Hi Adam,
not sure robust_register robent will work on this type of stuff (never tested it). I would recommend to keep the radius small (2,or 3), but switch on --entcorrection.
You can also try robust_register with normalized mutual information (NMI) or simple MI to see how that works.
You will probably be able to tell if a registration is completely off, but if you get something that looks similar to FLIRT it will be hard to judge which one is better.
Good luck, Martin
On 03/30/2016 03:41 PM, Adam Martersteck wrote:
Hi Freesurfer team,
I’m encountering a problem using bbregister on PET data:
-- When I attempt to register a PET volume to T1-space the asymmetric tracer uptake of the PET throws off proper registration. (Treating the PET as a T1 because of the non-specific white matter uptake > grey matter signal).
-- You can really see the bbregister mis-registration in this gif here: https://goo.gl/iM6yrT
-- It (expectedly) attempts to find the best match between signal intensities on both the left and right surfaces, but is inherently biased by the asymmetric uptake shifts the entire volume to the right.
-- Are there any suggestions for dealing with asymmetric scans like this? Stick with just a simple FLIRT? It doesn't seem to do that great a job for everyone.
Cutting left and right hemispheres apart and register separately?
Using Martin’s new robust register for multi-modal data? Any suggestions on the parameters of ROBENT?
Thanks, Adam
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer