The problems with ordinary mutual information registration usually relate to intensity inhomogeneities (bias fields) in the data. In principal, addressing this would be relatively trivial, but I haven't had a chance to work on including the appropriate bias correction (based on the joint intensity histograms) into the mutual information registration procedure.
There is a trivial workaround: first bias correct the images (and possibly skull strip them) via the segmentation function. Mutual information then usually works much better.
Best regards,
John
"H. Nebl" <[log in to unmask]> wrote:
>Dear everyone,
>
>Are there any plans for the nearer future to implement coregistration algorithms similar to the boundary-based registration in FSL (Greve & Fischl, 2009, Neuroimage, https://dx.doi.org/10.1016/j.neuroimage.2009.06.060 ) or the local Pearson correlation cost function in AFNI (Saad et al., 2009, Neuroimage, https://dx.doi.org/10.1016/j.neuroimage.2008.09.037 ), or some other "more advanced" and possibly more accurate algorithms?
>
>Best
>
>Helmut
|