 |
 |
You can indeed constrain the registration to be rigid. You would do this by setting one of the regularisation parameters for the diffeomorphic registration to Inf. This option was included so that I could generate averages of rigidly aligned scans for the evaluations using the OASIS dataset. One thing I haven't checked though is what happens if you disable the estimation of the nonlinear deformations, but ask the algorithm to write them out (or the Jacobians etc). This may crash. Best regards, -John On 23 January 2014 16:30, H. Nebl <[log in to unmask]> wrote: > Dear experts, > > > I'm working on a data set in which two anatomical volumes were acquired on > the same day. Previously I would have went with realign to create a mean, > which I then would have forwarded to segmentation. > > Now I would like to try the longitudinal registration in SPM12b. It seems > the two volumes undergo bias correction independently (?) before an average > volume is created, which seems to be preferable to me. However, I'm not > sure how to limit the registration to proper rigid transformations (can > this be specified with warping regularisation parameters, and if so, which > values? Should bias regularisation be kept with the default 1,000,000?). > > I would then like to segment the average volume. Should I adjust the > parameters for bias regularaisation? > > > Thank you! > > Helmut >