Hi David, Thanks, I didn't realise the spherical mesh also had to be in LPI orientation as well, I've fixed that.
The borders are now roughly in the correct regions after registration but they are nowhere near as good as LPR. It almost looks as the borders haven't actually been morphed - as if they've just been projected from sphere to another (which would explain my original problem). Is there a chance I've done something that has stopped the LVD morphing from working? Something that only gives me straight sphere to sphere registration? I've tried changing the parameters but it doesn't seem to affect anything (number of morphing cycles, forces, vector displacement). I've just re-downloaded the current version of Caret (64 bit version) and it's the same. I'm using OS X 10.6. Any help or suggestions about what is going on here would be greatly appreciated. Cheers, Tristan On Sat, Mar 24, 2012 at 02:04, David Van Essen <[email protected]> wrote: > Tristan, > > Registering from a left hem source to a right hem target (or vice versa) > should work for the LVD as well as the LPR algorithm. I've done it myself > on multiple occasions. > > Importantly, your source and target spec files must have the correct > hemisphere assignment. Check that 'Structure' is correct when you open > each spec file, and change it if necessary. If the hemispheres differ, > this leads to the requisite mirror-flipping, as is now explained in: > > > http://brainvis.wustl.edu/wiki/index.php/Caret:Operations/SurfaceBasedRegistration#Algorithm_for_All_Registration_Operations > * If the "structure" ("left" or "right") differs in the source and target > spec files, mirror-flip the source landmarks. > > > Another item is to verify that the left and right spheres are correctly > oriented in a dorsal view (medial wall is on the correct side, that is, if > viewing the right sphere, the medial wall should be on the left in a dorsal > view. For a left hem, it should be on the right.). > > It's puzzling that you had a problem only for LVD and not LPR, but I'm > guessing that one of the above will take care of the problem. If not, let > me know offline, as we might need to take a look at your dataset. > > David > > On Mar 23, 2012, at 1:15 AM, Tristan Chaplin wrote: > > Hi, > > I've been registering left hemispheres to a right hemispheres using the > landmark pinned relaxation algorithm and getting reasonable results, so I > presume process automatically mirrors hemispheres so they match. > > I wanted to try the new landmark vector difference algorithm. Using > similar settings the registration completes but all the borders are not > matched and are in completely the wrong spot - e.g. medial wall is on the > lateral side and appear to be upside down. Is there a chance it's not > mirroring? Or have I just not got the settings right? > > Thanks, > Tristan > _______________________________________________ > caret-users mailing list > [email protected] > http://brainvis.wustl.edu/mailman/listinfo/caret-users > > > > _______________________________________________ > caret-users mailing list > [email protected] > http://brainvis.wustl.edu/mailman/listinfo/caret-users > >
_______________________________________________ caret-users mailing list [email protected] http://brainvis.wustl.edu/mailman/listinfo/caret-users
