Hi,

just wanted to mention that the --cost NMI (normalized mutual information) and MI implementations in mri_robust_register are the same as spm_coreg and now mri_coreg (if the -nosym flag is specified in robust register, else it is registering in a symmetric midspace instead into the target space). If you have cases where robust_register fails, but the other tools work, let me know, because that could indicate a bug in my code.

Thanks, Martin

On 05/16/2016 03:19 PM, Douglas N Greve wrote:

It is my implementation of the spm_coreg program written by John Ashburner using normalized mutual information. I liked the spm program so much that I put the effort into replicating its behavior in C. My guess is that BBR was failing because the initialization was failing. Still I like mri_coreg for PET data because the blurriness can throw BBR off a bit. What kind of PET data do you have?

On 05/16/2016 03:06 PM, Elijah Mak wrote:

...

Hi Doug,

I just want to say that mri_coreg worked tremendously well for the registration of some very blurry PET-PK data. I was not getting any success with bbregister and init-fsl or mri_robust_register options.

How does this technique differ from the boundary-based registration?

Thanks a lot.

Best Wishes, Elijah

---

Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer