 |
 |
Dear Chris, when you say it doesn't look great, specifically what is the problem? Is it the final PET images or is it the structural MRIs after registration to the MNI? What one typically sees when registering atrophied brains with a 10mm warp resolution is that the resolution isn't quite enough and the widened sulci aren't "closed up" properly. It sounds like you are seeing something different? Jesper On 5 Jun 2013, at 08:50, Chris Leatherday wrote: > Hello, > > I've been working on warping a group of elderly (Healthy control and Alzheimer's disease) brains to MNI space using FSL. I've got a T1 weighted MRI and an FDG-PET image for each subject, and I've been registering the data in accordance with the FNIRT user guide > > So my process has been: > > Use BET to extract the brain from the subject MRI > > Use FLIRT with dof 12 to register the brain extracted subject MRI to the 2mm brain extracted MNI template > > Use FNIRT in conjunction with the affine transform from the previous step to calculate a nonlinear warp to the 2mm MNI template. I was also using the config file T1_2_MNI152_2mm.cnf for FNIRT, but with no inmask specified. > > Use FLIRT with dof 6 to register the PET to the subject MRI > > Use ApplyWarp and the output from FNIRT to register the PET to MNI space. > > This works pretty well in most cases. When the reigstration didn't look great, I manually edited the brain masks from BET (in FSLview) so that they were more accurate, which improved the registration. I have also tried warping the betted subject brain image to the betted brain template, which again works pretty well for some of the 'troublesome' images. But there are still some that don't look great. > > I've read in a past post that lowering the warpres value for fnirt can improve the registration accuracy, particularly for atrophied and elderly brains. Is there anything else I can/should try to improve my registration? The problem is pretty much always with the grey matter, the subcortical structures register pretty well. The brains that won't warp well usually end up a bit squashed, particularly in the parietal lobe. > > Also, is there any reason I shouldn't register a brain extracted subject image to the brain extracted template? I figured this was okay as it's only the brain material I'm interested in anyway. > > Thank you > Chris >