignment only runs for a fixed number of iterations. Regards, -John %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%