 |
 |
Were you trying coregistration on a resliced EPI image? If so, then this could cause the kinds of problems you report. Try it on an un-resliced one. Other things that can improve coregistration accuracy for some data are skull stripping and bias correction of the anatomical scans. Both can be achieved via segmentation and a little bit of ImCalc. Best regards, -John On Wed, 2009-10-14 at 17:22 +0000, Jeff Browndyke wrote: > Fellow SPMers, > > I'm stuck with about 9 subjects where we are missing upper brain slices in the EPI data and sadly we didn't collect coplanar T1 at the EPI resolution (4x4x4) on these subjects. We do have a high res T1 (1x1x1) on each subject and there was minimal movement in these subjects between the high res T1 and the EPI runs. In native space using "check reg" the EPI and T1 line up quite nicely, but when I attempt to coregister (estimate) the T1 and EPI there is a huge mismatch as SPM assumes that the truncated EPI boundary at the top of the head should match up with the present T1 boundary. I tried a unified segmentation approach with the T1 data and not doing any coregistration, but this failed predictably. Is there anyway to coregister the problematic EPI and present T1 without coregistering them, if that makes sense. Ideally I want to get good warping of both to atlas space without masking the EPI. > > Thanks, > Jeff Browndyke > -- John Ashburner <[log in to unmask]>