I'm not sure, but probably not since -applyxfm with no -init argument forces an identity matrix to be used for the transform. If you know both voxel resolutions, you could possibly create your own transformation matrix that upsamples the 45-slice subjects to 47 slices. Then you could do (trilinear interpolation is the default): flirt -ref subject1-47z -in subject2-45z -out subject2-47z -applyxfm -init mytransform.mat You could create the transformation matrix by hand (avscale can help) or use the GUI tool called nudge. On Tuesday, October 11, 2011 11:11:46 you wrote: > Hi Ben, > > thanks for your suggestion. Actually the scanning protocol had been changed > after a few subjects, so the voxel size is slightly different as well. > Would this approach still be acceptable? > > Thanks, > > Torsten